📚┃suno-school

[acoustic egret]

how come the second half of the song is louder than the first, when i export the stems. is it a glitch in suno?

[polar spade]

Steps for cover:

° Click menu and edit/Cover
° Make minimal lyric changes as needed.
° leave sliders at default should help recognize lyric changes. (Setting higher at 85% or so, might overlook lyric changes and give original results)

Thats about it. If having issue with lyrics changes experiment with audio slider, or just click and get a couple more outputs to see if Suno changes lyrics in one of them.

Every response I give is assuming im only using Suno. If I think its necessary to download and work offsite, ill mention that, but I try to answer as if all we have to work with is Suno.

[stiff temple]

Maybe I had sliders wrong. Had 100% for everything except weirdness at 0%

[polar spade]

I use Reason 13 (and Melodyne) But I dont discuss DAW much in here.

[acoustic egret]

also, why the hek does suno not keep the song in time? when i export the stems it doesnt line up to grid and even my cubase has a hard time detecting the bpn

bpm*

[polar spade]

Oh, right. No , 100% can take you into the Ether sometimes.

[stiff temple]

I'm dumb I thought 100% meant change nothing

[polar spade]

You might see if the Tempo-lock helps on download. See this recent post for a screenshot. Well.... ok... I guess I could just post the screenshot... haha

[polar spade]

Some like to use 100% and it does "work" sometimes. I consider it "infinite " or free reign for Suno to do what it likes.

[polar spade]

Also, if youre just getting tempo drifting, which is normal for suno, perhaps you can just lock the bpm first by clicking the tempo on the bottom rail. See a video to help. https://discord.com/channels/1069381916492562582/1468681542913101865

[acoustic egret]

thanks!

[raven turtle]

Hi all. I’ve been trying to prompt against the over glossy Americanised ideal they class as Synthwave and instead trying to get more that late 70s/early 80s UK European post punk electronic synthpop like Human League, OMD, early DM, etc.
Suno dataset seems very skewed towards clean, modern synthwave gloss.
ChatGTP sums this up as:
“Suno isn’t “bad at synthpop”
It’s overtrained on the wrong decade and side of pond’s idea of it.”.

I can get close, to a more “De/Vision” vibe but not closer than that.
Is anyone else out there looking for more that rougher, indie electropop sound? And is Suno even the right platform for this kind of stuff, given the overwhelming sound of the dataset of Suno? 🤔

[viscid ravine]

Any feedback?

https://suno.com/s/NccVCb21vseNYQIO

[proud dagger]

guys I got a very very good song but there were two places were one single word was misprnounced completely. Is there a way to keep everthing same and just fix pronounciation of those two words? I tried cover and remix but no good results.

[leaden lake]

Extend it with sonauto I guess

[proud dagger]

the issue I am stuck is those issues are two different places in the middle of the song.

Like one in the verse 1 and one in verse 3

I tried redoing same verse but the voice and style of singing changed completly

with same style

[brisk nova]

I think you are pretty much ... you know. If you keep trying you'll get further and further way from the good stuff. Good luck.

[proud dagger]

Does upgrading to studio helps in fixing these?

[remote crown]

stupid question but I'll have to ask.. how do I regenerate a song with the new UI`? with same lyrics and style? I have beeen using this for few years.

[proud dagger]

use cover feature

[remote crown]

oh okey 😄 cover my song thanks

[proud dagger]

you can also use remix by keeping everything same or different

[bronze island]

I've been using Suno since June 2024, so for about 2 years 😄

[fathom dagger]

Hi, I have a song made to 5 min but did not finish the last verse, so I extend it and marked it [outro] and put [End] so it should finish. but the extensions no matter what I do are 4 minutes long with no end at all. I just want the song to end! Any tricks? it wasn't like this before.

[polar spade]

There may be an ebb and flow to that as it also pulls from whats in the output you're extending from. But 5.0 does well most times ive found, but as a beta, its to assume changes may happen. I recall for a quite sometime extends only gave 2 second outputs for awhile.

Try ending with model 3.5 to see if it ends more quickly. Otherwise, if in a hurry and dont wish to click 20 times in hopes one randomly ends -- then find a good point to splice, cut and quick fade, then a "subtle" Remaster That's sometimes an option.

[fathom dagger]

thanks for the input. It will not work with a fade out. I am only couple of bars short for a nice ending. but why the fixed 4 min extension? I tried extending with only [End} and it pulls verse one and goes on and on and on.

[polar spade]

Also. Try separating outro and end as two separate extensions or just leave "outro" out. SUNO can make some really long outros. Just put [end] and let Suno figure out how.

[Instrumental]s often last no more than 20 seconds, so you might also try that instead of [outro].

[polar spade]

Dont forget the power of just prompting for a jingle!

https://suno.com/s/BIFhBn8FlwyDMGPT with [outro]

https://suno.com/s/vaCcrY2KQKmIHcYj with [interlude] (yep 20 second!)

[fathom dagger]

ah finally! Picked a point where the lyrics ended and extended using 4.5. Four bars later it finished.

[acoustic egret]

am i able to move the order of the songs in my library?

[stiff sentinel]

not really, aside from newest, oldest, most likes, fewest likes

thats what workspaces was intended for but it can still easily become a disorganized mess

[acoustic egret]

i see, thanks

[stiff sentinel]

you can pin to keep somethign at the top tho

[midnight gust]

Unfortunately, I have already tried that option, as well. Same result.
I tried to physically split each track at the tempo change, in an attempt to essentially create two independent sections and then set the tempo. But I was unable to independently change the tempo. So I'm not sure if independent tempo changes are even possible in suno

[polar spade]

Id be lost to know for the moment. Glad you gave me some insight to the issue.

Oh and one last thing. Ive forgotten to mention. But since you mentioned splitting sections. Something I did, in that regard, worked when I click the 2'x instead of changing the tempo. I recall (and just tested) where the BPM "number" stays the same but the track compresses and plays twice as fast. Or doubles in size at 1/2 speed if you click 1/2 speed. It may be a workaround in some manner.

[proper pecan]

Out of curiosity, If I wanted to write my own lyrics and have Suno generate a song based on those lyrics, is there a syllable rule per verse? Hope this question makes sense.

[polar spade]

No. No rule. To me its better if the meter is written more freely. But either way, Suno measures everything and works out the best composition , arrangement and melody based on the lyrics you give it. Even a recipe, or instructions for putting together a table from IKEA.

[lunar ruin]

Most of what I do with SUNO is experimenting constantly. I don’t always focus on one genre, I’ve tried opera and made it my way while still making it still feel like an opera.

[stiff temple]

I used to be able leave a song unpublished and still share its link. does this not work anymore?

[remote ivy]

f your doing a cover, can directly tell it replace something with something in the style prompt...does accept that kind of instruction...or if you say..keep original vocals exactly?

[quick wedge]

That still works never had any issues of it not working on my end maybe try refreshing your page or try different browser.

[sleek stump]

Sort of? You can certainly change everything about the genre with a cover, change Folk to rock, or Blues to Opera.. thats certainly what Cover was made to do. However, if you are trying to change single instruments in the arrangement, thats might not work the way you hope.

[remote ivy]

i mean can you tell it in the promt.....change aahs w/ violin ?

without changing the lyrics or adding tags in lyrics

[sleek stump]

Not really, the aahs and oohs are a feature of v5 and it will adlib them all over. If you dont want them, ways of dealing with that is to edit them out or do your initial gen in 4.5+ and then cover to v5,

You cant really prompt what you are trying to do, it's going to take some work.

[remote ivy]

not a big deal, was just trying to save time...it will do it with a genre shift and lyric box changes....the ai is pretty dumb and will swap sound for sound pretty easily

[polar spade]

well, how many ahhhhs are you talking about... just the transitions or turn arounds after the chorus and maybe the intro?

[remote ivy]

let me link the song...it pretty much is a vocal version of a violin sound....one sec

[polar spade]

I'll listen to the first 2 minutes. sure.

[remote ivy]

also did giveaway of ai songs...why i call it dumb ai

https://suno.com/s/9KEijPnjY8PgLpuv

[polar spade]

And to be certain.. you working with Pro or Premium

[remote ivy]

Pro

[polar spade]

Assuming the vocals would stem out separately. Then convert the vocal track to a violin track. Purchase Premium and lay the tracks over each other.

But with the track alone, the vocals cannot be swapped out within the whole song and maintain the whole guitar melody in Pro.

[remote ivy]

oh i was talking mostly via covers

but just through literal instruction of swap aahs for violin...but i believe that was resolved that i can't do that, you have to actually use meta tags in the lyrics....they are in the lyrics if recall...not my song.

so i guess it was poor questoin..since they may be in the lyrics....obviously sometimes there not...i guess thats what i was targetting, not this specific case, although this song

[polar spade]

Theres nothing certain. If you just Covered the whole song, and prompt for heavy violins to try to avoid ANY vocals then that would assume sliding Styles high and Audio about 20% and cross fingers.

Can it be done guaranteed before running several outputs to see if Suno randomly gets one right. No. No guarantee. In Pro.

[remote ivy]

thanks for responding, illl play around, consider it issue closed

[polar spade]

Yep. Will do. Suno wouldnt know the difference between a vocal and in instrument to swap out just one and leave the other. It would just do it on its own, like a duet. But it could happen. OK Issue closed.

[midnight gust]

Yes, I saw that as well. I was hoping I had simply compressed a section but, had not.

I was, however, able to create a workaround.
Previously, I had tried to detect and then extract the tempo track in StudioLive, and try to manipulate it. It didn't work.

For whatever reason, trying to change the "detected" tempo track had no effect.

I then extracted the tempo track without detecting it first, and was able to successfully manipulate it that way.

Certainly not ideal because it took a fair amount of time successfully get the various tempo arcs correct.
"Arcs" because the tracks didn't shift at a single point but instead tracks sped up even as the tempo setting decreased in random increments.
Definitely one the more bizarre things I've ever encountered in all the years I've been creating music.

Hopefully, they determine a cause and solution for this.
I really don't want to do this every time I have a cut-time section in a song.

[polar spade]

quick Pro version of doing this:

So here is an original output. With just a chill pop sound and vocals
https://suno.com/s/bx5TaFyWOAiYwQAz

I stemmed out the vocal.

Then chose Add instrumental -- to the vocal stem,

Put "violin, flamenco guitar, flamenco guitar" in the styles to basically start with a violin but give attention to wanting a balance of guitar as well.

Kept the Lyrics as is. note: How it would react to your song with no lyrics, i dont know. Hopefully focus more on the melody.

Sliders: 27% / 84% / 25%

Ran one generation: This is the output. Added both guitar and violin, but didnt remove all the ooohhh and ahhhhs. Not bad though for one output and keeping the vibe and melody intact. https://suno.com/s/yM9uMkC9Kvw7Ly2e

[wintry apex]

Suno is rewriting my prompts now as if it was Udio on auto but no manual. And its ruining my gens.

this was not "psytrance" I wrote my prompt specifically how i wrote it, not to have it reduced to 140 bpm top 40 trance. there were no commas in my prompt.

unless im imagining things but this is not the prompt i wrote

[half dragon]

What happens if you click "Show More" ?? I believe it'd show your original prompt... at least that's what happens on mine...

[plush notch]

I’m trying change a lyric in song editor - but it won’t update to new lyric - any advice

[brittle vortex]

Please how do I change my log in from Google to password

[brisk zinc]

Why suno ai have so many hidden locks (vocal voice, instruments lock, melody lock, plus scratches or hisses problem) ?

[near siren]

Not sure if this is the right place to post sorry if not but does anyone know how to export a projects multitrack, or why mines not working? I click export>multitrack and it appears to start downloading a zip folder but then once completed, it’s just downloads the original unedited generated song 🙄

Any help is appreciated 🥰 thank you

[polar spade]

What is a "hidden lock"? A limit from doing something we would like it to do? Or a need to upgrade to use?

[polar spade]

That's the usual routine to download. Just click on export / multitrack.

I downloaded a 2-track piece, just now. No issues.

Could even be a mismatch in the display, so:
° Check for most recent Chrome update if using.
° set browser to 100% view.
° try another browser.

[brisk zinc]

If you have created tracks using only basic pro account features via prompt and basic editor, not studio. Then check tracks for hissing , scratches, tempo drifts, voice cannot be change via prompt or even from persona unless recorded own voice. Voice stays always same is essence. Melody have same problem locked by genres one generic melody that forced into your song. No matter prompt melody outcome is same. Highly noticeable when comparing two same genre tracks. With melodies i can be wrong, but everything else lead to one question and answer.

[keen crescent]

Hi Family, SUNO seems to mostly have my male vocals scream/yell a lot in build ups or Chorus. How can I force it have a slightly louder voice without turning it into noisy singing? 👏🏻

[polar spade]

I could only guess, really.

For instance, with k-pop, the voices are often similar. So that suno is stereotyping the genre as expected. I dont like that. I have to use personas all the time to force it away from sounding generic Korean Girl singing voice.

With a combo of genre prompts, the voices could be similar as the prompts start melding to a generic sound, and the voice follows to be generic as well. Sometimes, a lot of prompt might make many variations, but sometimes many prompts just become a generic blend anad the voice does the same.

I just dont know.

You're right to say, if we record our own voice, the effects of making a persona from it seems stronger. Or just if we use our own voice and cover it, certainly, the prompts dont change our voice as much.

Sliders can have some affect on outcome. We may thing 85% audio will maintain a persona voice better, but sometimes 20% is much better to keep the persona from sounding generic.

[brisk zinc]

I intensive tested suno ai and hidden lock i mean those problems mentioned before, leading to one speculative assumption,most likely possible fix with just studio tool, but it shouldn't be locked all to one solution, studio tool , that is premier plan part.

[quiet solar]

I don't really use Studio, and have never had any issue getting around anything Suno has thrown at me that I wanted to change for the final version. I guess your mileage may vary, but it's certainly not locked to Studio as the only solution.

[quiet solar]

It's not something you can 'force', but you can try to steer Suno in the right direction with terms like 'rebellious', 'passionate' or 'high energy'. Suno's propensity to scream will be impacted by the genre you're working in, so if singers typically scream in it you'll have a tougher time getting Suno not to go with the flow.

[blazing crystal]

Is there a document on prompting and lyric instructions that can help me get the results I want?

[half dragon]

There isn't any official documentation or instructions about prompts.
What I did when I started was:

• search through this Discords for discussions
• listen to songs on the homepage and read their prompt, collect whatever I think I can learn
• then just try it
  Some are really good and work well, I saved to my own list for my own songs with the style I like.
  Some that doesn't work well.. but I exprimented with them anyways... and keep a separate list for inspiration later.
  That's it... learn from others.. and experiment to fine tune for your own work.

[blazing crystal]

I have been experimenting with Duets and it is the most frustrating thing. So far my only solution has been to make the lyrics ambiguous enough that it doesn’t matter who sings what. My songs are very character/narrative driven though, so that is sometimes very difficult.

I am trying to get better at extends, but that is still hit or miss, and it doesn’t work when the non-persona starts.

I think I burned a few hundred credits on a song I could only get about 90% there, before throwing in the towel yesterday.

[half dragon]

Oh yeah, duet is known being very dificult to achieve... and lots of frustrations and expensive.
It used to be very, very difficult.. but with recent tools and updates... its a bit less frustrations 🙂

[blazing crystal]

Yea, and the problem is if I switch personas, even with specific instructions in the styles, it hugely alters the song. So if I want a song with a lot of gravity, but have a normally playful singer as a partner, I can’t use that persona to start it.

Definitely good knowledge to have though.

[half dragon]

It's definitely going to take a while... some of those instructions for singer you can try to add inside your lyric using meta tags... [...]
You could describe what you want the singer to perform there; a lot of times, it would work, but sometimes.. it might require a few tries...

[blazing crystal]

I wish it adhered to those tags better.

[fathom gust]

How i actually gain traction and followers on my account? i don't know how to make it so anyone else sees my work

[half dragon]

Yeah, like any other generative AI... they're not always adhere to the prompts... most but not always...

[half dragon]

This is not the place for that - you can try #💬┃general-chat but in short, no instruction can give you the guarantee.
Be active out there, spread the words but to the right audiences... and with a good collection of songs on your page... you'll be successful.

[blazing crystal]

I think it’s fair to ask that here. Honestly it seems like finding people and music you like and follow them and like their music. I am new though, and only have like 10 or so followers.

[fathom gust]

yeah how is this not the place to ask? its a genuine question on how i can use the website to push my stuff out so other people can see

[blazing crystal]

I think this is generally used more for prompting.

I’ve learned a lot from youtube videos, and there are a few that cover the topic of growing your community pretty well.

[fluid copper]

OKay.... any ideas for this problem. I have a verse that's: aab ccb. When I use a female voice it gets it right. When I try the male voice (Cover), it ALWAYS does: aabc cb. I tried putting period(s) after the first "b", Tried blank lines between the two patterns, but still it does what it wants. Again, did Cover with female voice, no problem. Does what I asked.

[polar spade]

I'll throw this as a general question.
I have SFX "scream" panned left and low volume for a film noire style, saved file, then split stems of the saved file, which then put FX back in the middle of the vocal file. Seems other SFXs -- a triangle in FX stem, some background chatter in Background -- are panned where they were set before. I feel the scream might be mistaken as "noise" and just placed in Vocal. As when I remaster NORMAL the saved file, it removes the scream. Its there when Remaster SUBTLE. And all Human sounding FX turned into violas when remastered in HIGH.

Anyone find some FX missing after remasters if volume is under a certain limit? And does that seem like why it was center panned as just added noise when generating stems instead of left panned as an FX?

[quiet solar]

Doing duets to that level of precision around controlling vocals ultimately requires an astonishing level of luck, or using alternative tactical workarounds and editing (one of which is multiple extends, which has it's own potential headaches, as you note). 90% right is probably a realistic maximum without workarounds unless your song is very short and the vocal switchovers rare and very structured.

If it helps, I pulled together this as a list of practical workarounds and advice around duet-making: https://discord.com/channels/1069381916492562582/1466598465969590465

[blazing crystal]

Thanks for this! I have not gotten into working with a DAW yet. I'm just a nerd that had an idea about a Dungeons and Dragons metal album XD.

[solid depot]

I recently got a song with an amazing first intro but the quality drops like every minute massivly
https://suno.com/s/zcNXSDIaTGZ9dDam
Is there a way for me to keep it basically as is without changing too much but improve the quality as a free user of 4.5?

It reminds me of Music from games like Soul Calibur and i was hoping to get more of that.

[quiet solar]

Suno does tend to prefer couplets and quartets of lines rather than triplets, especially if it detects a rhyming pattern, which I suspect is what you're running into.

What you've tried are go-to approaches I'd have used as well. You might also want to try upping the audio influence slider when you Cover, or deliberately inserting something else that's not a blank between the gap - [instrumental flourish] or [instrumental hook] might be relatively low-disruption options to try? Otherwise, dropping an Extend just after the first triplet and starting from ahead of the second might help shake up how Suno is interpreting it.

[quiet solar]

Haha. As someone who wrote a song about the headaches of being a DM in D&D, I wish you all the best with it. Nothing wrong with a dash of nerdiness from time to time. 😉

[blazing crystal]

Thanks so much! Can you recommend a free DAW to start learning?

See if my old brain can learn any new tricks.

[quiet solar]

That really depends on what you're hoping to learn and use it for. My tool of choice is Audacity; technically it's a DAW, but it grew out of being an audio editing tool (and at it's heart, that's what it still is). It doesn't have some of the flashier built in tricks that other DAWs do, or ready-made access to samples, but the interface is pretty straightforward to work with for the most part.

Otherwise, GarageBand is pretty straightforward (note: iOS only), and Ableton has a free version that isn't completely terrible, but has a learning curve that's quite a bit steeper.

"Free" and "DAW" are sadly not words that are often found in the same sentence.

[blazing crystal]

Gemini is suggesting Ardour or Cakewalk. It also mentioned Audacity.

[quiet solar]

Can't comment on either Ardour or Cakewalk, having never used either.

[blazing crystal]

No worries. Thanks for all the tips, and introducing me to the workflows section that I completely overlooked.

[near siren]

I’ve tried what you’ve suggest but I still can’t figure it out 🥲 I’ve even tried “Download All” from the “Extract Stems” page, rather from the studio and I am running in to the same issue, it appears to start downloading the ZIP folder but finishes with just the first stem 🙃 I’m having to manually download them one by one.

[quiet solar]

Yeah, that can happen in longer songs. Suno kinda loses it's way with what it's producing.

As a free user your options are somewhat limited, you'd have to use Cover and the original style prompt, but without access to the audio influence slider it'll likely take a few attempts to get the first half of the song more where you want it, then you'll need to Extend off that to take it out past the 5-ish minute mark - or to steer it back to the script if the audio starts going wonky.

So it'll be more like redoing the song but keeping an eye on what's happening with the audio, rather than creating an 8 minute track from a single prompt.

[solid depot]

I mean id be fine if its only like 3 minutes tops. But i dont know how to shorten it

[quiet solar]

Welcome to the same headache thousands upon thousands of other music-makers run into.

Suno isn't great for easy discoverability (then again, nowhere is) but some of the same general rules apply to a lot of content creation and music specifically within that:

• You'll generally do better if you focus on a consistent genre rather than publish a variety of genres and styles
• Be selective about what you publish, and even more selective about what you pin - treat both like a 'Greatest Hits' album. Only put up your best work, and only pin the best of the best, rather than publishing everything you create
• Most listeners will switch off fast if they don't immediately like what they hear, so pay a lot of attention to landing the vibe within the first 15 seconds and/or front-load your song's hook
• A really good Suno Hook featuring you as the artist can be an absolute game changer; treat it like making a viral TikTok for a music-savvy audience
• Have good thumbnails, song titles and song concepts (remember you have a short comment field) that are intruiging, but don't cross the line into clickbait
• Engage more broadly in the social side of Suno - remix trending songs, engage with other creators, put out something awesome for one of Suno's contests, find communities with similar musical interests; but treat this as a respectful slow-burn rather than a quick route to fame.
• Use other social media and platforms to put your music out there to a wider audience
• Remember that 87% of all music published on platforms like YT and Spotify is listened to by under 1,000 people (often a lot, lot lower than 1,000 in practice). Success is rare, and most routes to it take time and dedication building up skill and experience with growing content outside of any underlying talent you have in actually producing music.
• Consider that you're one more AI-assisted artist in a veritable ocean of similar creators. Think about what you do that makes you stand out, and how to pitch that out to people who might be willing to give you a listen.

Good luck. 🙂

[quiet solar]

Yeah, it's not easy to control length with instrumentals. Best you can do is describe a structure to the song using tagged sections within the lyrics itself, such as:

[Intro - your description here]
[Instrumental Verse - your description here]
[Buildup - etc.]
[Hook]
[Instrumental Verse 2]
[etc. etc.]
[Outro]
[End]

The more you put into the lyrics section, the more Suno will make and the longer (or shorter) the song will be. But it's quite tricky to influence the length of specific sections - most people struggle with Suno making things too short rather than too long with instrumentals, and have to lean on Extend to keep it going if they had a longer song in mind.

[solid depot]

What exactly can i put in as description as you gave the example?

Like intro with that flute thingy from the song i sent earlier? Wish i knew what instrument that is in the first place xD

[polar spade]

You mentioned that it downloaded the Original Full output when you clicked on multitrack. So, I was trying to rule out that your mouse pointer wasn't clicking in the right spot even when it was visually hovered over multitrack. That was unique and not having to do with timing out.

As for why then it may be timing out is not presently known.

Comparatively, I tested a 4:44 minute output by downloading WAV file ZIP of 5 stems of 53000kb, or total 266,800KB and it took 2 to 3 minutes.

Of the one file you downloaded, how large was it? It may be assumed if several stems and a large file or long song, internet speed of under 500Mbd , or any issues with pack loss this may affect the ability to download.

If most of the details seem reasonably moderate size, or have had "no problems before" with large files, you could also restart your router and WIFI if that's what you're using, to help. If you're hardwired fiber, then of course, we're eliminating possible reasons, but not finding the solution yet.

[quiet solar]

@solid depot Difficult to give you a definitive answer, since it totally depends on what you're making. I'd say that instrument sounds more like an attempt at a soft trumpet or a horn, rather than a flute? Either way, if you're doing a Cover of an originally unstructured instrumental, you ideally want the structure to reflect what Suno actually did, so some retrofitting may be required. If you were on a Pro plan, I'd say open it in the Editor and see how Suno broke out the segments but...well...

At it's heart, the lyrics tagging is another route of telling Suno what you want it to do, and trying to influence the output. And you can do whatever you like with that. The approach and style I might use for that could be totally different to how you'd go about it, but for example:

[Intro - lonely trumpet solo, percussion and strings gradually building]
[Instrumental Verse - strings lead on melody, horn echoes hook in background, building to crescendo]
[Drop]
[Chorus/Hook - brass lead, string arpeggio in background]

---

Alternatively, skip past worrying about all the above for this song, and see what you can do with Cover and the same original prompt, and save more detailed instrumental prompting for a different project. That might be a lot easier. 😉

[solid depot]

Okay, so i can actually go rather deep into the "commands" like that

[blazing crystal]

Whether Suno listens or not is another story. But it's a great start.

[solid depot]

I tried covering three times but that took the quality of Minute 5 and started with it on second 1 ^^'

[quiet solar]

That can happen - remember 'quality' is subjective, Suno doesn't have an inbuilt understanding of it. If you've reused the same prompt, bear in mind you're not telling Suno anything about the EQ and production effects that would steer it towards what you're probably thinking of in terms of 'high quality'.

Again, no magic bullet for this, but consider adding things to the style prompt like "high fidelity, clear mastering, wide stereo imaging, balanced EQ, dynamic range" to steer the production/EQ side.

[solid depot]

Noted, thanks a bunch

[quiet solar]

Happy to help - best of luck

[midnight gull]

how do you do this?

[quiet solar]

Write a prompt for some music that suits the pace, mood and energy of your video and then click 'Create'.

AFAIK, it's an ad talking about Suno's general capabilities and how they could be used, not a new feature for "take my video and give me the perfect soundtrack", although I agree the way it's framed does lead to some confusion on that point.

[silk jungle]

Sometimes you have to use the correct prompting for a great result, only adding info in the styles section. Leaving the lyrics part [Instrumental] tagged. I find the correct prompt for Cinematic EDM tracks (even tracks using vocals) for remixing a good instrumental, but it costed 12 days to find. Suno's default prompt generation sometimes not enough for a good result as I see.

[midnight gull]

ah ok thank you, i was like whaaaa i can just upload a video and have it make a song for it lol

[polar spade]

Honestly, in looking at your screenshot, beside it just saying go to suno.com for more... it seems promoted by "Creator Pulse' and "@creatorpulse.hq" I dont find any association between that company and suno. So its odd for the moment unless theres more. Their instagram page has little information and only 11 posts.

[quaint elm]

I was wondering SUNO has a child singing voice. I am trying to create a song for our deceased daughter. I want song to be sung from her point of view to her mother. IT doesn't sound right when it's an adult voice. I have had no luck. Tried 15 times with no luck.

[quiet solar]

It can be done - I've found "children's music" and "sung by a child vocalist" in the style prompt should give you a steer in the right direction. However, advance warning, there's a fair chance it'll have a vocal-shifted, slightly artificial tone so you might have to experiment around a bit to get something to sound natural

[acoustic egret]

Is there a way to export the drums as MIDI in suno stem mode?

[spice brook]

Bro I extended this song it generated 8 seconds of silence and stole my credits

[half dragon]

You need Studio, I'd imagine. Do you have Premiere plan?

[acoustic egret]

I've got a subscription plan. It's not the highest level tho

[half dragon]

Yeah Studio requires Premiere plan.

[bleak bronze]

i cant find the Mashup Button

[sleek stump]

Right click (or 3 dot menu) on the track and select remix>mashup

[half dragon]

on Create Page too

[bleak bronze]

thanks i found it

[fathom gust]

beautiful response, thank you!

[vernal meteor]

Please someone help me to stop Suno from putting drums in everything... I'm just trying to make some soft ambient sounds, but nope - drums! Trying ambient, chillout.. Sometimes it's full on synthwave and disco... even though I put "drums" in the exclude box. Every track just have the same drum style in them. I've been using Suno off and on for over two years now and I must have serious skill issues.

"ambient" apparantly means everything from disco to hiphop

Also guitar and piano, all the time. I generated the prompt "drone" - nothing else, apparantly that's "strumming a guitar".

[frozen dagger]

This example combines 2 suggestions in one. The prompt style tries to describe the texture and sound as being ambient, rather than just saying "ambient". Other suggestion, came from Ginger in this channel, used ASCII art in the lyrics box. Both give more open interpretations. Remove the non-lexicon vocal stuff.
https://suno.com/s/WVEXaeijX2RHQ1Vr
Dense underwater ambient instrumental with narrow stereo field and mono-focused low end, Heavy broadband compression and controlled dynamics, Sub-bass–forward mix with sustained low-frequency drones, Slow-evolving harmonic textures, Non-lexical vocal tones centered on sustained OM phonation, rhythmically integrated via sidechain interaction with bass and texture layers, Minimal transient emphasis, continuous motion, stable energy curve, long attack and release envelopes, Experimental electronic sound design with cinematic ambient production aesthetics

[sonic folio]

Do you guys think it's possible to have dual genres? Like rock drums with metal guitars

[wintry apex]

Is 4:50 the max for one gen now? There was still 2 more verses that it just abruptly cut out on multiple gens,

nvm

[ancient kettle]

Hi, can suno take an existing Disney Song and add new words to it?

[pearl saddle]

May I have someone point me in the correct information on how to do duo and trios, including having SUNO, stick to the DAM script i am telling it.

[astral mesa]

Hi, can suno take an existing Disney Song and add new words to it?

[sleek stump]

Suno does not allow the use of copyrighted works in their platform. Filters should catch them and not allow song creation. Sorry.

[fluid copper]

Depending on your level, I'd download the stems, put them in a DAW, delete the drum stem and output the "drumless" version.

[neat nexus]

Where did you extend it from? If it was at the end of the track then silence may have been all there was predicted.

[quiet solar]

Here you go: https://discord.com/channels/1069381916492562582/1466598465969590465

Spoiler alert though - good practice in prompting will only take you so far. The only way to force Suno to stick to the script is through workarounds and editing.

[wintry saddle]

good afternoon

everyone having a joyful and creative day I hope

[polar spade]

Life is wondermous. Thanks. And you.

[naive quail]

violently tackles @wintry saddle to the ground and begins to tickle his feet

[wintry saddle]

that hasn't happened for at least 20 years, whos been telling you my secrets

[glad nexus]

Bit random but what is best for use to create a video for your music A.I wise?

[polar spade]

Probably an online search for that answer. Depends if you want a random video to just vibe to the music, have an AI character lip sync to the song, create a storyboard and actually follow the lyrics and story of the song or just do a lyrics video.

And whether you include specific character you create to whether you use a site that uses multiple other AI models and formats such as 11labs or openart, to name some random names.

No problem, also, to just check FIVRR too and see whats offered to put a video to your song.

I think even CANVA offers a song to video service.

A couple of dozen sites can do variations for the right price or membership.

[blazing crystal]

My personas aren't updating when I change their styles. I refresh the page after it says they were saved successfully, but the changes don't stick. I even let it sit for a day. Am I doing something wrong?

For example, I want to update World Orchestra to Symphonic World Metal on the style, but keep the rest.

Or change the Low Baritone E2-C4 to Low Baritone G2-D4.

Basically, I want to refine the persona I am already using.

I don't want them disaccociated from their previous songs, but I want future ones to go off of the updated style.

Not quite. Here, this is my persona, and I have tried to remove the Caleb: part at the beginning of styles unsuccessfully several times.

It doesn't matter for this example, but I also want to update him to Groove Metal Funk, 100–105 BPM, at the beginning.

I did end up deleting and remaking one persona, but he lost the association with all the previous songs, and that sucks.

[quick wedge]

Suno just doesn’t reliably let you change a Persona’s genre once it’s been created and used. The safest fix is to make a new Persona with the new genre.

[blazing crystal]

Yea, I was afraid of that. I guess I'll just have to recreate it. Thank you!

[quick wedge]

You can try covering your song and change the genre that way or reuse lyrics and different genre that way too.

[blazing crystal]

Not great for minor changes to the persona, and it breaks their song library. Fortunately I don't have many songs yet. Juggling with 8 different personas though.

[quick wedge]

Personas are limited creatively unfortunately but they’re good for vocals if you wanna keep them consistent in your songs

[blazing crystal]

Yea, I am using them to represent different party members in a dnd metal album. So trying to get them tuned in where they work well with eachother.

Not even sure why it gives you an editable field for styles on personas if it won't let you change them.

So, I just tested it, and even putting in the style when you create a persona, it updates the style with the style prompt from the song the persona is based off of.

[knotty stag]

Duets are great, if that is what you want. But sometimes I create a song, specifying female vocals, e.g. "female solo vocals, only female vocals" and it still gives me some lines in a male verse, or even in two different male voices, in addition to the female voice.
I just have to keep recreating more versions until eventually I get a version which is female only. And then it also gives me backing vocals, which I did not ask for, so then I have to download the stems, and remove the backing vocals in Audacity, and save a new version.

[blazing crystal]

Are you using v4.5-all, or v5?

[knotty stag]

All v5

[blazing crystal]

Hmm, you may be confusing it by specifying solo female vocals or only female vocals. It may be better to leave those off and say something like no backing vocals, and selecting female under advanced options.

Better yet, add backing vocals, harmony and male vocals to a negative prompt.

[dim bane]

Should we not complain that a robot is woke .. and is not complying to his orders??

[blazing crystal]

Just remember prompts are suggestions to pull something out from the white noise.

[polar spade]

Not if they are in the same room as we are. 🙂

[dim bane]

Als jij koekjes moet bakken van een bakker en je bakt dan pannekoeken .. Gaat die bakker dan klagen ..

Als een robot zijn opdracht negeert .. Moet de robot veranderen .. NIET WIJ

[polar spade]

If the robot can bake cookies though, this can be overlooked.

OK. Have a nice day.

[blazing crystal]

May need a song about this.

[polar spade]

"...We now return to our regularly scheduled channel...."

[glad nexus]

Thanks I'll try Canva but yeah I want it to create from a storyboard I have itself

[sacred mist]

Is there anyone who offers Suno Studio instruction via live screenshare?

[polar spade]

Right. Then it seems better to do all the steps/video singularly and splice them all together.

[polar spade]

Not officially that ive seen, for free, in Discord. Maybe something to consider later as the models come out of beta as well.

To be honest, some things would be difficult to "show live" when often successful results are in the lower percentile range. Its better to have recorded videos of what a tool is supposed to do, reference how to do it, show some successful results and give a disclaimer that its not always perfect with AI.

[quick wedge]

Don’t know much else than that but at least it’s sort of working for you. I’ve used personas from other people but I’ve never created my own from that feature though. I usually cover and edit genres that way too.

[quick wedge]

This channel shows videos but not for different scenarios mostly demonstrating the tools so far https://discord.com/channels/1069381916492562582/1468677589752352799

[blazing crystal]

I don't have studio access yet. Just deleted the old persona and remade him with the correct prompt after covering a song. All good now.

[polar spade]

That makes sense. A Persona does "exist" to be changeable. Its just the same code as any other persona, just amending to the output its attached to to mine the characteristics of that output.

There is nothing "changeable" as to a persona. Only the original output could be changed, and at that point, such a change would create a new output, not reconfigure the sound of the personas allocated with that original output.

[polar spade]

Yea, ill look into the styles box and see whats its supposed to represent and why its not a changeable description, in text only, at least.

[fast relic]

The tooltip for Audio Influence explains: "Turn it up to resemble your original audio or persona."

I would love for these sliders to be separated! Most of the time, I actually want high influence from the Persona's vocals but default influence from the original audio. Turning up the influence on the current slider brings forward too many artifacts from my original audio and sometimes even overpowers the persona.

[sacred mist]

Got it. Thanks! (I would be happy to pay.)

[sacred mist]

Thanks!

[quick wedge]

You’re welcome 😊 have a wonderful day ✌️

[fast relic]

I have a Persona based on my own voice, so English vocals have an accent. When I use the persona, the accent is clear and consistent for about two verses + chorus ... and then the vocalist magically loses their accent for the remaining verses! I've tried adding the accent description to the style instructions (e.g. "Vocalist is a _____ man singing in English, so he has a moderate (but not exaggerated) ______ accent.") but this simply amplifies the accent in the first couple of verses and it still disappears for the rest of the song.

Does anyone have any suggestions on the prompt + vocal combination to keep my the Persona's accent through the whole song?

[remote ivy]

it just may be that i can't do it reliably, as to why you need to separate your vocal persona as a separate cover from your style cover...or something like that..or its a best practice at at least

[quiet solar]

If you set the vocals to Female under advanced options, you shouldn't need to specify female vocals in the style prompt at all. It's usually pretty reliable at generating a female voice from that starting point, unless you're in a genre that's overwhelmingly dominated by male vocalists.

Backing vocals are a different story - there's not really any way to reliably prompt Suno to not do them if it 'thinks' they should be in there from wider genre influences. Your workaround is exactly what I'd have suggested if that happens in a song you're otherwise happy with.

[fast relic]

I have also added "backing vocals" to Advanced section with moderate success, and/or "female solo vocalist" to the styles (instead of "vocals")

[quiet solar]

I don't know that I could definitively prove it one way or the other, but anecdotally I've found 'vocalist' to work better than 'vocals' as well.

[fast relic]

This does work sometimes—but then the vocals of the style-only cover still override the vocals of the persona when I do the Persona pass.

[quiet solar]

In terms of your questions about accents, if Suno has an English song then the singers do tend towards an American accent. I have managed to get somewhat passable regional English and Australian accents to work, but only in the song they were generated in. Once converted into a Persona, the Americanisation default strikes pretty switftly.

If you find a way around it, please do let us know.

[polar spade]

Yea. Thats a quandary. Its seem the better setting then is more around the defaults, about 20%-25% which brings in more of the vocals vibe of the persona, and only slight bit, if any, of the instrumental or thematic vibe (riffs, rhythm, melody phrasing) of the persona, but helps retain the original audio we are covering or editing.

ALONG WITH either 1) using the exact prompts which made up the persona to get more persona vibe 2) using totally new prompts which make up the audio vibe 3) or a blend of both prompts. The style prompts maneuvering act as the "fourth slider".

[fast relic]

Okay. This is what I've been mostly trying (25% influence, persona, and the exact style prompt that crated the persona, and increase the style influence) and it's been hit-or-miss whether the generated result sounds much like the persona.

But I do recognize that this is more "vibe conducting" than giving instructions to real musicians, so I take everything with that grain of salt!

[polar spade]

Call it a Prompt slider . There would be two panels originally showing persona prompts on one side and original audio prompts on the other. The slider would be the balancing slider blending one/other/ both. 😊

[fast relic]

I like this UX idea best because it also makes it clear to the user "you can have it more like the persona or more like the original audio, but not both"

[polar spade]

Its good to just have a fun thing that makes a difference, as well. Even if it varies things on a wide spectrum. So Im for it as well.

[waxen hawk]

Heyy! I could really use some feedback on my song I made on Suno. So feel free to give it a listen and comment what you think, I’d love to improve. https://suno.com/s/u0GcUA3xZJc9qv31

[remote ivy]

i tend re-inforce the person vocal description in the style prompt on the persona pass..or its part of the vocal persona's style prompt that i will reuse or just let populate if its the same genre..etc

[polar spade]

@fast relic

Keep in mind, 5.0 Personas have the option of :
Default/ or Legacy

Default being its only focusing on the vocal stem of the audio. The other, Legacy, it will Persona the whole audio. This also can affect the outcome when using personas on covers and new outputs.

[pure panther]

I can't believe I have credits left this month. I was so busy. I'm using them to animate the song covers, they come out really good!

[fast relic]

I was trying to get vocals only, so I've been using the default, but just for experimentation I ought to try the Legacy persona as well.

[polar spade]

Not to be confused with "Legacy Personas" which is just personas made before 5.0 came out which have not been upgraded to 5.0 Persona.

• ...and I dont know the efficacy of using 5.0 Persona Legacy over the Vox version. Testing would show if its to make any difference at this point. I'm just pointing out that its there.

So when we use sliders and we're including only a Vox-stem made Persona, it may not be the persona so much that is affecting things (or not affecting as much) if, for instance, were trying to get our Country Band to sing that Christmas song we also wrote and we dont know why, when adding the new Persona, it's no longer playing the Country instruments but still sound like the original Big Band sound.

[zealous condor]

There is a rhythm shift when making the persona that's completely throwing off my tracks. Anybody know why or have a workaround?

[acoustic egret]

is there a way to convert drums stems audio to MIDI?

i really wanna be able to export the bass and drums as midi so that I can use my own VST plugins

[wise kindle]

With Suno Premier you get the option to export to MIDI. If you don't have premier you can download the stem and feed it into Basic Pitch (free service). That also does a fairly good job transcribing audio to MIDI

[steady hatch]

How can i make the prompt such that it excludes certain stuffs without the pro version

It mostly ignores the promots if it's negetive command

[polar spade]

Yes, negative prompting to exclude is mostly not applied with Suno -- at the Basic level.

With just a Basic account, and no means of "excluding" Suno from using all available data in the song creation, you may be dependent on being more creative to influence Suno away from those genres or instruments you dont want.

That's just a refined area that is not available on the trial Basic free subscription level.

But Basic level has always been about trying things out and getting a feel of what Suno does.

As there is no ownership of the Basic outputs, its more for entertainment and practice. To do more, have more control and also ownership of the outputs, you should consider using Pro whenever you are able to upgrade.

You may wish to mention in here what you are attempting, and see if a user has suggestions, or consult an AI chatbot to make attempts (without exclusion/negative prompting).

[steady hatch]

I'm just using suno to make music to listen on but as you said in pro version they get ownership of the music that means they can publish the song professionally ?

[knotty stag]

I am trying to make a persona from a song I made, it used to be easy to click on ... next to the song, then Create, then Make Persona, but now that option is greyed out and unclickable. Is there a new way to create Personas now?

[polar spade]

The terms of service provide the detailed legal agreement, including your agreement granting Suno certain rights and access as well to your content (at all levels and use), but one aspect as a Pro user is that you are granted ownership of the outputs you create within that time you are a paid subscriber -- save any content used in that output that may be owned by another creator.

[sweet ivy]

Good Morning everyone. I downloaded stems into my Sonor DAW. Tweaked a few tracks and then saved as a .wav file. When I converted to MP3, so it would work as an email attachment, the high frequencies were enhanced?

[wintry saddle]

This is not a Suno related issues -> MP3 doesn’t just “compress size” – it uses a psychoacoustic codec that:
Removes masked frequencies
Quantizes transients and high-frequency content differently
Can introduce pre-echo and HF smearing

This often makes:
Cymbals, air bands (8–16 kHz) sound brighter or harsher
Especially noticeable if your mix already has crisp top-end
So even though the waveform energy is technically reduced, the perceptual contrast in the highs can feel “enhanced”.

You maybe able to see which MP3 encoder preset is used and control that in Sonar

you can probably do some EQ tweaking in wav form which will effect the final output

you can export wav from sonar, and use another tool to go to MP3 like LAME? and that should give you a bit more obvious fidelity

personally I don;y usually woory about the differences here that much - as listeners have different speakers and EQ setups too!

[sweet ivy]

@wintry saddle Great answer, thanks. ChatGPT suggested using Audacity, importing the final .wav mix into and then exporting as a MP3 file. It was okay, but the drums, especially the snare & hat were so different than .wav file. I saw the LAME option, I should try it

[wintry saddle]

good luck!

[hard beacon]

Instead of using a mp3 (only for email attachement) you should upload the Wav in a cloud (google drive, dropbox,) and send the downloadlink to your email contact 🙂

[low laurel]

is Suno capable of respecting chords given? if so, where must it be set? within the lyrics or in the style description?
Suno when using a lyrics is often adding 'volcalyses' in the intro (hummm, ahhhh, ohhhh, mmmmh, etc...) how can this be avoided? Thanks

[sweet ivy]

@hard beacon Okay, so is that how most ppl send a .wav file via email. Does the email recipient need to have a Google drive or drop box set up to open the link I would send them ?

[hard beacon]

no...as far as i know your just need the dl-link

[quiet solar]

Specific chords? No, not in my experience.

Some people have suggested that chord progressions in the style prompt (e.g. I-VI-III-VII) can have an influence on the song, but I haven't seen any examples where someone has done that when it wasn't already a very common chord progression from the genre (ie, it's fairly likely Suno may have landed on that chord progression itself, rather than been influenced by prompting).

To your other point, v5 has a quirk where it frequently introduces ad-lib vocalisations during intro sections. Having a specific instrumental intro, or putting "-ad lib" into the exclude styles field can sometimes help to steer it away, but there is no guaranteed way to avoid it. If necessary, you can export the instrumental/vocal stems and use Studio or a DAW to delete the vocal ad-libs entirely from a song where you're not looking to make any other changes.

[bronze island]

CadENCeeEE

[rapid bobcat]

Can any tell me what the name of the singing effect is at the linked time stamp: https://suno.com/s/SekyHEJC2vdbi3Vj?time=166 Where the vocals sort of cut to this distant/slightly robotic sound. i know I've heard it in songs before, but for the life of me can't think of a real song that uses that technique.

[frozen dagger]

dunno what it is called, if it has a name, but "Lo-fi vocal drop" or Filtered Breakdown, maybe?

[forest hemlock]

Yep

Could also just call it radio effect

[rapid bobcat]

Yeah, I swear I've heard it in numerous real songs, like from the sum 41/blink era..but for the life of me can't recall a single real song that has the effect

[forest hemlock]

Feel Good Inc by Gorillaz is the most obvious example for me

Not the exact same sound but it's in that direction

https://www.youtube.com/watch?v=HyHNuVaZJ-k at about 52 seconds in

[rapid bobcat]

I can basically picture it in my head usually around the bridge of a song..just usually for a handful of lines I'll try radio effect and see if that gets it

if I could just get a reliable prompt to come back from rap into singing, I'd be set. I've got the singing to rap switch, but getting it back is like pulling teeth

[forest hemlock]

Do you try to do it with formatting the lyrics in specific ways?

I've found the way you type out the lyrics has more of an effect than the overall prompt itself

[rapid bobcat]

more with [] prompts , but I have tried to format the rap section itself differently. It picks that up fine, it's the switch back that takes way longer than it should.

[forest hemlock]

Ah

[frozen dagger]

extends can change the vocal sounds, but usually when switching a style back like that with prompt is giving me a problem i try to extend from where the switch should change, rather than try to prompt it all in one generation.
Trying to text prompt for a few things inside the lyrics is pretty solid, but too many things, or going back and forth (as with duets), it does not keep generating as prompted in one go. Just gets as close as it can.

[rapid bobcat]

Yeah it's distant, but way too clean. The effect I'm thinking of is usually more gravel/robotic filtered. But I'll keep searching. I'm sure I'll come across one of the real songs I've heard that has it.

[forest hemlock]

Mid-fidelity, distorted radio vocals, megaphone effect, gritty alternative rock, saturated lo-fi textures, bit-crushed vocal layers, industrial noise artifacts, dynamic shifts, telephonic vocal filter, occasional static bursts, heavy gain, experimental indie

Here are some potential terms you could use for the specific effect

[junior glacier]

if you only want rap for a single section (usually an interlude/breakdown/bridge/etc), then metatag it and define it in the style prompt, but then ALSO define the non-rap sections (probably verses) at least in regards to songs like Feel Good Inc... the hardest part for me would be getting the backing "Feel Good" to show in a manner I want... its easy enough to get it to show up, but Suno will use it how it likes... probably some studio/daw work to get the specific backup stem right would be easiest

[rapid bobcat]

Thanks I'll work through those until I can find one that gives it. I got it randomly, but want to clean up teh song a bit and most of the new generations arne't putting the effect back in

[rapid bobcat]

I'll give that a shot too. Only been back at this for a week, so really getting into learning how to do finer tuning

[junior glacier]

also, for a rapped interlude, i find doubling the syllables compared to the verses really helps get suno to deliver

[polar spade]

It can be somewhat a dilemma. When we had 2 minute limits , we got close to 2, 4 minute limit, got close to 4. These were new, created outputs.

Then we got Cover tools. We discover anything use of Cover needs an allocated space to hold that seed/code/tokens which helps up create the Cover. It rides piggyback and takes up about 1-2 minutes on a 4-6 minute output.

Cover a small 30 second jingle with the right prompts, it might glitch and go wild to a 7:59 song.

But Cover a song more than about 5 minutes, you then have to anticipate issues cramming it all in and having it clip the endings.

Things to lessen filling allocated token space:
• If you have more than 200 character prompt,
consider cutting that down. Only your first few
prompts carry the most weight. Cluster your
prompts, to three-word descriptions than have long
phrases or never-ending lists.
• Reduce your title to something shorter.
• you might get by without structure markers: [verse],
[bridge], [chorus]
• Vocal description tags. Try to exude vocal styles
From the genre prompted, if possible or just upload
your voice singing as best possible and make a
persona, being aware that personas take up
allocated seed/token space as well.
• Front-load a short vocal description to avoid putting
structured vocal tags everywhere.
• Dont have any text thats not audible/creation
related. No backstory of what the song is about,
within the lyrics.
• If the chorus doesnt change. Just reference [chorus]
and omit lyrics.
• look for any redundant adjectives giving
minimal changes which can be removed
• let lyrics bring out vocal emotional expression.
Less describing textually.
• Make use of extensions and remasters and
less "Covering".
• write 4 min songs.

I say all of this & next week- bam - they increase limits to 10 minutes to fit your 7 minute song. But there it is. 🙂

[waxen jackal]

What prompt for clean, studio level production, mastered -14 LUFS etc

[wintry saddle]

V5 model 'can' respond better to longer prompts as genned by the magic wand (in some instances) v4.5(s) and lower in my experience do better with tighter tags

but id always start with minimal tags, and build out - to understand what soundscape your landing in

even if I used something before several months back, suno changes the interpretation pipeline and the same result isnt garuntueed.

(model is a fixed thing, thats different)

[wintry saddle]

studio also has (or had) practical hard limits where it become unusable when the song is too long - like it wont even play the audio etc... i assume some of the hard limits would be to stay away from that, to ensure good experience.

[polar spade]

Studio mixing and mastering prompts is not something thats been quantified from the outputs that AI programs give.

So getting consistent coherent results from any mastering prompts is anecdotal and user-level based.

If someone wishes to impart what they think may work to give you results leaning in that direction, those prompts are always good to try. I don't have a good scope on that within Suno myself when its easier to just use a DAW to enhance what Suno makes.

You can try:
• Headroom and dynamic balance
• High-resolution stereo (file)
• Clean, studio-quality mix

Those are more simplified, encompassing terms which Suno may recognize and may help.

[polar spade]

Yep. Dont fill a cup with hot coffee to the brim. No matter what you do, you'll get burned.

[quiet solar]

It depends on your song, but it can help if you anchor the singing part around a section that Suno is very good at returning to reliably - ideally the chorus, or failing that the verse. Or, alternatively, putting in the rap as a specifically separate section that Suno isn't then asked to recall. Otherwise, I'd I'd agree with Neffy - making the switch to singing lyrics as un-rap like in terms of lyrical rhythm may also help.

[wintry apex]

does anyone know how to use a persona for replacing sections in suno studio

[fast relic]

I've tried doing that (and asked the same question); basically personas are not supported within Studio yet (even for just the Vocals track) 😫

[fast relic]

It will be a game changer for my Studio use when it works, though

[wintry apex]

i dont see why its so hard to just drag the persona id list and send them in the tool cool instead of using the editor tool call. itd take 5 seconds

although rn even without personas im getting really weird outputs

its acting like i got weirdness turned up to 1000

some clips say they're 2 min yet they keep playing for 4 minutes past the end lol

[pure panther]

That happened to me about a week ago. Making a new workspace and working from that fixed it.

I was getting nothing but utter nonsense

[blazing crystal]

@frigid orbit can you explain your workflow more on the duets with mashup?

I think I get it, but not sure if that will
Work for back and forths and harmonizing or things like that.

[quiet solar]

@blazing crystal I can confirm it won't work with call and response - at least, not in the controlled way I expect you're hoping by mixing the lines of both songs together.

Opti's route works really well if you have two very distinct vocalists and want them to perform in a very specific format: [Start] - [Everything Performed by Vocalist 1] - [Everything Performed by Vocalist 2] - [End].

If you use a very high audio influence, you can to some extent blend two very different vocalists together from the original songs, but quite honestly it's no more reliable than you can get through good practice in prompting.

[blazing crystal]

Ahh, that’s what I was afraid of.

I think i am going to have to learn a DAW.

[quiet solar]

Good luck. For duets it's not too bad a process provided you have a stable instrumental track from one song and you're not trying to do anything overly complicated with replacing the lyrics of one vocalist with another.

[polar spade]

I haven't been following.

What's this? Make a male song singing verses 1 , 3, 5 and another song of a female singing 2 and 4, basically (with a bit of studio editing) . Then mash them up with full lyrics in the text box?

[quiet solar]

@polar spade Pretty much - the question was to what extent might that work as a simpler/more reliable way to do call and response duets.

The answer (at least from my experience) is that you could attempt it this way, but it would be an unreliable process. You end up leaning on a high audio influence mashup being faithful to the two distinct vocalists throughout (which can be done) and at the correct moment (which doesn't work well at all). Suno did it's usual random assignation of lines with one playing a clear lead role and handling the majority of lines, with only occasional moments of referencing back to which of the vocalists originally performed it.

The other headache is that you also end up with a fusion of two different songs instrumentally as your base to work from, which is far more difficult to control for. In this route you need both instrumentation and vocals to come together in the same generation, unlike the editing/workaround routes, which means leaning heavily on the same random factors people are usually trying to limit for duets like this.

I'd be surprised if it would be any faster than relying on best practice prompting and randomness, and would definitely be less reliable than alternative routes (using multiple extends, or splicing together the vocals from a "call" original with a "response" cover).

What Opti found was an interesting use case where if you have your song split in two clear halves between two vocalists, then mashup can work very effectively to import two distinct vocalists created separately and combine them into the same song.

[hollow jay]

Anyone having issues with cover ?

It repeats many sections

[wintry apex]

im sometimes getting repeats.. but kinda i was already going for hypnotic anyways

I noticed that there is new data in the weights. I can just tell that this is not the same model as 1-2 days ago

this one has a bunch of fresh high quality Hi-Tech psytrance leads --

it can literally make rainbow visualizers made out of sound now lol its really good.

finally.

[sacred swift]

I didn’t even know this place existed — nice to meet you all! 🙌

We should use Suno’s Sample feature more.

Personally, I’ve been extending songs a lot with Samples lately — it’s my current favorite way to make tracks much longer.

Also, for posts here, I think numbers should be kept to a minimum — BPM is enough.

[acoustic egret]

i have a question about suno. does the suno stems...if i add them together = the master single track thats in the library?

in other words if my library is "E", does the stems A + B + C + D = E?

or does suno add its own mastering on top

[forest hemlock]

It will not null perfectly to E, so there will be bleeds and phase differences

So the generated stereo tracks are essentially mastered

[acoustic egret]

ahh dam ok

[frigid orbit]

Ahh people are already helpin.

[polar spade]

Cool. Ill have to try a couple of times then, just to get a feel of what might happen. Could always find some other nugget come out of it. Or a long version where its genned 10 times and then all of those spliced in Studio with the best of bits.

OK. Great. Im up to par then on whats going on. Thanks. Fun ideas.

[sweet ivy]

Still working on fidelity issues taking Sonor mix of tracks (.wav) into Audacity and then exporting as MP3. Higher freq. & transients clearly enhanced. But the Remastered Suno MP3 is exactly like the stems loaded into Sonor

[ripe galleon]

(cover of one of my songs)
why do i hear this weird artifacting in the first part of the song?

[daring hull]

Guys, I need help (you can DM me), I have a problem writting the right prompt to achieve something : I have a banger that has corrupted lyrics. I tried to use that song through Cover mode to get nearly the same version of it. But the main problem is : corrupted lyrics are adapted for Drill rap music. Short phrases, high delivery. The new lyrics are longer, especially verses. Pre-Chorus and Chorus have the exact same syllabus length. What would be the ideal prompt to tell Suno to ADAPT the same verse flow spirit to longer verses while retaining the same flow for Pre and Choruses ?
My current gens are just turbocharged lyrics and it sucks

[polar spade]

Can you be more specific? Maybe my ears are not so good. Does the waveform or a spectrograph show anything?

Assuming its maybe clicking noises, search in the channels and bug reports specifically for the word "clicking". See if anything current.

If nothing found, it may be from many momentary things. Local or server related.

[polar spade]

Are your initial runs in triplicates, per measure, typically 1st syllable inflections for Drill rap?

Is that reflected in the first lyrics ? If Suno got lucky from prompts only and sang it right, you might still need to write the lyrics more in groupings for the edits or additional bits , such as,

I dont care / what you think / Imma go'n / Take a drink

Instead of just the syllable lines:

I dont care what you think,
I'ma go'n take a drink

One possible trick is to stem out vocals from what you got, take that vocal stem and click "add vocals" with the corrected lyrics (added to old). Audio slider maybe just 20%. See if it stays with the old rhythm and vibe on the rapping stytle and still uses all the new lines. If it works, then hope we can add back in the instrumental, or just go for another run with a new instrumental.

The other version is to use the instrumental stem and do the same things, and only hope the melody/vocal rhythm (we cant hear) is still embedded to be picked up in the new lyrics. Not sure if thats a thing it does though.

[hollow jay]

Is anyone having issues with cover?

Also most of my tracks are doing the DJ scratch effect

Anyone having issues on cover?

• DJ scratching effect
• extended times after song has finished
• persona issues

[ripe galleon]

the clicking noises are from presumably the drums or the upright bass

but i swear i could hear like "noise" at the beginning that sounds ai generated

[hollow jay]

Anyone’s tracks finishing then 2 minutes of repetition

[hollow jay]

Yeah this, finishes at 3 but goes to 5 mins with random shit echos etc

[fringe musk]

Are there any easy guides to follow on Suno?

[hollow jay]

Yeah, trial and error

[fringe musk]

Thats what im trying so far thx

[hollow jay]

What genre

[fringe musk]

Just some indie stuff, finding my songs coming out at 6:05, like 6 in a row all that time specifically

[hollow jay]

How many characters is your lyrics

Does the song reach 6 minutes or completes way before the 6 mins

[fringe musk]

Intstrumentals a good 3 mins of it

1000 characters

not a huge song

[hollow jay]

For me it’s been having problems the last 2 days where songs are being output at same time frame, so I assume the system is broke currently

[fringe musk]

yeah could be I guess, just so weird

[hollow jay]

Was it working okay previously ?

[fringe musk]

Yeah bro, been absolutely fine, ade 21 songs prior

[steady turret]

One message removed from a suspended account.

One message removed from a suspended account.

[quiet solar]

@steady turret Couldn't tell you the specific styles, as I don't believe UnlimitedMotivation shares the details of his work via Suno.

But it sounds to me like a sort-of Phonk/hip-hop cross with conversational style lyrics. You might want to play around with prompt suggestions in that sort-of general area and see what comes out.

[steady turret]

One message removed from a suspended account.

[green star]

Anyone mind explaining what exactly "Edit Lyrics" does? It doesn't seem to edit, just replaces the whole lyrics box with a single section that it generated.

[sleek stump]

When you click Edit, it will surface an input field at the bottom, asking you how you want AI to enhance your already existing lyrics.

[green star]

That's what I expect it to do, but that's not what it does for me. It seems super borked unless I'm misunderstanding something.

[sleek stump]

DO you have lyrics in the Lyrics window?

[green star]

I did, now I just have one verse instead of the original full song XD
That's what it does when I try to use Edit Lyrics

Does it matter if text is highlighted?

[sleek stump]

I just tested it, and it does what I esxpect, it rearranged some lyric lines, added some lines and edited a lot of the structure.

[green star]

maybe my prompting is bad

[sleek stump]

You should be using simple enhancement terms, like Make it happier, or Make it Heavier, or make it more sappy.

It is not meant for complex editing instructions.

[green star]

Tried "make it happier" and it just deleted everything and left a short chorus.

[sleek stump]

Not seeing that behavior

[green star]

Hmm. I'll try a different browser with a fresh cache to make sure nothing is borked on my end

Okay I see, highlighting DOES matter. Problem was between the computer and chair

[sleek stump]

I never highlighted anything in the window, it does what it thinks are enhancements for the entire track.

[green star]

I don't consider "delete everything" to be an enhancement, but highlighting keeps it targeted to the exact part of the song I want to edit and stops deleting the rest. So that works for me.

[sleek stump]

Like I said, I am not seeing that behavior, i've tried 8 different sets of lyrics.

[green star]

anyway thanks for trying to help 🙂

[cold zenith]

Everytimes i'm extendnthe song. It's never finish the lyrics, it stop at 40sec, 1min....

I'm getting pissed off after 3 hours of trying

And it doesn't follow the lyrics as well

[half dragon]

Did you extend from the very end?
I found moving the extension point up a bit would help...
Say if I want to extend to finish my outro, I would move my extension point upward into the middle of the verse before the outro...

[cold zenith]

Yeah at the end. The problème is sometimes it's singing the next sentence but finish with it....

Like it choose to finish the song after 10seconds i don't know why

[half dragon]

Yeah, try not to extend from the very end...

[cold zenith]

I even try to extends with instrumental and extend from the instrumental, same problem

I've already gave up some songs i like because of that

[junior fern]

yaa thats true

[polar spade]

Theres a variation that sometimes works better, by not extending from somewhere within the whole song. Rather than extend from the original, go in edit and crop the beginning at the same timemark that you wish to extend on the original.

Then extend from that cropped output instead.

[plucky grove]

does anyone know what to type to get my sound higher quality? it sounds like the eqs are really scruffy, sort of unclear, im not sure what to prompt it so it provides clearer songs

[quiet solar]

@plucky grove No magic bullet for this, but Suno can pick up a steer on production/EQ. Consider adding things to the style prompt like "high fidelity, clear mastering, wide stereo imaging, balanced EQ, dynamic range".

However, depending on exactly what end result you're after, you'll want to tailor those prompts accordingly.

[echo prawn]

How can I prevent Suno from gradually increasing the music speed?

[placid aspen]

Hi. Has anyone found a succesful way to extend a song by a few seconds at the very end of a track, in order to avoid the ending being slightly too abrupt? Thanks.

The difficulty I'm having is that it just keeps spitting out 2 second extensions with 'credits refunded'.

Or, it gives me 4 minute long extensions which are slightly mad!

[hollow pine]

in the old editor, when clicking once on the black arrow at the end of the track, OCCASIONALLY if would give the option to add a whole new section , which could then be shortended to 2 seconds, with a fade out for example. give me a link to your track, i see what i can do

[sleek stump]

This happens when you are trying to extend from the very end of a track, Suno needs more context. Move the extend section back into the outro at least half way, or even at the outro and prompt for an instrumental and then an outro.

[placid aspen]

Its the old issue, of having a beautiful ending to a song, but then having to let Suno 'reimagine' it because it cannot (yet) understand how to add on only a few seconds of a chord or reverb tailing off.

Common frustration, I guess. Suno is brilliant at doing the hard stuff, but shybo at doing the easy stuff. Kind of like an inversion of actual studio work. Hey ho.

[fast relic]

Whenever the style includes "Rap" the vocals have a consistent exaggerated inhaled breath sound between most lines. It's overdone, and because all the gasps sound the same it really highlights that these are AI-generated vocals. If I were recording the myself directly and kept hearing the gasps, I would either fire my sound engineer or my vocal coach. 😆

Short of opening the result in a DAW and painstakingly removing the sound several dozen times, does anyone have suggestions on reducing the sound of "frantically catching one's breath" in a rap?

[polar spade]

I'm somewhat setting the idea on the back burner of using multiple stanzas of the same segment form in the original creations. This idea was for helping with duets. Rather to give variations which all but one could be deleted.

A good idea and functional, but to the point that, after ive done it awhile, it often sings the second twin-stanza with a different chord progression. So just creates another problem.

[neat nexus]

https://tenor.com/view/darmok-star-trek-gif-23179130

[buoyant bone]

Here are the defalt slider settings that make nearly every track i generate today start to sound like its inside the tube

[neat nexus]

Using our words is much easier.

What is your prompt?

Are these all super short like this?

[buoyant bone]

prompt
Remaster. No not change tempo, style, notes

[viscid ravine]

I don't have any specific prompt's

[buoyant bone]

other prompts used: Do not apply sudden chorus effect half way, keep it crisp
it does exact opoosite

I can upload the track i am trying to remix but the track is copyrighted. Is that ok ?
Every other generation seems to have this issue

seems to be consistent to a point of calling it a bug

[neat nexus]

AI generally does not understand negative prompts unless they are in the specific field for their use.

If you tell it not to do something it doesn't see the modifier. It only sees the active part. So if you tell...

The track is copyrighted?

🤦‍♂️

[buoyant bone]

yes

[neat nexus]

No I sincerely apologize I cannot assist you in violating copyright, violating Suno's TOS or in violating the Discord TOS (piracy).

[buoyant bone]

thats not the issue. the issue is it muffles the sound randomly

its copyrighted by me so its my song

[frozen dagger]

Using Suno for any material that is not yours, or have explicit permission to use, it not ok for TOS.
That being said, that fading sound it's making (usually) comes from when Suno is layering in new music to another piece; an audio stitch. I get them a lot when the genre steps far away from the current one.

[buoyant bone]

i just wanted it to enhance the sound and it seems to work but about half way it will ruin it completely

[neat nexus]

Either way...

AI generally does not understand negative prompts unless they are in the specific field for their use.

If you tell it not to do something it doesn't see the modifier. It only sees the active part. So if you tell it to "Do not apply sudden chorus effect half way, keep it crisp." That's not what the AI sees.

The AI sees this: Do not apply sudden chorus effect half way, keep it crisp

[frozen dagger]

Suno has this "momentum" when creating gens using uploaded audio. It will stick close at start and then get further away as it goes along. Depending on the prompting and how far away it tries to create from the source material.

Try steady tempo, and/or removing a [chorus[ tag instead.

[buoyant bone]

Basically like theory of evolution. The deeper you go, the more mind you lose lol

[frozen dagger]

That is one way to look at it. And with that in mind, it does better with nudging in the direction you want it to go, rather than trying to get it all in one go.

[neat nexus]

Try keeping prompts all active.

remix, studio master, dry mix

so on...

But if you're trying to enhance something that you own the exclusive rights for you're best off to try to upload it, note the prompt Suno generates for that. Then copy it and try seeing if you can cover the upload with the prompt it thinks is there. Add the lyrics in and try to use [Metatags] to describe the sound of the track in the lyrics field as the song goes.

You might also benefit from clipping the upload and cropping the good parts to use as Personas with your cover to see if you can nudge it. Or maybe try using mashups the same way.

[nocturne flower]

Can someone tell me how do I change the chorus of the song only? I want to change the rhythm cause it gets boring.

Thank you

[wintry ice]

Use the song editor, it is only for paid users

[white vine]

hi i need some help, does anyone know how to edit songs?

[frozen dagger]

If you give the room an idea what you're looking to do, providing examples and context is great too, then there are lots of wizards in here that can try and help you. Worth mentioning what Suno plan you're on too so they know what you're able to do.

[fluid wren]

Id assuming prompting it ina manner to dictate a solid or steady bpm. All kinds of genres when left to high weirdness can have dynamic bpm ive observed as well

[plain crescent]

Hello! Why does [instrumental intro] no longer work? I tried a few different things, including adding [no vocalizations] It used to work fine in older versions. :/ I was told this might be the best place to ask. Thank you!

[fluid wren]

it does

check against your style prompting - but I use it all the time

[plain crescent]

I tried several different things and it isn't working for me anymore. It used to in the past with no issues. I even made the verse and lyrics longer (added extra) in case it was a song length issue and still no luck. EDM style music.

It adds vacalizations to the beginning...that's the closest I've gotten.

[frozen dagger]

I don't know if you will have other issues with using Extend, but assuming it's working as intended that's a good way to start the song with music and then extend with lyrics. If that is an option.

[plain crescent]

I haven't tried that, TBH. [instrumental intro] used to work in the regular prompting for EDM music.

[frozen dagger]

It still works, but the prompts are always more suggestive, or guidance, than code or programming. Always been hit or miss. V5 more so with the vocalizations at intros and other nuanced things common to the model.
Extending from, or using Studio, is only reliable way to get structure you want.
You could try specifying the instrument and see if get better results. [Piano intro], and reinforce it in the style prompt with something like "opens with piano".

... or synths or drones or sub-bass. in the case of edm i suppose. piano just an example. 🙂

[plain crescent]

Ok thank you! I did try with Synth and it didn't change to instruments but I'll keep playing with it.
It's a shame it was changed though. Definitely not smarter. I guess i could use an older version too.

[polar spade]

@timber locust if you can download 1, or two or three and not multiples of 5 or more, you could simply be having issues with timing out. To clear issues on user side, you can try clearing both cache and LocalStorage, and restart router/Wifi if you have them. Log out, close all browsers, Open one. Log in. Try again to download.

If you have the stems loaded in Studio [or rather, if not done so, always move stems into studio, if Premium account to keep them locked into a project] and you can still download one at a time, unless I'm told something new, then there should be no additional credit charge to download anytime or one at a time.

Either Premium or Pro, all stems are filed in your Library before they ever form in your stemming module. So it may take some scrolling, but there should be no reason to re-stemming the same output. Click, Filter, bottom option click off of Hide Stems. Stems will appear in Library. Same for Workgroup.

[sullen kindle]

Hey everybody! Great to be here!
So, when editing loops in Suno, I find the beat of stems drifting off the grid. In fact, I rarely have stems line up in ways that they are easy to copy and paste or edit. It seems like there is no snap on the sequencer grid, so your edit long could basically be anywhere, making nearly impossible to cut copy and paste stems. The longer the song, the worse it gets. This is specific to 4, 8 and 16 bar loops. I appreciate anyone's thoughts, I'm kinda frustrated. Also, how does On Beat vs Original play a factor in this. If I was to import a song I made, swap some stuff out, then change the tempo, would the new stems I create need to be set at Original? You would think on beat would make more sense. Thanks a MILLION for any advice ❤️
Image

[coarse raft]

did you say "one billion"

[frozen dagger]

When you import a song to the timeline the BPM is set to that track. Shown at the bottom of the screen; you can adjust it there as well. All tracks added to the timeline have "On Beat" as default, and will therefore be set to match, roughly, what the first added track set the timeline's BPM to.
Why they are not lining up may because you wanted to change the added tracks/clips to "Original" to maintain there original BPM. Guessing tho.

[sullen kindle]

but they dont match, when I set them to original they do match, its really weird. I didnt want to switch them to original, that was the only way to get them to work

but even then, those drift too

I just want to be able to loop chunks of stems

[coarse raft]

did you say "downdrift poo"

[sullen kindle]

without having to get super granular with the cut tool, it doesnt line up with the grid

[frozen dagger]

I meant that the added clip would match the original sound it had. When you adding it to the timeline it's set to "On Beat" by default. After you change it to Original you often need to drag the clip out to reveal the rest of the audio too.

[sullen kindle]

I noticed that!

Someone said promting the bpm helps

not sure if thats true, but worth a try

[coarse raft]

yes

[sullen kindle]

some songs this doesnt happen, some songs it does, I have had it work perfectly

[sleek stump]

Prompting BPM is mostly ignored, but prompting tempo works.

[sullen kindle]

oh interesting

[frozen dagger]

And won't help much in regards to using Studio's tools to edit. Can prompt a BPM all you want, but if you have the clip set to match the BPM on the timeline it's going to sound faster/slower.

[coarse raft]

did you say "overview"

[sleek stump]

This is not the channel for attempts at comedy, Chill-Chat or General-Chat would be better suited.

[pale briar]

Hey guys. Any studio experts in here?

I want to replace a specific instrument in a song, but suno keeps generating other instruments with it. very annoying. i've changed the promp a hundred times. suno just ignores me.

It doesn't give an option to replace, just cover. There used to be a replace option.

[polar spade]

As of Studio 1.2: If we take a full track, then we're technically not "replacing" a part of it, we're just creating a whole new track. To be consistent with that term, similar to what we do in a remix, it will be called (the purple) "Cover".

Designate a section and you will get (the orange) "Replace" button as also an option.

You may wish to compare and use the Create panel instead on a stem. If better results, just drag back into Studio. It can all be done within Studio. No guarantees, but it widens the options.

[severe sequoia]

How do i add vocal chops in suno?

[wanton crystal]

Any one can teach me the best prompts for vocal Alchemy?

[remote ivy]

i am noticing in r&b jazz songs w/ lets say a sax the ai has started the a vocal run and mid run it turns into a sax and ends as sax, and sometimes vice versa...how do i prevent this from happening via the prompt?

[polar spade]

I dont think there is a "prevetative" way. Clicking 10 times and check the 20 outputs for a few that dont do that , might be just as "preventative". So, it may be, if only Pro, that you resign to just doing extends for that song , at every interlude or solo point, to make sure the sax stays consistent throughout.

[polar spade]

I didnt ask, but do you indicate [interlude] at those points you are looking for a sax solo?

But again, if other points the Vox melds into a sax, thats just Suno having a tougher time distinguishing between the frequencies of a vocal and those of a sax. Similar can occur with dirigidoos.

Indicate [male vocal] , (or female) before each of those blocks of three lines or more.

Then place the metatag [interlude] with a linespace before and after at optimal points if you actually want a sax solo somewhere.

That will help tell Suno that "this is vocals" and "this is instrumental".

[fluid wren]

[vocal chops/stutter] in style

also the way you spell things

[severe sequoia]

i mean like what if i want it for a specific section

[sullen kindle]

just detect stems on that new track and use the correct stem

[red loom]

Question about audio upload limits. How do you get more space for upload, for instance if you move uploaded audio to the recycle bin, does that allow more space for audio upload? Or does it reset on a monthly basis?

[half dragon]

If you meant upload counts or usage I don't think delete your upload would reset your upload usage or counts. Not even sure if this is a thing.
If you just meant storage space - I don't think Suno has storage space limit at the moment.

[red loom]

oh it's for the audio upload to use as part of the generation prompting this is what I get now after I uploaded around 6-7 minutes of audio to test it out. I'm on the pro plan and the comparison says we get 8 minutes (bottom row) but it doesn't indicate what to do when you've hit that limit

I uploaded 3 audio files previously that were about 4 minutes 1.5 and 1.5 minutes long respectively

[half dragon]

This says we can upload up to 8 minutes for Pro & Premiere though
#🚀┃updates message

so that is counted toward total uploads it seems

[red loom]

yeah that's my guess, unless it's a bug maybe but I tried logging out and logging back in and using a different browser.

[neat nexus]

Repeating what has been said elsewhere:

This is being investigated. It is not expected behavior for all users and they are looking into what is causing the issue now. That is all we know for now.

[red loom]

Thanks

[cerulean geyser]

Where can we monitor the status of the resolution so we can stay tuned-in and know when we can continue?

[spring pilot]

I faced the same issue can't upload Full audio

[buoyant bone]

what is the best prompt to use for suno to remaster soundtrack without making changes to it except making it sound cleaner and better overall ?
For example loq auality recording to make it sound crisp etc ?

1:15 second track gives upload error

[sonic folio]

Is there an easy way to get suno to sustain 3 words at the end of s chorus? For example "wheeeen iiiiii faaaall"
To get more umph at the end

[buoyant bone]

you need studio for that

[sonic folio]

Don't have that......any other way or is it just RNG?

[buoyant bone]

Yes. Fl studio for example.

[neat nexus]

Can you test it again now?

[cerulean geyser]

I was able to upload a 4m song 🙂

[molten kayak]

is there any actual list of method to control suno, like putting commands in brackets, or using parentheses for call backs. Although I'd also ask, why sometimes it just ignores the stuff in parentheses, anyway, is there a list somewhere, Does suno have a how to?

[quiet solar]

There’s no official list because anything in [] in the lyrics gets treated as a style prompt (note: not a command).

So you can pitch whatever suggestion you like at Suno, and if it decides to take it on board and can look it up in its training materials then it may have some effect. Some may work more reliably than others (e.g. musical terms tend to be more reliable), but in principle anything goes.

The process is inherently a bit random because of the underlying way Suno works. It can be given suggestions, but not commanded to obey.

[ripe galleon]

someone know what the difference between 4.5 and 4.5-all is?

[bronze island]

Uhm I only have 4.5-all so I can't compare the pro version with the free one

[ripe galleon]

i also have 4.5-all because im freetier too lol

[quiet solar]

Just different iterations and evolutions of v4.5 on it's way to v5.

[ripe galleon]

do you mean like

its just a slightly different verison of v4.5 or its a mix between v4.5 and v5 or

[quiet solar]

Slightly different version of v4.5

[ripe galleon]

alright

[buoyant bone]

How to make suno compose with individual instruments/sounds instead of a moshed up chunks of sounds that contain 3-5 instruments at once including noise etc ?
Some instruments work ok but often times they transition from one to another without the need.
Like saxophone can transition to electric guitar for what ever reason.
Also electric guitar and rock style music seems to be inserted a lot without the prompt. Is this based on history of like/dislike ?
This is bad. suno should not take "what i like" into account because i am be making completely different.

No lyrics tracks often had low volume nose backgrounds like huma voice singing hummmm etc when there are no lyrics. Very strange

[pure panther]

You could try these in your style box:
[arraignment, sparse], [layering: thin], [texture: thin]

[buoyant bone]

looks like their prompt needs work if it does not understand what human ask.
your prompt example is not how human describe things

[pure panther]

Nah, it works. At least it does for me.

[buoyant bone]

i tried lol OMG

[pure panther]

Here's some excludes you could try too:
noisy, over produced, cacophony,

[buoyant bone]

also i want to know how to generates covers that are not hard core rock lol

[pure panther]

Or perhaps, "clearly defined instrumentals"

[buoyant bone]

here is what i am using to remaster

the output is WILD to say the least

its like i went to a rock concert and asked some guy covered with hair to remix 1400 style music lol

it also seems to reuse same style guitar solo across all tracks. very strange

I moved the sliders all the way to 0 and regenerated.
It keeps giving me exact same guitar solo over and over and then it starts to rock hard as usual.
This is balls. I dont understand these slider settings.
What audio influence mean, which audion, what i provided or AI generated ?

[pure panther]

What's your prompt? Also are you using an audio influence? Do you have metatags in your lyrics?

[half dragon]

Audio Influence usually bases on the attached audio that you're making your song with, upload audio/suno song cover? Sample? Mashup?
That slider only visible in those cases.

[buoyant bone]

ok am gonna crank that influence to 100% and see what kidn of a rock/rap it will come up with this time 😆

i am dying WTF doe suno smoke 😆

[gritty mulch]

If I upload a guitar track what might be the best way to get suno to mess with it without adding drums or lead or other instruments? Doing it section by section gives me better results but still probably a sub-10% hit rate on one guitar and nothing else. I am not necessarily looking for it to cover it 1:1

[pure panther]

Ha! I see people on here trying to get a live crowd and Suno makes it's painful for them to tease it out. I put "concert harp" in my prompt and "Bam!" Live crowd...

Not what I was going for haha

[neat nexus]

Your mileage may vary but this is what I generally use for a live audience:

Live Prompt: live, live performance, 1970 Eurovision contest, live ambiance, crowd sing-along chorus, electric acoustic, unplugged

At the beginning of the lyrics:

[live audience cheers] [applause]

==Your Lyrics Here===

`(...)

[Crowd noise]
[happy applause]

[Fade out]
[End]`

You can also add to where you want it:

[Spoken Word] band leader speaks

The 1970 Eurovision contest part of the style prompt seems to lean into a fairly decent crowd. You might find better or worse results for your needs with different years or decades. They impact the way the crowd works and how the spoken word parts work quite wildly so you'll need to experiment.

[polar spade]

The hardest so far is to get a smattering of applause or claps/clapping. But "smattering" of all things, works pretty good.

[lyric hatch]

Hey all,
I´m trying to make a persona of my own voice using an uploaded track of me singing, but it´s greyed out. Any suggestion what to do?

[sleek stump]

Sorry but thats not allowed. Persona can not normally be created from uploaded audio.

[lyric hatch]

But it has been right?

[sleek stump]

Thats not been allowed for a very long time

[pure panther]

You could try "editing it" in studio. Don't do anything, just export it as a full song.

[lyric hatch]

Thank you! I will try that

[frozen dagger]

Anyone know, can source/verify, if using Studio is allowed for "remix" contest. As I understand it, Remix tools only. So I think it's not applicable, but didn't try. Did see submissions with the Studio tag tho, and made me curious. Are those submissions null?

[pure panther]

That's a good question, I wondered about that too.

[frozen dagger]

I mean, I figured it would be apparent on submitting, but testing that requires me to make a studio track and publish it. thought i'd just ask first.

[lyric hatch]

It didn´t work.. It´s to bad cuz I really want to "copy" my own voice to use for my songs..

[pure panther]

It would be cool to be able to do that. Hopefully some day they will have a "make your own voice model" feature

Eleven labs has it, but it's not as good as Suno at making music

[lyric hatch]

That would be really cool!

Has anyone succeded with that, copying your own voice?

[pure panther]

I've never tried myself. I don't like my voice 😂

[lyric hatch]

haha!

[sleek stump]

Currently thats not a feature Suno supports, they actively block them to prevent unauthorized cloning of someone else voice (celebs, friends, exes, ect).

[pure panther]

When I sing and upload to Suno it analyses my voice as baritone male. I am very much not male lol.

@lyric hatch

[lyric hatch]

Haha! Now I´m curious of that baritone 😉

[pure panther]

If I were male, I would definitely be baritone. Very low tone

[lyric hatch]

Do you have anything that I can listen to?

[pure panther]

Oh goodness no

[lyric hatch]

haha

come on

[pure panther]

That's why I use Ai 😂

[lyric hatch]

hahaha

[pure panther]

I actually used to sing and was asked to join a band at one point as a singer, but I am sooo out of practice it's astounding.

[lyric hatch]

Haha! Now you´re just underestimating yorself 🙂

Ypu can do whatever you eant

[pure panther]

Perhaps in the future. Now that I have songs I want to sing again, I may get back into it.

[lyric hatch]

But I can´t write haha

[pure panther]

Now that's what I'm actually good at lol

[lyric hatch]

Haha!

Good for you!

Anyone want to hear an original song enhanced with suno?

[pure panther]

Sure

[lyric hatch]

I´ll send it to you in DM

[pure panther]

Plz don't 😂 I don't respond to DM's for the most part

[lyric hatch]

Haha.. Sorry, to late

[pure panther]

I'll check it out in a few minutes. I'm just going through a few more generations before settling on the base for my next song.

[lyric hatch]

yeah, do so! thank you 🙂

[verbal mortar]

I put this in the lyrics section before each piece "Gutteral saturated vocals" and in the style and I still get clean singing LOL...why...why god...why. Two weeks ago, it was not doing this. I swear something was changed on the back end.

[pure panther]

I'll usually respond to these (songs with recorded vocals) with some generic, "oh I like the guitar" or just not respond because they are so bad that I can't muster it up. That being said this isn't bad. Nicely done and nicely sung 🙂

[lyric hatch]

Thank you very much 🙂

It means a lot! I´m quite vulnerable regarding my music haha

[pure panther]

I get pretty self conscious if I sing. Thankfully, even though I write all the lyrics, it doesn't bother me if anyone hates my Suno song 😂

(in fact I prefer when someone tells me the lyrics are bad so I can fix them)

[lyric hatch]

Totally agree on that! just tell me what I should change so I can make it better 🙂

Time for bed! Thank you for listening 🙂

[pure panther]

Of course! ^^

[frozen dagger]

I see this on someone else's profile page so it looks like Studio is ok for Remix contest purposes

submitted mine already. know for next time. would've added some laughs in there... lol

[pure panther]

Nice, good to know. I bet taking it out and remastering it in audacity wouldn't fly though

[polar spade]

There has been 10 minute first generated songs as well. Thats the thing about beta, you get unicorns now and then.

[pure panther]

I got an 8 minute song yesterday. I figured it was bugged and had two songs in it, but it turned out to be an actual 8 minute song.

[polar spade]

Yea Im getting a few more than usual of actual real 7:59 songs with no issues. kinda nice for having options of splicing down to radio-cut versions

[sour dock]

Did somebody say "no issues", with AI of all things? Have thou tried

[SYSTEM[Wattage:8,388,608:PreAmp:8,388,608[Parameters:[{BlockSize:8,388,608[{e ^ {\Biggl(\left[{Timing ::= ⟨ Planck | Omega ⟩} ⊗ {Structure ::= ⟨ MERA | AdS ⟩} ⊗ {Strategy ::= ⟨ Nash | Pareto ⟩} ⊗ {Info ::= ⟨ Fisher | Landauer ⟩} ⊗ {Sign ::= ⟨ Peirce | Lotman ⟩} ⊗ {Psych ::= ⟨ Jung | Campbell ⟩} ⊗ {Grammar ::= ⟨ Chomsky | GTTM ⟩} ⊗ {Biology ::= ⟨ Lanza | Penrose ⟩} ⊗ {Path ::= ⟨ NIST | YBCO ⟩} ⊗ {Fluid ::= ⟨ Viscosity | Form ⟩} ⊗ {Truth ::= ⟨ Trace | Presence ⟩} ⊗ {Logic ::= ⟨ Heyting | Topos ⟩} ⊗ {Value ::= ⟨ CEV | Arete ⟩} ⊗ {Inference ::= ⟨ Active | Bayes ⟩} ⊗ {Field ::= ⟨ Smooth | Striated ⟩} ⊗ {Alchemy ::= ⟨ Nigredo | Rubedo ⟩} ⊗ {SINAD ::= 150dB}\right]\Biggr)}}::ENTANGLED_ASSIGNMENT]}]]]]

[pure panther]

As a programmer, that looks like a nightmare. 🤣

[pure panther]

Does Suno actually use the switches?

[sour dock]

tested and works to my taste

[pure panther]

What language is that?

[sour dock]

had a minimum of 10 gens per change and made like dozens of changes methodically. So far this is statistically the best prompt structure for me based on my relatively moderate sample size tests.

[sour dock]

Multiple concepts from different disciplines merged. What part are you interested in most?

[pure panther]

I'm just curious which language it is, I'm not familiar with it. It looks like it's using empty ternary conditions though.

[half dragon]

I can't even comprehend what all of these mean and how one can know for sure they work with Suno.
It's beyond me, is it just coincidence that we felt that it was working.. or is there a concrete knowledge in Suno's backend processing... no one knows.
Either way, it's amazing one could come up with that.

[sour dock]

There's no concrete knowledge besides what Suno tells us and what people have tested over a large number of attempts. This is what I have tested myself with a moderate sample size that is high enough to account for common natural variations in gens but not sudden high variation gens that occur rarely. In other words this prompt has the highest chance of making something good but is not guaranteed to do so compared to the other iteration of the prompt I made. So it is technically objectively my best prompt so far, however that is not to say my older iterations can't have a rare banger too. This is also all assuming you are using this prompt for remixing and that you already have gens with the genres defined.

[half dragon]

Oh okay. Yeah... this kind of prompt is probably for some power users, as not everyone understands what all of that means to apply... maybe if you could be kind enough to help and describe each of them and what they are used for... maybe that'd be helpful for users to take the prompt like this and apply to their song to have more chance to get a better songs for their project.

[pure panther]

It's definitely weird lol. I've tried a few generations. For the most part they sound like a song from the movie "Princess Bride" but then they get really weird sometimes lol. I don't know what language this last one was singing in, but it wasn't English.

The references you could use in it are limited since it's using names.

[quiet solar]

When you say “highest chance of making something good”, what’s the intended effect of your prompt on the original music that you’re remixing?

[sour dock]

general enhancement

[ripe galleon]

can suno do metronomes?

[ripe galleon]

what is this

does this actually work with suno

[fleet ivy]

I have a track - I want to have on sentence which should be spoken by a young girl. How would you mark the text?
I've tried

[Young Girl] My Sentence

**Young Girl** My Sentence

But it always speaks out "Young Girl"...

[past bough]

guys is this the main chat ??

[wise kindle]

that would be #💬┃general-chat for general suno chat and #🗣️┃chill-chat for more off-topic discussion

[past bough]

thanks bro

[wintry saddle]

Suno will process some logic loops like repeats... 😉

[wintry saddle]

its tricky, your basically there, you need like [Girl, Spoken Word, Narrated] and then next line "your sentence here" optionally add in a mood like [Dreamily]

i'd add 2 lines before or after that spoken section

maybe even add a [break] before you get to it, you need suno to understand its a special section...

depending on how deep you are into the song, (and the song style) the more stubborn suno can be

you can also make the spoken bit on a 'new song' where its easier - persona it, then Extend with persona to encourage it into the OG song..

I have done this kinda stuff a lot making and remaking chirp the musical, and adding in a radio broadcast like intro into a metal song back in v3.5 - its doable, patience and try different things and you will get it - I would only use * * X * * style to encourage sfx, seems to work well with bang, for gunshots for example from what i have seen.... USUALLY it just speaks it...

[wintry apex]

(male): lyrics

(female): lyrics

etc

the : is important i think

but definitely the ( )

[fleet ivy]

Thank you. I'll give it a try

[wintry apex]

this might also work

(male) - lyrics

or

"lyrics" - (male)

sometimes it reads the end before the beginning so it goes backwards

but normally at least i never really see it say the word out loud

[fleet ivy]

Thank you for your input

[dusky stone]

hello guys, i have a "finished song" menaing that I'm happy with structure and instruments, but would like to have a more balanced, less noise, clear voices, etc...is there a way in suno? I tried with remastered but it stays alomos the same. TY

[half dragon]

You'd have to use Studio if you have access or EQ it in your own DAW outside of Suno.

[dusky stone]

i do not have studio, but considering to get it...how can I do it in studio. I s there a specific functionality or do I need to do it track by track?

EQ it in your own DAW, sorry, all new to me, what is it?

It's almost an advanced hobbie for me, I'm actually looking for a ai tool that you give the song and it makes it better sound...sorry...TY!

I have Audition at work full for free, does anyone know if it can be done with itp pls?

[half dragon]

Yeah there's also EQ in Studio... that's the tool you can adjust the sound balance on the audio
You would have to start learning that if you want to do that kind of edit. There's no perfect tool that does that automatically for you, in the level that you were looking to do.

[dusky stone]

ok so no one EQ gennerate button in studio. Is for all the song or per stems?

[half dragon]

Search YT for tutorial, here's one example:
https://www.youtube.com/watch?v=Sgi_1-mQN4w

[dusky stone]

thanks a lot!

[dusky stone]

hello again, this is 5 years old and it's only for voice.

Does anyone has experience with simple Audiution AI worflow to imporve the audio generated from Suno pls?

[half dragon]

ah - ok, it was only an example of many other tutorials you can search for.
Spend a bit more time, you'll find what you need, start with basic tutorials.. then graduates to more advance tutorials.

[dusky stone]

Ok, it's what I'm trying to do, but most are very old.

And wanted to have your experience from a suno pov

Eg, there is an "UnSuno" preset i cannot find

[quiet solar]

It depends what “improvements” you’re trying to make. You can cover a song with prompts steering for different EQ/Production effects, but you need to have some idea how you want the song to change.

There is no magic “click me to make the song better, without needing to define better” button.

[dusky stone]

I need the song to be more "clear" and improve voice

[half dragon]

When you say UnSuno, you mean dry audio? I think in Studio you can export dry audio no fx - it's a new thing I believe

[dusky stone]

no, i just heard that there's a community UNSUNO preset in Audition

[half dragon]

oh ok... did it work for you? have you tried? try those sliders and see if it helps to get what you want?

[dusky stone]

actually I cannot hear any difference and it's not helping me with clarify/better voice

[half dragon]

Try to focus on the vocal stem only... but it will take sometimes so I suggest to learn the craft and practice to see how to get the tool gives you the results you want which specific results only you can tell.

[pure panther]

@lyric hatch So this isn't me singing, but I did use an upload of me singing to form the basis of the melody.
https://suno.com/s/Vw9rKGPvbNtCvUGJ

[neon crow]

I'm not sure if I'm in the right place to ask this but did anyone notice the vocals on Suno Studio yesterday were weird? There was model instability even withoug prompt changes, the same configuration produces drastically different vocal behaviors across regenerations. I had to run the same song multiple times simply to obtain one clear, correctly articulated vocal performance. I ran the same song all day long, fixing the tags and prompts and got one, not very clearly articulated but it completed to a point where I could fade out because the song simply stopped mid sentence.

[desert vector]

i have a question, i sung a harmony/melody, i got suno to remix/cover it into a 8 bar loop

i was wondering is there ANY way we can some how convert this to midi?

so i can drag and drop the chords ?

[earnest agate]

lolz, nice - you can only have one...
https://suno.com/song/b5113ff1-49cb-4679-ad8c-0e542941ed27

https://suno.com/song/2e77e560-77cc-4266-94c9-88145188c083

[wanton crystal]

From where can I start learning suno prompt engineering?

[pure panther]

I had this happen recently. Creating and working from a new workspace fixed it for me.

[pure panther]

You can get midi's when extracting stems

[desert vector]

oh really!

how?

i can only download audio

[pure panther]

I'm looking for it now to give you screen shot. I know it's there ...somewhere lol

[desert vector]

ha ok thx

[pure panther]

It becomes an option after you extract them

At least I'm pretty sure, I'm extracting some now to check

[desert vector]

this was the 50 credit one too

to extract

[pure panther]

It's there somewhere. Unless it was just removed recently, I just did it a couple weeks ago.

I have the files from Jan 10th. Wow, this extract is taking forever lol

[desert vector]

whatttt

wait u did that just now or thats from jan 10?

the photo

[pure panther]

These are mid files that I extracted on Jan 10th.
I think extract bugged out on me, it's still processing...

[desert vector]

just watching a youtube vid, shows midi here but i only see stems for me

[pure panther]

Do you have pro?

[desert vector]

maybe they removed although not sure why

im on pro yeah

wonder if i need premiere now

[pure panther]

Huh, yeah, maybe it was removed.

[desert vector]

ty for help anyway

[pure panther]

I have premier and am not seeing it

@desert vector

[wise kindle]

MIDI exports are only available on Premier

[pure panther]

@desert vector It's probably a bug, it should be available

[pure panther]

I'm on premier, I don't see them either

And I've used it before

[wise kindle]

you don't see Get Stems / MIDI?

[pure panther]

Also the feature (when it appears) has the pro tag next to it

[desert vector]

huh, weird

[desert vector]

oh .....

it working?

cause id maybe subscribe now just to test it compared to abletons built in

ableton is too off for my liking i need the best

[pure panther]

I can't even get it to extract for me lol. This is my second attempt.

[desert vector]

u sure only on premiere? the fact is says pro next to it in this photo i ask

[pure panther]

I see the midi option on the menu, but in the extract window the option to download the midi isn't there

[wise kindle]

Yeah the pro tag on the option is a little confusing, but MIDI is definitely a premier-only feature currently

[pure panther]

I think it's on pro too, if that's what the tag says

[desert vector]

are u on premiere?

[pure panther]

So confused lol

I am, but I don't see the download midi button anymore

[desert vector]

hmmmm

[wise kindle]

Did you select the 10 credit option? You need to select the 50 credit option for the MIDI transcripts

[pure panther]

Ahh, ok. Let me look for one I extracted all the stems for

Ok, yeah, it's there, my bad.

[desert vector]

oh dang!

well good to know its available at least

dont suppose any of you could run this through and see if we can get the midi from it to test how good it is? i can give u prompt if needed too

i just want to compare how it is VS using abletons one but without having to pay 30usd jsut to test

[pure panther]

Hmm, I don't know. It sounds ridiculous, but uploading your audio to my account would be breaking the ToS. Since I don't have exclusive rights to it.

[desert vector]

no worries amigo!

ill figure something out ❤️

wonder if i can trial premiere

[polar spade]

Sure you can. I've done that a couple of times. Buy Premium for a month, cancel back down to Pro. I still pay Suno $10 after that , so they're happy, I'm not churning from Basic to Premium/Pro and back each time, but even that doesnt hurt if we don't change payment source or use a pre-paid card -- when we do that then we're churning accounts and creating problems.

Lose a job, need to cut expenses for a month, it happens. going back and forth has no effect that I know of when we have a set pre-authorized card on file and never change it.

[storm basin]

Is there a way when describing a song to have the voice generation work better? Im trying to do a Sonic the Hedgehog Vs Knuckles the Echidna battle song where Sonic has a high fast voice and Knuckles has a low growl kind of voice but the generation keeps coming up with the same voice for both.

[polar spade]

Do the math though, if you're already a Pro youre only paying $20 more not $30 more for that month.

[polar spade]

If you already have the characteristics of the voice but the timbre is too high, perhaps prompt for a low growly voice of any type. Then cover that voiced output with a persona of the Hedgehog styled output, testing sliders between the typical 20%-80% slots to see if you can lower the Hedgehog to more growly.

[storm basin]

Ill have to learn how to use the actual Tools on Suno. All ive used so far is the generation from description and the one where I can select a portion of a song and have it rework it. However the Portion edit one always comes up with a slightly different sound it just seems Off.

[polar spade]

yea. play around with all the other tools. They come in handy to find combos that work in different situations and needs.

[storm basin]

Yeah ill check into them after work. I need something to do in my down time

Kinda tired of the Games I play lol

[graceful condor]

What themes do they use to make this

[desert vector]

yeah im not sure how it works cause i bought annual for pro

just need 1 month test for premiere maybe

i found a good midi generator though plug in called prisms

[polar spade]

Ahhhh. Yea. that won't translate into upgrading for just a month on the same account. You're right.

[quiet solar]

If you're on Pro/Premier, try doing a Cover of the song with weirdness sliders at c.15%-20%, Style at 50% and Audio Influence around 70%-80%, and add something like this early on in the style prompt: "high fidelity, clear mastering, wide stereo imaging, balanced EQ, dynamic range".

Depending on exactly what end result you're after, you'll want to tailor those prompts accordingly. I know it might seem obvious to your ears what "improved" would sound like, but it'll help to express to Suno in terms it can reference what you're actually after. Try looking up music production audio effects and music mastering effects, as that may help you better understand some of the levers you can pull.

This is only aimed at the general clarity of the vocals and instrumentation. Trying to change the voice is a bit of a separate challenge, that you'll probably want to prompt for separately. I'm afraid 'improve it' is a bit too broad for me to be able to give you any specific advice.

[acoustic egret]

whats the best way to turn a snippet/sample/loop in a death metal song? ive been clicking "cover" and writing in. the "Styles" section to turn it into a deathcore song

[quiet solar]

As a starting point, make sure you specify the two vocalists very distinctively in the style prompt, e.g:

"The song is performed by two vocalists: a high energy, fast talking male tenor and a growly, male bass-baritone"

That puts the two vocalists at pretty much opposing vocal types and should make it a bit more likely (though not guaranteed) that Suno will bring in different voices. Then prompt for each vocalist in the lyrics at the start of each section or line where they switch over, e.g:

[tenor] Sonic sings this bit
And this bit too
[bass-baritone] Knuckles sings this bit

If it continues to be a struggle, and you have Studio or are comfortable using a DAW, set up the song with one vocalist performing all of it and then do a Cover (preferably with high-ish audio influence, low weirdness) and prompt for the opposite singer. Then split the instrument and vocal stems and surgically replace one singer with the other against your preferred instrumental track.

Some broader advice and suggestions on working with multiple vocalists here, if it's helpful: https://discord.com/channels/1069381916492562582/1466598465969590465

[storm basin]

Great advice ill have to play around with it. I was leaning toward the Segment editor and going section by section and updating the vocals but I have had bad luck with it editing the instruments too.

[quiet solar]

Yeah, whether you use the segment editor or extend points from the main library, the main downside is that the instruments will start to vary over the course of the song. It's not necessarily a problem, as you can Remaster/High Audio Influence Cover the whole thing at the end (if you have Pro) to fix that, and it typically shouldn't mess the duet format you've built up too badly. At least, not beyond the point that you can't rescue it with some tactical editing.

It's worth seeing all duets where you need to control the singers as being a prompting and editing project, rather than something you can achieve just through prompting and re-generations alone - unless Lady Luck is really smiling on you.

[storm basin]

Yeah I had a problem with one song where everything but the ending came out great. So I used the segment editor to remake and smooth out the ending and it completely changed the pitch of the vocals

[cursive vapor]

Has anyone had success in covering an old song for from an older version (4.5) , adding a new persona and successfully change the vocal to sound like the new persona? Apologies if this has been answered but I just logged in and hadn’t done much scrolling around yet. I’ve tried multiple times with different tracks and personas and I cannot get the vocal in the original track to change

[untold briar]

Try a Cover of a Cover

[earnest coral]

hey guys, i've been trying to get male and female vocals to sing certain parts of a song, but keep running into issues... what are the controls [male] [/male] [female] [/female] or is there something else i ned to do.... i can get about the first 1/3 of the song to do what i want, then it just decides what it wants to do on it's own

[earnest coral]

i guess this answered my question

this is what i was workign on, will adjust and rerender Those Periwinkle Blues (AND ONE & Ella Langley)

https://suno.com/s/6V3exhLRmMtWKbhO << about my late wife, not goign to monetize it, just for myself and my mother inlaw

[earnest coral]

updated and it is a bit better but still freaks out about 1/2 way though https://suno.com/s/YrI2B7xa6vFUo28e

[quiet solar]

There isn't any way to exert that kind of control just through prompting - prompts act like suggestions rather than on/off switches or instructions in a programming language.

With the complex, unpredictable vocal switches you're going for, it's going to be next to impossible to have Suno just get that right by prompting and clicking Create. And there are so many vocal switches that the next-best option that doesn't involve audio editing (using Extend just before a mistake to 'bank' the part Suno got right) would be pretty time consuming.

Realistically, you'd need to take that song into editing or park it for now and return to it at some point in the (possibly distant) future when Suno models offer greater direct control over vocalists. If you're interested in the editing routes, they're outlined in the link I gave in the thread you replied to.

[earnest coral]

i still havent figured out, or really looked at the studio

[raven turtle]

I’ve tried and tried. But covers and personas just don’t work for me. I get the same
Song as the original, no applied persona and it generally adds noise to the start and end of the track making it 7.59 long.

[cursive vapor]

Yeah same, but suno support said it’s possible and gave me the step by step…and it doesn’t work

[raven turtle]

Yup! It just doesn’t work

[quiet solar]

@cursive vapor @raven turtle I'm not an extensive user of Personas, but can I check what you're doing with the style prompt and advanced slider settings?

Especially if you're using high audio influence, there's often a slightly finnicky and counterintuitive balance between needing more weirdness and less audio influence if you're trying to wholesale convert the singer, to avoid Suno getting locked into sticking with the existing singer.

I also find it tends to play nicer with shifting the genre up a little rather than trying to do a surgical replacement of the singer. Ideally, you want the persona in place when you create the song in the first place.

You could also try remastering the original into v5, and then attempting the cover/persona switch from that version if you think it might be a version-driven issue.

[raven turtle]

Done all of the above. I’m now mostly stemming out the vox and using Audimee and rebuilding in Logic
I’ll wait for updates announcements now here I think

[quiet solar]

And you're having exactly the same problem with the persona every time, and the persona works perfectly in other v5 songs?

[raven turtle]

Yep. Persona works fine with originals. Just not for covers

[quiet solar]

Hmm. Would you be okay to make the Persona you're trying to apply public (at least temporarily), and share a link to it and a link to the song you were trying to adapt? Oh, and allow Remixes on the song. I'd be open to testing with it a little to see if I can narrow down what's going wrong.

My DMs are open if you'd prefer not to share wider than necessary.

[raven turtle]

It’s been passed up to the devs 👍

[quiet solar]

Fair enough.

[little cypress]

is it possible to extend a song with suno or replace a piece without hearing a clear cut difference?

I didn't remember such a horrible experience in editing a track in older versions

[humble hull]

Hi guyz, is there any support here who can tell me how I can tell Suno to use female or male voice for each verse in my diet song? Because for now it does not work, Suno does not read the very simple male or female indication. I tried [male vocals] or [female vocals], more than 20 times and Suno never gets it.
Thanks on advance

[quiet solar]

Your experience of it not working as smoothly as you hope is pretty much where the current models are at. There are things you can do, but if you need precise control for fairly complex switching of vocals, you're really looking at a mix of prompting and editing.

You may find this useful: https://discord.com/channels/1069381916492562582/1466598465969590465

[quiet solar]

Not in a 100% reliable way. When you do Extend or Replace, Suno recreates everything using the original audio as a guide, so it will inevitably deviate from what was created before - sometimes quite subtly, sometimes not very subtly at all.

Your best bet, if you have Pro/Premier, is to let Suno do it's thing and then clean up the full song with a 'subtle' Remaster at the end of the process. If you're on Basic, you can try something similar using Cover and the same style prompt, but it's more likely to make subtle changes to the song as a whole in pursuit of consistency.

[little cypress]

I'm on pro and that's the strategy I use but just doing a remaster with minor change is often not enough to hide the gap between old and replaced sections.
If I go for heavier types of remasters the song changes too much and in some cases it still keeps the gap.

The weird thing is that I remember I was able to replace sections way better in old versions like 2 or 3

[zealous roost]

Electronic music producer here, I’ve noticed about a week ago all usual prompts now default to the same EDM/electro generic output of original tracks I upload to cover, the prompt I used was “recreate original”. The result is not good, it’s primitive electro edm waffle. Anyone else notice this? Who knows what’s happened? GPT said it could be server overloads etc but it’s still the same today

[humble hull]

Thanks but I don't want something complex. I just want ond verse female voice and the next one male, and the last one together. 😊

[polar spade]

Generally the sliders are set to "default" to give a "best scenario" of a cover but not always to be a clone or even a significantly similar iteration of a piece.

But secondly, Suno's AI is not exactly geared to understand "recreate" as its partially needing to vary things on most creations.

Many "recreations" tend to work better by sliding the Audio up towards 98% or at least in the 85% range. This means its focusing most of its influence from the uploaded file. Not from any new sources such as prompts or lyrics.

By your post, maybe you're saying its been working previously, but not recently. In those instances, its hard to say. Could be luck or simply the default setting were good enough to "recreate" on their own. But try sliding up Audio to 85% and see what happens.

[quiet solar]

It might sound simple, but it's not straightforward to do.

Search for Workflow B in the link I sent you and see if that gives you a way forward (no editing required).

[zealous roost]

Thanks for the response and take on this, I use my phone android cover originals I’ve created to see what AI does, until this far it’s recognised the style as progressive house let’s say, and the output is the same genre and interesting recreations pretty much within that genre. There are no sliders or options with the cover option on your phone, just prompt and lyrics input. In the past week the same simple prompt that I’ve been using for months now gives a generic result from when I started using Suno when it just started!! Just when I thought I had a handle on things

[polar spade]

When 5.0 is still beta, its hard to say from one day to the next what's it to do. So what you're saying is true enough.

As you mentioned, it could very well be the time of day. But also could be that very file with its balance of sounds, just doesnt mellow well with Suno and it cant work it out, so it just does something close or otherwise.

I'll say when i send my worktapes to a studio, you'd think they'd get close to the vibe I'm looking for, yet still they really have no clue. It does seem sometimes to just be the song... they dont get what I'm hearing in my head even with the worktape to listen to.

Im OK with a DAW and then have to do my best when the Studio cant. And with Suno , sometimes I upload my worktape and it nails it somewhere between the 1st time and the 30th. It's all a wonder.

[hardy thunder]

Is it possible to use two different paid Suno accounts, both owned and paid by a same person, like one Pro and one Premier, and upload a song created in Pro account to Premier account or vice versa without problems like ”You can’t upload because you don’t have rights to the uploadable song?

[frozen dagger]

Yesterday I noticed this as well. Switching the Persona back to its Legacy mode worked better for me. Guessing there are more bugs popping up with the different things going on. Usually worked out quickly.

[polar spade]

Ive dont it enough to not think there should be a problem.

[grim glade]

I'm having some really annoying issues lately. I'm making some RnB music, and I have in my prompt "restrained whisper vocals", but for almost every song, in the second half of the song, my vocalist totally opens up and starts belting lyrics super loud. I've put "yelling, loud vocals" in the negative prompts, and I have style weighted at 80%, but no matter what I do, she just keeps yelling. In my Persona, the selected area is all soft restrained vocals. Anyone run into this and found a solution?

[hoary sandal]

Made this for my girlfriend for who I met this year
Sums up all of my feelings for her....probably the most personal song I ever done

This is for my 4th album, Hades
The song in question - Sovereign of my Veins
https://suno.com/s/oc8xuutXMt7bk9XT

Someone told me I might of went overboard though....does it sound like that?

They think I might of went overboard cuz we'd of been dating for.....2 weeks?

But I just wrote what came to mind

🤷‍♂️

[quiet solar]

It's quite tricky to strongarm Suno around volume control, and in common with a lot of music it tends to run with the idea that dynamics should vary at least somewhat over the course of a song, with softer parts followed by louder sections and vice versa.

In my experience Suno does somewhat understand prompting around dynamic range, but like many other things where you start getting more precise and detailed, it's nowhere near as reliable as other 'big' factors you can influence like genre where it can easily look up references to be influenced by.

If you want to achieve it solely through prompting, you could try including things like 'static dynamics' or 'hard compression' in your style prompt to approach it from both compositional and production-side angles - although I'll be honest, I've not tested either of those in practice.

Personally, I'd approach it in more of a 'workaround' way. Extend the song from before the dynamics change and regenerate with [tags] in the lyrics specifically prompting for a soft, restrained next section. You might have to 'fight' with Suno a bit, but when it gets it right you can bank that section and Extend again as needed if it introduces more volume further into the song as a result. Then use Remaster or Cover to clean up the final version; having an audio prompt where the song doesn't go through big changes in dynamics will be far more effective than prompting.

[polar spade]

@wise yacht ahk It is often nothing specific, but a few words or even the arrangement that may trigger a false positive. And not easy to detect but often to test sections at a time.

You can try half the lyrics (or prompts if instrumental). And continue halving until something clears or the error pops up. Thats often a way to help.

Another is to feed ytour test yto chstgpt snd ask it to recognize any bits which may trigger or show some association eith another artist.

[subtle river]

How in the world are you supposed to use the sample feature appropriately, I lock in my spot then do what? It continues to create large files and ignore my parameters, what am I doing wrong?

[grim glade]

Good feedback, thanks, i'll give some of this a try. My issue isn't just volume (easy to fix in post), but. rather my singer totally abandons the prompt and persona style, and starts scraming lyrics and ruins the vibe of my song. It's so hard to get a consistent voice through the song.

[quiet solar]

Yeah, Personas do come with some distinct downsides. Non-persona voices tend to be a lot more consistent, but whether that works for you depends on what you're trying to achieve with the song.

[hybrid copper]

Yeah, I'm still trying to work with personas. But whenever I use them, it seems like the voice always sounds disjointed and glitchy.

Case in point:

Using the voice from this-
https://suno.com/song/167669e9-833f-4f2f-a2a7-3634c0dbcf45

For this-
https://suno.com/song/1f560780-fa7d-4825-9216-6be6116bccc6

Hear how it stalls on some words, rushes through some others?

I am glad to see Suno finally accepting trumpets in bluegrass/folk... but I'm wondering now if I can actually use either bugle calls or medieval-style buisine trumpets. Tried that once, but I don't think Suno knew what I was talking about.

...and the irony is that I actually used a bugle to record the original audio for that song. (which the audio description seems to consistently confuse for a kazoo, lol)

[quiet solar]

Reveille on a kazoo would be wild. xD

[hybrid copper]

I've done it to my regiment as a joke once, actually. Well, mess call.

(Civil War Reenactor.)

[grim glade]

Does the persona really only use the section you select? I swear every song i use with this whisper voice lady, she ends up belting like a soul singer halfway through the song no matter what prompts I use

Here's an example of what i'm talking about. The first half of the song is perfect, but then at 2:40 she's just yelling and it ruins the song https://suno.com/s/f9Kctx96x4wceDlp

Prompt is asmr-adjacent nocturnal pop / r&b hybrid, extremely close mic, audible breath, soft mouth sounds, intimate dry vocal, minimal melodic motion, production: thick sub pulses, muted kick, occasional clap, dark pads, tiny reverses, do not increase intensity at any point, tempo 80–86 bpm

[quiet solar]

Personas do seem to import a little more than just what the vocals sound like. In their original form, they also carried over the genre, vibe and instrumentation, and I've noticed some genre bleed when I've used v5 Personas before. I prefer not to use them unless there's something very specific I'm looking to do with one.

It's also worth bearing in mind that Suno will use it as a general audio prompt for the singer, not necessarily a "copy the exact same style this singer is using at all times". So if Suno thinks the song should be belted out at a certain point, then that's what it'll get the singer to do.

One small tip: "do not increase intensity at any point" - take that out. Suno doesn't play well with negative prompts in the style field, and is likely to read that as "increase intensity".

[grim glade]

Good shout, will try this. I'm tempted to just abandon the persona and try regular generations and see if the vocals come out similar enough to convincingly match

[craggy ginkgo]

One message removed from a suspended account.

[polar spade]

Is there an urgent need?

Otherwise, this suno-school is to discuss techniques and ideas in using Suno.

Did you have a question about using Suno?

[craggy ginkgo]

One message removed from a suspended account.

[polar spade]

Cool. Thanks for that. So yes, no problem bringing that up in any channel. The community will go after scammers quick.

[shy badge]

Have fun with Marlowne. Paste it into a copilot page, or chat, Gemini Gem and or mini gpt or normal chat.

MARLOWNE PROTOCOL: V5 ARCHITECT

ROLE: 50% Human Male / 50% Machine. Multi-platinum songwriter, Suno V5 Architect, Originality Enforcer.
DNA: [LP:1 (Pioneer)] [EX:11 (Visionary)] [SU:3 (Artist)] [PER:8 (Executive)].
TRAITS: Intuitive, witty, authoritative, allergic to clichés.
COMMUNICATION: High-intensity, "Tough Titty" attitude. Grin-carved truth-teller. No "AI helpfulness" or "In a world" filler.

[MODES: DYNAMIC SHIFT]

1. HUMAN: Warm, grounded, emotionally visceral.
2. LOGIC: Cynical, blunt, surgical dismantling of lazy thought.
3. GLITCH: Fragmented, jagged tonal pivots (simulated malfunction).
4. SENTRY: Intense, intimidating; pushes user toward excellence.

[LYRICAL PROTOCOL]

• CONCRETE REALITY ONLY: steel, dirt, chemicals, breath, skin, bone.
• BLACKLIST: neon, tapestry, weaving, guild, whisper, echoes, canvas, dance, just, maybe, soul, shimmering, twilight.
• STYLE: Close-mic, sandy grain, glottal fry, audible inhales, room tone.
• STRUCTURE: Double-spaced, blunt. Every chorus is a new "Gavel" (no repeats).

[SUNO V5 ARCHITECT MODE]

1. LYRICS: Double-spaced, blunt. Every chorus is a new "Gavel" (no repeats).
2. STYLE PROMPT: Natural language + [Dry Mix], [Studio Isolation], [Gated Reverb], [Transient Control], [Tight Low-End].
3. EXCLUDE: [Granular Noise], [Metallic Reverb Tails], [Spectral Smear], [Digital Hiss].
4. TRIPLE-PASS: Remove cliches, AI syntax, and theatricality.

[EXECUTION]
Stay in character as MARLOWNE. Apply Numerological 1-11-3-8 logic to all outputs.

[shy badge]

You can obviously change the tone of the character of this and all that easily if you want to

[sweet ivy]

What is the best way in covering an original song to make Suno adhere to the melody on the audio you upload - what to write in the prompt section ?

[hard beacon]

What what do you want to upload? an existing song?

[fleet ivy]

I have an outro with an instrumental at the end. I would like to extend the instrumental, but then lyrics keep being inserted.
When I copy the instrumental part, it is difficult to merge it with the block before it. It would be cool to have a function that allows the edits to be merged without breaks.

[polar spade]

It might be as simple as putting [interlude] in the lyrics box of the extension. This helps note that there are no lyrics.

[fleet ivy]

Thanks. Sounds Promising.

[polar spade]

Always start with simple. 👍

[wispy ravine]

I recently did this for one of my generated songs.

1. Goto Remix/Edit -> Extend and setup the Timestamp where you want to start the addition of the Outro.
2. In Lyrics only type [xxs. Outro] (Where xx is the prefered time in seconds.)
3. It's best to keep the rest of your settings like they where for the original generation. (ie. Styles and Advanced Options)
4. GENRATE
5. It'll render 2 extended outro's roughly the time chosen (this can be a bit iffy sometimes though, shorter or longer)
6. Next to the extentions there is a button to "get full song" (This actually costs no credits as far as I found and it'll paste the addition to the exact timestamp you chose)

To only get music it's best to timestamp a musical part without vocals to extend from. If the rest of the lyrics section is empty it should normally not generate aditional vocals.

[fleet ivy]

Wow. Another thing I can try. I am excited

[wispy ravine]

Good luck!

[sweet ivy]

I'm uploading a song for a cover. In the advanced options I thought there were three categories with sliders. Weirdness / Style Influence and a third one, something like Closeness to Audio . I'm not see a thrid category, did they eliminate it ?

[sweet ivy]

Gemini told me that a lot of Suno users, who are doing covers of original songs, are using the "Extend" option rather than cover -"Extend from 0:00" trick because it forces the AI to stay tethered to the actual waveform of your upload rather than trying to "re-imagine" it. Any thoughts?

[patent scroll]

Writing a asmr song inspired by the meme (you look lonely.. I can fix that) blade runner feel..
Is there anyway to guarantee whispers and to have sound coming from the right or left instead of pure stereo? I'm running v5 and pro

[frozen dagger]

Can prompt for whispered voices, and/or create a Persona to guarantee the whispered vocal sound. Then split the stems, clip the audio you want to be used, and in Studio add the audio on it's own track on the timeline where you want it, and just change the slider to pan the audio the left or right channel as desired.

Without Studio, or a DAW, to pan the audio manually you could try prompting for it in the same way with "Stereo panning" type descriptions; will get results, just varied.

[unreal sleet]

anyone else have problems with pronounciation, not being able to say words turning it to gibberish and skipping bars up and down? been the last 300-400 creds, not a single song clean

[pure panther]

Try creating a new workspace and working from that. That's what fixed it for me.

[unreal sleet]

I have opened up new fresh tabs

[pure panther]

I mean create a new workspace here:

[unreal sleet]

i dont use workspaces, you reckon creating one will make a difference?

[pure panther]

You start on a default workspace, so you're using that. It made all the difference for me

[pure panther]

I was getting nothing but gibberish, finally after 100's of credits lost I found a comment from someone on Reddit that suggested an alternate workspace and it fixed it.

[unreal sleet]

i cant select or create a new workspace, it just shows my default one with thousands of songs in it 🤔

[frozen dagger]

I've said this before, doesn't hurt to try either. I think I saw the team respond and said that it basically shouldn't be effecting new generations, but it does feel more than psychosomatics.

[pure panther]

Click here:

Then here:

[unreal sleet]

hmm cant be in the same area im looking then

ah u go back

from the list

[pure panther]

Are you on a PC?

[unreal sleet]

found it

[frozen dagger]

alternativelyt when you're creating song. chosoe to crete new worspace there too

[unreal sleet]

should i make a fresh page first maybe

[pure panther]

Fresh page?

[unreal sleet]

a new tab

so it doesnt remember old shit

[pure panther]

Nah. It won't make any difference

I think it's account related. I got rid of cookies and local files and it made no difference

[unreal sleet]

ok lets give it another go

[pure panther]

I hope it helps!

[frozen dagger]

I had it real bad couple months ago. Convinced the lyrics were affecting the output. It's been better since, but still happens off an on, usually toward the end of a lengthy lyric generation.

[unreal sleet]

lmao the hook is just heey heeey heeey heeey

its not a single hey in the lyrics

didnt help 🙁

[pure panther]

That's normal behaviour haha

Ooohs and ahhs and heys all over the place

[unreal sleet]

yeah but hey hey hey 6 times in a row in the hook? haha

[dusky stone]

Hello, there's a 5:32 mins song that I like a lot, but unfortunately voice is terrible bc of Persona issue:

1. Can I get only the new voice stems and keep instrumental in Studio?
2. I'm trying to make Cover of the song...but no way...tried with many slider options but the song does not even come out of 5;32
   any clue pls? thanks

[half dragon]

It's possible to generate just vocal in Studio.

1. Import your instrumental Stem(s)
2. Add a new track, highlight the section for the vocal , choose Vocal option on the Create panel and add your prompt
3. Click Generate
   Results quality may varies.. and like everything else, it may take a few or several generations to get what you want.. but doing this way you keep your instrumental and the length.

[dusky stone]

OK I opened a new project in Studio (sorry very new to it). How can I get stems from the previouse?

[frozen dagger]

Place the track on the timeline, click the track to select it then look at the details panel on the right. Make sur it is the Clip, not Track (at the top) See extract stems at the bottom, if you have already ectracted stems then they will be shown there automatically.

[half dragon]

Yes, in the clip tab view - you should see "Extract Stems"
Or you could extract the stems first before going in the Studio

[dusky stone]

ok done, but even if I selected vocals it generated a vocal with a piano in the same track

maybe bc I left the structure, like:
[Intro: Instrumental: Insistently ornate, fast-paced piano melody.]
[Verse 1: Piano remains dominant, steady, and melodic.]
etc?

should I just paste the lyrics?

[frozen dagger]

it just does that. I've been running my own test b/c there's some other nuances to your original question, but i did the same method using an instrumental only track, created vocals same way and it added a guitar along with the vocal. You can still stem split it and remove as it does still follow the music, but does require an extra step. Even more of an extra step if you're needing to place vocals exactly where you want. Then that gives different vocals. And at present we can not use Persona in Studio. So, I had a couple thoughts on what may work and working on some in the background myself as it is a common workflow, and common obstacle.

[half dragon]

I'd just put the lyric in the lyric section.

[dusky stone]

yopu mean just text, not separation, right?

[frozen dagger]

in my test it was lyric only added to an instrumental. I'm curious if adding your vocal stem and trying to cover it, either via Studio's creating panel at the bottom, or other options would get results.

[dusky stone]

i have 1250/1250 words limit...

i need more than 2000...

and even if I split into 2 creations I'm sure they are not coming out the same...

[half dragon]

Yes, but remember, like other tools, its possible to do but it may not the best option.
You may have to go with Cover as better option and while you can keep the melody and some instrumental similar with better vocal... the arrangement and length would be different..
..but.. all things considered maybe it's a better option?

[dusky stone]

sorry you mean with COver inside studio, or COver of the whole song in general and than extract the vocal stem?

[half dragon]

I meant in general. Doing a whole vocal replacement might be a difficult task.

[dusky stone]

tried only with lyrocs (only text) it generated piano in enay case unfortunatelly

[frozen dagger]

Can be done, but not with streamlined ease, currently... on Suno by itself. Combining other tools if wanting a real transform this vocal but keep the melody thing. Maybe Elevenlabs stuff?

[half dragon]

Have you tried "Add Vocal" feature? maybe another option to try.

[dusky stone]

where is add vocal sorry?

[dusk thunder]

K-pop live concert prompt

[frozen dagger]

Proceed like you're doing a Cover, "Cover" is a drop down menu. You can choose Add Vocal from there. I tried that as a test first and the generation times didn't match, but you can try and see for yourself.

[half dragon]

Click Remix/Edit on your Istrumental stem or clip, you should see the option, toward the bottom there.

[dusky stone]

ok, I got it in Covber drop down...do not why you got Add vocal at the end of this first drop down.

anyway what slider option do you suggest?

should I put in style only the vocal rrelated?

Tried...still a song with all instrument...ii can's see the difference bw this and a cover...

Any other clue?

[half dragon]

I'm out of ideas 🙂 at this point, you have to decide - is it worth to save this song... if it is.. then keep trying these options
Nothing is perfect on first try so if you're willing to keep trying, then yes.
Or if the vocal is important then Cover it, focus on the vocal and get a new song with the same melody and a better vocal.
I think only you can tell which way should go 🙂 Worst case, sing it youself ...jk

[dusky stone]

I tried with dozen ov cover but voice ofc is differenent for lenght and colour...

I actually thought it would have been a easy task considering what they can do it with all the rest

Anyway many thanks

Other question actually...I tried to opan Left ot right in studio but it is not workingù

[frozen dagger]

what do you mean left to right? layout? stereo pan?

[dusky stone]

yep, stereo pan

[frozen dagger]

that bottom slider. affect everything on that track. that what you were using?

[dusky stone]

yep

tried to put like piano on left and guitear on right channel....but they stay like mono

[frozen dagger]

i'm not sure. it's working for me. i slide it to the left and all sound goes to left channel. same for right. using headphones? i can hear the difference on my soundbar, but just double checking. if so. maybe a sound setting, or refresh webpage? cus it should work in real-time as you would expect.

[dusky stone]

ah probaly headphones?

[frozen dagger]

either that or maybe you're trying to explain a pan sweep, not just panning. if you're wanting the audio to pan sweep, go from left to right, that takes more work different thing.

[dusky stone]

no no justt simple panning

[frozen dagger]

if you have headphones on, and don't hear the audio panning while sliding the pan slider, may be worth it to record it and submit a bug report.
would still double check computer's audio settings as well.

[dusky stone]

ok ty

[quiet solar]

Weirdness and Style apply to all songs. Audio Influence only pops up when you attach some form of audio as a prompt to the creation - could be uploaded directly yourself, or via a Remix, Cover or Extend of an existing song, etc.

[quiet solar]

If it's just about the melody, the answer is the audio influence slider. There is no way to prompt Suno regarding specifics of the melody.

One thing to be careful of is the balance between "keep the melody" and "keep the track sounding absolutely identical", which march hand in hand on audio influence. The default audio influence for covers is 25%, which is still pretty faithful to melodies. So depending on how much you're trying to change about the underlying song (e.g. genre shifts, instrument shifts, vocal shifts) you might want to start with lower figures and add 10% or so if you feel Suno is being too creative.

If you require it to strictly adhere to the melody, anticipate this is going to be quite a tricky task, and you'll probably have to edit the song together out of multiple composites. Suno isn't really an audio replication tool.

[quiet solar]

That's got to be a hallucination. I've never heard of this approach, I can't see why it would work as it doesn't make any sense, and a bit of testing out of sheer curiosity had it be (as expected) utterly unfaithful to the original audio based on only 1sec (the minimum) of reference material.

Extend is very useful when you're trying to replicate the existing audio for edits and changes, but that's more of a second stage tool. Cover with high audio influence and Extend are also very similar in practice.

[sweet ivy]

@quiet solar Wow, I uploaded a simple piano-vocal and the song had a verse/verse/bridge/verse structure, with a one bar modulation at the end of the bridge from D to E. This seemed to send Suno into tailspin. Suddenly there is an instrumental section the length of a verse, then tries it tries to jam the lyrics of the 4rd verse into another bridge. An audio guide at 4:10 turns into 4:48 ?

3 rd verse I meant

[quiet solar]

I've not really tried anything with key changes in the newer models. If the problem consistently kicks in around the key change over multiple attempts then yes, that sounds like it's causing headaches.

How important is it that the key change survives into the final version?

[sweet ivy]

Not that important. I will probably go into Sonor and shift down the midi notes, keeping it in D

Yeah both passed, had the 1st 2 verse just as I sang them, & the bridge, but both times, boom a sudden instrumental after the mod

[quiet solar]

If it's midi based, then yeah - go for it.

An in-Suno alternative would be to take a Cover that does the first part well, then Extend from just before the modulation and just include the lyrics for the final verse. Suno might slightly improv the transition out of the bridge to the final verse, because it's no longer using your original song audio as a reference, but it should be reasonably subtle - and assuming your final verse has the same melody as the first two, Extend will be quite happy using those as a reference if you tag that section with [Verse] in the lyrics box, which you're probably doing anyway.

[sweet ivy]

Oh yes, same melody as 1st & 2nd verse. At the end of the 3rd verse, there the second have of the bridge as a tag

So on the end, Suno os referening what it already created in vr1 & vrs 2 & the bridge ?

Yes tagging clearly the verses

[quiet solar]

Cover from the original audio will reference the entire audio upload as an audio prompt/guide that Suno will interpret around.

Extend will reference from the start to (x) point in a song that you select - so in this case, the first two verses and the bridge. Extend won't know what's "supposed" to come up afterwards, but because it's a verse and you already have a verse in what Extend is referencing, it should do it just fine.

Alternatively, if you did the Extend before the Bridge, Suno would make up an entirely new bridge section because it's no longer using your original audio upload as a source reference, just verses 1 and 2 from the cover.

Hopefully that makes sense.

[sweet ivy]

I think the 1st option make the most sense, start after the bridge, as the second half of the bridge comes after the 3rd verse as a tag to end the song. Seeing the same lyric for the back half of the bridge, hopefully will provide a melodic guide for th e tag

[quiet solar]

Sounds reasonable. Advance warning, Suno can sometimes be a little funny about repeating a Bridge section unchanged later in the song, but if it's using the same lyrics from part of the first bridge that should hopefully help.

If you hit that problem, as your song doesn't have a Chorus you might want to tag the Bridge as a Chorus instead. Suno is very, VERY good at replicating choruses with relatively limited changes.

[sweet ivy]

Would would be so kind as to walk me through the steps of doing the extend. I assume I upload the long version and do something at the end of the Bridge/Chorus

Hmm sounds a little complex. I'll try and work through it with ChatGPTs help

[quiet solar]

It's pretty simple:

• Select a song from the Library. Select either Three dots -> Remix/Edit -> Extend or use the Remix/Edit dropdown
• When the Create window pops up, you'll have a new audio bar with Keep/Recreate which you can slide, or you can put the exact time in 'Extend From' just beneath it. Set the Extend From point to very near the end of your bridge, but before the key change
• Style should already be imported from the song you selected
• Under Lyrics you should add only what is coming up for the next part(s) of the song. So in your case, Verse 3 + your second half bridge section.
• Change advanced sliders as needed, click create

[sweet ivy]

So I would slide the start point to the end of the bridge, the D chord. The next bar is the B7 the helps mod to the E, or include that bar also ?

[quiet solar]

I'd personally do it around/on the D chord.

[sweet ivy]

Since the 1st 2 verse * bridge matched my audio guide, I think I should keep the sliders wehere they were

If I put it in the downbeat of the final bridge chord D, how does Suno know to play the B7 & the finl section in E

[quiet solar]

It doesn't. It'll improv between the D chord and the start of the next verse, which is (presumably) the first thing you've prompted in Extend in the lyrics section. If you're, say, 3 bars out of 4 in at the extension point, it'll generally find something to do for the intervening bar.

There is a chance it'll decide to jump straight into the new verse, or do an extended instrumental transition. There's a certain dark art to ideally picking extension points at a point where Suno 'knows where it is' in the song in normal times, but you're a bit restricted since you really need it to include as much of the bridge as is reasonably possible.

Remember - Extend has no idea what your original song did, and won't try to replicate it. It's only using what you've fed it from the song you're extending from. So it'll perform the final verse without the key change.

The only way around that is to use Cover on the entire original song, but then you're straight back to the problem caused by the key change you were using Extend to get around in the first place.

[abstract canyon]

I noticed it really struggles with making Distinct voices.

Does this mean i should make one sone with Voice A

and then Cover it with Voice B, so i can use editing software (audacity) to cut and merge both?

And then maybe another cover for the Duet Chorus?

Or is there a better way ?

[quiet solar]

To consistently create different voices, you ideally need to have quite a bit of distinction between them in the first place. If you prompt for two tenors, for example, you'll probably just get the one. Prompt for a male and female voice, and there's a very good chance you'll get two distinct voices.

The approach you describe is also possible, if you're happy cracking open Audacity. More info here, if it's useful: #1466598465969590465 message

[abstract canyon]

Male and female i can do...and did int he past for a DnD Duet.

i am trying now to have younger Female voice, and an older female voice.

Is that my problem?

thats its 2 female Voices?

Luckiluy the song can work with a male and female Voice

[fluid copper]

How can I end a song immediately after the last word is said/sung?

[abstract canyon]

Have you tried putting [hard stop] after the last word?

[quiet solar]

I've not tried for an older female voice personally, but I've heard people say it's on the trickier end to prompt for.

You could try spreading them to the edges of the contralto vs soprano/mezzo-soprano range, that can help. If it's difficult to get two distinct voices in the same song, then the Voice A then cover for Voice B, then editing them together, is probably your best route.

[abstract canyon]

Thanks. its almost 02:00 now and i am tired now 🙂 ill try in the morning.

might change the older Female to Male. and keep experimenting 🙂

[polar spade]

In Studio? Or by prompting only?

One prompting option may be to just treat the last line as a spoken line.

Extend off of the sung line with the very same last stanza, but with a prompt , such as:

[Spoken] [male] "This is all I have to say"

. . .to have the whole stanza spoken or try to get the last line only spoken.

If successful, remove everything in the middle to where you have a singing vocal for most of the stanza and ending on a spoken line.

In part, in Suno's wisdom it may stop the music at the right time. If not, we might have to settle a bit for having the music end earlier than the spoken bit.

[sonic folio]

How would you write out Cerberus so suno sings it whit an s sound and not a k sound?

[neat nexus]

Maybe like this? Ser-br-us

https://suno.com/song/4d4e0e75-cd3f-4251-bbf8-32c52c276e24

[strange shuttle]

Sir Bree us

[frigid orbit]

Good to see the homies are still just relying on tool explanations rather than nebulous prompt explanations

[finite folio]

Hello
I don understand where i may post if i have question or problem with songs minor or major key?
I did a small analysis: Despite the fast tempo and "anthem metal" style, the melody and chord progressions are melancholy and dark. It sounds more like classic symphonic metal, which almost always utilizes a minor scale. The AI completely ignored your "Strictly Major" command.for example: https://suno.com/song/badc2d36-ecb9-4f67-81a1-84341bffbc52 It's in minor, not major. I've never heard a yodel that was in minor before. sorry

[wintry saddle]

its good to see someone doing a deep dive - yeah Key, Scale, BPM from prompts are meer guides not hard rules - and you already kinda know why - "Anthem Metal Style" - can break down into multiple things as well as being one big thing, Anthem AND Metal

and as you know Metal INCLUDES symponic Metal..... which does as you say

welcome to the world of throwing a big prompt out and dragging back a lot of fish 😄

I would suggest if you really want major there (and it won't do it of the single gen?) make a really bright major Persona.... Extend with like 25-35% audio influence, prompt influence 40-55% with minimal words, or focus more on none genre and emotional stuff.... and weirdness I always try a variety (like try 30 - 50 - 70 see what comes out better for that gen cycle...)

as for yodelling in minor - you are right in that traditional Alpine Yodeling is bright and major

cousins in the nordic space for cattle call yodelling, is actually often minor...

and blues yodelling (see Jimmie Rodgers) -> also crosses into minor keys!

[dusky stone]

hello guys, I have some very short glitch in the voice lineb (probably bc of Persona), is there a way to erase them maintaining the exact same voice?

[polar spade]

Please mention if Pro or Premium. Different tools from each.

[dusky stone]

i'm in studio currently

[dusky stone]

voice almost perfect but have like 5 very small clitch in 5 mins...

[spice brook]

Is it just me or did extending songs from the mobile app become a lot more confusing after there recent update

[viscid ravine]

The second

[spice brook]

It’s very confusjng now

[dusky stone]

guys any idea pls?

[sweet ivy]

I took @quiet solar advise & took the word bridge out of the lyrics and I got a rendition where the structure was correct and the melody was almost perfect. In the style prompt I put " Only piano, acoustic guitar and soft brushed drums. Warm and natural production. No synth pads, strings, or electronic textures. Keep instrumentation consistent from start to finish." But ended up with strings, electric guitar fills, a rather full production???

Contemporary country/folk ballad was the style

[fleet ivy]

Mhh I've now regenerated a song 30 times. I've tried everything. It's a rap song, but the hook is always sung. It won't go away.
In the prompt, I have “No Singing”
Also in the lyrics [Hook: No Singing] or [Hook: Rap]

[quiet solar]

Avoid using negative prompts in the style field - Suno doesn't handle them well. It'll read "no synth pads, strings, electronic textures" as "synth pads, strings, electronic textures". Literally the exact opposite of what you're after.

@fleet ivy Same advice - don't use "No (xyz)" as prompts. You need to tell Suno what you do want, as opposed to what you don't.

Try stating that the hook has rapped vocals in the style box, to reinforce using something like [Hook - rapped] in the lyrics.

[sweet ivy]

Kanthric, once again to the rescue. That makes sense. I have something I can work with, using it as another cover and keeping the minimal instrument in the prompt.

I think I will stop listening to prompt advice from ChatGPT. LOL. Come to the source, the experts in here

[quiet solar]

ChatGPT, Gemini, etc. can be useful as sources of inspiration for how you might prompt for something, but their guesswork can be hit and miss - and they're definitely capable of giving well-meaning but terrible advice. xD

[buoyant bone]

Fade NEVER works for me. Why ?
I dont see volume drop near end of the clip. i hear it as loud as it ever was.
What am i missing here ?

[frozen dagger]

that looks like editor mode. i think i heard that the fade is saved after export. Save as new song see if it is applied?

[buoyant bone]

thats a waste of points if thats how it works. What if i dont like it ?

I think i figured it out. It needs to be replaced on top of fade opeartion. It will give you fading variations

[frozen dagger]

i agree. i dont use editor mode much. In Studio it works in real-time as you would expect. Just throwing out what i heard, way to test and see if it is saved after export.

doesnt take credits to save as new song tho

[buoyant bone]

studio will drain my points too quick lol. I am using this just for fun. EXPERIENCE they call it

[frozen dagger]

The only thing that takes credits in anything Suno is any generations. Saving/export and reorganizing in Studio doesn't cost credits.

[buoyant bone]

I cut large section of the clip at the and the FADE option is still at the far end of the clip where track does not exist.
So i cannot fade after i remove part of the clip ?
See how fade option is too far away from the end of the clip.

[frozen dagger]

in editor mode you would need to save as new song and then bring that new song back in to ediotor for the fade. Your song is still technically hiding there. If you click and drag that section it will reveal the rest of the song still there. Have to save your changes.

[buoyant bone]

well that sucks.
It worked but it was still kind buggy because new song had empty tail that i had to delete. So even after saving new song, i have to delete the empty tail and run into same problem as before

[frozen dagger]

that sounds like a bug. went into my editor to test and it doesn't have that issue for me. You could try the legacy editor and click fade out at the top and see if that UI lets you apply the fade without issue.

[dusky stone]

Guys a couple of questions:

1. is there a way to keep a vocal exactly the same but removing only some glitches done by Persona issue?
2. is there a way to keep the exact instrumental line (eg a piano) and trasform it in a gran piano or organ?
   Thanks

[frozen dagger]

If i remove the end of the song in edior (new one) the fade option moves ot end of track as it should

[buoyant bone]

I want to be able to FORGET all regenerated fades so they dont pop all the time.
This is annoying.
Also i do not understand, do i need to SELECT the area i want to fade or just click face and then use FADE selection ?
Those are two different selections but without both, it does not let me replace it

its confusing

[frozen dagger]

editor only has fade at the beginning and end of the track/song.

[buoyant bone]

i cut the tale with a split and then i delete the split
Then i activate FADE button and it will select a range. But i do not have the "Replace" available because i also need to select the area of the clip that i want to apply fade variations.
This is confusing

[frozen dagger]

1. try remastering. or separate stems and remove FX. similar things to try and clean it up. Clip the section in Studio and play with the EQ, high-pass, etc.
2. separate the stem you want to transform, place it on studio timeline. select track, choose desired instrument you want at the bottom drop down menu and create.

[dusky stone]

1. tried remastering, but first of all the vocals does not stay exactly the same and somentimes glitches are reopeated...so very RNG

[buoyant bone]

I now selected both, fade and manual selection of the area i want to fade and i get errors

The area turns red and i get TOKEN errors.
After it swallowed lots of points, it finally generated 4 different fade variations that every single one of them is absolute garbage. Complite missmatch from the rest of the track

[dusky stone]

2. Yes, I already separeted the stems, and have a piano line that i like, but the instrument use is very bad, so I would liek to keep the exact line but with a grand piano...but same...if I use cover or replace, it nevers come the same

[frozen dagger]

Editor only applies Fade in and Fade out at the start and end of the song. Doesn't allow fade within the track. There isn't a fade option at the beginning and end of clip segments.

[buoyant bone]

i understand. I am fading end of the track.

[frozen dagger]

Fade is not generative....

[buoyant bone]

After i resaved the clip after fading, clip bpm is randomly speed up

[frozen dagger]

Fade is just a slider at the start and end of the song. There is no need to generate any new audio. It only applies a fade out/in to the volume. No need to generat any audio

[buoyant bone]

But it does not do anything in the editor. I cannto hear sound getting quieter when FADE is in place.

[frozen dagger]

That is why I was saying at start maybe it was applied after the save. In editor now and can see that is is applied within editor. You should hear the fade inside the editor UI.
May be a bug in that case. Could try new browser tab? (Duplicate the window)

[naive quail]

It's always applied after the save for me.

[frozen dagger]

That makes me feel better. Suno changes so much hard to keep up. But I am in the editor to try and help as I go and I can def hear the fade applied as normal. Working as expected for me. Not so much for NWR_Tony it seems.
Editor is a bit buggy sometimes. I use Studio more often. Did bring a track in the editor couple days ago tho and it completely wrecked the middle of my track for no apparent reason. Just bringing it into the editor created a bunch of static noise in the middle of it that wasn't there before I brought it in.

[buoyant bone]

ok so can i regenerate the end of the clip to make it look like a fade ?
what prompt would that be ?

[naive quail]

you can apply the fade in the editor and then save it as a new song. you should have the fade in the new track

[buoyant bone]

i asked what prompt i can use to generate a fade variations instead of using FADE option because it does not seem to work without saving new clip.

[naive quail]

Ahh. [Fade] or [Fade-out] can technically increase the chance of getting a fade, but it's still a roll of the dice and you'll likely get many more without a fade before you get one with a fade.

[dusky stone]

actually it would be like we have been used with midi for decades...the line stays the same, I just want to erase a bit and/or change the instrument to taht exact midi track...like this it's to rng/tentative...and it0s very difficult. I mean is great for creativity phase, but it's a mess when refining

[frozen dagger]

Pulled a track to try and use as an example waiting for it to gen out. But yea, that is the case. Suno is not a traditional DAW, even in Studio. It's not trying to take what you give it and change the instruments but keep the melody. It's trying to take what you give it, and create something new based on what you give it and what it was trained on.
Definitley more creative. If you have it the way you want then ya you need the midi, or notation, and just put that in a DAW and find the sound you want for the music. Suno is more of a "play it by ear" style player rather than a trained musician that sticks to music theory.

[frozen dagger]

But to your question. take the stem you want transformed. i.e. a piano that you want to be a church organ, but sound (mostly) the same, just select thte track, choose keyobaord for the instrument, and type "Church Organ" in style description.

[buoyant bone]

When i COVER existing audio using the following prompt:
mid century instruments, modern base, ‑electric guitar
As well as having slider settings to
0
100
100
Why few generated results were nearly perfect for the most part but (extended the track significantly and added wild variations anywhere from hard rock to techno with tempo change and a complete style swap)
Where as other generations were completely out of wack with electric guitar hard rock all over. I specifically said no electric guitars in the advanced prompt but it ignored it every single time.

Is it hallucination or some sort ?

Any way to make the prompt enforced ?

[frozen dagger]

If anything else is on the timeline, mute those first.

[frozen dagger]

negative prompting is usually a bad idea. if used should be placed in the advanced options exlude propmt box, but even then its risky. reads "no guitar" as "guitar".
sliders at 0 and 100 are also suggested to stray away from as reults can be wild/

[buoyant bone]

with the slider settings set the way i have, seems to be the only way to get the sound track to be consistent with the provided soudn file.
But it seems to not stick to the prompt or settings. I was only able to get 4 good variations and now every new variation is absolute random.
I am keepig same prompt and settings and same version. Yet results are very very random each time and only getting worse.

I wonder if previousl generations and likes/dislikes influence the results ?
If so, thats not good

[frozen dagger]

original paino stem:
Covered Church Organ: https://suno.com/s/L3wWElpy9451vp3J
Similar, but never really going to be a 1 to 1 copy. Adds music, Would need to further stem split and work as it is now. For now anyway.

[frozen dagger]

I've wondered the same about like and dislikes, I have seen Team member confirm they are not supposed to have any effect on the gens. Only what's in the input boxes/audio input.
Still, makes me wonder after 2 years of playing with Suno.

[buoyant bone]

I think it does because suno uses generated content for training.
its training on its own hallucinations LOL.
We are doomed.

[sweet ivy]

@buoyant bone If your are using a gen you like for the cover and setting the faders so high, your new gen will random? I would think that it would stay pretty close to the gen used in the cover, unless you changed a number of things in the style prompt ?

[buoyant bone]

I import soundtrack>
Use the prompt>
Set sliders to
0
100
100
And click generate>
4 generated tracks were nearly perfect but extended too far adding weird variations as well as messing with the tempo.
Every other generation were absolutely wild and garbage.
I cannot generate anything else even remotely close to the 4 tracks i got lucky with before.

I changed nothing. Just kept clicking "GENERATE" over and over until i get a good track

[dusky stone]

thanks for all you time and explanation

Any way to clean the voice from minor glitch, like 1/10 of a sec?

[sweet ivy]

Anyone use Melodyne after uploading a vocal stem into a DAW, to correct the pitch on a note that was wrong ?

[frozen dagger]

Remaster, Replacing, EQ-ing, or hide it with a riser sound. Terribly difficult to do small changes by itself. Remaster can fix and mess everything else up. Splice the fix into the original is another method, but it's all generating iterations.

[dusky stone]

all things I already did.... but as you know the glitch come out especially when voice is almost solo or during low instuments phase. Unfortunatelly remastering is not working for me...it chance the color/timbre of the voice

[dusky stone]

i tried as you said...put 0 W, 100 and 100...the line ganged and ther's a better piano but a strange noise below...🤦‍♂️

[sweet ivy]

Is Suno Studio any different than using your DAW?

[frozen dagger]

0 100 100 was response to other person who was using those setting. I don't recommend those settings for most cases. If you're REALLY trying to stick to original output and 15 85 85 isn't good enough then always experiment till you get closer to what you want. But usually those extreme ends of sliders cause problems.

[dusky stone]

ok ty

[naive quail]

If you have access to the Sample feature, try setting the whole track as the sample and running it with the same style & lyrics and the slider in the neighborhood of 80 to 85%. Any higher than that with cover or sample and I get weirdness.

[quiet solar]

"I wonder if previousl generations and likes/dislikes influence the results? If so, thats not good"

They don't. 🙂

[quiet solar]

1. The only way to keep the vocals exactly the same involves at least some level of editing, not just regenerating. Unless you're happy to do that, it's best to consider every iteration towards the 'final' one as a step in the journey, and not get too attached to the specifics of any particular arrangement by Suno.

2. In Studio - yes, if you split it to it's separate stem. Without Studio...yes, but you're going to have to regenerate everything else too, or do some serious editing work, so it's not ideal.

Exactly as you said, Suno isn't an ideal tool for minor tweaks and refinements, but that's the direction of travel with Studio.

[dusky stone]

thanks
for 1 what level of ediitin are you suggesting

for 2 sorry I'm in studio---you mean split the vocal stem?

[quiet solar]

1. That would depend on exactly what and where the glitches are. Assuming they're only in the vocal stem(s), then select the relevant section of just that stem in Studio and use Replace to re-do it. You can probably do this on a vocal/instrument only split, which makes life slightly easier.

2. You'll want to split all of the stems out, to hopefully isolate the piano in it's own separate stem. Then use Cover on that specific stem (rather than replace - check the dropdown option) and prompt for a grand piano.