📚┃suno-school

[fleet ivy]

Okay I give up. Sepnd many Credits for one track. Can't get rid of the singing. [Hook - Hard Rap], [Hook - Rapping], [Hook - Speaking]
The Styles does not include anything wierd. The Style worked for many tracks before.

[frozen dagger]

if you're trying to get it to work by single generation. try just recreating the part you want to be rap. recreate that section with editor, or studio, and just use a heavy rap style prompt. nothing else. you already have the audio there to act as a guide for the song, if you just want that part to rap and not sing, recreate just that part with Rap prompting only.

[fleet ivy]

But it's hard to get then perfectly 4 times the same into the track. Copy and past sounds awfull

[quiet solar]

It's unlikely to be a specific problem with what's in the Style box, but you can use the Style box to reinforce what you're asking for in the lyrics tags. As a general rule, anything in Style is more reliable than lyric prompting.

[quiet solar]

Remastering the track on 'subtle' at the end of the process should fix copy/paste issues (assuming you have Pro/Premier).

If you're on the free version, Cover can do the same - just expect you'll probably have some other changes creeping in as a result

[fleet ivy]

Oh remasting. 😮 haven't seen that. Have Pro.

[frozen dagger]

meaning there's 4 stanzas for rapping that currently are all singing? yes, you are correct. Covering it with the rap part in place tho can make it easier after it has rapping in there as well. Sounds like it's missing context from either the audio or the style prompt, which have more influence on the track's output than prompting inside the lyrics box as Kanthric said as well.

[fleet ivy]

I try now the replace in the editor. If this will also not work 😄 then I just should not make this song.

[quiet solar]

Also, if you're using a Persona, you might want to try turning it off while you work on getting the rapping part fixed, then reapply it when you're doing a final pass to fix up the song.

[fleet ivy]

Marked the Hook. Styles "Deep hard Rap", or "Deep hard Rap -Singing". It offers singing. 🙃

[quiet solar]

Could you share a link to the song and make sure you have 'Allow Remixes' turned on? I don't mind a challenge.

[fleet ivy]

https://suno.com/s/rkga0SzxYVy5kVGw

[quiet solar]

Haha...well, editing the style prompt is going to be a test of my German skills. xD

[fleet ivy]

don't learn more german. awful language.

[sweet ivy]

@quiet solar You are amazing, still here helping ppl. Getting back to my situation, I didn't realize adding No strings, No synths, was why I end up with them in the first gen. On the cover using that gen, if I enter piano, acoustic gtr, soft brushed drums, subtle strings. Warm and natural production., that will remove all the synths, the big drums & electric gtr fills?

[quiet solar]

Possibly. The problem is that Suno will 'hear' them as part of the audio prompt, and the fact you don't mention them in Styles doesn't automatically mean it'll exclude them. Obviously, a tricky balance to strike between wanting the new cover to be faithful to the first gen, yet not so faithful it carries across the instruments.

You might want to try using that modified prompt on the original instead?

[finite folio]

yes but moustly Yodeling still are Major Key. example Celtic vs. Nordic: Why AI Gets the Tone Wrong
AI models often confuse Celtic and Nordic folk, defaulting to a dark, minor-key "Viking" atmosphere. Here's the main difference:
Celtic (Central/Western Europe): Inspired by fairies (Aos Sí). It's characterized by light, speed, and joy. Musically, it should be high BPM, rhythmic, and strictly in a major key (mixolydian). Think faun: uplifting and captivating.
And, Nordic/Viking (Northern Europe): Inspired by trolls and giants. It's characterized by rock, heaviness, and cold and harsh climates and long winters and dark, dark nights. Musically, it's low BPM, atmospheric, and in a minor key (Dorian). Think Wardruna for example : dark and gloomy.
I think that when it comes to Suno is more creative on the basis of words and makes major, then he ties them together and makes more minors, but wouldn't it be more correct to distinguish two cultural bases, for example, that Celtic metal is more major and Nordic metal is more minor?

[sweet ivy]

@quiet solar That makes sense. On the 1st gen, there was a part where the background got a little funky. On the piano-vocal audio track I uploaded. I had used quantize to square up the piano. I thought I read somewhere, it might help Suno to clearly see where the bars begin & end. where chord changes occur. Was that a mistake, does quantizing hinder Suno, rather than help

I'm playing to a click, so it's pretty close.

Maybe I should redo the piano track

[wintry saddle]

kulning is in scala tone which isn't strictly minor, but equates to...

[quiet solar]

@sweet ivy I've not tried using quantize. Suno is pretty good at adapting what it hears to the tempo it's working with, so it shouldn't be a necessary step unless the piano was audibly out of sync with the vocals. Equally, I'd be surprised if quantizing led to significant problems in the background unless you could hear them in the quantized version, since Suno is only using it as an audio guide.

[sweet ivy]

@quiet solar No the vocal I put down lay perfectly on the track, the piano was smooth, especially after quantizing. Very simple, mostly block chords on the quarter with some eights every now and then. No melody on the right hand or anything fancy. On the note that was wrong in both versions of the 1st gen, I might have been a little flat, will check and tighten that up

[finite folio]

But I recommend listening to one of the most well-known Estonian folk metal bands, Metsatöll (they sing in Estonian, but the music is still 90% minor, it's also a style, Estonian folk metal, which also belongs to the Nordic metal style).

[frozen dagger]

Found this while surfing around. You can see the guy that is in all the Suno official videos seated in the back. I thought I'd share the link here in "Suno School". Using real instruments, Suno Studio, and other production apps to make music.
https://youtu.be/TL5j2K8k0o4?si=WMUtfILV92HRbMBH

[quiet solar]

Wow...that's like pulling teeth.

I'm guessing that because the rapper is somewhat melodic in his delivery to start with in the verse section, it's priming Suno to have him be even more so in the Hook section. That was really hard to steer away from. Closest I was able to get in 20 separate attempts trying a few different things was this, which you're welcome to use for further experimentation and editing:

https://suno.com/s/2Rrl2iOGcyGKYgnE

[fleet ivy]

It's still somehow singing. Mouth with 100 teeth. But I do my best to bring it into the right form.

Thank you very much

[quiet solar]

Yeah, I think there's just too many influences pushing it that way - the melodic delivery in the verse, the girl singing over the top, the instruments being more soaring and melodic behind it, there not really being enough lyrics for the amount of space it takes up musically (so Suno wants to draw out words, which doesn't play nicely with rap)...

Maybe someone more familiar working with that genre in Suno might have other, better ideas for getting around a problem like this.

[fleet ivy]

I keep my eyes open for skilled HipHop Producer. Anyway thank you for all your input

[royal rampart]

Has anyone managed to remaster songs made in V2/V3/V3.5 and V4?

These are songs that usually have artifacts, but I haven’t been able to find a workflow that helps achieve an almost identical generation with good quality.

[buoyant bone]

You know what else is interesting ?
the 4 tracks that generated nearly perfectly (which also extended the original sound file too far and added wild variations in the same output) never generated a thumbnail like the rest of them.
Its very strange.

All i did was press generate new and nothing else.
No prompt change, no slider change. but 4 generations were very different from the rest.
See, they have no image generated at all

[frozen dagger]

"It do be like that sometimes." The image generation been bugging out for several days now. Feels (imo) internet traffic has a big effect on outputs as well.

[buoyant bone]

its very interesting how its linked to 4 generations being very very good while everything else that was before and after were complete trash

i hope that bug will trigger more often

[foggy solstice]

Hi. Can anyone help me? Every single track I am making has hallucinations. Including the extended tracks that I have done to replace said hallucinations. The voice is fine and around the end, it starts speaking/singing as if it didn't recognize the language of the text (it's English, btw). 🙁

[quiet solar]

Hmm. Sounds like a pretty unusual glitch. Can you share a link to the song?

[rain lance]

Has anyone ever figured out how to stop the singer from going high pitched and singing the last words of lines for 10 years? I've tried blocking so many things like sustain and portamento, lowering weirdness, none of it ever works. I'm using an audio upload that does not do that at all. This has always been one of my biggest issues I can't seem to fix.

Getting the spanish accent to work was easy enough on it's own, but using the persona really hit it home, thanks~

[foggy solstice]

That's an example (from 0:44 seconds). It's an extension of a song, so it's short. But on long songs, I have the same. It's great and then suddenly, it speaks gibberish English: https://suno.com/s/BY2X1CEK6ZSMwivS

[frozen dagger]

better if we can hear and see examples, if possible. Negative prompting is a last resort for me, but try putting melisma in the exclude prompt in advanced options.
Positive prompting, phrasing what you do want, is, imo, almost always better results.
Difficult to prompt for it. More curious what the prompting is that may be making it do that, since you mentioned the audio it's covering doesn't do the higher pitched sustained singing at the end.
Could try "through-composed vocal phrasings" or even "dry vocals" things that lean away from the more aria styled vocal expression.
Guessing without hearing what you're hearing, seeing what you're seeing, but guessing that V5's expressive vocals and the prompt leaning towards a more vocally expressive genre is probably why it's trying to steal the show with the vocals.

[earnest coral]

i've had some success with style prompts like "main vocals by Luke combs, chorus vocals by a Megan Moroney and Ella Langley duet, dark country with elements of Spanish guitar" but how can i try to distinguish better

[frozen dagger]

surprised the name drops made it past the prompt filter. but if you want to take those and make them more descriptive just use the magic wand in Suno. Take the part of "Luke Comb Vocals" and place that in the style prompt box and click the magic wand and let Suno turn it into a more descriptive prompt. (Using artists names don't always work due to blacklist filter, try song name or similar ballpark if it doesn't change anything when you click the wand).
Copy and paste that descript and keep it off to the side. Repeat the method and place the other parts in there to turn the artist you want as influence into a descriptive style prompt that (hopefully) aims towards their style. Add them together for a prompt that mixes those vocal styles etc.

[frozen dagger]

if youre looking for distinguish better as in Sunoknowing which lines of the song are sang by which voice. There is workflows in the discord for Duets. Try searching in there. (CTRL + F)

[earnest coral]

some do, but like Morgan Wallen gets changed ro bro-country

?

[rain lance]

Woodhouse won't let me type my example because it's too long. HOW DO YOU THINK ABOUT THE SONG, WOODHOUSE?

[frozen dagger]

i didn't get to see what it was you posted before it flashed, or Woodhouse response it all happen so quick, send it in a DM if its directed at me @rain lance

or link it?

[frozen dagger]

there u go big brain

[rain lance]

Never let bots get the upper hand

[earnest coral]

https://suno.com/s/b6Sf22SuNV3Jhfrt see the style tags

song is about my late wife

i'ev also been turing some of my favorite industril/goth songs into country songs 😉

[frozen dagger]

has a chorus tag on it? if so, remove that to start.
If it's a song where it was ok to start with but then started going wild with the fermata, then is clipping the vocals from the part of the song that works and turning that into a Persona to finish the rest of the song an option?

[rain lance]

I created a persona of one and planned to cover the other, I can do the clipping. It does have a chorus tag as it's the final chorus, I want it to match the other 2 but this has different lyrics. I'm okay with it having a grander sound but I don't want it trailing off. It's even missing a bunch of the lyrics in a lot of generations. It's always with the ends of the songs lately. The first versus and choruses are usually great

[earnest coral]

i tried to recoird myself singing, and it would not let me use it as a persona

not sure what else to do

any suggestions so that i can make songs with my voice?

[frozen dagger]

I've been in practice lately of not using tags for song structure. Mostly...
Yeah, there has been several odd behaviors lately. Either time for a refresh or v5.5? >.> lol

[frozen dagger]

Not currently an all-in-house way to use your voice with Suno as a Persona or ai voice.

[earnest coral]

ok, great

[frozen dagger]

You can use your voice to create instruments, or sing and add them to music, create songs with your voice or singing, but there are limits on all of that, with publishing to Suno, making Persona's and more.

[earnest coral]

ugh

grok was suggesting somethings, but it wanst working

[quiet solar]

That's...interesting. While Suno is using the lyrics, it's absolutely mangling the pronunciation.

I'm about 99% sure it's because your Style prompt is generating weirdness because Suno is struggling to reference it amongst musical sources. It's worth remembering Suno isn't an LLM, so throwing in metaphors about goblin parties is going to have...unexpected side-effects.

As an example, cleaning the style prompt and lyric tags in this version instantly fixed the problem: https://suno.com/s/EqXf6iK7jgRzTSIU

[quiet solar]

Yeah, just to repeat, Suno has deliberately disabled making personas from anything linked to uploaded audio to prevent people trying to creatively voice clone established artists. A side-effect of that is that there's no way to make a persona inside Suno using your own voice.

If you extract the vocal stem (requires Pro/Premier), you could try putting it through a separate AI voice replacer, or just record your own actual singing, and then reconnect it to the instruments in Suno Studio (requires Premier) or in an alternative DAW / audio editor of your choice.

[foggy solstice]

Thank you for the explanation. I'll keep it more simple 🙂

[dusky stone]

hello guys, do you do something after exported the final WAV (eg remastering in Premiere or Audition, like Hard limiter for Youtube, etc)? Is that really needed?

[rain lance]

I've create an example for you to hear if you're still interested. https://suno.com/s/rVc3zThjliV0H8Iv

[quiet solar]

That's entirely up to you. I'd say remastering is less necessary in more recent models, and if you're only working with a single download with all the stems combined then there's a limit to what you can really do, but it depends very much on the song and what you want from it.

If you're asking if mastering can make improvements? Yes, it absolutely can. Is the payoff worth the additional time and effort? Only you can answer that for yourself. 😉

[quiet solar]

In that specific example, you've got three words in the lyrics and no song/instrumental structure, so for Suno to make a song rather than a <30 second sample, it's making those three words do a lot of heavy lifting; drawing them out to slightly ridiculous levels and slipping in a repeat.

I presume this was created purely to demonstrate what Suno is doing that you'd prefer it didn't, but if you're looking for specific advice it would be better to see the song you're trying to fix and the specific problem in that context.

FWIW, I'm not aware of any reliable way to steer Suno to specifically not do something in general when it comes to taking on board genre influences and trying to fit lyrics to the available space, so unless you prompt it to specifically do something else which wouldn't involve a sustain (e.g. [Spoken Voice]) which might involve struggling against Suno's preference to stick to genre norms and what it's learned makes for 'good' songwriting musically over your prompted suggestions, then potential solutions for this are likely to be on a case by case basis.

[slender glacier]

Trying to keep a Spoken part as female voice and male voice everywhere else, but it keeps mixing them, is there a special prompt or something?

[dusky stone]

Thanks, I agree that most of the outcome are good from Suno. My only concern is to verify the -14 hard limiter required by Youtube for example

[raven turtle]

I never use Suno alone and mostly record, mix and master my own stuff in the box in Logic Pro. I tend to master my dark electropop stuff to about -10LUFS with about -1 true peak, which translates to YouTube absolutely fine.
The true peak headroom and clipping the peaks in the master is very useful for pushing loudness without LUFS

[drifting viper]

Quick Question guys. First off hello and thanks for any support in advance. I have writen my own songs some in 2006, 2007, and recently so say 2022. I have uploaded my track with my singing and it has supprised the shit out of me on how its produced a version of the song. Excellent. However i didnt pay attention to the lyrics that it tried to convert to text and some were missed out but i loved the version it created and havent been able to create another one similar since. I tried studio and tried changing the word blue for the but it doesnt do this perfectly when i preview it before inserting it it skips strange. Any fixes or am i doing something wrong.

[polar spade]

Only using Pro? Its good to mention Pro or Premium, or even Basic, often having different tools for the same job or a combo maneuver.

[drifting viper]

yeah i tried pro and thats where the i changed the word from blue to the and it didnt like it

[polar spade]

I had replied to bloodmage. But no worries

[drifting viper]

ah soz

discord i cant get my head around age dont help lol

[raven turtle]

You could try dong a mash up using the same Song twice and adding the correct lyrics to the prompt?

[drifting viper]

yeah its asking for a secondary song as well ill have to try something. I dont like the fact that once you have a song created you cant keep that one and change the lyrics using that same one that is great.

[raven turtle]

You can drag in the same song to the second slot

[drifting viper]

ok

[polar spade]

I use New Edit UI for my lyric changes often.

But in studio, you can add a vocal track , using a full stanza to see if it just sings it all correctly?

Thats just one means. There are about 5 good ways to do this. Even Covering the segment as an extension or covering the whole song, can give you a segment with the aame or quite close melody to the edit into the vocal track.

[drifting viper]

ok so pop into studio first right?

[polar spade]

Use the reply button , right click on their comment, so the threads are chained properly and with a reference.

[drifting viper]

ok

[polar spade]

New Edit UI is a separate mode. Found under the menu , then "Open in Editor". If you've not used, its a good practice to at least try to work out the tools of the Edit interfaces. They can be used hand in hand while using Studio.

[drifting viper]

ok ill hunt for that im sure a pic popped up and disappeared here

[polar spade]

But sure. Just using Studio alone and their replace feature is fine. Did you need a video?

[drifting viper]

yeah sure sorry im not sure why i am not getting this...

[slender glacier]

Sorry using pro

[drifting viper]

im in there now

[polar spade]

When there are a few ways, its good for one helper to just talk about one way only. Hahah... so OK. We'll do Studio only then. Ill go make a video.

[drifting viper]

How its sang is great, just needed to change some words and add some

[polar spade]

Right, you just type in the lyrics as needed. working on vid now.

[drifting viper]

yeah im kinda stuck i got as far as changing the lyrics on 1 section by highlighting it all good there. But saying i must have at least 26 seconds in fixed which its 36 seconds long. its also changing how it was sang as well as the melody which i havent asked it to do

[polar spade]

OK. sounds like you are moving along anyway. But check the videos. The second ended at 5 minutes but with enough info to help I believe.

[drifting viper]

where do i find the video

[polar spade]

coming up. Grab a soda while waiting. make some popcorn. get a haircut. read a book... about 60% uploaded so far

[drifting viper]

lol

[hard beacon]

This great service is brought to you by Dasvon. ©2026 all rights reserverd

😄 😄

[drifting viper]

ok that helps sooo much more

[polar spade]

So That should get you enough in steps to make some replacement tracks in Studio with new lyrics. Always depends on the style and tempo and honestly the lyrics and their pronunciation, so exchanging more lyrics for less can be an issue, when you're squeezing in more lyrics than you had, but replacing less lyrics in place of more, then it shouldn't be as bad -- expect the melody might change as you've now completely changed the lyrics and the metering.

[drifting viper]

yeah i appreciate this i will doa number of versions and see what i can do.. What confused me was where you type or paste the lyrics where it says lyrics lol i didnt know u could paste there

[polar spade]

Glad it helps. I should buy a video recording program... or learn to make 4 minute videos. haha

[drifting viper]

well explained i apprecaite it alot ty

[polar spade]

Im trying to do more live so we can see the mistakes as well.

[drifting viper]

that will be handy! snag with this is every mistake costs coins lol

[polar spade]

Good point. Thats encouraging. Make videos including mistakes to avoid.

[drifting viper]

yep absolutely..alot of the work is in the bin its a throw away coin burner

[polar spade]

CC+ 2026 (creative common + ) 😆

[hard beacon]

saved

😄

[polar spade]

You also saw me going into editor. When you have time, Use editor and play around with the different options to get a feel of it. It can come in handy as Studio is "limited" per se, as you saw, I didint choose the full area I needed to contain the lyrics I wanted to replace -- using Studio. The Editor UI is more accurate in that regard. So I use them hand in hand to do editing of lyrics.

[drifting viper]

yeah i will look into that if i cant crack this, its still putting in the melody, i think what i am looking at doing is keeping the style how it was sang the same just add the odd word here and there and remove the words that are wrong. So keep everything as is but just change 2 -3 words that is it, but this doesnt like to do that

Its actually creating another cover which is what you can do in the library by creating a cover of cover its not different, just need to find a way of isolating the vocals only and change the words as i dont want how it was sang changed

[polar spade]

@drifting viper In editor, we get a Edit in Studio which can throw the edited clip into a new Studio field. We will have to choose the Project we were working on, and then just drag the edited clip into a new track. Another video, I fear needed on where to find that Project locator... haha...
Well, thats another time.

[drifting viper]

ill try that now

[rain lance]

Update: Created a new persona by cutting the end off with the editor, no luck. Dled, edited and reuploaded, covered, turned that into a persona, still no dice. It's just doing the thing where it hardcore remembers what it already has planned regardless if I have it in any input. Even turning weirdness way up, which changes a lot of the song, maintains the huge dragging out of vocals at the end. I just want the poor dude to sound human, but the machine insists on flexing its superhuman capabilities.

I've also messed with in-lyric prompting a bunch and no meaningful change.

[drifting viper]

I mean the apps fantastic at what it does, but i just feel at the moment you can do the same thing 3 different ways and all costs. But yea give me just the vocal on its own and how it was sang, and change the lyrics in the same style and bingo but its not

[polar spade]

OK. cool. Good luck with it. Ask again in here if things get frustrated. Expect the lyrics field to get wonky with each edit.

Pro tip: have a copy of the lyrics pasted somewhere in a notepad or I even stick it in an email frame. Everytime Suno worbles the lyrics from exhaustion (they will look funny, double-up, overlap, miss some lines, add some lines, etc) , I just highlight them and paste fresh lyrics back in before saving the edit.

@drifting viper said enough. Start with what you got. Experiment. Have fun. Burn credits!

[drifting viper]

lol

[polar spade]

I now have no issue to then STEM OUT the NEW VOCAL track cover. Yes, more credits, but way easier to then get a solo stem of the new vocal cover we like.

Caveate: Stemming out again the vocals stems with some bits of music in them, we may still get a bit of instruments show up, I've often done it twice to get 4 versions of which one usually is a clean vocals only. Then, if needed, right-click on that new stemmed track and find "dry vocals" to create a dry clean vocal to work with similar to a DAW.

Point: Studio has no tools to add reverb to a dry track, so we can play with doubling or tripling tracks, panning and adding EQ to fake a reverb --- but again, another day, another video.

[drifting viper]

is that in editor or in studio

[polar spade]

You do that separately when all is done... but Studio is fine.

Just choose the new created Cover/Replace track you want, go to the right Clip panel , scroll down and click extract stems for THAT one new track.

You can just use vocal+instrumental and then drag over the vocal stem only and 🤞 cross fingers it doesnt still stick in some instrument or have bleedover artifacts.

[drifting viper]

ok ill check it out just burning my money throwing it on the fire lol

[polar spade]

Yea... new learning curves will do that.

[drifting viper]

One lesson here is to ensure every lyric is right before commiting to your first cover, cause if its wrong you will never get it exactly the same as the original man damn

[polar spade]

My reminder lesson today to myself is always do Foley sound efx last , after all edits and splices are done.

[drifting viper]

i havent entered that realm yet i need robot voices saying shield up, phasers anything hjere that can do that

[polar spade]

sure. Try vocoder or vocaloid in prompts and try to create one yourself. otherwise, see if the Sounds will make one for you by requesting just that "robot voice saying 'shields up;"

https://suno.com/s/PJdvWMwwEfJGS718 decent one on first generation

[drifting viper]

hehe will check that out also ty

[polar spade]

and going old school and prompting for it, we get ( a more human sounding one)..... (and would need to edit out the lines)
https://suno.com/s/WOF5PsSGetysPe8L

[drifting viper]

Suno just took a dump on my song. It was perfectly what i wanted and then it skipped and totally went into a different song> lol

[drifting viper]

Would u happen to know any good AI that would do Videos 3-5 mins worth?

[polar spade]

Nowadays, I think we need to be specific about what Video means... You can look at those AI sites that combine 20 models and you pay a fee and pick and choose such as 11labs now has introduced/added to their repetoire .. I know of Openart.ai as another, about 20 more I think.

So then we have the older Canva that does videos to a point, as well.

Even we can find someone on Fiverr who can do one for $40 that we might like.

In the end, my answer is "no" not really. As its more of a read and research kind of thing to find your taste.

[drifting viper]

yeah fair comment! its tricky I will check fiverr

[quiet solar]

No magic prompt I'm afraid. There's some best practice ways to prompt for multiple vocalists, but it will be reliant on a certain degree of luck in whether Suno performs it as you want - or you'll need to use workarounds and/or do it through editing.

This was written for singing duets rather than singing + spoken duets, but the principles are all the same, if it helps: #1466598465969590465 message

[median harness]

@stuck junco Tried to do it as briefly as possible. It all started over two weeks ago. I created a song and the cover art seemed strange to me. It literally had white noise in the image. I thought it was incredibly conceptual, but subsequent cover generations were clearly worse than before. I spent another 200 points and realized something was wrong with the generation. I contacted support, and your bot said that it must be some rule violations which ruined generations.
Then I called my friend and ask him to generate the exact song on his account and it was ok.
I deleted the song from which it all began and all its covers. In trash folder too.
Welp, my tracks contain mature words it was ok like for a year, but nothing racist, suicidal, no names of other artist or promts like “do as lady gaga does”.

I was doing happy-themed for 4 days straight and nothing happened. These songs were still corrupted.
I wrote to support and said that it doesn’t get better and the bot said that my account isn’t flagged because I have no such email. Then it said to wait for a human help, it was 10 days ago and no one appeared.

My account link: https://suno.com/@adamovladislav659

These are "positive" variants, because I deleted "bad" songs.
The song examples:
https://suno.com/s/VnTzuPMPYvfMrYPX
https://suno.com/s/Rzomk2JnheH9Zdf1
https://suno.com/s/4fKUMidiXDlVFsCt

[frozen dagger]

The Persona didn't help even when trying to create on an instrumental stem of the part you want (remove the vocals already there so they don't try to copy)? Sorry you haven't had luck with getting results you want, yet.
Could try a different model and see if can get better results, can cover with v5 if you want. Once it has the singing in place the way you like it. Or maybe you've tried that already too?

[frozen dagger]

That is an absolutely wonderful trick.

[polar spade]

Once I find a good screenvideo maker I may make some how-to's on the "things we should know" of anything that takes less than 4 minutes long, rather than the tricks and hacks .

[frozen dagger]

"Expressive Vocals" and all the energetic prompting is naturally going to be more, well... expressive. If the song is good and just the one part want sang without all the melisma then when you recreate that part, or extend, or whatever method, going to want to omit that type of wording.

[frozen dagger]

OBS?

[polar spade]

Sure. I havent looked at any to know.

[frozen dagger]

That "trick" tho, I never would have discovered that. But that looks like it would be great when I'm crafting instrumentals to fuse styles. Tagging the sections and bringing into studio with that method could break the sections with little effort to swap around and work with much easier.
OBS, open broadcast software, free open source may be a good option.

[polar spade]

" ... OBS, open broadcast software, free open source may be a good option...
Thanks. I'll look into it.

[frozen dagger]

I know I'm not the only one looking forward to seeing more videos from you.

[polar spade]

As you see in the second video though, the tags are not perfect yet, just by tagging without checking alignment. I just did that on the fly, so two words "in love" were noted as part of the Verse 2 but were over in pink section leading in to the next segment. Suno blended it correctly anyway, so no sweat. But just have to check those things in Edit first before going back to Studio.

Speaking of.... dont change the first couple of word or last couple of words in the lyrics we are editing, so that Suno still has no issues blending if it is out of alignment.

But yea, I've often gone to Edit first before Studio just to have segmented sections.

[rain lance]

Yeah tried it all. Each new song is costing more and more credits to get to where I want it, and it's entirely been the last third of the song that completely switch things up. At least the singer isn't changing into another person this time, but he's acting so different. And I tried with different prompts too. It's hardstuck on this wild sound, no matter what I do unless I change it up entirely by starting over. The spanish speaking persona did it first at the end, now it's forever in the song if anything is based off that start, unfortunately.

[unreal sleet]

Okay I'm getting really sick of every single song that sounds good have word slip-ups and can't pronounce some, so I try different words etc. I have tried to create a new workspace as suggested but it did nothing. Please someone tell me what I can do before I pull my hair out. Wasted hundreds of credits to gibberish as if it no longer understands english

And editing to make the word clear just creates background noise or something off-level that doesnt fit the section of the song

Im not asking for much, just pronounce the damn words

illusion = illusor
negotiate = negodiade
denial = deniar
everything = everybeen
everything = everythring
noise = noir

[stuck junco]

Hey daLenochka

First and foremost, thanks for reaching out with all this context. Super helpful. Sorry you've been bumping into issues.

You may not have links to where these issues happened since you've deleted them, but if you bump into this again, please reach back here or to me directly via DM with a song link example and I'd be happy to take a look.

Looking forward to figuring this out!

[placid raft]

Hoping for a bit of advice myself. Been using SUNO for nearly 2 years now, but this is the first time it's generated a track which just gets louder and louder on its own, to the point that the waveform is maxed out and the actual audio quality suffers as a result (see the image). I also generated it in 6/8, but that section when I try to replace it just rushed through it and seems to force 4/4 again.

So my questions are:
1: How did this even happen?
2: Would remastering help here at all?
3: If I have to replace it, how do I get the editor to recognise the original 6/8 format? (I'm a pro user, not premier)

[stuck junco]

Hey Rav3n. Thanks for reaching out about this. Do you have a song link for this by chance? Looking forward to seeing what's going on here

[placid raft]

right here (obviously not a final version xD): https://suno.com/s/y8JQniCRlp8pNNB4

it starts getting weird from around Verse 3, but especially noticeable from Chorus 3

[stuck junco]

Thanks for sharing that! Talked with @unique stone (s/o to them btw for the assist) about this and they provided a few good suggestions:

- possibly remaster will help (there's a chance it might, at normal/high strength)
- try extending, outside of the editor, from a point before the start of the problem

_ _
Hopefully this helps!

[wintry saddle]

This used to be a common issue, but has been resolved a lot since v4 came out (ISH), that said it can happen... The 6/8 stuff is really hard, I think your gonna need to extend from a really clear 6/8 chunk and have a legacy super strong 6/8 persona to give suno a kick past most it's training data, have a play with all the sliders to (stronger than 50% audio influence might actually help here!)

[unreal sleet]

what is it even saying

To name the throne, an hourglass bleeds in a lightless room. not really close

[quasi mango]

Is it just me or has Suno made some unanounced changes/updates?

• Remastering using 'subtle' now makes changes to time signature and can also change the voice much more than it did before.
• Prompts seem to create a better result when it comes to following. Example: Setting one word in caps seem to finally trigger more emphasis on the voice used. - This is something I've wanted for a long time and previously I only succeeded with it three times
• Blips and glithces. I have 20 songs I'm trying to finalize. Recently there's a new issue I've noticed: Almost all of my tracks have some glitching sound in them. I've tried extracting stems to fix the sounds / missing sounds needed to fix the annoying issues. Anyone Else experiencing this? After days of testing creating covers and remasters t

[unreal sleet]

look above, ive had something like that in the past 6-700 credits

[quasi mango]

To me the vocals are good though. I'm getting artifacts or sudden loss of sound for a fraction of a second

BTW: What I have done in your case is to do one cover with slightly edited words. Example:

Illusion = illusor: I'd try: "illusionn" or even illyouzion
denial = deniar: I'd try: "deny-all"
noise = i'd try: "No ice"

The idea here is to do a second pass cover on the new version you get. This time you have the original and correct lyrics (as long as the words are actually changed).

This tactic has worked a lot for me

[quasi mango]

Another suggestion: If you never have gotten the lyrics to sound good then you should try shortening the lyrics or write them differently.

"So name the drone so glass leave the lively throne" (😂)
Just an example: If that's how the lyrics are written, try writing them like this instead

"So name — the — drone,
So glass leave the lively throne"

Just changing the arrangement with the same lyrics can help too

[unreal sleet]

its differnt words everytime thats the problem

[quasi mango]

what parameters are you using? 50/50/25?

[unreal sleet]

different every 10 gens

[quasi mango]

just randomly testing, trying to find something that works? Been there, done that. In my experience, the default is the best to get consistency.
My experience is that the higher the Audio Influence, the more fucked up the vocals become

[polar spade]

Keep in mind something Suno always does is try to find the rhyme. It seems a thing to look for rhymes to set meter or actually force rhyme if something looks close.

So since Throne and Room positions would often be an end-rhyme and they both have "o"s in them, Suno will force a long o sound in 'room" to rhyme with "throne". Until it gets better at words, that seems to be the priority.

Always good to look for these potential spots and phonetically re_write end words (such as [rume] instead of "room")that dont rhyme on paper before clicking create.

[quasi mango]

Agreed. I've found my biggest enemy to be my stubborness and unwillingness to make changes to the lyrics

[unreal sleet]

how does rume sound when its supposed to be room

[quasi mango]

I have a track that I find utterly perfect, except for one artifact that appears from nowhere.

• Remaster at 'Subtle' replicates the glitch every time. Remaster at the other levels changes the track waaay too much.
• Splitting it into stems and setting it up in Studio has not worked for me - I get new artifacts.
• Remastering every single stem that sound decent removes/changes the stems too much, even on Subtle.

• After a whole day trying new ideas, I ended up with a new track that sound worse than the original, but.. without the glitch.

Anyone have any ideas?

I hear you have the same quick glitch sound I'm talking about, in Castle In The Panic during 1:17, @unreal sleet

[unreal sleet]

ofc. glitches everywhere, at some point u just gotta accept it if its small enough lol

[quasi mango]

Maybe.

[frozen dagger]

Is this happening throughout the whole song, or mostly at a certain area, like towards the end of the song? Happens with other lyrics/songs not the same audio or lyrics, still experience same issue?

[unreal sleet]

different songs, different words everytime, no consistent pattern of words it doesnt like

sometimes it can be the easiest word

[frozen dagger]

and what about the timing? does it occur during the end? or indiscriminate?

[unreal sleet]

usually the hooks, 1 hook can be good and the other bad, usually both bad

so sometimes i can just fix it with my daw with the vocal stem

which im doing atm with 4 different edits of 1 song

[frozen dagger]

That's no bueno. The speaking in tongues usually occurred towards the end of tracks for me. It is very frustrating. I hope it works itself out soon. Wish I had some idea to help, but the only thought I have is to try doing the smaller section by itself with no vocals if it's a cover. I know it does this from gens it's creating on it's own tho. So not sure how we can avoid it.

[unreal sleet]

yeah it is what it is. idk how to get out of this

[quasi mango]

@unreal sleet Are you okay with sharing your Style description? I can try other lyrics, just want to check if the problem might be there (if my vocals gets scrambled too)

[unreal sleet]

can we do it in dm?

[quasi mango]

sure

[quiet solar]

Is it a single glitch at a single point in the song, or a recurring sound that hits a few times?

[quasi mango]

I have one song where it's recurring, but in the particular one I was talking about above it happens only once.

Do you think this can be the result of a combination of sounds getting intertwined? This is my new theory, that maybe this can be solved by adding EQ to splitted stems

[quiet solar]

Very difficult to say without listening to the song itself.

But if it's a single issue, and you have studio, splitting out the stems purely for that part (sticking with the original combined for the rest of the song), isolating the stem that it's in and then regenerating just that stem is one of the least instrusive ways you could fix it. It may also be possible to EQ it out.

[quasi mango]

The funny thing is that I didn't notice it until I mastered it using masterchannel.ai. That's when my headphones suddenly popped.
Testing with EQ now, but man that's really not my strong suit hehe

[quiet solar]

In terms of why the glitch happened in the first place, it could honestly be down to a number of things and how they interact (or don't interact as they should), or just Suno having a bad moment. I'm not sure there's a general all-purpose answer for guaranteeing a glitch-free track.

[quasi mango]

I totally see that. What I wonder, however, is why the glitch remains in every cover, in every remaster using 'subtle'

I tracked it back - it's been there 5 versions ago, lol

[quiet solar]

Most likely because while it's obviously a glitch to you, to Suno it's just an audio prompt. And if it can hear it, there's a pretty good chance it'll try to replicate it.

[quasi mango]

Yeah, that's what I find weird. It replace and add other stuff, but that particular one that only lasts one second and even spikes in the waveforms, that one it wants to replicate

[quiet solar]

This is where EQ work, or just straight up deleting that section of the stem before remastering, can be quite effective tools. But the more you want to keep the track totally unchanged, the more you're really going to find yourself pushed towards an editing fix over a regeneration fix.

[frozen dagger]

The new Clip editor in Studio with the Time stretch may be useful to squish that bug too. (For the glitching sound artifact)

[sweet ivy]

I did a true vocal match on the first gen, after taking the word bridge out & replacing it with Chorus. Tried to run another cover using that version, suggested a few things in the style prompt & Suno "had a bad moment". Even though I left the same prompt in, there were some melody changes and jumps? What kind of things does one do in a cover of a previous version ?

[quiet solar]

Yeah, it's Sod's Law. I had a random whine introduced at the start of a verse in an otherwise perfect cover, and it was clear and distinctive enough that Suno decided it was definitely intentional and should be replicated from that point forward.

I eventually managed to nudge Suno into doing a cover where it replicated the whine as a woodwind sound. And since it was the only woodwind sound on the track, at that point I could isolate the stem and just delete it out of existence. Wasn't the easiest process though. xD

[quasi mango]

Haha, that was a smart move though!

@frozen dagger I just tried that. First I tried the "squeeze" it, then I instead deleted it and dragged out the second before.

[frozen dagger]

Did the latter sound ok? That was what I was thinking, or stretch the glitch out and mask it with audio.

[quasi mango]

It worked for one of the stems I suspect cause it. There's too many sounds in the main audio I'm afraid

Here's the "glitch" at the 4 second mark

[quiet solar]

So it's worth bearing in mind that Cover expects to make variations on the original, rather than being a clone option. If you want it to clone the original closely, you can tweak the audio influence up and try to find the sweet spot between having Suno retain everything melodically while still respecting what you're asking for it to change in the style prompt.

Remaster 'subtle' is also a fairly good option to try if you're not trying to make big stylistic changes to the underlying song.

If you're looking to make small surgical changes, then using Extend to create a 'sample' stem for the fix that you can use in subsequent editing in a DAW is my go-to route. If you have Studio, that gives you another option as you can do a surgical regeneration on a specific stem using Replace while leaving everything else alone.

[frozen dagger]

Curious, haven't used the new Remove FX feature, wonder if it would be helpful tho. It's barely noticeable, you're kinda lucky lol. But once you hear it ruins the song knowing it's coming. =\

[quiet solar]

Okay, might just be me but I'm not hearing a glitch on speakers. Do you mean the echo on the vocals?

[frozen dagger]

it's a very minute hiccup as i hear it

[quasi mango]

It's after "put a smile on"

[quasi mango]

yes, better word than glitch :p

[quiet solar]

Ah yeah, I can hear it more clearly with headphones on.

[quasi mango]

Based on the stems right now I believe it might be the fx, drums, vocals and bass fighting for room at the same time, as it's not present in any stem alone

[sweet ivy]

@quiet solar Thanks, I thought doing a cover of the close version, using the same prompt, would change things. I had the audio influence pretty high - 95. I can't imagine bumping it to 98 ot 100 would do much. I did have a side-stick land off beat in the 1st bar of the 2nd verse. Not sure a regen would fix that, rather than me moving it to the 4 in my DAW. I was a little flat o the high B in the key of D and the AI singer was also a little flat.

Would there be any reason to do a cover of the 1st gen that has the melody correct

[quasi mango]

If you have a version you want to keep in every aspect, try using the default value (50/50/25) if you don't change anything anywhere in your prompts/instructions

[quiet solar]

I think it's honestly a glitch in creation - part of the issue is that the timing goes out of sync for a moment, which is why it really stands out to the ear.

This hasn't perfectly fixed it, but you might be able to use this as a replacement to then cover/remaster without reintroducing the problem.

This was done by repeating the unglitched audio to a new track and crossfading it with the next section after it had been realigned to where it should be with the glitch removed.

[quiet solar]

As a general rule, stay away from the redzones on the sliders (<15%, 85%+) as Suno can get a bit glitchy when you put it in that much of a straightjacket.

If you're using a DAW, I'd strongly recommend trying Extend to create a fix that you can import for further work rather than Cover. Cover works far better if you want to redo the entire song, not just a single section.

[sweet ivy]

@quiet solar I was watching a YouTube video on Extend, & first of all, it looks like it creates a new version of the section you selected, say the second verse. Then creates 2 options to select from. I assume if you don't insert one of them into your track, the tracks remains the same. I want to make sure using extend doesn't change my original, if I don't select the new verse, don't insert it. The new verses will be in the library & I can pull one into the DAW to work with

[quasi mango]

Oooh! Thank you!

[quasi mango]

It's definitely less present here. I'll try to use this!

[sweet ivy]

Curious, can you create an extend for just a 4 bar section or does it have to be larger, like a whole verse or chorus

[quiet solar]

Nothing you do in Studio automatically replaces what was there before, nor does it edit the original song from the Library. But if you have any lingering worries, you can do an Extend from the Library and import extensions into Studio separately. I find that makes things slightly easier to keep track of, personally.

The way Extend works is that you select a section of audio (marked 'Keep') and Suno will use this as an audio prompt. It will then continue the song starting where 'Keep' left off, based on what you put in the Lyrics box - and by default it'll tend to be more faithful to the original than Cover.

From the Library, you can use Get Full Song to attach an extension to it's 'Keep' audio and create a new combined song of the two. It's a bit more messy in Studio, but once you get used to how it works it is fairly intuitive.

What you're describing is Replace - which uses the same underlying function as Extend, but takes the audio prompt from the selection you make and recreates just that section.

[quiet solar]

There isn't a way to specify how long an extension will be - as with creating a new song, Suno will take as long as it needs to cover what's in the lyrics box. In my experience, though, it doesn't like doing super short extensions as much and will often pad them out with improvisation.

[sweet ivy]

That replace options sound like what I want, hoping a better section. I need to research that. For now I'm going to master what I have and export the stems so I have it locked down

[quiet solar]

I would agree, except that when we spoke before your experience was that the key change itself was causing problems in Cover. If that is the case, it's highly likely it'll cause problems in Replace too - although there's no harm in giving it a try.

The thinking behind using Extend from before the key change is that Suno will then have no memory of it, and you can ask it to carry on with the rest of the song. Since it's very unlikely to do an unprompted key change, that should get you around the problem.

[sweet ivy]

Oh, I passed on the key change, went back to my guide track - piano-vocal and finished in the same key

[quiet solar]

In which case, Replace sounds like what you're after. 🙂

Honestly, no need to research it. Just select a section, hit replace, and it'll present you with two new options you can use or not use as you see fit.

[sweet ivy]

@quiet solar

Suno should be paying you and a few other experienced users in here. You are the person who understand what I'm trying to do, creating a nice "demo" of an original song. As you stated, something Suno isn't really designed to do

I don't know if anyone from the Suno staff is in here. I assume they would peek in from time to time. Maybe they are already working on this, but if they created a whole new work flow for creating a great demo of an original song, it could be a goldmine. I'd be happy to pay an additional $ 20 a song if there was a simple process - upload a simple guitar-vocal, piano-vocal, select s style and Suno generates a very presentable demo following the melody almost exactly, with some flexibility for minor changes and follows the structure exactly

Could easily gross $ 20 million a day

[quiet solar]

Suno is generally quite good at taking a recording you give it and then giving it a glow-up, or converting an acoustic song into something performed by a full band (or vice versa). If anything, I find it does amazing things with audio prompts - it just takes a bit of experience getting used to steering it the way you want via prompting.

What's more troublesome is making very small, deliberate tweaks. But as you get more experienced with prompting and editing, you should start getting a feel for what might be the best way to work around any problem you run into.

[sweet ivy]

@quiet solar The prompts is clearly the secret sauce and setting the three faders correctly. I see some ppl advertising a book of prompts, wonder if any of them could help

[quiet solar]

I wouldn't give them a cent, personally.

There isn't a fixed list of prompts you can use - it's more a case of the more easily Suno can reference what you're asking for in it's training library of music, the more reliable the prompt will be. But all style prompts and lyric tags are treated more like suggestions than commands - on top is the additional factor that any [] tag in the lyrics may not be picked up by Suno, especially if it's unusual for that genre. No book of prompts will change that.

If you wanted a source of inspiration for prompts, I'd suggest taking a look at the 'tag reference' that Stayen put together on GitHub. As with all such sources, take any individual thing said with complete confidence with a pinch of salt, but it's a pretty solid set of suggestions: https://github.com/stayen/suno-reference

[sweet ivy]

@quiet solar Thanks I went there and see the list that comes up after clicking on Tag References

You talk about lyric tags, I assume tha's using a tag in the lyric section, [Intro acoustic guitar with piano accents]. I haven't put any on the lyrics I upload, just placed a prompt in the Style section.

[quiet solar]

That workflow is exactly what Cover is designed for - it just sounds like you've had a rocky experience with your first attempts.

[quiet solar]

Yeah, that's the one. Some of us unofficially refer to them as 'tags' to distinguish between a prompt in the lyrics box vs the style prompt, to avoid confusion. Tags can be used to prompt Suno to do something specific for that section, or even line of the song, or to help Suno understand the song's structure.

The Style prompt is typically better at applying what's in it to the entire song, although you can ask it to do something specific for certain sections. Style prompts are generally more influential and reliable, but lyric tags can be more precise. You can also steer the vocalist by the way in which you present and structure the lyrics themselves.

[sweet ivy]

@quiet solar I'm remastering the songs as is, will select one of the two and export. Then will try the Edit/Replace tomorrow.

Goodness, didn't like either remaster, lost a bit of the warmth in the vocal & higher frequencies popped out & I selected "subtle",

Gonna export tsem from the original

@quiet solar There's a version1 and version 2 of the stems. Thought there would only be one, a straight transfer of the stems in the track ?

[quiet solar]

Suno does everything in twos.

The stems aren't really stems. Suno doesn't actually create and save individual stems and them compile them into the final version when making a song. It creates the whole song at once as it goes along. When you 'get stems', it retrospectively tries to break the audio out into it's component pieces.

What it's given you are two slightly different attempts at that exercise - often they're very similar, but sometimes you'll find one has done a slightly cleaner job at separating what you're interested in.

[sweet ivy]

Okay I listened to the first. I assume of I don't save the stems in version 1 and click version 2, I can go back and download version 1 if I like it better, or is it gone

And I would have to select download again. I assume I get both versions for the 4 credits

[buoyant bone]

Can we get credit reversed for disliked generations ?
i have 95% dislike ratio and i ran out of monthly points + additional credit + earned points in just few hrs.
😵‍💫

I feel like i pay for work done poorly.
If i hire someone and they do crappy work, i pay them accordingly or pay someone else to redo their work.
At the very least, 50% refund for terrible resiults would be accaptable.

[quiet solar]

@sweet ivy Since you mentioned workflow, I was looking over a song that I took from acoustic to electric, and it went something like this:

4 Covers on 40% Audio Inf.
6 Covers on 40% Audio Inf. with modified prompts

Picking one of the covers as a good starting point for further iteration

4 Extends to create 'Part 2' of the song, since it was slightly too long for a single cover.

Picking one of the extends to serve as the starting point for part 2.

6 additional Extends deliberately aimed at creating samples to fix minor issues in the vocals or instruments from the two candidate picks.

Putting Part 1 and 2 into an external DAW and replacing bits with the 'fixes' from the additional extensions.

Remastering the track in the DAW (not the Suno Remaster function)

Done.

[quiet solar]

They're saved forever. You should be able to find them in Studio easily - if not, they'll be in the Library. You just have to uncheck "Hide Stems", which is on by default

Generating the stems costs credits. Downloading them doesn't, as far as I can recall.

[sweet ivy]

@quiet solar Okay. I'm downloading both versions any and will decide which to bring into my DAW. That work flow sounds very complicated, lots of editing, extends

Goodness, I listened to the keyboard track, more little errors than I thought, once a total hole, one out of tune little accent, volume drops. Gonna be a lot of tweaking. LOL

[quiet solar]

This was pre-Studio, so there wasn't a very good 'replace' option at the time.

It's pretty rare that Suno creates a perfect cover straight off the bat, and even a really great cover usually has something in there that I would ideally change - or an outright mistake that needs fixing.

This workflow is basically a balance between using regenerations early on to establish a 'master' version that's say 60% - 90% there, then switching to Extend/Replace and editing to make specific fixes. For me, that gives the right sort-of balance between amount of time spent vs amount of control I want over the final version.

I'm going in expecting an editing phase, so I don't get too stuck on covering over and over again trying to address minor issues. If the problems are small enough that I can fix it in editing, or I'm willing to compromise with what Suno has done, that's great - on to the editing phase. If the problems are too big or too numerous, then I'll do another cover.

[sweet ivy]

@quiet solar Sounds like a solid plan, Guess I'm going to have to play around with Replace, extend, edit and get good at it, it I want to produce good demos of original songs

[quiet solar]

I've also learned to expect that sometimes Suno will just have difficulties. It won't like the lyrical flow of a certain line, which will trip it up for the rest of the verse. Or it'll struggle with a bridge or verse that does something deliberately different, etc.

So again, I know going in that the early versions are partly about flushing out those issues as well as getting a feel for the sound, so I can make changes to the lyrics or structure and the prompt as needed. I don't expect to get a good 'master' version until several covers in - though I won't complain if one comes along earlier.

[sweet ivy]

I got lucky on this one, 2nd pass/gen and I had the vocal I wanted, with a few pitchy words I can fix in Melodyne

[quiet solar]

In many ways, there's nothing worse than getting really attached to an early version that happens to sound great, but is riddled with problems that make it a devil to fix.

[sweet ivy]

I don't have studio. I have 74 VSTs in my DAW and think I can work on the tracks/stems better. If there is a big difference using Suno Studio, over an external DAW, maybe I consider upgrading

De-Essers and things like that

[quiet solar]

I've only been using Studio fairly recently - prior to that I did all the editing in Audacity. Some things I will still take to Audacity, because I'm so familiar with it and the plugins I have there.

Studio has been progressing really rapidly, though. Being able to regenerate or create a specific stem is incredibly useful.

I guess what I'm saying is to expect creating a demo to be a multi-step process. Especially if you're learning Suno as you go along.

[sweet ivy]

YouTube guy Tim Makes Media has some very good videos on Suno, one on Removing Artifacts from Suno Stems I plan to rewatch.

Well it's getting late & I want to thank you again. You are so kind to spend all this time with me tonight and all the other times I came on here as a newbie.

[quiet solar]

It's what this channel is for. Best of luck with the demos!

[hoary sandal]

Just needed opinions in general, whether about prompting or lyrics
https://suno.com/song/b295e132-fdec-4022-8283-c621f3f30489

[frigid orbit]

I’d cover it with the persona

[frigid orbit]

Don’t think this is the place for a second opinion. Is there anything you want help with as a starting point? Anything not working how you expect it?

[frigid orbit]

Hey, aslong as they don’t spout odd nonsense about a golden grail prompt, any secondary point of perspective is useful. There’s a lot of methods to remove the uglies.

Depending how set you are on the song. You could rough chop and delete whole sections, and run a cover or a remaster over it to smooth over the empty spaces left behind. Small tactical replaces. Running the same track twice in mashup with a slightly altered prompt. Or running two stems that don’t have the issue in mashup with the whole prompt.

There are countless methods. Sorry if I’m repeating, or getting your issues wrong. It’s early and a little too much to scroll back.

[neon hound]

Sup, people! How would you consistently make vocals stretch out like here (https://suno.com/s/nv12MtmtoGHQXzQ8), past 3:53? Mashup made a banger, and now I would like to reproduce it without generating 40 versions

[shrewd knoll]

Hey! Anyone have any clever ways to get the solo vocal from a track as high quality as possible? Suno´s stem seperator messes up vocal quality (understandable) - i tried to make it generate vocal only tracks using a reference i gave it but doesnt work, or even generate the reference i gave it just only vocal and chords (So i can bounce .wav instead of stem seperating). Excluding "drums", "instrumental" before generating doesnt work either.

Anyone have any clever tricks/ways to get the highest possible vocal quality either with beat or solo?

All help appreciated!!

[hoary sandal]

I just sometimes wonder if my prompts work like they should...both in styles and in the lyrics themselves

Takes me about 30-45 minutes to finally get one that sounds good.

Also slightly annoyed because most generations will either leave stuff out or blur one line into another

Was instances where it started halfway through the first line of the Breakdown and went into the other half of the Outro...skipping everything in between

[frigid orbit]

It is a matter of your interpretation vs how Suno and its LLM interpreters interpret it. Prompting for me goes relatively quick, let’s say 4-6 prompt revisions. But I don’t expect Suno to fully adhere, and I don’t prompt with natural language.

Each term I add I see as a nudge and not a robotic instruction it’ll follow like a hardcoded machine.

I hope this helps more in more of a perspective way rather than giving you the golden grail of prompts (which there isn’t)

.

Like that car game. “I see what you can’t see and the colour is Stretched LFO Tuning.” And then it’s up to Suno to guess my intention, and if it fails, I didn’t give it the right hint.

[frigid orbit]

Lyrics tho and their adherence often messes up with wild Slider settings (and dependent on the tool used, think cover or mashup). When in doubt, reset your sliders and see if that helps.

Sliders imo only add more uncertainty under the guise of more control.

[fleet adder]

https://suno.com/s/0arLNpDRjiTeJ9Ye
omg nah guys this one sound like kill this love or bangbangbang kpop thing
but it taiwanese
rate this guys pls

[hard karma]

Hey I'm a country pop rock artist lookin 4 someone 2 collab with

[unreal sleet]

If you like a song but want another take on the lyrics because of errors, will making a cover of it and turning all sliders to 0, and keeping lyrics and leaving style empty should be the best?

Keep lyrics and remove the style

[nocturne flower]

Hey guys. When I did covers I would get different melodic line, these past few days I get the exact same thing as the cover over and over.
Does anyone know why this is happening?

[polar spade]

Which do you prefer? A Cover normally should be close the melody. Suno can do either sometimes. And lyrics can affect a cover as Suno tries to vary things in general but focus on NOT changing the melody.

[nocturne flower]

Yeah I got amazing results by just covering an existing song with my lyrics. I would get similar song but different melodic line

last 3-4 days I just get the same results over and over no matter what I change

[polar spade]

Yes. I got that. Which do you like better? Cover should give you similar results, ultimately, just change things like tempo and instrumentals.

[nocturne flower]

the old method ofc

I'd get different vocal lines and it would be a brand new song

[frozen dagger]

create a track that prompts for arias, melisma, singing with sustained notes, prolonged notes. Then separate the vocal and layer it in the song with Studio or a DAW.
You can use it as a Persona to recreate the effect later as well. May be able to use that section as a Persona to recreate the effect as well. Particularly if you are wanting the singer to sustain the note while remaining in the back layer as the singer continues next verse.
https://discord.com/channels/1069381916492562582/1473392311563063456
Tips on layering vocals.

[unreal sleet]

How do sliders affect covers?

[polar spade]

You can save any new or additional prompt settings by clicking the bookmark found in the lyrics box.

Copy Lyrics if they are changed or edited new.

Go out of Create panel and refresh the browser. Come back in.

Click Cover. Choose prompts from the save Library icon. Paste back in lyrics.

Style Sliders to 85% prompts if new prompting (assuming you're prompting for a newer sound or adding a persona)
Audio Slider to 4% to vary everything else.

Click create and see if that helps.

If same, clear cache, close Brower. Open. Try again.

[unreal sleet]

davson u seem to know about covering

[quiet solar]

That's a tricky one - if I understand right, you basically want it to have another crack at the lyrics but leave everything else alone?

I'd be wary of setting sliders to 0, because that tends to introduce problems. I'd try something more like 15/15/60-85. I'd copy across your current style prompt rather than include a new one, unless you think Suno significantly deviated from what you originally prompted for.

The reason for upping audio influence is that you want to keep the song fundamentally similar, but leave enough space for Suno to correct the lyrics rather than stick strictly to what it hears to the point it ignores your lyric changes.

[unreal sleet]

ok that makes sense, ill give that a go and thank you for the great reply

[dusky stone]

Guys I have a good song with very bad voice and trying to cover with a Persona I like a lot. I tried like 20 times and it change the structure of the song exactly at the same point and mispell exactly the same word...even if I tried crazy slider options...nothing

Any idea pls?

[frozen dagger]

said misspell, but guessing you mean misprounounced. what is the word it consistently messes up?

[dusky stone]

Yes Sorry

The world was not mispronounced for 30 takes few days ago. Also tried to insert pronunciation in text...

[frozen dagger]

Others reporting similar errors. Can't be sure if it's similar issue without knowing exactly what it is, but possible it is having buggy errors.
Depending on the word and it's usage could be the word as well. Phonetic spelling is usually the best route for that. Sounds like you tried that. Not sure again. (wouldn't copy and paste the dictionary stuff into lyrics)
Have you tried doing an extend before the error with the correct lyrics?
Not my preffered method. Especially if you want to keep the song as is mostly. In which case it's trying to cover that section and splice it back in. Either trying to cover with the vocals removed so it doesn't negatively influence the new gen with old, incorrect pronunciation, or covered with a correct pronunciation in place (even if artificially inserting the correct pronunciation of the word, via Sounds if needed).

well you say u have the correct pronunciation, splice that back in there? If you're doing cover's of covers it can degrade and do weird things if you make modifications to the lyrics as you go as well. as it tries to realign new lyrics.

[dusky stone]

Tried to start from very first take where pronunciation was ok and structure too...but it messes up over and over again

[sweet ivy]

@frigid orbit LOL. No I'm not into golden grail prompts. LOL. I just mentioned a video I thought might be helpful. Guess there a number of options to handle artifacts.

@quiet solar Given what I'm trying to do & your solid advice, I will upgrade to Suno Studio. Like the idea of regenerating a stem, pitch adjustments, editing. It's just way more work downloading in my DAW and trying to do everything I can do in Studio.

[quiet solar]

I'm pretty sure that was the logic behind Suno wanting to introduce a built-in DAW with AI generation features attached. Best of luck with it!

[quiet solar]

People have been reporting a few more pronunciation issues recently.

If the Persona is tripping up over a predictable part of the song, that suggests there's something about that section that's giving Suno grief when it tries to replicate it. It's possible that's directly a result of the persona, or that the persona is aggravating it - it's equally possible there's something more fundamental about that part of the song.

For example, I've seen it happen when there were too many / too few syllables in that line - Suno landed on a way to handle it once, but trying it again throws it and it tries to improvise, breaking everything in the process until it gets back to somewhere it recognises.

As ever, if you share the specific song we might be able to give more specific advice.

[hidden yarrow]

Any tipps what to do against audio quality degradation towards the end of the song using v5? Sometimes loss of details & depth can already be heard starting at ~1:30 and gets worse and worse until the end.

[frigid orbit]

Extend before it messes up

[hidden yarrow]

Makes sense. But what if it is a cover? It would not follow the original but only the previous context...

[frigid orbit]

If it has enough audio context from the earlier 1:30, should be fine. Gotta put some effort in yourself.. Or make the original that messes up into a persona, and run it in tandem with an extend after 1:30.

Run it as legacy tho, not the new persona.

[opal orbit]

Adding "Dynamic Instrumental" has saved many of my covers. It makes Suno give clarity to individual instruments. Drawback is sometimes it doesn't sound as "lush".

[hidden yarrow]

Was it a real instrumental without vocals? Or just the prompt?

[opal orbit]

With vocals usually. Still works on straight instrumentals though. And yes, just adding that to the prompt.

[frigid orbit]

It's always worth going back to default sliders, incase you're on some wacky settings.

[hidden yarrow]

The music itself is often great. It's just that the audio quality gets so much watered down that a remaster would not help. I think this all started with the Mashup-update (not using it, but has very often the same problem).

[frigid orbit]

Remaster is a trap.

It's honestly a degraded tool.

[hidden yarrow]

Yea, but even with all the flaws the remaster-function has, it cannot change the sound enough to fix this type of sound degradation.

[frigid orbit]

It''s either too little, or too much atm. Prior to all the updates it was a wonderful tool for slight variations. Never found a use with cleaning up clarity/degradation.

It's better to avoid the degradation, than trying to fix it.

[sweet ivy]

Is there a way in Suno Studio to select a small region on a track where a few notes pop out too loud and then lower the volume, just in that region

[polar spade]

Select the region.
Right click and "split"
Click the selected segment's header.
Look top right-screen and click Clip.
Find volume slider there.

[sweet ivy]

@polar spade Appreciate it. Knew there had to be a way. So it's saved that way when you play the clip. Suno has a memory of the adjustment?

I mean the track, play the whole track back

[polar spade]

So far, it seems there's an ongoing current status while the tracks are showing on screen. But just like wordprocessors and vid games, its good to periodically save if you're seeing you are going to do about 10 edits.

Presently, for me this week, Studio suddenly who's the tracks as blank and if I refresh, I get a white screen with only track labels visible. Eventually things clear, but definitely reminds me to Save.

[frozen dagger]

Saving is always good practice. There is a variable control in Studio as well that should allow you to go back to previous versions from the auto-saves it does while working on projects.
Seems like loading information on the timeline isn't working properly. Whether it's the white screen with track labels, or dragging clips from Studio to the Cover area on the left. (Says clip is still loading, but it's not you can load it there other ways rather than dragging.)

[opal orbit]

So, your question got me thinking and I started experimenting a bit. So yes, covers appear to have degration issues since mashups, but found an alternative. (Works great for high energy electro-pop anyway)

Use "Add Instrumental". You might need to fix some vocal glitches, but the quality difference is night and day.

[hidden yarrow]

Interesting! I will try it out.

[hidden yarrow]

Uhh... this assumes that you have made a Persona from the song? Cannot find that option.

[plush notch]

I’m having issues changing lyrics - I’m doing it same way as before - anybody having same issue

It’s not generating new lyrics when I regenerate

[opal orbit]

I didn't use persona.

[frank matrix]

yo admin, its my birthday today, lemme access full features only this day🤣😛

[subtle plinth]

How does one get Suno to stop putting la-las or do-dos in intros when you have a female vocalist? I’ve tried basically everything I can think of in styles, lyrics, and negative prompts.

[hard beacon]

Hi @subtle plinth ..about what style of music are we talking?

pop, rock, rap, edm?

[subtle plinth]

‘female white British rapper with a rapid-fire delivery and a deep husky voice, indietronica, fast pace that precisely matches 170 bpm’

And while I’m at it, I can’t get Suno to get real contralto range

[hard beacon]

can you post a link to an output with a la la la oh oh oho intro ?

i have to take a look into the prompt and the metatags 🙂

[subtle plinth]

https://suno.com/s/JZT846xN6PCByiq2

[hard beacon]

ok.. hold on 🙂

[subtle plinth]

Thanks!

I ditched the negative prompt at that point since it wasn’t doing anything

[hard beacon]

first of all.. v5 likes to add some adlibs in the intro... I think the "oh oh oh" in the intro even fits with the rest, but that doesn't solve your problem yet...try [instrumental intro] in the metatags instead of [synthesizer, bass, and drums intro with no vocals]... try to set style influence slider to 80%

and never forget: it's AI... Getting the right result may take a few outputs.

[sleek stump]

Its like rolling a D100 hoping for a 80 or higher 😄

[hard beacon]

😉

[neon hound]

Hello again! A question is this: what's a better way to make a persona from song? For example, I have a song in which vocals go from calm to normal to loud in span of 30 seconds. If I make a persona from these 30 seconds, will that persona just BE ABLE to go from calm to normal to loud the same way it did in original song, but only whenever necessary, or will it ALWAYS go from calm to normal to loud even if it's not required?

[sweet ivy]

Have a timing issue, where a bar moves ahead maybe a 1/16th. Tries to fix it with selecting the bar before it and the bar with the push, doing a cover. but that didn't work. I assume I have to select all the track and shift them slightly. Would I have to select every track from the push and shift it all back? Any ideas?

[hard beacon]

@subtle plinth : and??

good results?

[subtle plinth]

One sec. Had some other stuff to do

I have definitely tried [instrumental intro] before

Ad libbed vocal intros in the first two gens so far

[hard beacon]

hmmmpf

[polar spade]

You can slide the 30-second marker anywhere along the song and pick the bit you like best.

To answer specifically, it can be possible that Suno picks up the change in volume as a character of a persona. I would have to think though that volume is quite a low priority of most songs in general to other points of music/songs and thus has little weight with personas as well.

[lusty rune]

That would vary on the song you attach the persona too. If you direct the sections [Chorus - Soft whisper] The persona will adapt. Personas aren't copy of song you created and captured the voice from. It's a copy of the vocal range, key and resonance of the vocal itself.

[neon hound]

Got it, thanks. At the moment, I'm trying to pull off a strange thing: I have an old song made back on the v4.5 model. I like the way it's performed, but I'm not very happy with the lyrics. I rewrote the lyrics and now I'm trying to repeat the performance. I made a persona out of an old song, copied the style in Creation window, put new lyrics, added the old song as inspiration and attached created persona...

And I can say this: Everything that could break did so.

[dense ice]

Guys hello, What best setting can I set sliders to get the most appropriate results

[bright haven]

Is there a good way to force a male or female vocal for a for a song using v5?

I had a great idea for a duet and no matter what I do it doesnt listen to my tagging to keep it female or male

[pearl topaz]

Have you tried [Male Vocal] [Female Vocal]? I haven't tried duets yet, but these might work?

[bright haven]

ive been using things like [Man, sung] because that worked back in v4

but it seems that it just says "Nah, think I'll do it this way, thanks though" 😂

[stuck junco]

Hey Ken! Thanks for reaching out about this.

Here's a really cool guide that could help: https://discord.com/channels/1069381916492562582/1466598465969590465

Additionally, if you'd like to share a song link, I'd be happy to take a look to see if anything sticks out

[pearl topaz]

Has anyone found a way to set a vocal range? Would like to limit my range to baritone or low tenor. Suno tends to like to throw in some falsetto that I'm not very good at. Poked around a bit but couldn't find anything. That being said, I like the songs that were created so just trying to edit the vocal range a bit. Thanks!

[bright haven]

somehow suno takes this as make them alternate

[Female]We don’t say it, but we both know
[Female]There’s a ghost here when you get close
[Female]You hold my hand, but your eyes drift
[Female]Like you’re somewhere I don’t exist

6 different generations and it just says nope Im doing it my way lol

[bright haven]

https://suno.com/s/HWlnFxI2fwM1TB4N

This was so close but it screwed up and had the man sing a line the female shouldve sung and that kills the whole message in verse 2
Lines 3 and 4

[deft dust]

Good afternoon, did they quit the option to separate a song made in Suno to extract the instrumental? Or its only for pc app?

[lusty rune]

If you are doing a cover for v5 make sure you pull the weirdness and prompt adherence to 0. The move the audio adherence to 100. Then attach your persona. That way it grabs the track and Persona and doesn’t try to add anything new. Then you will have the same track new lyrics plus you new vocal artist.

[bright haven]

you cans till do this on pc by clicking the ... and hitting extract stems

[deft dust]

But it doesn't allow you to do it on mobile phones, right?

[mystic siren]

Getting it 100% female and male to be correct is very hard. I have had a couple of good attempts like this one: https://suno.com/song/2276460d-893b-4291-8f83-277dfb817c95

[bright haven]

what does Studio do exactly?

I've used editor but the results have never been great lol

[lusty rune]

Everything and I mean everything. It’s a browser based GAW/DAW in every imaginable way. Truly in my view a technological marvel.

[bright haven]

Im 90% sure its pc only

[polar spade]

But was this just an experiment?

You wouldn't have a need to prompt individually in each line for a female to sing, normally. Just once.

So, a line spacing (often double spaced) is helpful for changing the voice.

for example, if you have the following:
[male] "We’re fine"
We’re always stuck
In the same red line

[Chorus]
[male] "We’re halfway in the headlights"

The second [Male] in the chorus is likely going to be female as the previous stanza was sung in the male voice. I assume that Suno feels the designation of a one gender means that the next lines (spaced apart) is likely best as an alternate gender regardless of the metatag. If that above chorus has alternating female then male prompts, its is already starting on the wrong foot.

2. A recent one I use more often now is to indicate the gender, and then provide a quoted line. As seen above as well.

So you may have something like this, with or without spacing:
[male] "We’re halfway in the headlights"
[female] "Halfway changing our minds"
[male] "Arguing over nothing"
[female] "Till the silence feels kind"
[both:] [accompaniment] "But I reach for your fingers"

It feels as if the quotes themselves take the place of a line space and show "completion" or "change"

[bright haven]

yeah it was an experiment because I've tried it a bunch of ways and it didnt work so i thought maybe ill try to force it

[polar spade]

But, saying these, I anticipate there will be changes within the song somewhere. ultimately, I use other tools of replacement and editing to tweak through about 30 iterations to get it how I want, or sometimes re_write some of the lines to fit what Suno gives. I never worry about what I want in the song, but know that once I publish it, its about what the listener perceives. Makes songwriting much easier.

[polar spade]

yes. forcing can be a bit fun. If suno keeps singing male when I put female, then of course, just put female in some places I want male. Why not. But I'm spoiled with having Studio now to just splice in places to be changed.

So with Pro or even Basic, its better to stick with extensions and dig into each stanza individually until you get it right. With the benefit of personas in Pro , you can avoid trying to metatag and prompt the heck out of it.

[bright haven]

So, Ive had NO luck with replacing in editor and maybe im just doing it wrong. Ive been thinking about studio

but every time I do a replace or something it just completely changes the voice, the background it doesnt even sound remotely close in any way

[polar spade]

Some concepts such as Suno believes chorus are often sung by both at the same time, is consistent with duets in real life, especially from the 40s, 70s and 80s. So its often nothing to do with prompting, but Suno training. Once thats worked in the equation, we start seeing patterns.

[polar spade]

I know many will write the song first, maybe an epic metal that someone wants each singer to sing their appropriate lines. So its a harder struggle.

For me, often if its a duet I end up with, I'll go in that direction and rewrite the whole song to fit the duet. Just something to consider. Again, no one knows but me, and when someone listens to it, they are impressed and wondered how I did it. 😊 So, sometimes more about being effective and efficient than perfecting an idea.

[opal orbit]

I always make 2 covers, one with female vocals, and another with male. Then splice together in Studio.

If you cover the studio export, it should then generate the correct vocal genders where you need.

You can use a normal DAW too. Studio is just more convenient.

[bright haven]

My covers always change the vocal style a little too much

[opal orbit]

As in the way its sung, or you dont like the voice?

[bright haven]

the way its sung.

little of the voice too

[opal orbit]

Dont worry about the voice as much, cause you can reroll on covering till you get what you need.

For the way its sung, regenerate using Studio or using the waveform slider to fix it. (Again, how it sounds doesn't matter as long as its the right gender and sung correctly)

You can also record yourself singing the line. Even if its bad, Suno is really good about regenerating a nice voice over the notes you were hitting. Ive done this before.

[bright haven]

Is it better to put weirdness on 0 when generating?

Like If I want to make a acoustic cover of something I've already made. How do I go about that?

[honest heath]

I would not go above 80 or below 20 on any slider, here there be dragons...

[subtle plinth]

I'd love to know if anyone's figured this out

[quiet solar]

It's worth bearing in mind that tags in the lyrics are more like suggestions than instructions. Suno might take them on board, it might not.

It's not really possible to get perfect control over which singers sings which part through prompting alone. You can get close by using good practice in prompting, but If you need specific singers to perform specific lines then it's worth seeing it as an exercise that will involve a mix of prompting and editing - controlled duets are a bit beyond the capability of Suno's models as it stands.

[bright haven]

I mean doing covers with 0 weirdness and a female persona then splicing might be the way to go

[quiet solar]

I'd personally suggest not going into the red zones for any of the sliders unless you're absolutely certain around the effect you want to achieve. Stick between the 15% - 85% marks. A 15% weirdness, 15% style, 85% audio cover should be close to the original - just keep an eye/ear out for Suno changing the key signature and tempo of the original song. If you know them, put them into the style prompt as well.

The splicing method works really well if you don't have too many changeovers between the two singers.

[neon hound]

Thanks again! Covering with this helped to make almost what I wanted, and I don't think it'll be any better (and I'm not spending the rest 1700 credits to try). Now I just need to extend to add two remaining choruses, outro and solo, because suno failed to generate them. Shouldn't be that much of a problem, hopefully

[quiet solar]

Also, to add, you don't want the audio influence too high if you're looking to switch the gender of the singer. You need to give Suno enough breathing room on audio and weirdness to be able to make that change rather than sticking to what it hears, which will be a male singer.

If you find Suno seems to be struggling, just dial down the audio influence a little.

[neon hound]

Ok, wrong, without song in cover source it completely messes up... So I guess I'll have to regenerate cover until it generates the whole song... 🤕

[frozen dagger]

I don't think you can split stems with the Moible App, however, you can go to Suno.com on your Mobile device's browser and there you can split stems from your songs.

[bright haven]

Can suno generate sounds like rain?

or do we just edit them in

[frozen dagger]

made rain and thunderstorms with Sounds. ..but yes, then u need to edit the sounds in after.

[polar spade]

Its happened. But with Sounds and Studio, thats obviously better.

Im all about finding workarounds, even odd ones:

You can generate a random or real song tagging where you'd like [thunderstorm] . Suno likely will randomly put it anywhere, if at all.

But once done, then edit the whole song around the sound effect. Keep verses you want before it, remove others, add lyrics, try different edits.

The issue is that with rain, suno likes to continue it , and wind, or waves, throughout the songs or at least in the background of a verse. Great if you can stem it and do something with it, like try mashing it together with a song.

But again, the easiest is then to just sequence it in as a track. Or course thats with Studio and a Premium account

[bright haven]

sometimes i swear it doesnt listen to style at all

This cinematic pop punk sad song features beautiful, expressive piano lines carrying the core melody, set against a gentle backdrop of falling rain for atmosphere. Minimal drums offer subtle urgency, perhaps using rim clicks or soft snares. Rich strings slowly build in the chorus, heightening drama.

somehow turned into full on screaming , heavy distorted guitar and a high BPM

[neat nexus]

🤔 Does impossible have another meaning I am unfamiliar with? I've been reliably informed that these two tracks I just generated are impossible. What am I doing wrong to accomplish such a feat?

https://suno.com/s/bWWpglpRssy4JMeH
https://suno.com/s/lT2sVDnr09EKT6vq

[sweet ivy]

Having a tough 1st day with Studio, even though I'm following the instructions I've seen. The most bizarre thing that in the Pro plan I had 2,340 credits left. I upgrade to Premier and suddenly I only have 1,000 credits left? Anyone else have something like that happened.

I tried to "Replace" a two bar section on the guitar track and in the last 2 beats of the second bar, it played a note outside the chord for that bar. Clearly a wrong note ?

[lusty rune]

You’re more than welcome.

[bright haven]

Love when I get a almost perfect song but it mispronounces one single word

🫠

can I regenrate a song the EXACT same way and then have it try to fix that one word?

[pearl topaz]

You can try to put the audio at 100 and the rest to 0? Maybe it'll work?

Someone else might know if it's something that can be fixed in Studio too.

[polar spade]

That helps to get close. I never 100%, but Cover the song with the edited lyrics. Set audio to 98%, and style 4%.

Thing is, for just one word, it might be pronouncing off because the sibilance is too close between the words, or its a bit tongue-twisty for AI or its just phonetically wrong. At the least, spell the word phonetically so it helps to pronounce the way you wish.

[sleek stump]

My setting for a 90-95% retention COVER (I use cover to fix lyrics) are 90% Audio, weirdness at default (50 is really 0 here) and style at 5, and delete the the contents of the styles prompt(leave blank). These are the settings I have the best results with, YMMV. It also may vary with the genre you are working with.
And as Dasvon noted (you never noted what the actual word problem was) use phonetic spelling for mispronounced words.

Additionally, refrain from using 100 or 0 for settings, Suno tends to do really weird and unwanted things when sliders are at those values. My safe values are 10-90 for all sliders.

[quiet solar]

@bright haven I've found it more tricky to prompt for slower, melodic pop punk songs as so much of the genre by volume is geared towards a more rapid, rebellious feel, and genre is a very powerful influence on what Suno looks up as a reference.

It can be done, but expect it to take multiple attempts. You could try moving pop punk to the back of the prompt, leading with 'acoustic pop punk', blending in adjacent genres like acoustic emo or post-grunge, depending on exactly where you're trying to take this song. Since it's expecting guitars, you might want to pre-empt that by asking for clean electric guitars, and consider more "mood music" prompts like 'vulnerable', 'somber' and 'emotional'.

[neon hound]

6:14 is some magical number or something?

[stuck junco]

hahaha

🐉

[sleek stump]

This can happen when you have a pending charge that is working its way through the banking systems. Once Suno gets the funds through the processor, your account should show the proper credit balance. Keep in mind that Suno is NOT the processor, it's Stripe, a 3rd party processor.

[sleek stump]

Were these all generated within minutes of each other? There may be a back end issue happening...

[neon hound]

no, the oldest was 4.5h ago

[sweet ivy]

@sleek stump Thanks, it processed on mycard right away but that makes sense about having to clear through another party. Can I ask you a question about trying to fix a timing issue. I have a bar where it seems all the track start slightly before the downbeat. Timing is fine by the next bar. I asked Suno support and they said it can't be done in Studio, I'd have to use my DAW.

[neon hound]

I had a bunch of generations in between them. It seems that if I cover a song and use persona made from that song, it generates 6:14 songs with last part missing

the moment I remove either persona or cover source, it generates songs with different lenghts

[sleek stump]

Thats sounds like an issue, you may want to report that in https://discord.com/channels/1069381916492562582/1082775596767117312

[sleek stump]

Processing from Stripe can take time, on top of that we are now heading into the weekend.
On your other question, they were pretty spot on, that would be something that would require external tools, such as a DAW.

[sweet ivy]

@sleek stump Okay, they check the posts in the Feedback area - bugs. Loading tracks into my DAW & see what I can do

[sleek stump]

Yes, Suno does have team members watching the bug channels, best to post issues in them.

[neon hound]

Using both cover source and persona results in 6:14 tracks with obviously missing last choruses and sometimes messed up lyrics
Removing persona breaks tempo, stressing and vocals, also missing last choruses
Removing cover source breaks melody, vocals, missing last choruses
Removing both persona and cover doesn't even make sense because the whole purpose of what I'm doing is rewriting old version with new lyrics.

So, with that first version... Did I accidentally create a masterpiece that cannot be surpassed? Am I ahead of the technology of my time?

https://giphy.com/gifs/snow-in-the-laying-on-stairs-15esG75tRFTGg0QBFs

[sleek stump]

One detail to note, if you are using a large persona or long song you are covering, you will not get up to 8 minutes. Suno will reserve some of that track time to process the audio you are using, so you may get the same time render over and over due to that.

This article https://discord.com/channels/1069381916492562582/1422252931943501935 notes this. @neon hound

[neon hound]

*looking at new lyrics*
Yeah, sounds right, then

first one was 6min as it is, so now... well, I'm screwed, then

[sleek stump]

You can extend from near the end with the missing parts and see if it will give you something close to what the original was.

[neon hound]

tried that. Results in 4m tracks with wrong/broken vocals, tempo and melody

probably because extending removes cover source, and removing cover source = breaking

I have only 2 options I can think about. I have perfect generated piece with 3d verse and following pre-chorus and chorus performed. Maybe I can:

1. Try to generate another piece that's starting from 3d verse, pray that it generates the perfect version, then glue them together in DAW
2. Same as above, but put them in mashup and pray

[pure panther]

If you do option one, remastering on subtle is a great way to make the splice seamless.

After re-uploading the edit

Sooo, question. Does putting a song in Studio and exporting as full song no longer give you the "no styles" tag? I've tried it a couple times and it keeps adding an Ai generated style to the song.

[sleek stump]

Seems that was a bug that recently was brought to Suno devs attention. I do not know what the status of that is.

[pure panther]

Aww, that stinks. Not a bug, a feature.

Unless they give you a way to customize the style after words.

[sleek stump]

Back in the day, we did have the option to edit the styles field in a track, but that went away with 4.0 if I remember right.

[frozen dagger]

We have that, an option to use it in place of the style prompt used is what is wanted it seems.

[sleek stump]

You can not edit the auto generated styles section

[pure panther]

I wonder why? It's seems ridiculous to force users to keep the Ai generated one, it's often very very wrong. It consistently analyses my female alto singers as male baritone...

[frozen dagger]

No, you can add your own "style summary" in the song details however.

[pure panther]

Does that fully overwrite the automated one?

[frozen dagger]

I' mnot sure that it shows anywhere to anyone, just to us privately?

[sleek stump]

You can edit the styles you used, but Suno also does an auto styles interpretation, and that you can not edit, and it is viewable by anyone.

[frozen dagger]

If they opt for a "user generated" style prompt it will almost certainly you choose pre-chosen tags to add. To avoid abuse.

[sleek stump]

I dont like it either, but it is what it is.

[pure panther]

That might be ok if it were accurate. Unfortunately it's way off at this point.

[crude void]

is this where Suno noobs ask questions that can easily be found elsewhere?

[pure panther]

Sure is! 😄

Ask away 🙂

[crude void]

ok so i made a song, i tried to set it up for 2 male verses, then a female verse, followed by male last verse/outro.... song is near perfect in what i was thinking in my head but the vocals in verse 3 turned out male and the last verse/outro is female vocals.

whats the best way to fix it??

so its still decent but misses the mark of what its supposed to be

[pure panther]

There's a lot of past discussion here on methods. I have my fav, but there's other ways too if you wanted to search the channel.
I make one song of all male vocals, then another of all female, then spice it togther roughly in audacity, reimport, do a cover, then remaster on subtle.

[crude void]

and that will keep the same melodies/meter between the male and female versions?

[pure panther]

To get the sameish melody and meter, I find a version that I like of either all male or all female, then use that as a cover with high audio influence for the other gender. It won't be perfect, but it's usually good enough to splice roughly and then re-cover after re-uploading it to Suno.

@crude void

[crude void]

thanks a bunch 🙂

[quiet solar]

It's not guaranteed - expect it to take a few attempts. If you're doing it through cover, it can help to specify the BPM and key signature of the first version you create when covering it for the second one with a different singer, as well as doing a high audio influence cover.

If you have both vocalists in the same song, and a version that's fairly close to what you're looking for, there are other routes you can take as well based on Extend (which is a bit easier to work with than Cover for surgical changes, IMO).

More info including workflows here if it's useful: #1466598465969590465 message

[pure panther]

I wonder if my account is bugged or something? I've never once gotten a good output using extend and I've tried to use it several times over the last 4 months or so. It's often gibberish and doesn't sound anything like the original (even if some of the original is grabbed). Like it will be an acoustic guitar song and the extend is like disco house or something ridiculous. I've never understood how it could be a useful tool.

[quiet solar]

I've been using Extend since v3 in the workflow for the vast majority of songs I've made. Before we had the audio influence slider, it was hands down the most effective way to create samples or alternative takes for editing - high audio influence covers can now get into the same space, but extend has the advantage of a lower risk of introducing tempo and key variations, changes to the singer, etc.

On the flip side, it has more chance of changing the rhythm of the lyrics, taking the song in slightly different directions, etc.

My core workflow for a new song made in Suno would be Create to start with, then Cover as-needed on the best version for general improvements with a view to getting a 'master' candidate version, then using Extend as-needed to generate specific samples for fixes or alternative takes, then editing those in, then doing anything mixing/mastering related.

I find both Cover and Extend can go haywire on occasion - sometimes that's the nature of things, but if it's predictable then in my experience there's often something causing Suno to struggle that might need addressing in the underlying song or prompt.

[quiet solar]

@pure panther I'm trying to find a good working example that I don't mind sharing from the mess of what is essentially the cutting room floor of past songs. xD

[lofty merlin]

Hey, all. Does anyone know how to get an instrumental to generate without any percussion? I'm trying to make an old-school 8-bit chiptune track with no drums, but no matter how much I put "drums" and/or "percussion" in the exclusions prompt, it always includes percussion anyway.

[pure panther]

Thank you for the tips. Another thing I struggled with is that none of them seemed to stitch with the original, is there a way around that? The ending point of the original never seemed to match up even remotely with the start of the extended bit. Also, is there a way to get the whole song? it kept giving me clips, but there didn't seem to be a good way to add it to the existing song. (I was using extend in the creation tab, not in studio)

[quiet solar]

Yeah, from the Extended version in the Library there's a "Get Full Song" option that will combine the original and any prior extensions, including the one you select, into a single new track.

I'll be honest, that's not really how I've used it - as you say, the stitching often isn't 100% perfect. Outside of rare cases where I did actually need to extend the song, I typically used it to facilitate editing.

For example, say I have a song I'm totally happy with, but the singer mangled one of the lines. And what I want to do is a surgical replacement of just that line by the singer, leaving everything else unchanged.

Cover and Remaster can be problematic, since they'll break consistency with the original in subtle ways. It may be consistent with itself in the new version, but that's not ideal if you only want to change one thing.

Replace in the Editor/Studio seems an obvious choice, but exactly as you pointed out, that can create issues with whether or not it seamlessly blends with the existing song at both ends, and whether it gets the fix right, and whether the fix has got consistent instruments and vocals. Studio also gives the option of just regenerating the vocals, which is very helpful, but if you don't have Studio...

The solution I came up with was using Extend to create multiple new takes of the problematic section. So for example, putting the Extend point two thirds of the way into the song, just after a chorus, so it had all the reference material it needed for existing song sections - then getting it to reattempt the problematic section two or three times in each Extend.

What I'm looking for in this case is a vocal line that does what I need. I don't mind if the Extend is seamless, or if the instruments went a bit wonky, because what I intend to do is split the vocal stem out and take it to Studio or a DAW to replace that line (ideally) or that entire vocal section (if I can't do a single line change seamlessly).

[pearl topaz]

Could you generate it, then mute the drums in Studio?

[lofty merlin]

Yeah, but then the generation depends on the drums to drive the rhythm, so muting it just kills the rhythm. I'd like it to generate a track where the synths drive the rhythm instead.

[pure panther]

That makes sense. I'll have to keep that in my tool chest. I'm usually able to fix mispronunciations and slurred lyrics using the Song Editor, but sometimes it won't cooperate.

[quiet solar]

There's no magic bullet for that. Drums and percussion are so common across music that Suno will often look to include them.

Shep's idea above is great, if you have Studio. If not, adding 'acoustic' to the prompt may be helpful. I've heard other chiptune tracks on Suno where people have evaded this problem, so it might be worth looking over the prompts they've used to see if there are other keywords or phrases you might want to lift and experiment with.

[lofty merlin]

Thanks. I'll check out other chiptune tracks and see if I can piece together some prompting approaches for this.

[pure panther]

Perhaps the new stem cover feature would work. Maybe you could take the drum stem and switch it to a synth or something else?

[quiet solar]

See it as as step in the process, rather than a one-and-done fix. If you can get a track you like with drums forcibly removed, a high audio influence cover on the same track should be fairly proofed against introducing drums again. You can then focus the prompt on what you want the synths to do differently, or recreate the synth line specifically in Studio.

[keen crescent]

Morning Family - I am having this issue where I am trying to create a Song but Suno keeps adding a Live Audience cheering and clapping even if I include the fiollowing in my Prompt right in the beginning - Help Please

((Gentle crescendo, soft spoken voice - studio recording ONLY))
((NO live audience, NO crowd, NO cheering, NO clapping))

[quiet solar]

Change your prompt to: "Gentle crescendo, soft spoken voice - studio recording" in the Style box.

Suno doesn't handle negatives and conditionals well. If you put "NO live audience" there's a good chance it'll read it as "Live audience".

[balmy hound]

How is persona used in Suno. I want to translate German song to English. Persona successfuly "cloned" German voice in English. But when I apply it, it will rewrite my new "updated" lyrics within new cover attempt. As if persona would just copy entire old generated song style.

[quiet solar]

@balmy hound Have you tried covering the song focusing on the language change first, then covering that version with the Persona separately rather than making both changes at once?

[ember sky]

anyone created a santoor (indian instrument) instrumental song? suno is generating santoor sound but not santoor instrument in custom or simple modes

[polar spade]

Where is it generating the santoor sound, in a random "Indian" styled song ? But not if specifically prompted for a santoor?

Is this sounding like a santoor, for the first 30 seconds, then it goes into a something like a sarod after that?https://suno.com/s/aXToFXeCagagec7J

[peak socket]

I want to create a song in the style of Ivoxygen's “Spaces Out” in Suno, but I can't get the style right. Has anyone done this before and has a suitable prompt for it?

[sweet ivy]

Good morning, I had a chord progression 6min - 3maj7. In one verse Suno played a 3min. How easy or hard is it to copy and paste in two bars from Vrs 1, where it's correct ?

[quiet solar]

Depends on a few factors - primarily your comfort level with editing and the tool you're using, and secondly whether Suno has kept a stable tempo between the two sets of audio you're working with, and whether it's split the stems cleanly (unless both are purely instrumental to begin with).

You'll probably want to take a bit more than the 2 bars so that you can crossfade the tracks at the join marks rather than do a straight copy paste. Assuming the instrumentation is the same in verse 1 and the verse with the problem, I'd personally do it by aligning a copy of verse 1 and it's transitions in and out on a new track against the verse you're editing. Then it should be a fairly straightforward task to pick a point during the verse before the edit where the tracks look very similar for your first crossfade, and (hopefully) same again for the second afterwards.

It's the kind of exercise I'd say is very doable, but may potentially turn out to be a little more challenging than anticipated in practice. If you're not otherwise wedded to the audio, I'd be tempted to try to fix it by regenerating rather than editing in the first instance - but again, it depends what you find easiest.

[sweet ivy]

If I did a regen of those 2 bars or 4 bars, what are the odds the Suno makes the same error, it's not relying on the audio guide I used to created the track where the progression was correct in each verse

[quiet solar]

It probably wouldn't be a case of trying to replace those two bars specifically (e.g. using the replace tool), but either;

a) repeating the process which created the instruments in the first place and hoping this time Suno doesn't improvise a 3min chord in the wrong place

b) using Extend from somewhere before the 3min chord and getting Suno to recreate the rest of the song based on the song up to that point as a template

Really hard to give specific advice as to which route makes the most sense for your specific situation, as it'll depend on the song itself.

[sweet ivy]

Yeah. I'll just have to try a few things. But I still have that push in the second verse, where everything starts an 1/8 early. I may just have to use the same prompt, gena few new versions. Save stems and past something together

[ember sky]

if i choose 'sounds' option on top and input "santoor" it create mini clips of santoor instrument .but when i input " santoor instrumental" in simple or custom mode its not able to generate the santoor instrument sound

[sweet ivy]

I thought I posted this. The regen didn't have timing issues, but the arrangement was so different, even though I used the same piano-vocal audio, style prompt, set the three advanced settings close to where they were. That 8th bar is giving Suno a fit. Instead of 3min7 followed by 6min, it's plays a 3 major chord. Where did that come from, wasn't on the piano-vocal I did.

[potent maple]

Hi there , I was wondering if there are any people out there who seriously produce with Suno. I need some advice.

I sell beats for rappers, but I've noticed that when I try to mix or master a Suno track's stems, the sound, after spectrum analysis, is cut off around 15kHz, even in WAV format. Yet my EQ in my DAW shows that there's content up to 20kHz, so I don't really understand. If I sell a track and the guy analyzes it, he'll think it's crap, even though the mix actually sounds pretty good, even if I'm not a pro (except that I feel like a lot of the kicks sound different throughout the track, so I've had to invent some techniques to get around a lot of problems like this with Suno tracks).

But then, is a Suno track actually marketable by labels? Can you really make clean mixes and masters of it?

[quiet solar]

Suno interprets, influenced by genre norms. It's a generative tool rather than a word processor for sound - we can nudge it towards being more or less creative with sliders, but can't stop it introducing creativity.

[quiet solar]

I suspect that's a reflection of the stems being retrospectively extracted from the final version rather than being actual stems - Suno is probably using EQ ranges to determine where some of the sounds are expected to be for stem splitting.

There are things you can do to make cleaner stem-level music in Studio, and to clean those 'stems' afterwards, but whether it meets professional standards, or whether there may be other complicating factors around copyright for AI-created rhythms, melodies and sounds...those may be conversations you'd need to have with potential buyers about their needs and whether a Suno-made track can meet them.

[strange mural]

Has anyone here had success with time-stamping to hit certain cues or to get into a specific track length. I am having mixed results and just wanted to see what other techniques are working for people.

Currently if I am creating a 30-second mix of a longer song (Usually this is me creating underscore for a radio spot or something like that for a client) I will add the length in several times in the prompt, add time stamps in the prompt (like [VOCAL HOOK] at [00:00:27]) and then in lyrics have time stamps [00:00:00 - 00:00:03 - INTRO]; [00:00:03 - 00:00:26 - INSTRUMENTAL]; [00:00:27 - VOCAL HOOK].

It works - ISH.

If I could nail it down it would save me a lot of cleanup time and might be able to generate multiple different tracks for use on radio/social/etc. Right now its a bit labor intensive so once I get one I tend to ride that puppy into town. At least now I can do it in Studio instead of having to take into Audition (even though I'm faster in Audition for now).

[graceful condor]

hi

[quiet solar]

Others have mentioned the same 'ish' around trying to specify for times in various different ways.

I'm not aware of any reliable way to specify exact timestamps to the level you're looking for - closest I've got to has been based around creating something at about the right length and uploading it as audio for Suno to iterate on.

[static wren]

hello, got a quick question. made a song where i like the instrumentals and the lyrics, but don't like the voice. is there a way to replace the voice's tone without changing the way the 'singer' sings? i tried to do it through the stems, but adding the lyrics back in changes the way the ai sings the song

[sleek stump]

Not really, the best you can get is to cover the track changing the vocalist in styles but using high audio influence slider. For this kind of change I set sliders to Wierd-50 (default); Style-70; Audio-90 and only use a vocalist changing prompt in the styles window. You will likely have to do a bunch of gens, but you should be able to get something "close" to the original.

[static wren]

mourns for the potential loss of points orz

thank you for the help, just wanted to be sure.

[pallid pulsar]

Hi! Someone could help me? How could I do something similar to that? Like using almost the same melodic structure and "vocals"? And, I'm having trouble figuring out exactly where the "Ohh ohh ohh ohh" and the "Hinatazaka! 46! 3 - 2 - 1!" would go. https://youtu.be/bG9m2uZ6UJA?si=7907hMYFB5M1Rlbm I'm creating this for a personal project, an incarnation of Doctor Idol-style, but keeping the overture idea. So.... I made a few that turned out okay, but they're not quite the way I was aiming for yet. For exemple:

https://suno.com/s/KrRegZNZiouh1801
https://suno.com/s/LPDnCbeFgBUzXdIc
https://suno.com/s/7FClzzjkTKlbktQ6

[main hazel]

been struggling to find a matching suno promt for this type of song, any idea what i could use?

[polar spade]

I and others seems to do pretty well with using an AI genre analysis program. There are a few online. Just starting with writing down all genres it reference and trying about 8 of those at a time can get some close approximations. Audio AI Dynamics has been a reliable one for me. But others area available to try for free.

[main hazel]

cool

so far the AI genre analysis gives me just that?

if use this in suno?

cant rly do much with it it seems

[polar spade]

You're right to doubt. Thats just showing more of a mood than genres.

[main hazel]

thats the only thing i could get out of it but still sounds different

i mean, that song is def not country

nor electronic lol

[polar spade]

Are you using similar lyrics, apparently Russian, Ukraine, eastern europe... or just going for an instrumental

[main hazel]

yes Russian, and east

[polar spade]

Im trying a couple... similar prompts... I'll see what I can come up with

[buoyant bone]

Unable to upload this file to SUNO.
Says its less than 6 seconds short.

Had to convert to WAV

[polar spade]

I've attached a playlist with some waltzes and one in 2/4, which are getting close to the sound. Add a kick drum, bass drum, punchy sub bass kick, whatever works for you.

https://suno.com/playlist/1bd13140-6f18-42d2-baf8-6a0fafa1a0a1

[main hazel]

sounds very promising

have you also tried Chanson promt yet?

i didnt get far with it

but seems to be in that category of the uploaded mp3

[polar spade]

Not Chanson. No, i didnt try that. But I'll leave it for you.

Yes. its getting closer!

For me to go further is just to keep juggling the prompts around and add one occassionally to get it closer and closer. Of course, now I would start taking three or four and doing mash-ups or inspires , with more instrumental prompting to see what happens

[polar spade]

Oddly I clicked inspire in the Playlist and it didnt add the files to the Create panel. So I've reported the glitch.

[main hazel]

any changes u done here?

coz i m trying to use my similar chris rea voice i made a while ago with it

i did get too happy schlager tunes with your promts lol

like oktoberfest

here take a listen lol

[polar spade]

My sliders on the inspiration of the first three songs in the playlist. And attached results song. Vocals come in at 30 seconds

I left songs in playlist open for awhile for a remix if you just want to use the prompts.https://suno.com/s/Q3v3VciWGoRg6X5G

I thought you were going for the female voice, so i kept putting in female. hahah.. OK.. whatever works in your direction, just make the changes on your side.

[main hazel]

hmmm

so just these 2 promts u used?

oom pa, bass kick drum

[main hazel]

for that last song

[polar spade]

No, sorry, plus all the prompts from the first three songs in the playlist.. mainly the first one so lets see... ummmm

Laiko waltz, russian dance, russian rai, oom pa, euro pop, slow schlager, disco polo, volksmusik, ensemble, turku, syrtos

Then I did an inspire from all three... and added the drums only as prompt.

and exluded balilaika and accordian, it sounded too Greek. Still got Accordian in this one above... but it's Suno

[main hazel]

ah yeah thats what i previously tried

[polar spade]

yes. Everything requires tweaking usually. And the Inspire may have helped to find the similar elements and emphasize better towards a nicer result.

[polar spade]

I"m sure you're almost there then. Good luck.

[neat nexus]

A better place for this

@empty saddle

[spice brook]

I noticed something really interesting recently Suno has had a habit of mid way through the song just going back to the beginning of it and generating the same lyrics and basically restarting the song and I recently tested the ai music generator sonauto and it does the same thing

[rough yoke]

Suno making some cool songs today

[main hazel]

tried and tried, so far horrible luck lol

[main hazel]

maybe u can try somethin with that voice i sent? would be curious to see. i ll be on 2morrow

[bright haven]

will studio allow me to do a regenerate on only the lyrics to fix mistakes?

[quiet solar]

Yes

[fast relic]

I get so frustrated when I put in public domain lyrics — for example, Latin text written in the 13th century that happens to have been used for many classical music settings.

I’m not using any copyrighted material (and not uploading any recordings or music). The copyright protection engine needs a LOT of work, since it so easily misses actually protected lyrics yet blocks things that are clearly public domain.

[bronze island]

Not really sure how I can help with that... Sorry.

I can't because I'm not a mod-

[fleet portal]

From my experience That only happens if you click on the song immediately after generating it. Refreshing the song or the app fixes the problem me thinks

[polar spade]

Ive found that it is not just about getting it right. That could take me 20 iterations and maybe it sounds perfect.

But then you try the exact same prompts without going through your own 20 iterations and you can only be lucky if it's the way your ears want it.

So in the end, its about just learning the tools and trying them all. As long as it keeps getting closer, youre going in the right direction.

It can be frustrating I know. Ive just "almost" finished a great song yesterday, but Im still struggling with replacing just one word.

Looks like we got somewhere with the music, so now its getting the voice to cooperate.

If you are in Pro. Try all the tools from:
• adding vocals to instrumental
• stacking an instru track and a vocal stem as a mash up.
• combo a vocals stem you like with a persona of the instrumental (legacy option)
• run the instrumental with a vocal persona
• likely even extending in small increments per stanzas with the instrumental prompts we used, but with the influence of the segments, to keep it consistent.

Hey, have fun learning and trying what you can. I have faith your persistence will pay off.

[wintry apex]

big question.

When you use Legacy Personas - does that count as Style Influence - or Audio Influence?

cause it says 'style' persona

vs vox

[polar spade]

It is not about "copyright" though that is the generic notice. Its about Suno already having that audio stored in its database that the censors ping that it is clearly not something created and exclusively owned and solely controlled by the users.

As you mentioned, yours is something that is often used in classical music. Which likely means its been recorded somewhere. So somewhere out there , its assumed, someone already holds a derivative copyright of a public domain piece. Our attempt at making another derivative of the same thing and adding repetition to the database is not within the scope of Sunos practice to continue creating unique outputs.

When others have equal access to use an audio or lyrical piece, as well as recognized as publicly available, then it is not permitted in Suno's database and is blocked.

[wintry apex]

You know eventually after avoiding existing music so neurotically we will end up with musical mush that sounds Horrible.

music only sounds good because we have a small number of sounds, melodies, and themes that sound good to our ears. if you avoid it, its not music

maybe not now, but some day it will be unlistenable.
Music is a narrow corridor of culturally reinforced patterns.
If an AI bans every familiar pattern, it eventually bans… music.

Suno’s logic taken to the extreme = “anything pleasing is too similar to something.”

[polar spade]

Possibly. That's Suno's error to make, I guess. 😁 So far, some people still like the results.

Making music out of existing material does seem counter-productive to me. Its fun though, id agree. Thats why we have "traditional" music because we all want to feel attached to something we cherish as a culture.

[wintry apex]

I like the results when it does resemble something recognizable 😛

when it doesnt its garbage and i gen again

not 100% similar obviously

for me its more about existing material being the canvas that you paint over

the texture of the canvas but not the color

or vice versa. it just gives the ai a boost in creativity so it doesnt all sound like commercial pop (but thats my workflow cause I feed in futuristic psytrance and turn it INTO post-pop lol)

[polar spade]

As far as I can tell, (according to the devs) personas are always influenced by the audio slider. So using the Legacy just includes the instrumental or "whole" song rather than the Vox portion only.

Theres IS though the concept of , if using a whole song persona (any which are deemed Legacy -- whether old personas or new) then having audio near 20% seems to emphasize the vocal characteristics over sliders near 90% which really pumps in the instrumental influence in addition to vox -- when adding personas to a Cover.

Style prompts (and lyrics) then become an important variable.

[wintry apex]

When i use legacy personas I get this weird thing where it starts reading the Style prompt out loud as if it were lyrics

[wintry apex]

ill experiment with the 20%.

So far a good setting for legacy seems to be, but which is definitely not tested that much yet, is

50% weirdness
80% style
50% audio

[polar spade]

Its hard to tell someone else, "this will work" when they find something works well, also. So, even genre might have some influence, but theres too many genres to test every combo of sliders. Each person just has to find their sweet spot. I agree.

[knotty stag]

That has happened to me, it seems that Suno does not distinguish between copyrighted and public domain. I get around it by rewriting some parts, and using phonetic spellings for the public domain lyrics I might want to use. Obviously I never upload copyrighted lyrics.

[sullen kindle]

does anyone ever have issues downloading their work? I exported the whole song, went to download, it opened the download screen, but clicking the download file button does nothing, as if greyed out

[polar spade]

It doesnt "react", at least for me when clicking, so its odd, yes, I just know that in about 30 seconds to a minute, depending on file size, typically about 35MB, the save file panel pops up. So i dare not click twice.

[sullen kindle]

Thank you, you're right

[knotty stag]

I have that too, and if I click twice, I get two downloads.

[halcyon sundial]

is there some way to speed up only a section?

[frozen dagger]

you can double the speed of a clip, or can adjust the BPM and have one clip follow the BPM you set while other track follows it's original (prolly the quickest) and then there is the option to adjust the speed of clips with the right click remix/edit and then adjust speed. (but this works for saved clips, would need to export and import back into studio, but depending on your needs that's a couple options within Suno.

[bright haven]

Is it weird that suno is using a different voice when Im using a persona but making a slightly harder track

[frozen dagger]

if i use a calm talking persona in a heavy metal song i can expect some screams that aren't present in the persona.

[polar spade]

Yea. Dont do that. I burned and learned as well. 😁

[polar spade]

hmmmm... "CALM" in Suno means, "Channel All Loud Moments" through the vocals?

[plush notch]

Is anybody having issues with changing lyrics in song editor/studio? It will not change the lyric.

It seems I’m just wasting credits for nothing - ugh

[polar spade]

Over time, it seems some rate it fine for the Editors to change lyrics. That can include Covers making small changes.

As for Studio , it seems to do OK for me , personally. Moving Slider high on styles might help if not done.

I;m trying a quick test.. hold on

[plush notch]

Ok

[polar spade]

It changed lyrics quite well in Studio.

I randomly selected, literally 17 measures of lyrics.

With two Covers, they both replaced the selection with lyrics I typed, same voicing, and instruments, it picked a melody from the song approximating , sang for lines of lyrics through and the remaining was instrumental and both finalized/ended.

Two Replaces actually blended the melody right into the change beautifully and just replaced the equivalent lyrics replaced and blended back into what would be the next lines in the original song.

Sliders Style and Audio: 85%
Weird at a random low 6%

[plush notch]

Ok I’ll try 85% and 6% thanks

[polar spade]

What is happening to "change the lyrics" when something does happen? And what does "change the lyrics" mean. Are we replacing a full stanza, some words in the lines? Rewriting the whole section?

[polar spade]

Ahhh.. OK. .Try that first then, if preferred.

This is the lyric change on a Replace, It did its best to fit the lyrics, about 4 measures, into the 17 measure segment. Not what we may want, but again, I did a quick and random lyrics and segment choice, so it did quite good.

[plush notch]

Trinity change one line - the version I like sand the same line twice back to back - I want to change and fix the second line

Oops trying to

[polar spade]

Preferrably you want the lyrics to be as close to the syllables and metering of the lyrics being replaces so that it can blend back into the song properly.

[plush notch]

It sang this lyric twice back to back (Spanish) cuando el pullso llama / desde interior
But that line should be followed by a different lyric which is
Se junta las almas/un solo latidor

So I’m trying to take out that repeated lyric and out in the correct one

put in the correct one

[polar spade]

Is this, changing sliders up, an improvement over other attempts.

[plush notch]

I’m try the sliders right now

[polar spade]

The two lines appear to be slightly less in syllables, so when Suno tokenizes them, the changed words may fit withing the space, but the melody may change (or be similar), we'll see.

[plush notch]

12 syllables in each

13/12

[keen fog]

anyone else have an issue with suno hallucinating lyrics and a description for an uploaded audio clip? I'm having an issue with uploading a plain beat that has no lyrics of any kind, no vocals. When uploaded, suno is hallucinating lyrics and adding a permanent style description that says the clip has male vocals. When I try to use this clip anywhere, extend/remix/inspire/sample that permanent male vocal tag poisons any outputs that include this clip and converts a song that has female vocals, to now male vocals. I've even tried remixing the clip and stripping out the style description leaving it blank, along with blank lyrics but the remixed clip still somehow has the original style description embedded somehow

[frozen dagger]

sounds like you just want to select the gender of the voice you want in advanced options. (unless you're free plan, then that is not an option)

[keen fog]

yea it ignores whichever gender i choose, even adding male vocals in exclude, still male vocals

pretty annoying, never had an issue with suno hallucinating non-existent lyrics into an uploaded audio clip before

[frozen dagger]

you have tried choosing in advanced options?

[keen fog]

yep

still outputs male vocals lol

i wish suno would let you edit the description it tries to add to your uploaded clip before you save it rather than permanently tagging a clip with whatever it wants

[frozen dagger]

that is interesting indeed. the gender selection tool is usually pretty good. especially if there were no vocals in the original audio. (negative prompts can sometimes add instead of exclude, presume you've tried different combos before coming here tho, and already tried not having male in the exclude)

[keen fog]

yes ive tried every combo i can think of, ran out of ideas and was hoping i wasn't the only one

I think its because in extend mode, it really trys hard to preserve the original parts of the clip you are extending from

because the clip is tagged as male vocalist, it doesnt want to change it

[frozen dagger]

It is an odd thing if it is recurring. You tried re-uploading the audio to see if it will auto-generate a different description?

[keen fog]

yea I've tried a few clips, they all get hallucinated lyrics and a male vocalist description

This is what suno tags the clip with

A high-energy, modern pop-rock track with a driving 4/4 beat at 128 BPM, The arrangement features crunchy, overdriven electric guitars playing rhythmic power chords and a prominent, melodic bassline, The drums are punchy with a crisp snare and active hi-hat patterns, A male vocalist delivers an enthusiastic, mid-range performance with slight grit, supported by layered vocal harmonies in the chorus, The production is polished and bright, utilizing subtle synth textures to fill the mid-range and a slight reverb on the vocals for depth, The song follows a standard verse-chorus structure with a dynamic build-up leading into an anthemic, wide-stereo chorus

wrong genre, wrong instruments, and absolutely no vocals of any kind

[frozen dagger]

approximately how long is the audio?

[keen fog]

20 seconds

[frozen dagger]

have you tried running it through Samples first?

Just to maybe get teh sound with a different description is all i was thining.

[keen fog]

as in using it as a sample? Yes, so as a sample I can get it to use female vocals, but, it loses a ton of fidelity from the original clip and is more of a close approximation. Its what I'll end up having to do for now since it doesn't seem like there's any fix here without Suno allowing style box editing after generating

[frozen dagger]

Would reccommend the old remove style descript with Studio but that it patched out.

[keen fog]

yea...bummer

[frozen dagger]

i mean try running it through the sample feature, does better with shorter audio and you may be able to adjust sliders to create similar sample sound with new description on it

[keen fog]

yea that worked

silly extra step but whatever, got mostly was i was aiming for

thanks!

[frozen dagger]

That's awesome. (Not the extra step, although that is needed sometimes.)

[bright haven]

I am convinced the style thing does not listen to any instruction at all

tell it to avoid crowd chants

almost the WHOLE song is crowd chants.

[frozen dagger]

Even mentioning what you do not want, risks having it appear.

[polar spade]

OK. mention tomorrow then what the results were so far.

[bright haven]

anyone got any good suggestions for making a demonic style voice

almost like possessed style tone, Trying to do a song involving anxiety (even if not suno i'd like pointers if possible lol)

[frozen dagger]

i have an idea that is more fun than prompting [demon voice] but i wanna test it and see how well it works.

[polar spade]

satanic, seems to ge a good prompt. But you can also try recording an audio of anything you can muster that sounds similar and see how Suno treats it on a cover, along with prompts like satanic, demonic, growl, menacing, etc.

[bright haven]

so if im only trying to add that, what should I set weirdness style and audio to?

the only thing I can think of that matches what Im thinking about is Ice Nine Kills - IT is the end where he says just reach your hand down here and TAKE IT

that like.. demonic low voice while still having the normal voice as well

[frozen dagger]

im experimenting with what i hope works. in mean time. can you share a link with timestamp for the part in INK IT youre refering to. scanning through it to try and hear an example

[bright haven]

35 seconds in

Or actually Tom Mcdonald did it on Best rapper ever in the beginning of it when his alter ego is talking

[frozen dagger]

there is another trick to achieve a similar effect. not what im currently working on, but i did kinda... Transpose. in studio take the vocal stem and transpose it down a few semitones.

if you want more tom macdonalds version you can transpose one vocal and later the other on top so it creates the clear understood vocal with the possesed voice behind it.

[bright haven]

Ooooh

[sweet ivy]

Still looking to fill a gap in the tracks after I moved them all to the right, to get them back on the beat. Could time-stretching or elastic audio capabilities help, or will I need to stretch the audio from the preceding bars, do cross fades. Anyone have any ideas to do it?

[tidal dagger]

Whoops wrong chat sorry y'all

[frozen dagger]

well... my other idea is not working as intended thus far. used sounds to create a demonic voice, duplicated and saved into a persona. meh. Persona been acting wonky for me lately. But the results so far do not make me hopeful this will work well. Using "Sounds" to create some cool demonic voices for adding at when you're at the end of production might be cool. Turning it into a Persona (at least for me) did not work (so far).
Still, could use Sounds to put to create some 1-shot demon vocals.

[shut wave]

This is what i wanna do when suno doesn't use the right words

[strong sandal]

Why can't I upload audio? I thought maybe the file was too big but I compressed it down (438 kb from 4MB) and still nothing. Do I need the paid version or something?

[neat nexus]

How long is it? Free accounts can only upload to a minute. Paid accunts can upload as much as eight minutes of audio.

[strong sandal]

Oh then that's why. Its a minute and a,half.
Thank you

[fresh spire]

can someone read this because l'm trying to see if it good to make a song about it?

2:07 A.M.

(Soft piano intro, distant echo)

Verse 1 (Sad + Vulnerable):
The clock keeps blinking in the dark
Like it knows I’m falling apart
Your side of the bed’s still cold
But I still reach for you in my sleep

I told my friends I’m doing fine
Said your name like it don’t rewind
But silence hits the hardest when
It’s just me and what we’ve been

Pre-Chorus (Emotional Build):
I’m tired of pretending I’m okay
Some nights I still break

Chorus (Raw + Honest):
I still think about you at 2 a.m.
When the world goes quiet and it sinks in
I don’t call, I don’t text
But your memory’s heavy on my chest
I don’t want you back like before
But I miss the way you loved me more
Yeah, I still think about you at 2 a.m.
And I hate that I still care

Verse 2 (Healing Starts):
I deleted pictures off my phone
But healing ain’t a switch you turn
I’m learning how to sit with pain
Without running back to your name

I’m not weak for missing you
I’m just human going through
A love that felt like everything
Now it’s just a memory

Pre-Chorus 2 (Stronger):
Every tear I cried alone
Built a backbone I didn’t know

Chorus (More Power):
I still think about you at 2 a.m.
But it don’t break me like it did back then
I don’t beg, I don’t bend
I don’t lose myself again
I don’t need you to feel complete
I found strength in missing you quietly
Yeah, I still think about you at 2 a.m.
But I’m choosing me instead

Bridge (Dramatic + Cinematic):
One day I’ll hear your name
And it won’t hit me the same
I’ll smile and mean it
I’ll breathe and feel it
The love we had was real
But so is the way I heal

You were a chapter, not the end
And I won’t lose myself again

Final Chorus (Confident + Anthem Energy):
I still think about you at 2 a.m.
But I love myself more at 2:10
Yeah, you taught me how to fall
But I taught myself to stand tall
You were once my everything
Now I’m my own healing
I still think about you at 2 a.m.
But I’m stronger than I’ve ever been

[eager spruce]

hi after a little help. style box is 200 chars lyrics 3k and neg style 200 but is there a cap to the box with the lyrics mode set to auto for the lyrical vibe input? i know in the ai write full song part its 200 char is it the same when not in that window?

[restive pebble]

Make it!

[sonic folio]

Has anyone tried to get faint sobs before?

[polar spade]

I agree. Just make it. No one is going to critique your poetry/lyrics. If you like it, thats what counts. And then after the music , melody, genre is added, you may have a new opinion about it first.

[frank mortar]

Hey guys can someone dn so I can send them my lyrics for feedback?

[neat nexus]

Sorry, perhaps I was unclear....

I meant you could try posting your lyrics here to see if anyone responds. (No one is obligated to if they don't want to.)

But please do not keep requesting that people DM you. That is not something we encourage. Thanks.

[frank mortar]

Oh sorry

[neat nexus]

Whoops. I see you ran afoul of the word filter. Try removing the swearing in the lyrics? Otherwise the bot absolutely will eat anything with those in there.

[censored] probably would work in their place.

(Sorry about this, but we do some things to make sure that Discord itself doesn't make this an adult only Discord and force everyone to have to jump through hoops to be here. So Suno wants this to remain at the level of PG\PG-13 discourse on the chats.

Website is their own and has different standards, but with Discord we have to follow what Discord will let us do.)

[frank mortar]

Oh sorry

[neat nexus]

Glancing it over I don't see too many issues with it. Is this intended to be rapped? Might need to use more rhymes. Otherwise two issues I see off the bat are lack of organization and maybe possibly issues with pronunciations on some of the emphasized words.

[Intro]

[Verse 1]

[Chorus]

[Verse 2]

etc are very useful for organizing the flow.

You can even do things like:

[Verse 1]
[emotional male vocal]

or

[Verse 1]
[rapped vocal cadence]

and so on...

The pronunciation issue can only be worked by trying and adjusting, trying and adjusting.

[frank mortar]

I have dyslexic so my spelling not great

[neat nexus]

Oh! My apologizes, I thought you were using phonetics and emphasis to try to show emotional queues.

[frank mortar]

And I’m a bit slow but besides thar

No I just am slow

[neat nexus]

I do that sometimes to make my songs do vocal effects and I thought maybe you were doing that too.

[frank mortar]

Hahaha no sorry

[neat nexus]

Well then the best advice for that is to use a spell checker. Or if you have access to a free one, use a LLM that you have instructed that you are checking for spelling and ask it to interactively help you correct the text.

Because trust me the AI will try to sing it as written.

That's why the phonetics trick works at all.

And why sometimes to get that extra bit of vocal I'll change out words that have homonyms, like using Mourning instead of Morning to make sounds shift a bit when I try to push emphasis into the vocals.

[dark kraken]

Is there a technique to add sound effects into the lyrics still? I used to use asterisks and it would work SOMETIMES but can't seem to make it work now. Brackets don't work either for me.

[neat nexus]

I think that was always only a luck thing. I still prompt for it but rarely achieve it. You might have better luck generating sounds with Suno separately then using those in a DAW where you want them.

[dark kraken]

yeah i was trying to have traffic and car sounds weaving into the instruments...

[wild pine]

Hey guys. How to make cover version, but 2-3 times longer? Without extend function (it's quite buggy and like to show error, sometimes)

Sometimes it making 6 times longer, sometimes same length. Maybe there prompt command for that

[glossy idol]

Hey, can i send my lyrics here for review? thanks

[high snow]

Go for it

[high snow]

Have you tried saying that in brackets that you want the length a certain length?

I’ve never tried it, but it’s good to try and see.

[stuck kindle]

Does anyone have any tips on applying short lyrics in the editor? I'll open my track in the editor, highlight a specific section let's say 4 seconds then add lyrics, something small as an example This is the life, a lot of the time it will cut off some of the last word and will add This is the l, other times it will only add half of the lyrics so This is then other times it will just straight up replace half of my track instead of the highlighted 4 seconds. I've wasted a ridiculous amount of credits trying to get this to work but no result, I've tried playing around with the selection filter changing it from smart to fixed but still nothing. Any tips or guides would be very much appreciated

[solemn topaz]

SUNO BEGINNER GUIDE — PART 1/2

​SUNO BEGINNER GUIDE — COMPLETE STARTER CHEATSHEET
[Custom Mode • Structure • Syntax • Ad-libs • Tweaks]

[Note: Based on personal experience using Suno. Hope this helps fellow creators!]

​How to Think About Suno?
Suno works best when you treat it like a singer and producer following directions. You are not just writing lyrics; you are guiding how the music should behave. Clear direction equals better songs.

​1. Always Use Custom Mode
If you want control, use Custom instead of simple generation.
​You control lyrics and structure.
​You influence vocals and mood.
​Results become consistent and intentional.

[Auto mode = AI guessing. Custom mode = YOU directing.]

​2. Always Use Structure
Label sections clearly to prevent skipped parts and improve transitions.

Example Structure:

[Intro]
[Verse 1]
[Pre-Chorus]
[Chorus]
[Verse 2]
[Bridge]
[Final Chorus]
[Outro]

​3. The Two Main Brackets

​[ ] = Strong instructions [AI listens seriously].

​( ) = Soft ad-lib hints [can be ignored].

​4. Using [ Square Brackets ] — Strong Control

Use [ ] when you want a specific action to occur.

​Use for: Song sections, vocal style, instruments, mood, energy changes, or delivery.

​Examples: [Intro – soft piano], [Verse – intimate male vocal], [Chorus – full drums], [Beat drop], [Whispered vocal].

[Do not use square brackets for lyrics, as the AI will treat them as instructions rather than lines to be sung.]

​5. Using ( Parentheses ) — Ad-lib Hints

Parentheses are mainly for ad-libs or vocal flavor.

​Examples: We rise tonight (oh-oh-oh), Stay with me (yeah).

[Note: Suno sometimes ignores parentheses; they are suggestions, not commands.]

​Reliable Ad-lib Method:

[Step 1: Use instruction in square brackets.]

[Step 2: Write the ad-lib text underneath without brackets.]

​Example:

[Crowd response]
Oh oh oh!

SUNO BEGINNER GUIDE — PART 2/2

​6. Describe Sound, Not Just Genre

Avoid vague terms like "Make a sad song."

​Better: Slow tempo, warm piano, intimate vocal, minimal drums, emotional atmosphere.

​7. Control Energy Changes
Tell Suno when intensity changes so sections do not sound identical.

​[Verse – minimal instruments]
​[Chorus – full energy]
​[Final Chorus – layered vocals, bigger sound]

​8. Repeat Important Instructions
Suno can forget earlier directions. If a specific style matters, repeat the tag [e.g., Deep male vocal] before major sections.

​9. Keep Lyrics Rhythm-Friendly
AI sings rhythm better than dense poetry. Use short lines and natural speech flow. [If you can clap to the beat of your lyrics, Suno can likely sing them.]

​10. Use Vocalizations Smartly
Avoid writing (humming). Use rhythmic sounds like "La la la," "Oh-oh-oh," or "Hey! Hey!" for better melody generation.

​11. One Idea Per Song
Avoid mixing too many styles. Stick to one core style with small influences [e.g., Melodic techno + cinematic atmosphere].

​12. Iterate Like a Producer
​Generate.

​Notice what worked.
​Change one specific thing.
​Regenerate.

​13. Common Beginner Mistakes
​Not using Custom mode.

​No section labels.
​Using [ ] for lyrics.
​Trusting parentheses too much.
​Expecting perfection on the first try.

​Safe Start Template:

[Intro – ambient pads]
[Verse 1 – soft vocal]
Lyrics here...
[Pre-Chorus – tension builds]
Lyrics here...
[Chorus – energetic, layered vocals]
Main lyric Oh oh oh!
[Verse 2 – stronger rhythm]
Lyrics here...
[Bridge – stripped instruments]
Lyrics here...
[Final Chorus – biggest energy]
Main lyric Hey! Hey! Hey!
[Outro – fading atmosphere]

​Golden Rule:
If something MUST happen, use [ ] plus a written line. If it is just flavor, use ( ).

[solemn topaz]

This reads like something between modern emotional pop and therapy-core storytelling.

Big strength: emotional authenticity.

Big weakness: predictability in imagery and phrasing.

In short: strong heart, slightly familiar vocabulary.

[robust fable]

In Studio, what is the most effective way to get re-takes of a solo lead vocal stem from the original vocal stem separated from a cover? I just want to get the same exact (SOLO) vocal but without the digital noise created by the stem separation process AND without sounds from other stems. Studio keeps creating lanes in my vocal track that have the entire multi-track mix on it and/that sounds nothing like the original vocal. There's no audio similarity slider in the Studio tool bar.

[solemn topaz]

Hey, I sent you a quick DM with some formatting suggestions for your lyrics — check it when you get a moment 🙂.

[quiet solar]

@solemn topaz Some really good advice there.

The only thing I'd add is that [] instructions are also suggestions that can be ignored, and will be - especially if they deviate a long way from what's typical in the genre(s) you're working with. General suggestions in the style prompt are often the most reliable way to influence the song, but stage directions in [] can be an effective way to attempt to influence a specific part or line within the song.

Although Suno can interpret () as being style prompts, because it also interprets it as backing vocals and ad-libs, it's generally a good idea to get into the habit of using [] for prompts and () for vocals in any event.

[solemn topaz]

That’s a really good addition — and yeah, I agree.

[] definitely aren’t hard commands either; they’re more like strong suggestions. I’ve noticed they work best when they stay within the logic of the genre or arrangement — if an instruction fights the style too much, Suno usually ignores it and keeps musical coherence instead.

Also completely agree that the style prompt tends to be the most reliable overall influence since it shapes the song globally, while [] feels more like local steering for specific moments or lines.

What I was mainly focusing on in the guide was tweaking inside the Custom lyrics field — using [] there to nudge delivery or structure for particular sections, and using () mostly for vocal ad-libs or responses. That separation has made iterations easier for me when troubleshooting generations.

Appreciate the extra insight 👍

[quiet solar]

Extend is the best way to do that, since you can steer how much the song is extended by the amount of content you put in the lyrics box. Don't extend from the very end of the song, and anticipate there may well be some variations from the original despite your best efforts.

What you're looking for is for the song to continue on with a reasonably similar vocal and instrumental vibe, so that you can use Get Full Song once you're done with the Extend(s) and use that song as a starting point for covering or remastering to smooth out the differences so that it all sounds cohesive. It's a good idea not to get too attached to the specific sound in your original song as part of this process.

Note that if you extend the song beyond around 6 mins or so, you will likely run into some headaches with the cover/remaster process as Suno allocates some of the maximum generation time to audio influences and may not be able to cover longer songs in full. If that's the case, then your 'cleaning up' process will be quite a bit more complicated, and likely require some direct editing work.

[quiet solar]

You should consider putting it up under workflows in Suno Hub as a guide to using prompts in the lyrics box.

Actual solid advice based on reliable experience as opposed to LLM guesswork is very much appreciated by people looking to learn more about how to use Suno effectively.

[solemn topaz]

Thanks a lot — really appreciate that, means a lot honestly.

I’d actually be happy to put it under Workflows in Suno Hub, I just don’t really know where or how to post it there yet 😅 I noticed some people asking similar things in general chat, so that’s why I shared it here.

If you could point me in the right direction, I’d gladly format it properly and put it up as a guide.

[quiet solar]

Sure - the workflow channel is here: https://discord.com/channels/1069381916492562582/1252267444773326848

Just click on New Post and put things in mostly as normal. The UI is slightly different to a standard Discord post as it adds a title that can be seen for ease of reference by people scrolling through the section, but the body of the post is the same as everywhere else.

[solemn topaz]

Thanks for the suggestion! I’d really like to post this under Workflows, but I noticed there’s a 2K character limit while the guide itself is around 5.9K characters.

What would be the best way to share something this long there — should I split it into multiple posts, shorten it, or link it externally instead?

[quiet solar]

Splitting into multiple posts is probably the easiest way, since that'll fit neatly into three.

[chilly wharf]

Pardon my interruption, but is there any way I can tell Suno the difference between "live" (short i; alive, animate) and "live" (long i; as in "not playing back a recorded performance")? I'm not sure how to approach this challenge, and neither is Suno, apparently... - Thanks!!!

[remote ivy]

i'm getting lots of restarting the song w/ a different verse take on verse 1 rather than just ending the song cleanly...all renders on my last 2 creates...is just a straight up bug, or can i avoid this w/ some magic instruction....i already have outro fade out, etc.....and it sometimes inserts it between the final chorus and outro

[solemn topaz]

Thanks for the suggestion — that worked perfectly! I split the guide into three parts and posted them in the workflows channel as you recommended. Really appreciate the help 👍

[quiet solar]

Live vs Lyve should give you the distinction. As should li've vs l'ive.

Have run into that problem many, many times before. xD

[chilly wharf]

Thank you!!!

[frozen dagger]

Pronouncing "Live" when it rhmyhes with hive is where I had Suno problems. "Lyhv" worked. It is probably the single-most problematic word (heteronym) in Suno. This problem is exactly what brought me to Suno discord last year. lol, otherwise "lihv"

[quiet solar]

Yeah, Suno does sometimes get the context from the sentence, but more often than not it'll speak Li as you would in "Lift" and Ly as "Lie".

[quiet solar]

Just to add, you can put the second and third parts as further posts under the same workflow as the first one, so that it's all in one place and reads like your original did here. 🙂

[solemn topaz]

Really appreciate the tip — Now everything’s together under one workflow instead of the 3-part hassle. Much cleaner and easier to read 👍

[bright haven]

If I upgrade to Premier while on the pro plan will I only play 20 instead of 30 and will I get the credits?

Also side note. Can someone help me come up with a style prompt for a Similar style to Avenged Sevenfold - Doing Time?

[quiet solar]

It pro-ratas the cost, so you should effectively get the unused part of your Pro subscription refunded against the cost of Premier.

It'll also reset your billing cycle, and your credits will reset to the new Premier level starting from the upgrade.

[quiet solar]

That might be a better question to put to Google Gemini (or your preferred alternative) to give some starting suggestions to work from. It's quite difficult to write style prompts to a song level, rather than a band/genre level, so expect it'll take quite a bit of tweaking to get into the right space, and you might have to look for associations outside of the obvious genre/sub-genre prompts.

[bright haven]

I cant even get it to be avenged sevenfold style lol

[quiet solar]

Yeah, Avenged Sevenfold are quite distinctive, and Suno isn't really intended to be used to mimic existing artists (hello, gigantic copyright headaches). Getting into the right sort-of ballpark space around genre/sub-genre, mood and instruments is a challenge in it's own right.

[hollow jay]

There’s a new feature OG users don’t get

• on new accounts you get to merge multiple covers in one go including persona

You can’t do this on OG accounts

[frozen dagger]

Sounds like Inspo+ / Inspiration. Is this what you're referring to?

[hollow jay]

No, it’s remix when I generate (label)

Maybe they’ve merged it, but when you add cover on new accounts you can use the slider to select the timeframe too

[hollow jay]

This doesn’t show a slider to select time frame

[frozen dagger]

Some of the UI features are rolled out. That looks like the same features with new UI arrangements and elements. Things are always changing. Seems mostly the same just rearranged. Sounds like you're saying the Cover has a selectable range as well now with the UI there, similar to Sample. That's cool.

[quiet solar]

That looks like a small UI change rather than a new feature?

Hah...OBLMOUS, we need to stop sharing the same brain cell. xD

[frozen dagger]

I don't mind having my brain cell being compared to yours good sir.

[quiet solar]

Yeah, I can see some potential use of being able to cover a portion of an existing song rather than having to cover the entire song, crop it and then finnick around with the start and end to get it to naturally start and finish.

I think the merging multiple covers together is the Inspo feature, though possibly with a UI tweak. Being able to use Inspo to affect a Cover would be an interesting twist, though.

[hollow jay]

Grok came late to the LLM game but is the most accurate …

I don’t even like Elon

But grok is 🔥

[wise moss]

If I wanted certain sounds in the song instead of typing them out would I just put [Cricket sounds]?

[frozen dagger]

You can try, but it is not a reliable way. If you want crickets best way is to make the crickets with "Sounds" and then add them in with Studio or DAW of your choice, outside Suno.

[wise moss]

Ok I will mess around on those things

Thanks

[quiet solar]

Yeah, at it's heart Suno isn't a SFX generator - the new Sounds feature is a major steps towards that, but it currently needs to be edited in and can't be called mid-song through prompting.

I'd probably try something more like [Cricket chirping] rather than 'sounds'. The more specific you can be to help Suno attempt to look it up the more likely it is to work (although the average chance it will work is still low).

[frozen dagger]

I've done crickets so I know it does well. Cicadas also was nice. lol

[sweet ivy]

Well my "challenges" with Suno continue. I upgraded to Suno Studio Saturday and paid on the same credit card, but my credit which were at 2,300 left until March 5, dropped to 1,000 ? Suno support ( the robot) said they would pass it on to a human. Still have heard nothing or seen any adjustment

[ebon sand]

Has anyone any trick how to align lyrics in the editor that exceed 7k symbols, i cant edit anymore because it doesn't let me align it

Managed to align the spot i'm editing but didn't figure out a method that works, i've tried around for 2 hours to align it

[polar spade]

What does your billing history in your payment history show. You are describing what is typically an immediate upgrade and prorated account, not an update to activate at next refresh date.

[sweet ivy]

That's right a prorated upgrade. They charge the card, but some mixup with the credits. Hopefully someone gets back to me soon

[polar spade]

@sweet ivy You also describe the use of a new payment source. Have you refreshed at least once since being on the Pro plan, Ill assume on February 5th. And you describe as if you changed your card /payment information to do the upgrade? Otherwise, it would be a quick charge to your existing card on file.

If you've used a new payment source, keep in mind there would be delays from banks holding funds for about 5 days, not counting weekends and bank holidays in the U.S. #1417459973780148316 message

[sweet ivy]

@polar spade Appreciate your response. I used the same credit card on file. The prorated charge was there. I'll refresh again tomorrow, maybe the credits will be straightened out.

[polar spade]

It should be an immediate activation, as you suggest. Recently the 1,000 credits are "complimentary" while a users new payment info is being authorized, then bump to more credits after that. So, if no changes to the payment info, its a bit odd, but yes, log out and back in and check tomorrow. Refresh may not update all files to current, but try one then the other. Again, for whatever issue, it may have triggered the 5 day delay to review, if the expiry date on card changed, something different on profile doesnt quite match and so on.

OK. Good luck.

[small bridge]

is suno down?

[glossy idol]

(Chorus)
All of that for this?
I did all of that for this?
Bought you candies for all this
Now I'm frantic for my sins
Won't you stand me from abyss?
Lend me hand nor give a bless
Now I'm chanting from my place, saying
All of that for this?
I did all of that for this?
Bought you candies for all this
Now I'm frantic for my sins
Won't you stand me from abyss?
Lend me hand nor give a bless
Now I'm chanting from my place, saying

(Verse1)
standing in the dark, understanding all my part
Found a magic you're a star
Sound of static from afar
started sanding all my scar
Then I grab it try to run
Pack of candies from my aunt
They say all I do is hunt
They may call me for I'm young
Good thing, could they for the gun?
Won't they shoot me when I'm stunned?
Told me "put it and have fun"
But I couldn't, didn't learn
You'll do it be so stern
Called me stupid for my turns
Maybe we should go and learn
What they wish so what they want
See their vision but we can't
Feel the fusion of the bar
This the mission raise the bar
Miss the season chase the sun
Now we see the sunlight's waves from afar

(Chorus)
All of that for this?
I did all of that for this?
Bought you candies for all this
Now I'm frantic for my sins
Won't you stand me from abyss?
Lend me hand nor give a bless
Now I'm chanting from my place, saying
All of that for this?
I did all of that for this?
Bought you candies for all this
Now I'm frantic for my sins
Won't you stand me from abyss?
Lend me hand nor give a bless
Now I'm chanting from my place, saying

wow lyrics

[halcyon sundial]

is there a way to only translate my song to another language?

like i want the instrumental, vocals and melodics exacly how it is

but in another language

its a close language so the lyrics wont change too much

[quiet solar]

Simple answer: get your translated lyrics, cover the song with the original style prompt on high audio influence and low weirdness and replace the original lyrics with the translated ones. This will probably change up the instruments, vocals and melodics slightly, but if you run through a few versions you should get one which is highly similar.

Complex answer: if you want to keep the instruments and melody totally unchanged, this becomes an editing job. Do the same as above, but you're looking to extract the vocals stems from the translated version and attach them to the instrument stems of the original song using Suno Studio or a DAW. You may need to run through a few versions if you also want an almost identical singer - using a persona when covering the song may help with this.

[polar spade]

I will leave this for @high snow to discuss, as they noted, but I would like to say, Suno can add music to any text for free. SO, just create a song by noting what style you want, or even leave the style prompts area blank and and then run it. See what you personally like or feel you need to change.

[glossy idol]

i made the lyrics myself. Theres also a second verse but server doesnt let me send it. I js need to know how to effectively use styles.

[polar spade]

Are you Basic or Pro subscription? If Basic , you are limited by just 50 credits a day, but that means going at this slowly, perhaps.

You are asking what type of style or genre these lyrics lean towards? So a quick and free means is ask directly to an avaliable chatbot such as Grok, ChatGPT, Gemini or any available. Its not so much needing a human touch to read the lines. An AI would do just as well and maybe better.

[glossy idol]

am a basic user. and thanks maybe i should use ai for style

[polar spade]

Honestly, AI would be very inciteful and remain objective. Good idea.

[glossy idol]

aight i see thanks

[elfin wraith]

may i ask why i can't find the daily tasks where you can earn credits?

[polar spade]

Click "More..." found bottom of your left menu panel in Suno under Labs

[elfin wraith]

ah sorry that's not what i mean
there's only "listen for credits", but i think there used to be tasks like "publish a song" or "write your own lyrics"

[elfin wraith]

ah ok, thanks

[quiet solar]

They were a temporary thing for a while which isn't happening at the moment, as far as I'm aware.

[elfin wraith]

oh i see
thanks!

[sleek stump]

I didnt see it noted, but lets not forget cadence.
Different languages can have wildly different cadence and trying to just do a 1:1 replace rarely ever works like one wants. Suno "will" change something, the melody, arrangement, make the alternate language lyrics too fast or slow trying to fit them to a melody that was designed for a different rhythmic flow, or a combination.
Expect melodies to change, sometimes dramatically, when trying to just do a translate.

When I write songs with Gaelic lyrics included, I have to gen the song with them because they have a very different cadence from English. I have tried to do translations after the fact and it never works out as I hoped.

[ebon sand]

Does anyone have tricks to align the lyrics in the legacy editor when they exceed 7k symbols, i need to edit things everywhere but have trouble aligning anything by now, it often will wipe the lyrics when i try to align

[echo citrus]

Is there a feature that allows you to set a max duration for a song?

[half dragon]

Unfortunately there isn't.

[echo citrus]

Thank you!

[quiet solar]

They mentioned that the languages were fundamentally similar, so I didn't bring that up - but you're absolutely right. In most translations, you can't really aim for a literal translation from one language to another because it'll tear apart the rhythm and rhyme scheme.

[ebon sand]

I think it worked to some degree to remove all the consonants to fit the lyrics so that it can align them, at least for one more replacement, i hope it works for all after

[quiet solar]

Best bet is probably to do the edits in an external document and then paste them in.

[ebon sand]

I have everything in a text pad but my lyrics are 11k and i need more space, i have trouble to align them when text is missing, it tries to spread the 'too little lyrics' over the whole song and doesn't align correctly

and then i cannot edit

the limit used to be around 6-7k when it started failing but now it's down to ~4,8k, when i paste more it will wipe them

[quiet solar]

You have 11k character's worth of lyrics... 😵‍💫

Are you just trying to edit how they're displayed in the final version? Presumably you can't do meaningful amendments to the song when you're thousands of characters over the lyrics cap?

[ebon sand]

no i'm working on the song, it's more of a concept song and currently 30 minutes long, i'm using the editor

but you can only replace when the lyrics are aligned

[quiet solar]

That is genuinely one of the most interesting ways I've seen someone make life incredibly difficult for themselves in Suno, and I speak as a veteran of inflicting that on myself. xD

Do you have access to Suno Studio?

[ebon sand]

yes, but studio doesn't work for replacing at all and doesn't have a model selector for replacements or any sort of alignment

[quiet solar]

True, but it might hold the key to doing this in a simpler way. One moment.

[ebon sand]

i have tried it just yesterday but i'm all ears

[quiet solar]

Yeah, just did a quick test and looks like it should work.

If you load the track in Studio, you should be able to select half of the song and move it apart, splitting it in two. You should then be able to save the versions as two new and separate songs, with the lyrics (hopefully) intact. You can then do any edits you need on those in the basic editor, and then stitch the whole thing back together in Studio afterwards to create a single in-Suno song.

[ebon sand]

I don't really want to use studio, it degenerates the song quality

[quiet solar]

That'll keep both versions under the 7.5k cap, so it should fix the overarching problem you're running into

In this case, you're only using Studio to split the song in two and then reattach it. Nothing gets generated in there, so there's no obvious reason why the song quality would be affected. Still, it's ultimately up to you

[ebon sand]

it degenerates the quality even when you do only save and make no changes, it does a reprocessing round on the song

edit v3 is the same

i wish they had an import in edit v2, that one leaves the songs as they are

[quiet solar]

I've just tested a no-changes save and .wav export from Studio and compared it in a DAW to the original song that I uploaded. The waveforms are identical.

If there is a change, I'm not sure a human would be able to hear it.

[ebon sand]

i invert and layer, if it doesn't cancel out to silence it's changed

[quiet solar]

I guess your mileage may vary by the song itself, but in this test there was some slight percussion echo and a tiny bit of distortion hiss after inverting. Listening to the two tracks with headphones I honestly wouldn't be able to tell you that they were actually two different tracks, though.

[ebon sand]

Yes that's what I mean, it also renders the songs incompatible for any further processing with everything else than v5 🙂

i'm creating on v3.5

[quiet solar]

Ohhhh, right...okay, that puts a different spin on things. I was about to scratch my head and say "Whaddya mean Studio is incompatible with further processing in v5?!"

[ebon sand]

I wonder why the limit is lower, it used to be able to align up to 7500

[quiet solar]

I'm guessing that odd things may happen as you approach the limit. That seems to be a common theme in Suno elsewhere.

Sticking with the same underlying principle, you could try using the editor to crop the song into two different pieces and then reattach them in an external DAW or audio editor afterwards. Although I don't know if the lyrics alignment would correct itself if it's broken in the version you're cropping from, it's been a while since I used the legacy editor and I never used it all that extensively in the first place.

[ebon sand]

The lyrics alignment is always fixable, it doesn't permanently break, it's this feature here, the problem I'm having is that it either removes the lyrics from the box or it has a hard time finding the words at a certain point, but it was easier when there was ~7000 symbols because if you leave lyrics out it can find the correct spot better if there's enough 'around it'
Splitting generates new problems, for example that the parts aren't finished and that I would remove the context window, and I'd probably need 4-5 splits in the end
if there's a better way to get it to align for example a string of dots or minuses as replacement for words that would be my preference

[sweet ivy]

@quiet solar. Suno seems intent on frustrating me. First the billing issue when I upgraded to Studio on 2/20. Yesterday I went into Editor to change a word that was wrong. Followed the instructions, using the whole line, not just the wrong word at the end. Only that line was highlighted. Hit Replace and both gens had the singer singing the same word, incorrectly. Saw someone on YouTube saying using V4.5 was better than V5 for Replace ?

[ebon sand]

Try it in the legacy editor if you used the new editors, the old one has a higher success rate for replacements
If you used legacy I can likely help, i'm in the editor all day

[sweet ivy]

Legacy is that V 4.5 the one just before 5?

[quiet solar]

I'm not sure there's any way to give Suno something that isn't the lyrics and yet have it correctly align that against the lyrics. Your idea of using just the vowels or consonants was a pretty smart potential workaround.

If what you're trying to do is rewrite the lyrics, I'd personally be tempted to re-do the section using Extend, take the vocal stems and then stitch them back in using an external DAW, but I think our preferred workflows are so far apart that I'm not sure it's a helpful suggestion. Best of luck to you though. 🙂

[ebon sand]

it's actually hidden or you change it in the address bar manually to edit-legacy
alternatively you have to click there when the new editor loads, no kidding, there's an invisible button

[quiet solar]

What's the word that's giving you grief? Because it's quite probably a phonetics issue that you'll need to solve alongside re-attempting the line.

[ebon sand]

I appreciate it, but stemming 3.5 isn't a good idea haha
I'm very happy with the results, just wished the editor wouldn't have been neglected, it's the most fun I have on suno but it's frustrating when you spend an hour trying to align the lyrics again somehow

[sweet ivy]

@quiet solar @ebon sand It's only one word at the end of a line, the AI singer sang dreams instead of dream, surely there is an easy fix using replace in the Editor mode

[ebon sand]

Try in the legacy, it's worth an attempt, it's easier to navigate for such replacements, you can try to tick manual lyrics replacement if automatic doesn't do it correctly, i always use manual, not sure in what state automatic is

[quiet solar]

That's...weird that it repeated the same mistake. The word was always 'dream', it wasn't 'dreams' in the original version?

[sweet ivy]

@quiet solar No, Suno changed the word in the lyrics. It read the vocal on my piano-vocal as dreams and put that on the lyrics it uploaded. I didn't read it carefully enough, Lesson learned, never let Suno grab the lyric off your guide audio, paste them in yourself

[polar spade]

I always thought the initial Create (and maybe then , Covers) is limited to 5000. Its extension that allow 7500?

I havent looked in awhile.

[ebon sand]

I have been working on this song for 4-5 months now, I only got the problems yesterday, didn't have them in... december i think was the last time I worked on it
here's a screenshot of the lyrics counter i pasted in december, now it's below 4800, at 4800 and more it reliably will delete my text from the box

[polar spade]

Thats not good then. Wonder whats controlling that. Hope it's not simply a new thing, but something fixable.

[sweet ivy]

@polar spade That billing issue we discussed last night. Nothing changed, so this morning I paid the same amount as I did on Feb 20 for the upgrade. That seemed to work & I got all the credits I should have, but it changed my renewal date from 3/5 to 3/20.

[quiet solar]

Ah, you can edit that in the displayed lyrics of the upload before you then do any further iterations on it. Always worth doing that, because Suno regularly mishears lyrics when it's trying to interpret from an upload.

[sweet ivy]

@quiet solar but surely I should be able to change that word now, with the Replace option ?

[ebon sand]

i hope so too, but also a bit unlikely that they will look at legacy and invest time haven't tried edit v3 lyrics alignment since they are separate but could maybe verify over there... if that one also should bug maybe it's up for a fix

ah nvm i cant load the song there, it has a 10 minutes limit

[polar spade]

OK. Glad you are current at least. If you felt you paid for premium and didn't get service for , I guess about three weeks, you can mention to Support when they respond. Request some extra credits .

[high snow]

Sorry, just getting to this now.

[high snow]

I like the lyrics but just wanted to let you know for the parts where you have verse 1-2-3, etc. I would put those in brackets because with the AI it’s a hit or miss of whether it will accidentally read that as lyrics. I would just put in brackets to be safe cause that’s what I do.

[high snow]

You could use ChatGPT to translate your song into another language. Because although the AI can read in other languages and singing them because I’ve done Arabic, Greek, Serbian, etc. those all worked fine but Suno itself does not translate songs. You hav to do that yourself.

But if you need help on translating things, I’m more than welcome to help you out. Just let me know or send me a DM.

[ebon sand]

yes you can do that with replace, when you do and are on legacy pick minimum duration for the replacement, no need to select the whole line, it's also cheaper, 4 credits per try

[glossy idol]

thanks i wrote it myself

[high snow]

There is a GPT that someone sent me and it helps with creating the correct styles that the Suno AI actually reads correctly. It’s basically it’s code language the best way of saying and it puts/formats the styles in a way whatever you give it to something that the AI will be able to read. Do you want me to send it over to you? Cause it’s helped me because I’ve always had a tough time with styles.

[glossy idol]

sure of course i'd appreciate it very much

[high snow]

I’ll send it to you in DM’s.

Also, you can also use it for reformatting your lyrics as well.

[glossy idol]

alright, i will be waiting till then

[high snow]

I just sent you the DM.

Let me know how it works for you.

[high snow]

Also, when you use that GPT for the styles and how it helps you out. You can also use the magic wand and sometimes that works wonders as well.

[sweet ivy]

@ebon sand So you are saying in Editor go all the way back to version 3.5 ?

[ebon sand]

No, the legacy editor is a completely different editor, the selections you're looking at are just the models

[quiet solar]

Yes - although consider it part of the learning curve that replacing lyrics is not always a straightforward exercise. If you can get them right first time, it'll save future you several headaches!

I'm not sure why it didn't like switching dreams for dream. If you're trying to do it through Replace Section / Replace Lyrics, then try adding an apostrophe (e.g. dream' ) to see if that shakes Suno out of whatever rut it's getting stuck into.

I'm not personally a huge fan of the legacy editors (credit to ΤΑΘΑΜΕΙ for sticking with them). If you open the song in Studio and split the vocal and instrument stems, you could run a replace on the vocals that way too, which I've personally found to be a lot more reliable.

[ebon sand]

You should use it and bask in its ancient glory

[quiet solar]

It literally drove me to doing all editing work in Audacity for the best part of a year. xD

[sweet ivy]

@quiet solar Yes, I thought about trying the lyric correction in Studio on the vocal stem, somehow got the impression in was a function handled by the Editor mode. Maybe watched too many YouTube videos on "Replacing a lyric" all illustrating how to use the Editor mode

[ebon sand]

You can forget about the standard editor, it was dead on arrival, I was so disappointed

Either legacy or studio although Studio doesn't work for replacements for me at all

[quiet solar]

Yeah, the UI from the main Library loads the non-Studio Editor to do it - which I guess is a lot more convenient for non-Premier users. But it's an option in Studio too, if you select that section of the track after you've split vocals and instruments.

[sweet ivy]

@ebon sand Someone, probably on YouTube said to use version 4.5. I might try that on a vocal stems as Kanthric suggested

[ebon sand]

I cannot really confirm that the model selection makes a big difference for which word is spoken, it just would sound differently if you use a model of lower version

[quiet solar]

@sweet ivy If you made the original song in v5, try sticking to v5 in the first instance.

Use the apostrophe route in Editor or the Studio route I suggested (use the apostrophe approach there too, it won't hurt). If the Studio v5 route doesn't work, it's pretty easy in Studio to switch the model to v4.5 and give that a try too.

The potential headache with switching down to v4.5 is that you risk changing how the vocalist sounds, and needing an additional step to correct that back after you've made the lyric edit.

[ebon sand]

Where do you switch, for replacements it always v5 for me in studio and the model selector on the create panel doesn't change it for replacements, is there a separate option?

[quiet solar]

Maybe I completely imagined that, one sec...

Ah yeah, it only works for Cover rather than Replace, because it needs to shift across to the Create panel. My bad.

[ebon sand]

No problem, I was hopeful, I have a V4 song that needs fixes but I remembered something was not working with that in studio

[sweet ivy]

Thanks @quiet solar & @ebon sand . I think I will try the ' on the vocal stem in Studio first. Do I highlight just the word dreams and change to dream' or a include a few words before it ?

[quiet solar]

Remaster to v5, Open in Stu- alright, alright, you do it your way. xD

[ebon sand]

Tried, doesn't work haha
It was a V4 remaster of a v3.5, sounds bugged when it does another round on it

[quiet solar]

I'd select at least the line rather than a single word - Suno works better when it has a bit of space to interpret what you're asking it to do. Surgical replacements isn't it's strong point.

If it messes up the timing in the fix, then provided it sings the word on the correct note you can always manually adjust it to be in the right place.

There are other ways around this problem, but they're starting to veer into direct editing techniques, which I always feel a bit bad dumping on someone new to Suno unless they're had a bit of separate experience in audio editing. If we can avoid having to go there, great!

[quiet solar]

Haha. Fair enough.

I have 3.5 and 4 tracks that would be great to unleash v5 on, but have had the same experience - they just won't make the transition to v5 neatly, and it's just too much effort to rebuild them piece by piece, knowing how I want it to sound.

[ebon sand]

I think the persona and sound I'm using is fundamentally incompatible with v5 somehow, i've never been able to get any remaster, cover, persona, vocals, guitar or anything I like on v5, extensions also change the whole sound immediately so that it's unrecognizable, hopefully the next model is more varied in its sound capabilities

and it can't speak the language i need from it

[sweet ivy]

@quiet solar I have stems up in Suno, what are the steps I take, from here if you don't mind sharing

[quiet solar]

Personas are...their own unique challenge. I'd certainly be wary of bringing across any non-v5 personas into v5, it's very hit and miss as to whether it plays nicely with them or not.

I've never had any huge beef with Extensions, but then I usually use them to generate samples for editing rather than to actually extend a finished or part-finished song. And if I am genuinely extending something, I'm expecting to remaster or cover it afterwards to fix the seams.

[ebon sand]

I used uploads actually with curated vocals from elevenlabs and chosen instruments and sections, persona is basically my prompt, but all after v4 werent able to reference it

[quiet solar]

Click on your original song to highlight it.
Under 'Clip' on the right, Stems -> Extract -> Vocal and Instrumental
Once they've generated in the box below, Insert All.
Click and drag over the bit you're wanting to change in the vocal stem
Expand the 'Replace' box, put in the new/replacement lyrics for the selected section
Under Advanced Options in the Replace box, set Weirdness down to 15% and Audio Influence to maybe 65% or so
Click Replace
Click on where it says Version (n) with a pink dot and press play to preview it.
If one of them is perfect, click the up arrow to copy into the main track.
Save as a new song, and you're done.

If it doesn't work that neatly with getting new versions, we'll cross that bridge when it comes. xD

@ebon sand That's an interesting way to do a slightly different kind of audio prompt. I admire your creativity!

[ebon sand]

Thank you! I'm just a bit somber because i know it will end in a few months, they have already announced last year they will remove all the models and with it my favorite

[sweet ivy]

@quiet solar Put the stems in studio, highlighted the area in the vocal track and ended up with a context window, two lines one either side & don't know how to get rid of them?

I close Suno & came back into Studio expecting those Context Window lines would be gone, but they are still there???

[ebon sand]

Studio autosaves everything

why do you want to get rid of the context window

[sweet ivy]

@ebon sand It's not the right size. Was it instantly created when I highlighted the section of the vocal stem where the lyric mistake was ?

[ebon sand]

the context window just tells you how much the ai takes into account for the replacement, your selection is the part that gets replaced, not the context window, it needs the context window to fill something into your spot

[dusky stone]

Hi guys, what is the purpose of REMASTER and should I use it anytime before to go into Studio? TY

[sweet ivy]

So I move it over the area I'm highlighting in the track, even though it's the wrong size?

[ebon sand]

the part in between the blue arrows is the one that gets replaced

[frozen dagger]

The context window should match your selection. It must I clude the selection, it can be dragged to include more context of the song, but not less than your selection.

[ebon sand]

leave the context window as it is preselected, it will be right for everything you'll do and you only move it when you have a reason to do anything special, I wanted such a feature but you really can just leave it as it is

[frozen dagger]

Remaster applies different techniques to try and "master" the track. Not really necessary before taking into studio. Most cases I use it to try and get small noise artifacts removed and then spicing them into the track if I need to.

[dusky stone]

ok thanks, and so can I remove artifact in studio and if yes how pls?

[frozen dagger]

Inside Studio most corrections are replacements, covers, EQ-ing them out, using the clip editor for time-stretch, spicing, or masking.

[sweet ivy]

@ebon sand This is getting bizarre. I happened to open My Workspace in Studio and it's filled Replacements from the 1st song I did. I assume I didn't see them before because I maybe didn't open My Workspace, just worked from the Library window. I assume I can delete them and not affect the final version of the song I ended up with ?

[quiet solar]

Yeah, there's not a single method. It depends what the artifact is as to what's the best way to remove it.

[quiet solar]

Yeah, Studio operates in it's own separate space. Nothing affects songs saved in the Library. When you're done with Studio edits, you can save a new song into the Library with them in place.

[frozen dagger]

You can change create a new work space and work within it so your replacements and gens for that project will be kept together. My Workspave is default folder. The only folder thay doesn't allow pinning tracks to the top as well. Mostly useful for Create tab tho. (Pinning I mean), but always good to try amd stay organized.

[dusky stone]

guys how can I tell suno studio not to start from 1 beat but form 0...because it's cutting part of the piano intro, ty

[small bridge]

is the site having issues right now?

[high snow]

Go to #🪲┃bugs-web

[high snow]

Yes it is

[covert terrace]

anyone have info for how to use 'inspo'?

[regal flame]

Does Suno integrate w/ SoundCloud?

[bronze island]

Idk

[half dragon]

https://help.suno.com/en/articles/6882753

[half dragon]

No, it doesn't.

[sweet ivy]

this is driving me crazy. I went into Editor, selected the 4.5 version, highlighted the entire lyric line. That line became highlighted on the track. Then typed in my one word change in the line. Selected Fixed, the Replace. It produced 2 versions where the word was correct, clicked on Insert. Another button came up that said Commit, Clicked on it. Saving a little circle to right going around "Saved". Thought I was done. Went back played track, the singer sang "dreams" again even though he sang dream in both replace generations?

[polar spade]

If about posting to Soundcloud, you would just download your MP3 and upload to Soundcloud.

[quiet solar]

Yeah, it's not the most intuitive. Remember, Studio doesn't amend the original song at any point - you have to save your edits to a new song:

Export -> Full Song from the top right of the central screen

[sweet ivy]

I exported as Full Song and it moved to my Library. I figured out what happened, even though both replacement generated had the word correct, just dream, not dream, when it was dropped into the track, at the point is ended there was still a little "s" on the original vocal track. I maybe should have included the first word of the next line, when I highlighted the lyrics . Could try and eliminate the hanging "s" in my DAW

[quiet solar]

Ah, man...that's harsh. xD

[sweet ivy]

@quiet solar I think I may have found a solution for Suno playing Bm instead of B7 in two of 3 verses. Odd it used the B7 in Vrs 1. Wonder if anyone else has tried this. I can upgrade to Melodyne5 Editor. It has the polyphonic option. I should be able to spot the Ds and move them to D#.

[meager totem]

Any tips to reliably have vocals sound correct? I uploaded a song but I want female vocals and I've literally tried hundreds of times even with 100% audio influence and it just won't replicate the vocal rhythm right

[gritty mulch]

My best efforts at crafting vocals more precisely is to sing them in in studio, even if you suck it'll take it where you want it.

[polar spade]

Yes. Makes many things more simple.

[quiet solar]

If the issue is rhythm, you can also try giving Suno a more explicit steer on the phrasing within the lines using punctuation.

However, if you're not getting it from a direct audio prompt (which I'd also agree is the most reliable method) then it suggests Suno is struggling to parse the line in the way you were hoping for. It happens - Suno is a talented mimic, but it's not designed to be an audio replication tool.

If it's essential that it follows the exact flow, you might have to fix that on a case by case basis with editing in the areas where it typically handles it differently by default.

[dusky stone]

Hello guys, I'm sorry but I did not get how to tell suno to start a bit before beat #1, thanks

[unreal sleet]

when i extract stems it happens quite often that i cant hear the track or the stems?

[quiet solar]

Beat #1 is typically the start of the song, so although there's a tiny amount of space before it in case a sound is transitioning in, I don't believe it's meant for extending backwards towards a hypothetical beat #0. If that's causing problems, you might want to try something like realigning the current song to beat 2 so that you get a full spare beat ahead of the existing track to work with when generating something new.

[quiet solar]

Can you give any more context to that? Where are you extracting stems from - and when you say you can't hear it, how are you playing them?

[unreal sleet]

the detected stems have no volume

[dusky stone]

Ok, thanks, bc I ask to create a new string track and it added before #1 so i was asking if it had been possible to tell to start forom 0 instead of moving all (risking to creat misunligments)

how can I select all?

[sand hollow]

Why is it that i get 2 COMPLETELY different results from these 2 prompts? Like high-energy phonk is different from phonk

[quiet solar]

Zoom out so you can see everything horizontally, click and drag to cover everything horizontally, use mouse scroll to move down as needed, release when everything is highlighted.

[unreal sleet]

even the full song above goes mute when i extract stems

[quiet solar]

Based on how you're describing this, I'm assuming this is in Studio?

[unreal sleet]

no this is extracting stems from the workspace

[quiet solar]

That's...very odd. Never run into anything like that before.

Given that it's happening on multiple songs, it might be worth flagging to support. I can't think of any reason why the original song would be muting as a result of extracting stems - in principle, nothing should happen to it.

[unreal sleet]

i mean these are all without any volume and its happened many times now

[quiet solar]

You've tried turning off the 'solo' marker on the main track, I presume?

[lone nova]

Hi guys, how do u guys properly define 2 voices: 1 for main vocal, 1 for group response.

Lets say my lyric is currently
Hack the code. (The golden road.)
Heavy load. (The secret sowed.)

I think the 2 voices sounds too similar, if I do not clearify.

[quiet solar]

As a starting point, you want to specify that the song has multiple vocalists in the style prompt, and define them both. For example;

“Male rapped verses with a female-led melodic chorus”
“Vocals are sung by a high energy male tenor and a smooth male baritone”
“Vocals are softly chanted by a choir with a female contralto soloist”

It helps to make them as distinctive as possible. For example, male vs female vocals are highly distinctive, as are soprano vs bass. The more different the vocals, the more likely Suno is to introduce two separate vocalists rather than pitch-adjusting the lead vocalist.

In your case, because you're relying on calling the backing vocals using () to implement this, you probably want to specify a lead and backing singer (or group). Depending on your genre, adding 'gang vocals' or something similar things into the style prompt might help as well.

There's some additional info about working with multiple vocalists in a workflow here: https://discord.com/channels/1069381916492562582/1466598465969590465

It's tailored more towards duets than lead/backing, but many of the same principles apply.

[dusky stone]

thanks!

[neat wolf]

Okay, Personas…

I originally used them, thinking they were basically “keep the same singer’s voice” so I could do consistent branding/story for my little AI idols.

But I think Personas are closer to “song DNA” than a pure voiceprint. Since a Persona is created from an existing song, it seems to carry over biases from that source track, like:

• Vocal/performance vibe (timbre, cadence, attitude)
• Production palette (drums/synths/guitars, mix color)
• Arrangement habits (where the hook lands, how sections build) ← this one is a BIG deal

So when I apply a Persona to a new song, it can feel weirdly cover-adjacent, because I’m implicitly referencing that Persona’s original track.

This is awesome for consistency (and you can be lazier with prompts), but it can also hijack a song if your style prompt + sliders are fighting the Persona.

Also: Cover + Persona feels like “3 bosses at once”:

1. the cover reference (structure/melody),
2. the persona source DNA (voice + palette),
3. your prompt (genre/structure rules).
   If those don’t align, the output turns into a Frankenstein compromise.

How are you all using Personas without them taking over? Any workflows/tips?
(Like: vocal-only Personas vs lane/style Personas, or “cover first then persona later”?)

[quiet solar]

Vocal only personas are a much better fit for how you were originally using them - although they also pull across subtle elements around vocal performance as well as just attempting to repeat a similar voice.

Would also 100% agree with the 'multiple bosses' problem. Where I've used Personas, I've generally created/covered the song first so that you're only dealing with two out of three at any given time.

Having got the song into roughly the right place, the next step for me is to do a fairly high audio influence cover to switch the vocalists. In that instance, your style prompt isn't so important as you're relying on the audio feed, so you're only focusing on the cover source audio plus the persona.

[neat wolf]

That's a smart and common-sense idea Use a vocal stem rather than a song to attach the persona to

[quiet solar]

Or just use the new-style Persona, which only goes off the vocal stem and saves you a step. 😉

[neat wolf]

For me, I noticed a few other things shift with the new persona style
Earlier Hook
Bigger chorus lift
Brighter Presence
MOre dynamic
Overall, a quieter voice on average

I made specific styles and Lyrics for each persona, like an audition that was meant to have them sing in a range and try to capture the personality of the persona...

so i can see why it did it even if it was slight

[sleek stump]

Using a Vocal stem for a new version persona is effective in maintaining a vocalists core. It is however still Generative AI, so it might take a few rolls of the dice to get the consistency one expects.

[hybrid vine]

How would you guide Suno to make an intro similar to "Physical" by Dua Lipa or "Tattoo" by Loreen? I'm trying a lot of stuff, even asking some LLM to help but it keeps doing something different.

[sleek stump]

The genre you are prompting likely has a lot to do with that. Try adjusting genre styles.

[neat wolf]

I have spent way too much time trying to chase a specific voice... but Suno still has RNG on output 🙁

[hybrid vine]

well, the genre is exactly what I want throughout the song, just the intro lol

[sleek stump]

But is that kind of intro common in the genre you prompted?

Suno will actively fight you when you try to do something it didnt see in its training, unless you are doing a "fusion".

[sleek stump]

And that is the nature of Generative AI. Suno isn't a word processor for music. Anything you prompt is a suggestion and not a command.

[neat wolf]

And thats why AI will take over the world! If i tell it to stop, it will take it a suggestion...

[hybrid vine]

I don't really care about the vocals since I'm gonna use only the instrumental and sing myself in the end. The only prerequisite for the vocal is that they have to be the same range as mine and I never had any problem with that (sometimes it still gives me a female voice even when I select Male and specify male vocals though lol)

[dusky stone]

guys I have always Piano with noise in the background, do you have the same? I tried to recreate in studio but more noise pianos come

[shut wave]

Tip in suno: if you want your song to go viral, all you gotta do is have a cat picture on the cover of the song and get it to new songs sections

[viscid ravine]

i need feedback on this

[slender glacier]

Dam, is there a trash section to undelete?

Found it, in the desktop version

[polar spade]

yep. You got it. "Restore"

[lusty rune]

Use the exclude box for a clean output.

[unreal sleet]

I know how that works and sometimes I use stems simply to listen to the vocals to see if there are errors so yes I am aware ;p

but as you can see in the picture im "listening" to the song, which isnt giving sound

[neat wolf]

It's hard to give feedback unless you say what your goal is, or else it's all opinion
high level, vocal was clean and consistent, hook hit early enough, pacing was solid and the lyrics delivered.

My personal opinion it felt too safe and bland, Id want to hear greater peaks and melodies i felt like it was restrained somehow

[lusty rune]

The main track is always on mute toggle it on.

[dusky stone]

what is it? In studio?

[unreal sleet]

I just told you I know how to mute or listen to specific instruments

have extracted gigabytes of stems its not something i randomly forgot how to use

I'd record it just to prove I'm not an idiot but I wont sink that low

[dusky stone]

i'm in Studio and have already worked a lot on the single stems...but the piano stem has bkground noise

[neat wolf]

side note i found something pretty interesting

When writing lyrics, if you use feminine language, Suno will assume a female is singing, so you have to be cognitive on languages! I kept getting a duet because I translated something wrong. Style, sliders, personas, do not matter 🙂

[humble sleet]

Quick newbie question, when in SUNO Studio mode, how do I get stems from a song? Or do I have to be in regular Suno, click “get stems” first AND THEN enter Studio with the song? Or is there a more efficient way directly from within Studio?

[unreal sleet]

that happened by chance lol

[neat wolf]

i mean i guess im lucky because it hit 100% of the time until I changed it

[unreal sleet]

what happens if you write in a non-binary language

Yes like my first message stated

Strange

Having a lot of weird issues lately that hasnt been showing before

[lusty rune]

The noise is on the piano stem? I would simply redo that stem.

[dusky stone]

yes

i also tried to extract stems from it...but it's there...also in other versions

[lusty rune]

How many stem did the tool detect?

[dusky stone]

I tried to recreat but ofc the outcome is different and most of the cases with noise too

[dusky stone]

you mean from the main song one?