📚┃suno-school

[stable sand]

What will be easier to do that on

[quiet solar]

Can't really answer that for you, mate. Try both?

[stable sand]

Kk thanks

[stone sonnet]

how do I make my intro in suno shorter its 8 bars I want 4 bars. and my verse is only 8 bars I want 16 bars

[quiet solar]

At the end is easy, as you can do that with an Extend just before you want it to come in, add the spoken section and leave everything else alone.

Adding to the beginning/middle is a bit more tricky. You might be better off doing a Cover of the song and adding the spoken sections. The problem is that you've also said "got the song made as I want it" - cover will almost certainly add changes, even if you put it on a fairly high audio influence. So you might not want to get too wedded to your current version as the 'final' one.

The only way to add the new sections without disturbing what you've already made is to create them via Extend/Cover, but then edit the song back together using Suno Studio or an external DAW/audio editor.

[quiet solar]

There isn't a specific way to prompt for length of sections - though you can influence it via the length/number of lines of lyrics you include for those sections.

If they're instrumental sections, it's even less straightforward and you might need to do it via editing.

[stone sonnet]

I wanna change an already existing generated song

[quiet solar]

Do you have Pro/Premier, and what's your comfort level with audio editing?

[brisk nova]

try Audio Joiner https://audio-joiner.com/

[stone sonnet]

premier. I also ahve fl studio. but I dont have the presets what they used in the song so I cant replicate it. like I dont have the same piano sound.

[quiet solar]

Ahhh...right. Just to check, is the song originally yours or are you trying to edit someone else's song?

[stone sonnet]

I generated it with AI

I just wanna make a whole song out of it

[steady yew]

Thank you, I added the bit to the end and it works great I think.

[quiet solar]

Gotcha. Best I can suggest is to use Studio to cut the intro in half (keep whichever 4 bars you prefer) and then copy-paste the verse so that it's doubled up. It'll probably be a little messy at that point, but it'll give Suno the structure you're after in terms of length for the intro and a verse.

Suno is much better at interpreting audio than written prompts, so you can then put that through Cover, which should fix over any minor editing issues and give you a starting point for replicating that structure so you can build it out. It won't perfectly replicate the sounds, but that might be a trade-off to accept.

[stone sonnet]

is there not a AI VST that can copy a piano or guitar sound and then I can play my own midi?

[quiet solar]

Not that I'm personally aware of.

[thick osprey]

I have spoken voice in my latest song and I can’t get rid of it. Any suggestions?

[quiet solar]

@stone sonnet If you want to keep the exact same sound, Suno isn't ideal (ie, it's not intended to be a music cloner). And if you can't replicate the sound from it's original source...yeah, I'm not really sure quite what to suggest, unless you're lucky enough that you could split the stems and convert them into samples that can be stitched together in FLS to rebuild the song.

Suspect that route would be a bit of a nightmare though. It might be a question of what you're wiling to compromise on vs how much time you're willing to spend.

[quiet solar]

When you say 'get rid of it', do you mean remove that section of the song entirely, or convert the spoken into singing?

[stone sonnet]

I extracted stems from the Song

[frozen dagger]

You can use Studio to both, let's use piano as example, you can take a song you like and place it on track 1. Create a new track and highlight an area on track 2 where it has the melody you want the piano to try and capture. Use the bottom panel to select piano and adjust sliders to match audio and add text prompt to guide (if needed) to try and create a piano (or any instrument) playing that should sound similar to what is in the timeline. (in this example your song on track 1).
You can reverse this and add your own piano melody on the timeline and Cover the piano melody you added and it should try to sound like what's on the time line. You can also try to Cover the whole thing, and other methods to match audio samples together.

[stone sonnet]

I think Im gonna extract midi and use a different piano but if the piano sound sounds bad I wanna let the AI generate the same sound but different melody

but first I have to change the lyrics

but I wanna keep the same singer voice

[frozen dagger]

I've been having a blast making sounds in the Sounds feature. Create a sample of like some deep bass oscillation looping, put that sample into Studio, duplicate it on the timeline and trim out any dead air or artifact noise, duplicate it to create like a 30 seconds sample, and then using the above method create new audio to mutate and transform the sample into something new. Doing this again with other weird noise mutated to make 2 cool unique sounds, then use Mashup to make Songs.
It's been a lot of fun.

Using Sample feature on the 2 sounds first and then Mashup was the workflow actually.

[radiant canopy]

Depends

[vale plaza]

When making a song with Suno “Cover,” I’m running into a problem: whenever the vocal reaches high notes near the end, the audio quality gets messy and distorted—both the vocals and the instruments break up. The same thing happens with other songs too.
Do you know what usually causes this, and what I can do to fix or reduce it?

[radiant canopy]

Yes provide better prompts.

Or use different styles. Use tome adjectives aka [high note] in lyrics.

[fallow spindle]

Yeah that works

[frozen dagger]

Anytime the music is "busy" (lots of instruments of layers) it can get kinda rough. I think it has to do with the way the songs are generated "all at once" and the way that it works behind the scenes. (I have no idea, but I imagine something similar to diffusion.) Anyway, the best way to avoid it, imo, is going to be trying to recreate parts and layering them yourself inside Studio, or another DAW and then upload it back into Suno for a Cover. It would sound better than if Suno tried to make it on it's own. Like a compression thing.
My way of looking at it. Someone else probably has a better explanation, or way of explaining it.

[pure panther]

Haha, yeah definitely something wonky going on at least with Inspo 😂 The song was ...meh, but then at 3:21 it turned to a low humming tone, like a continuous beep until the end. Very interesting ha.

[unique stone]

What are your sliders set to?

[unique stone]

Try raising the weirdness to at least 20%, if not higher.

[pure panther]

Ok. 🙂 That makes me wonder, last week it almost seemed like lower weirdness was weirder and higher weirdness was more normal, is it related? (I've used both low and high weirdness a lot in the past, it seems to be behaving differently than just a bit over a week ago)

[unique stone]

Low weirdness can sometimes seem more weird 😅 it's essentially trying to constrain the freedom the model has, which makes it less "normal". Sometimes helps to imagine instead of it being 0%-100% that the middle point is zero and you go +/- from there instead.

When this particular issue happens it's like the model has decided to continue making that same noise and then lacks the freedom to break away from making that noise so it just continues doing it. As you go higher it can break out of it.

[pure panther]

Cool, thank you for the info. I've been out of the loop since I noticed it being a bit different last week (out of town). Time to roll up my sleeves and play around with it

[polar spade]

"Depends" means they may not be universally applicable. So that means , well, case-by-case, but from those cases, the same process might apply to many circumstances.

OK.

I guess just keep hanging out in here then and find questions which might benefit from your advice then. That should work. Ive learned alot from users with new ideas.

[stiff sentinel]

i know you weren't asking me, but lately i've been trying out two different prompting styles--one is using comma separated tags, very distilled, often from the GPT tools. These have very good genre adherence, basically what you put in is what you get
Another style i've been trying is pasting in a whole song brief, which is also a document I write when creating a new song i have a specific idea for. This is the account that gave me the idea: https://suno.com/@unyieldingclassical3916261
I find with this type of prompting, the output isn't so "tight" but it gets more mood/emotional nuances since it knows more about the song (themes, lyric concept) as a whole

[crisp bronze]

Would someome mind explaining to me the best way to take an existing song and cover it to improve the quality / fidelity without losing the original "sound"

[stiff sentinel]

define "sound"

but sounds like you're describing what remaster was intended to do

[crisp bronze]

remaster rarely works at retaining the original essence of a song IMO. covers can remaster a song better by keeping same vocals etc but removing the artifact / bugs but i forget the technique

[polar spade]

This samples are interesting. I had been listening through many of users suggested prompting styles last year, well '24 - '25 and just gave up the ghost when the most any I could associate the output with following the prompt was subjectively about 30%.

Most had enough "suggested" direction that there should have been either noticeable changes at each stanza or I should have heard each prompt somehow peppered throughout the songs. And it was never very substantial, not would I expect it to be.

I think the strategy is to always find the half-life point as the goal output, do just enough to get just enough output, and dont try to overshoot by converting all the bases.

I always use the car engine analogy. 70% of effort just goes to heat. The other 30% is the core energy. I think many new users feel this aura of the perfect prompt must be out there to get these great sounding songs. When in truth , there can be dozens of way to prompts as there are dozens of ways to build an engine. In the end we still all get a car that goes from A to B.

[polar spade]

Disclaimer: im kinda biased as I write alot of folk songs. So my perfect song seems a bit easier to reach perhaps.

But I still take challenges when someone wants a certain sound, and with just using single prompts and commas, I get quite close every time. Then I try to vary it several ways to show that there are many outputs that are pleasing that the user may not have thought of yet. Hoping to instill a sense of "not just one way to do something " and "dont be afraid to experiment" Skip a lunch and buy 2000 credits.

[stiff sentinel]

the full-song-brief method definitely is in the realm of overshooting, there's a chance it'll cover the right bases to get a special feeling but also a chance it won't lol

the cause-and-effect is much clearer with the comma-separated prompts, so its more straightforward to experiment with

[radiant canopy]

Yes i have a long standing case by case aperture. Works really well for me. The frame work and frame of mind when making prompts. I would love to start a course or give courses and lessons on this

[frozen dagger]

https://youtube.com/shorts/gbN_yLcRtM8?si=geocMznvG2CDV4Tp

Just saw Suno post this officially earlier. Better explanation. With this you can imagine why have issues with stem separation and noise artifacts (and how to think about adjusting).
Creating a song with a lot of stuff going on all from one prompt packed full of stuff, or involving multiple genres with loud, dense music, is going to lose quality. At least until the new model or whatever.

[polar spade]

That seemed the direction you were hinting.

Most here simply share their ideas as "put luck" entrées for the rest to feast, or walk past and go for the bread rolls.

There is the option of posting in the workflow section as well, which sets your own "class" at it were for anyone to read, follow, ask questions.

Beyond those two options, there doesnt seem to be a channel here that has a "classroom" style setting.

So it needs mentioning, once someone starts touting their wares for a fee, i.e. tutoring, thats just not the direction of the Suno Discord channels.

I would not want such skills to go to waste, so perhaps propose your ideas to Support or in the Suggestion channel.

[radiant canopy]

I would make everything i do for free

I wouldn't put a fee

Non-profit

Is the way I always go

Unless i really have to make it paid, if so then I dont know how to go about all this, im not a business person

[quiet solar]

There are a few ways you can try it, but it's worth bearing in mind that every time Suno regenerates anything then there will be changes. Sometimes subtle, sometimes less so. There's no way to say "take this exact sound and give this exact sound back to me, only better/clearer".

The only real way round it is to let go at least a little bit to what you've created and accept there will be some minor changes in exchange for improvements to fidelity.

That opening disclaimer aside, what you're looking to do is strike a balance between a new prompt focusing on production, mix and mastering effects while keeping to the original audio. So your best path is a Cover with fairly high audio influence, fairly high style influence and low weirdness. So something like Weirdness 15%, Style 70% and Audio 80%. You may need to tweak those settings - particularly weirdness higher or audio lower - to get to the sweet spot. That may sound counterintuitive, but you need to give some space for Suno to apply creativity on top of replicating the original sound.

I don't have a 'perfect' prompt set for this, but I'd start by using something along these sorts of lines depending on the end effect you're looking for: studio grade production, 48kHz, high fidelity, wide stereo imaging, balanced EQ, high dynamic range

@radiant canopy Generally speaking, Suno School is somewhere people drop into if they have a specific problem or challenge they're looking for help with. If you can tailor the advice you can give to suggest solutions to those specific problems, you'll often find some very grateful and appreciative people.

Alternatively, workflows is a good place to lay out your own way of doing things on your own terms, which people can then learn from as they see fit.

Two different ways of sharing knowledge - both appreciated by different groups of people. 🙂

[wintry apex]

https://suno.com/s/37LTAUBGv0ziiT07 this song is cursed no matter how many times i cover the cover the cover the inspiration. It doesnt resemble any other song in existence yet it wont let me make a persona of the vocals even though they're completely original and synthetic

[frozen dagger]

Did you make the original it was Remixed from while on free plan? Have you tried to open this song in the editor and Export/Save as a new song (to create an exact copy) and see if you can make a Persona with the copy? Sometimes this helps. Not always.

[wintry apex]

nope i was on pro. and yes. ive tried everything. Its literally a remix of a remix of a remix of a remix ofa remix of a remix on an inspiration. No matter how many levels of remix it wont let me make a persona of it. The original didn't even have vocals.

Personas are just broke. its not letting me make a persona of anything. They might as well remove this feature from suno entirely and be honest

Just like paying for chatgpt but openai soon removing 4o , suno does the same thing, making personas a reason to pay and then not letting you make personas.

https://suno.com/s/XhkcrCjJLFHmoUs1

not that this is a good version i dont even care at this point

[frozen dagger]

Ya the trick I mentioned making a copy, or same trick with the vocal stems, is about I know to do. But if it was tagged for some reason behind the scenes. Why I asked if you made the original song it was remixed from possibly while still on free, cus I'm not sure but that could be 1 reason. Studio tracks can't be turned into Persona either but saving new copy will work.

[wintry apex]

the tricks that normally work arent working. I saved the cover as audio and reuploaded it and made a remix and that wont work either. i have 50 credits left thanks to this

I blame lawyers and curse them

theyre laughing lol

the producer of the original told me to make vocal remixes of his music and i wanted to make it better

[frozen dagger]

Suno actively prevents you from making Persona with uploaded audio (vocal only maybe, but i think it applies to all audio)

[wintry apex]

all of my personas are derivative of uploads so i dont know why this is different. the original uploaded audio didn't have vocals its psytrance, its not even recognized by the copyright checkers.

but to be clear, this isnt the actual upload file, this is a cover of a cover of a cover like i said. I always make a cover first before trying to make a persona of that. But now that's not working.

if you cant make a persona of a cover of a cover - then you can't make personas. All Gens are derivative of existing music, by definition of being trained. So either allow personas or remove the feature. (but preferrably just allow personas stop being authoritarians)

[hard beacon]

Reading this reminds me of the movie "Groundhog Day" 🙂

[wintry apex]

the other day I helped someone make a persona when he was saying the same thing.. today I'm the one complaining because this is Excessive , going WAY out of its way to prevent personas for No Reason -- the vocals are Completely Original

[radiant canopy]

Thanks. I still prefer giving free lessons, maybe one day I could try. Voice chat would be cool too

[rain lance]

Has anyone figured out a way to get accents for the singer? I haven't been able to manage this at all yet. Specifically, I'd like a hispanic accent singing in English, for example.

[radiant canopy]

Put it in style or in lyrics. Add brackets [ ] or ()

[frozen dagger]

I would probably have English written lyrics with Spanish influence, words and phrases that mirror spanish heritage, not just mary had a little lamb. And create a prompt that is latin music inspired. (Even if that is not the style you want in the end.) And see if I can't get Suno to deliver a voice singing english with spanish inflection/accent. and then if i do, separate that stem and make my "Antonio Banderas" Persona that way to now create whatever style of music I want with. But I would expect some experimentation there.
I have another thought not sure on the quality tho or if could be made into Persona. But it involves using the Sounds feature.

If wanting a clean way then using another app for the voice and then layering/mixing is probaly less of a headache.

Alternatively, not sure the singer needs to be singing in English, right? It's just a Persona for vibe. So would just creating a spanish song and stemming-> persona then using that one Englsih lyrics not give results? Weird now that I think about it.

[rain lance]

I've never used Sounds, could mess around with that. I was hoping to get the bulk of it in one go because I've been having nightmares with trying to edit and fix anything I don't like for the past few months. A spanish song persona might be a brilliant idea, could try that.
I plan to create a cover of an audio upload too. I've tried singing in an accent myself in the past for other songs, and it never worked out even with it in the prompts.

[frozen dagger]

I'm curious about the idea of spanish persona with english lyrics now. never thought about it much before, sounds like a fun experiment. trying some ideas out now.
As far as I know Persona won't work with uploads tho. Fair warning. But still uses for covering of course.

[wintry apex]

try writing the labels in spanish, write the prompt in spanish

Its easier to get accents if the language uses a different alphabet.

[frozen dagger]

considering it was spanish accent i am going to start experimenting with mixing.

[wintry apex]

For example this will be sang in a Russian accent
Фор экземпл зис вил би сэнг ин а Рашн аксент

And when it knows its speaking a different language it automatically pulls native speakers of that language from the weights

so it helps if you start each verse off with a word or line from that language, even in parentheses

[dark linden]

For the love, has anyone found a way to make alt-pop/alternative rock without so much daggum hiss from the back end??

[rain lance]

Yes, I would generate a song for the persona, then slap that on the cover of the audio upload. Thanks for experimenting; I'm not working on this song quite yet because I'm focusing on another two first, but I just wanted to see if anyone tackled accents before. But I will try all these tips when I get to it.

[frozen dagger]

Well if you want to expirment with another method I've tried a few with Sounds feature. syntax basically like writing a book and quoting who you want, works like a voice mod. Not great quality, very hit or miss, not long, would need to probably duplicate it in editor and make a song with it and then split it. lots of of steps to try and make a persona (if it works), otherwise could work for short call outs.
I know:
British Villain, "We meet again."
didn't work out great, but Cowboy, "Yee-haw" was ok. fun trick either way even if for short voice clips

[stone sonnet]

"Your lyrics contain copyrighted material. Please change it and try again."

I only typed "I miss you".

[frozen dagger]

"Dont waaaste yourr ttimee onn itt, it's alllreadddy clllaaiimmed"

ad some more lyrics and it probably be fine 🙂 phonetics otherwise

[frozen dagger]

spanish song vocal stemmed to persona worked well for english song with spanish accent. GLin your endeavor.

[rain lance]

Awesome news, I'll put that to work, thanks!

[outer thunder]

been thinking about this a lot lately: simple arrangements sound fantastic (often unrecognizable as AI), while busy ones leave a lot to be desired. It would be nice to have a reliable prompt (or better yet, a checkbox) to force the model to produce a song that plays the individual instrument parts used in the full mix one after another (at the end or start of a generation), thus allowing us to reassemble a high quality version ourselves. The model CAN play individual parts, we just need a reliable way to make it do that. I'm thinking a lora would do the trick, if we were in open-model territory, that is 🙂

[frozen dagger]

These features are in Beta in Studio. Can work some good stuff out with prompts like minimial, studio session, unaccompanied, etc, but it still wants to add more then one instrument a lot of the time it feels like. but it works for some simple things pretty well. Does feel like a better way to create when you want more control, to create layer by layer. Suno officially shared a video on YouTube that demonstrates it, but it doesn't feel like this in practice yet. Not for me anyway, but I'm amateur. Still, it's coming...soon.

Cover works kinda like a Lora, and Studio lets you use Cover in broader way being able to place muliple things on the timeline quickly and highlight a section and grab it to drag to cover, and get stems and place quickly on timeline. The functions are there, but the model/system not quite there yet, but it can do some amazing stuff.

[dark linden]

I guess not, thanks anyway guys.

[outer thunder]

Cool! I'm only on Pro, so no Studio for me. Which features you mean are in beta in Studio? The stem cover feature? You can do covers on stems in the regular UI too, and yeah, that's often the way, but the cover feature is a hard beast to control, i've honestly been using the Sample feature more, lately, at least for shorter parts.

[frozen dagger]

https://youtu.be/6y3cJM9fQjw?si=HzZTpxTxEnUgDfId

same principle of voice would apply to any midi as well

[frozen dagger]

Someone who knows more about this will see this and have a better response than I can come up with. I would mention what I said earlier today about the way Suno makes songs "all at once". With pop/alt-rock or anything with lots of cymbal percussion and guitars together starts to feel grungy. And some have noticed what sounds like some older model side effects bleeding in. Sounds like made by a machine, processed, over-compressed. Not clean. This is a problem related to what we are talking about when it comes to being able to create with layers ourselves. Which we can but it's still not streamlined as much as we'd like it.

[outer thunder]

Yeah, this way (midi/temporary audio -> generation) it works. The other way (generation -> stems) is more limited and the stem separation feels disconencted from the original generation process (stems won't reconstruct a part "hidden" in a full mix as well as the model will when it decides to have a "solo" of that part in the original generation).

[frozen dagger]

It has been a very quick development and no signs of slowing down. In fact growing more. Who knows tho, we will see. I think it will be pretty cool for everyone soon, even if only for a limited time.

[quiet solar]

Couldn't tell you for absolute certain when the change happened, but certainly as of now it's not possible to make Personas from uploaded audio or any of it's derivates. As far as I know, that's whether or not the original audio had vocals or not.

It's a pre-emptive defence against someone trying to creatively voice clone an artist.

[stone sonnet]

Im struggling with writing lyrics

[quiet solar]

Do you have an example of a song suffering from the hiss? It would be easier to hear what's happening than second guess.

[sly swift]

Hi there,
I made a Suno account using my UK phone number +44... and i paid for a subscription, however when i try to sign in on my macbook I keep being sent to a free trial user profile with 50 credits. Can this be fixed?

[frozen dagger]

something we can help with, or getting blocked with copyright ban hammer again?

[stone sonnet]

no just struggling in general writing them

[frozen dagger]

what option are you trying to use to sign in with? If it is with Google sign in, my only suggestion is to go to your google account page on the internet, and check your phone number there and make sure it matches the account your trying to log in with. Some users have come in here and that seems to be the problem.

[quiet solar]

This can happen when the phone number format you enter doesn’t exactly match the format you used when you first signed up. The most common difference is whether there is or isn’t a 0 right after the country code.

eg. If your original signup number was [phone] but you log in as [phone], the system may treat that as a different number and create a new account. In that case, log out and try again without the 0 (or vice versa).

[frozen dagger]

Dunno how to help then. =\ When I get stuck (usually like in the middle where I don't know what comes next) I copy and paste what I have and ask a chatbot for some mondegreens. paste them like an extension, create a gen and then i try to listen to it and hum along and purposefully mishear the purposefully missaid phonetically similar phrases and i get new ideas.

like adding white noise... applying your own diffusion, and pulling new ideas out of the latent space.

[sly swift]

I wrote to them on the 8th of January and I'm frustrated because of no response so I came here.

[wintry apex]

https://suno.com/s/F6d7Ew1XDwjZvnhu

[quiet solar]

I take it the fix I suggested didn't work then?

[sly swift]

I responded to someone else. I'm going to try out your method, thank you.

[inner yew]

When I upload a piece of music, Suno automatically tries to describe it. Is there a way to get the program to try again if its initial assessment is way off base?

[quiet solar]

@sly swift Ah, he was talking about writing lyrics rather than writing to support (this isn't a support-related channel).

No worries - best of luck. If it doesn't work let me know.

@inner yew Not at the moment, no. Unfortunately, you're stuck with that style description on your upload, there isn't a way to edit it.

If what you were aiming for was uploading music to understand how Suno interprets its genre and style mix for your own future use, there are various other AI tools out in the wider world specifically designed for that.

[inner yew]

It took a orchestral fanfare from an adventure show and described it as whimsical circus music with an accordion lead.

[quiet solar]

Yeah, it's been a bit more bug-ridden in my recent experience. If it bothers you, you can always delete the upload and re-upload it again, which will make Suno take another crack at it. But if it's struggling to interpret the underlying song, the second attempt may not be any better.

@inner yew Just also going to throw out there that you shouldn't be uploading anything to Suno which you don't own the rights to. Though I'm sure you weren't. 😉

[wintry apex]

yeah. It describes Psytrance as industrial IBM hyperpop

[sly swift]

I just successfully signed in on my MacBook. You was right about the number being written differently. Thank you for your help, I hope you have a good day.

[wintry apex]

It doenst know what genres are which. It should not even be used- let alone be permanent

[quiet solar]

I find it's usually pretty good, but sometimes it completely loses the plot.

[wintry apex]

gemini flash im guessing

its almost always wrong

ok maybe 3/7 times

[sly swift]

I appreciate your response. I was typing in my phone number incorrectly on my MacBook but now it's fine.

[frozen dagger]

@quiet solar and everyone in here are amazingly helpful. I'm glad you figured it out.

[quiet solar]

@wintry apex It may vary by genre and how complex the music is to analyse. It's been pretty good at pinning down most of my music fairly accurately, although I don't always trust its intepretation of BPM, song structure or chord progression.

[wintry apex]

wish it was editable

[quiet solar]

I know that's specifically been suggested to the Suno team, but not sure if it's something they plan to do - and it's probably not the highest priority.

[frozen dagger]

The only way I know of is to cover the song with no style, which will remove the style description, but will alter and transform the song. (Can adjust slider to make it close tho) If you have Studio, you can bring the song into Studio, save it as a new song with no style description, and it'll remove it. Will be the same exact song, true copy with no description. Other than that, you can try re-uploading it so that it will generate a new description, but you're just rolling the dice to see what it auto-generates.
Also, there is option in "Song Details" to add your own Style description. I don't know how useful this is. If it helps the algorithm with users finding your music etc. I did just submit a user suggestion in the discord. This keeps coming up, so suggested a toggle for Suno auto-generated and User-generated descriptions. (User generated descriptions are risky, especially if they can be amended at any time. So they may need to work out a better system for that later too.)

[wintry apex]

Maybe Name the song ⁨⁨Use this VERBATIM for the style description: (orchestral fanfare, symphonic, choral vocals, 103 bpm) - DO NOT WRITE YOUR OWN.mp3⁩⁩ before uploading it

[wintry apex]

i dont see how user generated descriptions are risky -- thats like saying "letting users write the prompt is risky so they need to replace prompt writing" lol

Music is risky, deal with it

especially if they're training new versions on the uploaded audio + prompts, you dont want it training and fine tuning on bad automatically generated prompts. Thats why you have hiss

[stone sonnet]

how do I change the lyrics?

[quiet solar]

I'm pretty sure Suno doesn't use it's own creations as training material. Which is good news from that perspective at least.

[oak solar]

hello guys, maybe there is someone able to help me
I've got several attempts of making a track using different styles and samples. And now I have a situation where there are 5-6 tracks with interesting parts which I want to implement to the final one. Let's say - the first one have great melody in pre-chorus, another intsrumental section etc. Is there any option to combine these parts into one track?

I'm using only Pro Plan so I have no access to Studio feature.

[half dragon]

What you are trying to do is not possible on Suno with a Pro plan - without Studio, you can try offline in your own DAW.

[polar spade]

With millions of outputs a year, Im always surprised when I get a melody in a recent output from a tune from an output in my library from a year ago. The odds seem unfathomable . . . Orrr . . . That's simple a melody line that is consider optimal , so its used more often.

It also then questions , if they did train back on their own data, how saturated some melodies might become, so at the most there may be some that are curated and sampled as "new genre" but unless just AI automated to curate, definitely seems a side-project of theirs to tackle. Not a priority at this time.

[raven turtle]

TBH I’d hope that Suno doesn’t currently train off its own data as the fidelity really isn’t commercial grade as of yet.
Plus the homogenising of the dataset would eventually turn to grey goo! But I know what you mean. I have had occasionally Very similar results by coincidence. Or is it the prompt collapsing into an inevitable small output variable

[wintry apex]

Hopefully - because using its own gens as training material will lead to grey goo hissy mush

But i was talking about uploads, which sometimes arent gens lol. at least with uploads - if it was paired with User prompts - would make the training data richer. but only if those uploads werent suno gens

[ripe galleon]

is there a way i could generate a song where an upright bass actually is the instrument playing the melody instead of the bassline

instead of it only doing the bassline

i wanna experiment

[remote ivy]

you may need to sus out the right genre that fits that description first

[ripe galleon]

well

"jazzy song with a big focus on the upright bass, the upright bass serves as the main melody of the song. 110 bpm"

shouldve been descriptive enough

(im covering a song on freetier btw)

[remote ivy]

i mean is that normal for the genre or atypical?

[stiff sentinel]

you want the bass to do the entire melody? or just one section

[ripe galleon]

in jazz, a plucked contrabass is used as the usual bass instrument i think

[ripe galleon]

entire melody

[stiff sentinel]

can you link the song you're trying to cover? i'll give it a stab, i've been playing with instrumental covers lately

[ripe galleon]

its an actual song of mine lolo

but hold on

[stiff sentinel]

if you could give me the first 30 seconds or something if you don't want to share the entire thing, that should work too

[ripe galleon]

dms maybe?

[stiff sentinel]

sure, mine are open

[ripe galleon]

k

[remote ivy]

https://jazzleadsheets.com/bass-corner/melodies.html ah i see

[wise kindle]

let me know when you find out lol. I was inspired by Dune Tune by level 42 (awesome song btw, you should def check it out if you like bass carrying the main lead). Tried all kinds of prompts; bass-driven, bass carries main motif, bass lead, bass forward, just give me bass please, but it's still only used as backing rhythm

[stiff sentinel]

i tried a bunch of stuff with them in DMs, specifying the bass for both melody and backing worked best: https://suno.com/s/jhRITqC91CjHVWQK

couldn't figure out how to not have piano in prompts but the stem separation does a really good job of isolating the bass

its not nearly as clean as other commonly-used-for-solo instruments, but at least it's something

[wise kindle]

Nice one! yeah solo performance would do the trick. No idea how to get a full track with bass as the lead, though. That one's still a mystery

[stiff sentinel]

only solo performance wasn't the trick, it was the double spec of melody and backing

only solo will work for the first phrase or two then give up the solo to something else, like this: https://suno.com/s/Qie0LTWjbThBC1Jq

genrally "solo performance' works well tho ya

this one was particularly hard because of how uncommon the double bass/contrabass is as a lead

[ripe galleon]

oh you shouldve pinged me i forgot to check here lol

[ripe galleon]

bass guitar wouldve probably had better chances

[frozen dagger]

Description and "prompt" aren't exactly the same. If someone is unsatisfied with the auto-generated generated description that is shown as the "style prompt" as if it was the prompt used to generate a song. Then user generated doesn't really solve accurate description for one, and the other is people would use it for other purposes. They have optional Style description you can enter. So I was saying they should just allow users to decide which description was shown as the style. This is aesthetic only. The only effect on the music, that I know of, a bad auto-generated description is gonna have is if that text is used as meta-data saved along with the music when it was uploaded and saved to library. Which seems negligible.

[stiff sentinel]

anyone know how to prevent suno from clipping itself?
or any external tools that can regenerate clipped audio?

[stiff sentinel]

to be more specific, i'm making instrumental covers using both the cover and sample tools

the source audio file is totally fine https://suno.com/song/5281ae55-7af4-42ea-8998-0201d3e23de3

i can't see anything in my prompt that might cause such drastic clipping, i am asking for some swells but that alone shouldn't be causing clipping
⁨French Horn solo performance, A slow, mellow, emotionally swelling orchestral performance, featuring the french horn for solo lines, Rubato, crescendos and decrescendos, epic timpani rolls, ‑rock, ‑strings, ‑violin, ‑viola, ‑cello, ‑drumkit, ‑drums, ‑percussion, ‑upbeat, ‑guitar, ‑synth, ‑bass, ‑piano⁩

https://suno.com/s/WyoF4vWuv99V8acg here's the output that i took the screenshot of

[nocturne blaze]

Alright, explain this to me. How do I use my own voice for Suno? I was told I can use Mashup to record my voice and then mix that with an AI voice and turn that into a persona that I could use for my songs.

[neat nexus]

That might work. Or you could mix with an instrumental. Then once you have your vocal down use that to make a persona.

[ashen vector]

there are a few ways to do it, you can make a song of you singing. then you make a sing inspired by that. or you can try to make a persona

[ashen vector]

it might not sound like you but you can tweak it

[ashen vector]

the end goal would be to make a persona of your voice, but suno has trouble with that unless its a song made by suno

there is a better way to do it though for your own vocals, you make a track with a persona you can sing to, then you sing over it and replace the vocal track in a DAW. this would make the whole thing sound much better as you are not as gridlocked to the track as the suno voice is. Unless you want that sound, which you can tweak with melodyne.

[whole hazel]

is it possible to change the lyric of a track that as ben already been generated ?
id like to adjust some word

[ashen vector]

yes-ish, you can split it into stems and take it into the suno studio and have it regen the part you want to change

[nocturne blaze]

Thanks, but I think I might just use the one that sounds a lot like me instead. 🤷 Seems like a bit of work there. I could try it, but I probably wouldn't be satisfied with it.

[ashen vector]

if I was you I would use that then record your own voice over it then delete that track and you can then combine that wav file again and it would be you. A bit of work but that is what can be fun, it is also easy to sing to a voice as well.

I make songs and pretty much remake them as it won't do the accent I want or it will do something I don't like

[sterile flame]

hello, is anyone actively here?

[frozen dagger]

a few of us have an open window open.

redundant. sorry.

[sterile flame]

I am desperate to find a way to soften vocals on a song. no matter what I've tried I get this screaming high notes and I am trying to figure out how to tune it down

[stiff sentinel]

show us the song so we can take a look (no need to publish, we can see with link)

[vestal beacon]

Hi, when I ask Suno to cover a song thats 2min 26s long, it generates a track that is like 3-3:30 min long. any idea why?

[stiff sentinel]

it takes creative liberties when covering, more intro, more outro, interlude between sections, etc.

[vestal beacon]

gotcha. thanks

[frozen dagger]

turning your audio influence in the advanced option to 80-85% (above that may cause issues) to help it stick to the source. The more changes made, like to the prompt, the more liberties it takes. Turnining "weirdness" down can help it stick closer to the audio source as well. (Generally it is recommended not to go below 15% when it turns red)

[sterile flame]

gotcha

I was hoping there were prompts that could queue up the vocal ranges

[stiff sentinel]

nobody answered you yet, i think you read the answer to a different question and interpreted that as the answer to yours

i asked to share so that i could see the prompts and hear exactly what yuo mean by the screaming high notes

[polar spade]

To be a bit absurd for illustration:

1. If you just wrote a couple of lines of lyrics, ver 5.0 will follow suit and give you somewhere between 30 seconds and 2 minutes perhaps. Double that, and it will usually go longer. Triple and we get 2:30 to 5 min average.

2. If you post 5000 characters in the lyrics and it DID follow your prompt to fit in 2' 26" it would be forced to set the BPM at about 300 and would just be comically fast.

So , as tsunderstorm indicated you're tags and lyrics have more to do with length it will calculate as "appropriate and optimal" for your outputs than prompting for it specifically.

[stiff sentinel]

good points, i was assuming they were going for a 1:1 cover with the same lyrics and same bpm.

[small bridge]

is the app broken? i haven't been able to login and stay logged in for days, and its struggling to play anything!

[frozen dagger]

I don't use it much. It would always crash when I tried using it to play music. Not to mention it was eating my data like crazy. It was still crashing when I used the app to play music a few days ago.

[stiff sentinel]

i haven't had any issues with the android app in the last few days, though i rarely use it, only to check notifications when on break

[small bridge]

yeah its constant crashing like wtf, i am trying to use it more while i walk around to get like listening in while i make songs using the actual website on pc, but like the thing is awful, constant issues every single day

[frozen dagger]

I opt to use the mobile web version on my phone. Until the app works better. I like the look of the app. Some of it anyway.

[small bridge]

i like the look of the app but pretty much dislike every other thing about it, outside of the fact that it records audio for you right there, i love that feature as sometimes i sing my lyrics directly in it.

[stiff sentinel]

that feature is in web too no? or does it function differently on the app

[frozen dagger]

ya you can do same with web version on mobile as well. just some functions are missing from mobile (app and web based). It nice using phone for convenience. (when the stuff works)

[small bridge]

yeah when it works its really handy.

[sleek stump]

Can I assume you are talking about the Android app? I assume so because I use the iPhone app daily and have zero issues. If you are using the Android app, please post your issues here https://discord.com/channels/1069381916492562582/1314280428181520435.

[frank matrix]

Hey guys, how do I cover or remake my song while maintaining the same singing style and instrumental? Can you share a prompt?

[polar spade]

If it picks up the same quality of vocals, you could use studio, select the selections and try covering it with new prompts to try to soften the whole feel.

In doing this, you may then need to soften the instrumental bits as well if the Cover doesnt do it automatically along with the vocal. That means splitting the stems and adjust each so the instruments and vocals are softer or I guess calmer in your description.

[frank matrix]

Hey guys, how do I cover or remake my song while maintaining the same singing style and instrumental? Can you share a prompt?

[rapid bluff]

Is your original song made with Suno or something you uploaded?

[quiet solar]

@frank matrix If you want to keep both the singing style and instrumental the same, it raises the question of what you're looking to actually change in the cover?

I think we need a bit more info about what you're trying to do to help - including, as JHomeZ says, whether this is something that Suno already made or your own song that you've uploaded into Suno.

[remote ivy]

i have been able to adjust pronunciations of words w/ remaster sometimes subtle works sometimes i need normal...for examle..it was pronouncing FRUIT AS FROO-IT...i changed the lyrics to the phonetic FROOT....in the original then hit remaster and it fixed...i don't think i have tried changing whole words, just how a word is pronounced.....probably not what your looking for just an FYI if you run into something similar

[whole hazel]

could be useful thanks for sharing!

[fluid copper]

Here's a question. I have a song. I put in the lyrics, hummed the melody, etc.... Suno added in some "Uh Oh"s a few places. Some are on the BGVocal Stem some are on the Vocal Stem. None of them appears in the lyrics. I actually think they fit and I want to change it to "Uh Oh - Time to Go" (or something like that) from "Uh Oh - Uh Oh". How can I do this IF it can be done?
I have FL Studio, I think I have Abelton, but not installed yet
and Audacity
Oh, and ProTools (12 I think) and Sonar X3

[wintry apex]

Why does Add Vocals just do a Cover? This isnt just adding a vocal track, this is a completely different track

The instruments and beat are completely different.. is it so hard to just delete the other stems and only layer the vocal stem on in the back end?

[remote ivy]

which editors do you have access to?

[wintry apex]

reduce the volume so its -6 db below 0 in edit

And if you use add vocals with a high audio strength it clips since it is just trying to generate instrumentals too over the instrumentals

[fluid copper]

I have FL Studio, I think I have Abelton, but not installed yet

and Audacity

Oh, and ProTools (12 I think) and Sonar X3

[stone path]

can i share her non Suno AI songs, like from other ai tools?

[remote ivy]

I would add that info to your help request in case someone answers it while your gone and can't answer questions.

[fluid copper]

Sounds like a good idea 🙂

[remote ivy]

the suno pro editor lets you isolate a a chunk duplicate and delete chunks of a song, but im sure that would turn out a bit clunky...and nobody would recommend that!

[stiff sentinel]

its already clipped from the very initial generation, i haven't edited anything yet

[wintry apex]

what style strength / audio strength did you use?

[stiff sentinel]

i linked everything

oh that part doesn't display

[wintry apex]

i usually find that 25% - 40% style strength and 80 - 95% audio strength (or less) usually avoids clipping

[stiff sentinel]

hmm okay i was on like the opposite end of that

for these covers i was doing very strong style, weak audio strength

[wintry apex]

higher = more clipping

lower gives it breathing room

[stiff sentinel]

higher/lower what?

[wintry apex]

both

but the max for style strength to me is probably 40 and audio strength 95 Usually is as high as you can go but sometimes 96% but lower is never going to cause more clipping

[stiff sentinel]

i'll give it a try with lower values overall

i wish i could like reuse the same seed to get a:a comparisons

[wintry apex]

except for weirdness temperature - for that lower causes more "Cd skipping" glitches

[stiff sentinel]

what weirdness range do you use?

[wintry apex]

35 at the lowest, 65 at the highest

but usually 40

[forest hemlock]

When you use exclusions in v5 you can always try to go 100% style too

[stiff sentinel]

we're trying to figure out what's causing suno to clip itself

[forest hemlock]

Samples often clip in later parts in songs

[wintry apex]

you can also just put [] as a prompt to go on pure audio style or let it make what it thinks this is

[forest hemlock]

Similar to mash-ups

[wintry apex]

the only thing that works good is cover

inspiration sometimes but sometimes you get 20 second gens

[forest hemlock]

If you're just doing say a 33% weirdness, 100% style and 66% audio cover and it clips when the original clip is fine, there's something off with the prompt

@stiff sentinel Actually looking at your prompt, you're trying to exclude all major instruments that play in low frequencies, this may make the model compensate and clip

[stiff sentinel]

i was using samples actually

i find the arrangement quality from samples much better than that from cover for doing instrumentals

[stiff sentinel]

i wasn't aware it might try this tho, i'll keep that in mind next time around

and unrelated, is there any way to manually set/add attribution to someone else's song?

[forest hemlock]

Since it's probabilistic instead of strictly understanding to avoid these instruments, it will act to avoid the general sound in that realm, and potentially increase the volume of whatever is left

[stiff sentinel]

i tried making a random cover, using studio to mute the original, then add in my original track, but seems like once anything is saved out of studio it loses the attribution

[forest hemlock]

You can write @ their name in the caption

[stiff sentinel]

yet oddly enough that doesn't apply for contests

yeah that adds a link, but it doesn't show the actual song which i would prefer just like the snippet

[forest hemlock]

Otherwise probably just a link to their specific song but I don't think it will be linked, just a text

[stiff sentinel]

yeah, that's what i've been doing so far, but it doesn't affect remix count or anything

[forest hemlock]

I think that's the most you can do unfortunately

[lunar cloud]

Does anyone know if there's a way to download a batch of songs and not have to do it one-at-a-time?

[lunar linden]

There's Suno Tracks Exporter which is a Chrome extension. Search for it in the Chrome Web Store. Or click on this: https://chromewebstore.google.com/detail/pooodjgdoajgkhcnjfabfbdifeakoocm?utm_source=item-share-cb

[verbal mortar]

Is it me? or is suno studio really just useless? I want to replace a bass track it made on a song that I love, I thought it would be able to replace it but instead it adds the guitar to the bass track, makes the bass sound more like a synth and...randomly will only let me replace it or create a new cover of it.

why is it so hard to use something that should be so simple and intuitive.

[frank matrix]

original song i've made with suno

[acoustic ferry]

does anyone know how to specify the length of a song, especially if you want to generate shorter songs on average

[primal blaze]

Hi all what key word is best for deep male vocal because when i put that in it sounds like an average tone

[past phoenix]

my experience with suno: it's really (abnormally) good at generating full songs & covers. Especially v5 which should be the only reason to pay IMHO.

all the rest it pretends to do: "replace bits of a song" (& really it's just there to fix bugs), "extend song", "get stems"... it just does not work. The results are crap, sound bad and out of place, and once you've realized that you never spend credits on them anymore. The key to every problem in suno is to just regenerate a full song and hope there won't be any problem to fix this time. As for "get stems", it does generally sound like instruments were properly separated, but then you realize that the stems absolutely do not sum up to anything close to the original.
IMHO suno should ideally have worked on separated layers and keep things separated all along the chain.

[verbal mortar]

@past phoenix you may have nailed it. it is HORRIBLE at fixing stems, regenerating stems and even making stems. drums are the only thing that usually come out right. Especially for anything heavier than Alt Rock. It does Rap and dance music no problem because its always been synthetically made (synth, drum loops in fruity loops or samples, etc) but it cannot do distorted guitars, it cannot do true bass guitar if its outside anything that is rock or Alt Rock. anything heavy and it just falls apart on itself, which is a shame cause sometimes, you strike gold but it takes 10 or 20 iterations to get it

i just spent an hour trying to fix the guitars for this song. https://suno.com/s/T2V1f1mE0j5XU0Dh and it still messes it up. it will make the guitar 70% of the way then add drums into it for some reason, even if you say don't add drums

[past phoenix]

if this is low enough: https://suno.com/s/5gepzhG5yY2Kz3tk
-> raspy voice, deep low voice, caveman voice, guttural (not sure the last 2 do help)
and then you wait until you get a low voice and you make a persona of it (that's one feature that does work well!) & reuse it

[past phoenix]

yeah I've experienced that and always ended up regenerating more songs
even 20 more songs will cost less credits than trying to fix, and eventually you'll get something good again

[primal blaze]

oh thanks

[past phoenix]

I vote for that too
yes when you start doing stuff on a cover you do lose the link to the original, but there are even bugs that makes the link go away while you've done things the right way
we should totally be allowed to manually re-link to the original

[verbal mortar]

@primal blaze in case that's not low enough https://suno.com/s/KDfZUhILqfaC6iOu here's my guy from Goliath

[frank matrix]

hey guys, how to remove noise from song? im free user btw i wanna remove the noise from song version 3

[verbal mortar]

The only way I have been able to do this with success, is to make stems, download them and in a DAW like Acid Pro or what I use Ableton, remove it that way, that's it.

Suno will continue to pollute anything you make, and studio is hot garbage as i'm finding out today. it doesn't help, it just makes things worse

[frank matrix]

how to remove it in DAW? u converted into MIDI?

[verbal mortar]

I'm sorry, I wish I had better advice for you. Sadly SUNO isn't ready for "primetime" like that yet and you are 100% better off doing it manually using Audacity or a DAW

noooo the midi tracks are worse than normal stems, just make WAV files

the MIDI tracks are useless

[frank matrix]

i did normal stems but still can hear noise in the background of every stem

[verbal mortar]

hmm, that is odd, it shouldn't be in EVERY stem. you may need to go in and surgically remove them, listening to every stem solo and finding the issues. if you have a good DAW, you can use Waves plug ins or Izotope Ozone 12 to fix it

[frank matrix]

i downloaded my song from suno into MP3 coz im free user and i seperated the stems on another website. but the noise still can be hear. btw i only have fl studio 20

[verbal mortar]

share the link to the main song, let me see if I can see what's up. i'll make stems

[frank matrix]

ok i will DM u

[verbal mortar]

just post it here. it's ok.

[frank matrix]

https://suno.com/s/RJPPERWfaIOLCUR6

😅 im not effort to upgrade btw

[verbal mortar]

oooo i hear it, its the synth

you need to remaster it or do a cover I think. all the synth is dirty as hell

[frank matrix]

i tried cover but suno changed everything

[verbal mortar]

its a shame cause its really good. Try to remaster it.

[frank matrix]

not afford to remaster😅 paid feature

[verbal mortar]

I can't get the stems cause i'm not the author. I thought it would let me but yeah I hear it right away.

[frank matrix]

if u download that song and upload it?

[verbal mortar]

I can't because if I download it, it will just be the main wav. it needs to be individual stems

but its the synth its got a lot of artifacts

[frank matrix]

hmm.. if any website to fix that too much synth?

haha

[verbal mortar]

Yes. https://www.lalal.ai/

you can make the stems here. then with that you can use a DAW to clean it all up

but it's gonna be hard man. the synth is really dirty I am not sure it can be fixed

I can't do it. no time

[frank matrix]

i've tried but it can't be fixed

[verbal mortar]

I thought this was gonna be an easy pass, but yeah you really got to get in there and be a surgeon, would take a few hours and I can't commit to that. I got my own songs to fix

[frank matrix]

im afraid if i subscribe suno to remaster and the result might be worse e.g suno change the singer's vocal, singing style, instrumental

[verbal mortar]

only one way to find out, you can always cancel the sub my man

[frank matrix]

but suno can't give a refund, right?

[tidal prism]

im in the suno editor changing the chorus of my song cause the vocals were terrible

ive found a good replacementn it blends the end of verse 1 to the chorus perfectly but the end of the chorus to the start of the verse doesnt blend at all and makes a terrible jarring transition

is there a way to smooth it all out?

[past phoenix]

simple, you made it using v3 (I thought 4.5 was free for everyone now?), and while v3 was pretty impressive for its time, it was pretty much spitting out 1 good-sounding song after 10 bad ones. V5 is WAY above it. But if you have access to 4.5, even that should produce much better results.

(also suno has a feature to remake a song using a newer version, but it doesn't do an amazing job at it)

[lunar cloud]

Sorry for the slow thank you, but THANK YOU! That's awesome 🙂

[urban forum]

I know there’s a dedicated channel for #1420792719969878028, but does anyone have a recommendation for a good YouTube video explaining the new Studio features? I just upgraded my account, and honestly, my first impression is that it’s easier to just regenerate a whole song.

When I try to fix parts using the extracted stems, the quality feels awful, like v3.0 instead of v5, Plus, the instruments are producing sounds that aren't even physically possible for those instruments (high pitched sounds for Drums).

Is it just me, or is anyone else finding the stem generation doodoo?

ahhh shit, i just read above, seems everyone is having issues generating stems

bless living in EU i guess since i'm entitled to a refund 🤣

[near siren]

I completely agree with your thoughts on separated layers. Suno has been everything I’ve needed as a neurodivergent musician. It takes my lyrics, melodies, chord progressions and demo productions and creates the perfect final product essentially acting as my music producer. However this isn’t ideal for me because I can’t remove the vocals and replace them with my own without it messing up the instrumental.

The ideal situation would be for Suno to generate each individual stem rather than a single finished audio file. Even two stems instrumental and vocal would be great. I really hope this feature comes to Suno ASAP because until it does, it’s only a novelty rather than an effective tool for me.

[wet minnow]

I do that in studio. Add a layer and generate alongside it. There is also the option to import stems and for example eq them or clean them up.

[near siren]

Additionally—does anyone have an effective strategies currently they are using to replace the AI vocal with another vocal? If you don’t mind sharing, that’s would be appreciated 😊

[quiet solar]

"male bass vocals", or "bass-baritone vocals" are often a good starting point

[near siren]

Do you mean like for example, I could open studio, import my acappella and generate the rest around that? I didn’t consider this. So far I have tried to take the AI isolated stems and regenerate them, but that did not work out at all lol

[past phoenix]

other than generating an instrumental track to start with, not really

[wet minnow]

Exactly. Studio works in layers. You can for example add a beat or a vocal and then make a new layer and generate for example strings or sound effects or whatever you want to hear. you prompt for the things that should get generated. studio takes everything into account that is in the timeline where you generate. its quite intuitive.

[past phoenix]

if that's what you wanna do, you don't need studio for that:

[past phoenix]

I've never tried studio - do those layers sound perfect, or just better than extracted stems?

[near siren]

@wet minnow @past phoenix Ahhh thanks — will check out both.

[past phoenix]

I find the basic editor pretty lame btw, especially with its built-in timestretching (which fortunately we can switch off, but it shouldn't even be on by default) that makes vocals awful
I only use it to readjust the volume after the -3dB reduction when you import anything

[wet minnow]

they are direct generations, so they have the full fresh quality and no quirks of stemming.
What i usually do is to generate a song that i like, then i stem it apart, import the stems in studio.
Then i see that i can clean the stems, or i remove a whole stem and generate something different instead in the existing song. for example take out the synthesizer and replace it with a saxophone... things like that. or different voices... adding background vocals. all that.

[near siren]

And does this create the instrumental individuality or does it only add it to the vocal?

[quiet solar]

Hmm...even if the stem splitting put them in the wrong place, it's likely Suno will associated those ad-libs with being backing vocals when it 'listens back' to it as an audio feed for Cover or Extend.

There's a few ways you could go about this - to start with I'd suggest editing the displayed lyrics of the song to add in the backing singer lines where they actually appear, as you want them, e.g;

Your lyrics as before
(Uh-oh - Time to go)
Your lyrics carryin on again

...then cover the song (ideally on a high-ish audio influence if you're a Pro/Premier user. Try something like 65%-75%) and see if it naturally blends them in to the new version.

If that doesn't work, or it changes too much, there are slightly more complex ways that you could take out to make fixes in FL Studio or Audacity, but it's always worth trying the easy route first. If you need a workflow for the editing route, let me know.

[past phoenix]

I didn't know that
now I'm wondering if the non-studio song generation isn't also working like that, only it then masters everything to present you a full song, and then ditches all the good layers. That would be pretty bad if that's what it really did.

[past phoenix]

it adds it
without studio you don't get layers

[wet minnow]

Studio doesnt do things by itself. it needs you to make it do things.
The usual suno songs are a "blob" of sound. you get what you hear, no stems, just the thing that it is. The stemming is afterwards.
Where in studio you would build the song layer by layer. Starting with just one thing, for example a beat or voice, and then you add to it...
The other way is to start with a set of extracted stems and then make it sound nice by removing quirks or stuff you dont like in the music.

[past phoenix]

btw we now have this, I've briefly tried it and the results sounded really bad. That's why I thought that suno was just not able to produce clear single layers

[wet minnow]

the sounds thing is for short sound effects... i make these and use them in studio. here is an example where i generated a ton of sound effects and then layered everything in studio ... you will notice that it stays super clean through the whole song. because i take just the best sounding things and use them from start to finish... consistency. https://suno.com/song/2561dd43-40db-480a-9244-529335c32571

[quiet solar]

Suno can produce a clear single layer, but it wasn't originally designed to do so. It generates everything all at once and tries to create an entire song, and that 'thinking' can sneak in if you're working on a single stem outside of Studio. One of the notable benefits of Studio is that you can actually, to a large extent, leash it to just produce that single stem.

It's not perfect, as that feature is still pretty new in the grand scheme of things, but it's good enough to make things by layering and will almost certainly get better in future given Studio is one of Suno's flagship features.

[past phoenix]

but an AI generating a fully mixed song and generating separate layers are 2 very distinct processes
it'd make sense to me that the process would work in layers from the very beginning. But then suno would not need that awful stem detection as it would already have those stems in the first place

[quiet solar]

It creates the entire song as it goes, rather than making different layers and building them up. Even the mastering effects are mimicked as it goes based on what the end result is intended to sound like, rather than applied afterwards.

[past phoenix]

yeah that's what I'm thinking, it hasn't been designed to do so. Especially when it has been trained on fully mixed songs, so it already relied on (flawed) stemming for the analysis part.
I believe that a -good- process in layers would have to be trained differently, on solo instruments.

[quiet solar]

All of the stems are Suno retroactively breaking the pieces of the song apart into what it expects to hear in each stem, much like any other AI-based stem splitter.

[past phoenix]

doesn't it at least generate vocals separately though
looks like a more logical thing to do

[quiet solar]

Yeah, as you say, it would be two totally different ways to go about producing a song. It's a good, sometimes astonishingly good, audio mimic rather than AI version of, say, FL Studio which can generate it's own samples and sounds. Although if you look at the new features inside and outside of Studio, I'd say they're indicative of a direction of travel which is going to support both routes over time.

[wet minnow]

would make sense, but seems to be a comletely different approach from how it works. for layers you need context. so it would have to be made layer by layer, after each other, instead of all together..

[past phoenix]

yeah but sadly it's black magic (& I'm not saying it's badly done, it really is a very hard thing to do) that simply isn't working yet. Before AI, it wasn't a perfectly working process in Melodyne either.

[wet minnow]

I would say the stems are quite usable.
It also depends on the base song how well stemming works.
If you use the stems to clean a song, there is a chance its difficult as the things you want to remove usually "hide" in stems you still wanna use. The art is to filter that out or replace that.
Stems are stems 😄

[past phoenix]

well on "my" last song I only wanted to do some P filtering on the vocals, gave another chance to the stemming (so only extracting the vocals & the rest of the track), and both sounded pretty bad remixed together.
So I ended up surgically editing the spectrum, but it was too busy to do anything good and I just left the popping lol

[quiet solar]

Black box rather than black magic, but yes. 😉 Generating or working at the stem layer outside of Studio is not one of Suno's strong points.

That said, I've been editing at vocal vs instrument level in Audacity since v3 to fix obvious errors, and dipping in and out of individual stems if there's something specific I'm looking to add or remove. So it's not as if it's impossible, it's just not at the "I click Create and it's all just done for me" level yet.

I find it helps to think of Suno as being a toolkit of flawed-yet-still-often-very-effective options that you can use to make music. Whether it makes more sense to do something using stem layering piece by piece, extend (or the new sample tool) to create samples for later fixing/editing, or cover to try to attack the problem as a coherent whole affecting the entire song, will all vary depending on what you're trying to do and how much you're trying to leave untouched.

[past phoenix]

no, more black magic than black box.
Stemming was done before AI and was hand-made algo. It's black magic because (especially in a song) harmonics do cover each other and the information is normally "lost", you can only make guesses and it's already amazing that it works as well as it does.

[quiet solar]

If the vocals need fixing at the popping/artefact layer and you can't do that in a simple pass after extraction, I'd suggest doing a tactical Extend at some point after they first appear in the song and re-doing that section in the Extend. It's unlikely the artefacts will be repeated, and if you're only looking to raid and replace the vocal layer from your original song with the new one, that kind of edit can be pretty seamless.

You'll just need to do it in Studio or an external DAW/editor.

[past phoenix]

to me, "extend" does the same as replacing a section in the editor, it sounds awful and out of place, like it just couldn't regenerate the same song. Each time I used those I regretted it, and when I didn't ditch the results and kept the song, the replaced section stood out almost as badly as the original glitch that I was trying to fix

So now when I get a really good song and it misspoke something, if it still works musically I just keep it.

(or I misread, what you described seems to require stemming again)

[restive pebble]

Not being that skilled in DAW or studio, when I find vocal issues in a track I really like, I make a cover of the song with a persona from the original. It isn't a perfect solution but usually it resolves the vocal issues in a few passes provided I find the right slider settings.
It's a gamble that pays off with enough fiddling about. Maybe that'll help.

Maybe this is why I have a hard time releasing anything lol

[past phoenix]

that really is the thing to do, it's what I do too
persona really works well

[restive pebble]

Not all the time lol

sometimes it takes too much of the audio instead of vocals

but yeah it is the only way I can get stuff done in here

[past phoenix]

what I don't understand though is that when you do a cover with a persona, you only get one "audio influence" fader, so you're controlling both through one value? It's unclear to me

[restive pebble]

Hence my issue. finding the right spot for that slder isn't always easy

yes it controls both

[past phoenix]

I'm no even sure it does affect the persona as the hint says

I'm suspecting it only applies to the original song

[restive pebble]

it does, turn it down low and it's barely recognizable. At least thast's my experience

[restive pebble]

it does apply to the original song. It is supposed to influence how much the original influences the current cover

or mix

[past phoenix]

or is it just that the persona takes "more or less" the same voice but is still a bit random regardless of that fader

[past phoenix]

yeah but the hint says "or persona"

[restive pebble]

I truly can't say. I always get the voice but sometimes the audio is too much in front and the vocals too far in back. Not in the mix but in the influence

[restive pebble]

The other sliders have an effect too.Weirdness I always push to 17 or below to get my persona and style is either 50% or lower

not always but usually

[past phoenix]

not sure that has to do with the fader
it's more in the mix it's trying to do, sometimes the mix is just awful, perhaps a result of having been trained on badly mixed songs lol

[restive pebble]

The faders have influence but if the original mix is poor quality, Suno won't fix it with a higher slider value. It will amplify it IMO

[past phoenix]

yeah they surely do, style influence is pretty straighforward
as for weirdness, it's still unclear to me, but I noticed that if it's a remix I'm doing, it seems better to lower the weirdness

but if it's really the case that audio influence applies to both, then I vote for a new persona influence fader

[restive pebble]

Yes the weirdness slider is important from what I found. If it is not lowered the AI adds all sorts of changes that I never asked for

[restive pebble]

Ideally that would be great. Two sliders, audio and vocals

Wonder how that'd work. Suno would need to separate the stems before hand/
Idea!
What about pulling the vocal stem and making a persona just from the vocal stem?

that would cause the slider to affect only the vocals

[past phoenix]

I also don't quite understand how the personas work. Yeah it does take the timbre, that's clear enough, but it also does seem to keep a personality, which isn't something you can guess from a vocal sample.

[past phoenix]

well no, it's already regenerating everything

[restive pebble]

is it without being asked to?
It's bringing the mix itself for the cover/persona or how would it be causing the vocal differences to the new generation of the song being made?

[past phoenix]

what you described (making a persona from the vocal stem) is already how personas now work after the last change, I believe
but you're right, before that, using a persona on a song also took influences from the musical track. I'm not sure it still does that now

[restive pebble]

Maybe I missed somethign

[restive pebble]

yeah but it still has the audio not just vocals

at least that's what I see

[past phoenix]

but now when you make a persona (or upgrade an old one) you select a part of the track with the vocals you want
and it shows this, so it's already doing it

[restive pebble]

I've an acoustic persona. Vocals are what I like, acoustic is only for specific songs obviously, but if I use the persona on something not acoustic, pushing the slider (audio) gives me more acoustic influence where the song isn't needing it, just the vocals. Hope that made sense..

yes I have that selected in my persona.... but that above still happens. More audio not enough vocal

it's not a science, more like magic lol

Also noticed that during peak usage times Suno behaves differently. less quality as if it's multitasking too much

[past phoenix]

well in the documentation it does say that personas save "the essence of the song"
but they might have changed that with the new version
you can also drag & drop several songs for "influence" and I would suspect that it's exactly what the old version of personas was doing

[restive pebble]

True and maybe so about the influence/persona relationship.

Without a definitive this is what the 'thing' is doing and how to use it from Suno, we are becoming detectives on how to use it and what it can do

Little tips help but only. A little

[past phoenix]

not sure about the quality, seems random to me
however the bugs like empty songs, songs cropped after just 5 seconds, or garbage sound, seem to be when they're overloaded
but there are also "bugs" that aren't random, like if you do a cover and your audio influence is too high, sometimes it looks like it can't keep up with the beat and ends up skipping beats to catch up and the result is just garbage

[past phoenix]

it's even easier to say what does work in suno
at first I was about to learn all of the possible lyrics tags
but the reality is, they almost don't work. Most of the time (but not always) they're just ignored. Except for adlibs which do work well (well you can't control who sings them but it generally makes a good choice)

[restive pebble]

yes, too high and Suno adds things or misses others.

quality isn't just clarity it's all the things you mentioned too. I've extracted stems that are completely unusable. A song I made has no synths mentioned in the style and the stem for guitar is flat but the synth stem holds the guitar.
Usually this happens when Suno is olverloaded. IMO

[past phoenix]

even just switching male/female vocals during a song is so unreliable, you just can't control it

[restive pebble]

I must disagree. The meta tags are crucial for my sound be it vocals or instruments..

[past phoenix]

well, plz share the ones you know to be working then

[restive pebble]

I seem to have success with male/female vocals using tags and the male/female button in advanced

[restive pebble]

I've been workign on something but it isn't completed yet. It is never 100% but 85% is about the best I/U can get from Suno

[past phoenix]

yeah obviously the switch does work
but the tags, I've browsed discussions and everyone seems to have the same problem, switching often works, but you can't control which one will start first, and it's generally random
so if your song has bits made strictly for one singer or another, you're screwed most of the time

and whenever I read the lyrics of a song I liked on suno, generally NOTHING that's specified in the tags seems to appear in the song lol

[restive pebble]

In my lyrics I always use vocal tags. Suno does a good job of choosing when to add feeling to a song and when to lessen it's vocal output but nudging it with the vocal cues make a big difiference in my stuff.

Styles describe the vocalist, Male,Baritone, Husky etc..
Lyrics I usuualy call the vocals just under the vers if the song is all one gender. If multiple I call out under the main header tags.
[Verse | Vocals: Male etc]

or

[Verse]
[Vocals: female, etc]

[quiet solar]

Extend can be hit and miss on that front (anything involving a regeneration is), but if you're only looking to extract the vocal layer from the extracted stem and replace the original vocal stem with it then it works just fine; and is far better at keeping the singer than a straight cover.

If you prefer the cover/persona route and are happier with smaller-scale changes across the whole song, that's also a solid workflow.

[restive pebble]

my song taut was done using multiple vocalists in the song and done in a single pass from Suno. (not a one shot workled perfectly) many generations for the vocals to be exactly what and wehre. Fine tuning helps and the sliders did make a difference.

[past phoenix]

yeah that's what I've been doing, I even tried lots of different formats (like [Verse - female vocals] or putting [female vocals] before/after the verse block), suno always does whatever it wants, and it seems to apply to most users

[restive pebble]

let me see if I can find the tags I used for the whole song maybe that'll help you some. Like I said it ain;t 100%...

[past phoenix]

well it has always been a miss for me
it just has never worked for me, really

[quiet solar]

Guide to best practice and workarounds for duets here: https://discord.com/channels/1069381916492562582/1466598465969590465

Spoiler alert - what you're trying to do requires workarounds and editing

[restive pebble]

here's a snippet of the song When We Said Hello, a duet I did in a single pass.
[Verse 1]
[Vocal: Male]
Where did we go wrong, (my love) [Am]
Was it when we said hello? [E]

[Vocal: Female]
Never learned to play this game, [C/G]
Always moving way too slow. [F]

Suno caught on and made the song how I asked it.

And @quiet solar is right. That is the only way to make a duet and guarantee Suno does what you want.

[past phoenix]

well that sums it up, yeah

[restive pebble]

I use this method, (mine) because I am not great at editing

[quiet solar]

Yeah. And the rest of it is how you work around that limitation, built up from bitter experience. xD

[restive pebble]

I wonder if a mashup of the male and female versions would make the duet easier

[restive pebble]

Bitter... never saw that in you brother lol

[past phoenix]

ah, perhaps it's because in your case you start with the male
from my & others experiences, it's starting with the female that's unreliable

[quiet solar]

It's just as random, in my experience.

[restive pebble]

This is what makes Suno such a challenge. We do not know what is hard-coded to work or not

[restive pebble]

no idea. I could run it with a female first and see.

[past phoenix]

well it's AI so most likely it's not even hard-coded and the behavior is probably as hard to understand for the devs

could be a hard-coded pre-process but I'd suspect it's based on AI training as well

[quiet solar]

My assumption is that it's not hard-coded, just that some tags are easier to reference - particularly in the genre(s) it's looking up.

If you tag [rapped vocals] in a Nu-metal / rapcore song, 95% of the time Suno gets it without complaint. Do that in Country, and it gives you a funny look, ignores you and does it's own thing.

[past phoenix]

yeah

[restive pebble]

I believe this is due to genre limitations being so different between them Suno is like, "this guy is nuts' rap in coutry

[past phoenix]

well it's based on what the training has been likely to encounter, what exists out there

[restive pebble]

I;'m running a gen of fem,ale first vocals in the duet

slow gen this morning

[quiet solar]

Female first is definitely doable. You might be right in saying it's slightly less reliable than opening with a male singer, though. Anecdotally I used to find that - haven't really tested in v5 though.

[restive pebble]

https://suno.com/s/SQ5ijjwZwwnfYhOH

one pass

very close

[past phoenix]

reliably?
I've even tried forcing a [male vocals] right before [female vocals]

[restive pebble]

reliable = laughable... lol
works but not always

[quiet solar]

Nothing is perfectly reliable. Even [Chorus] goes wrong from time to time.

[restive pebble]

but it caught the vocals very close. they're too close in timbre so my ears don't like it but it is male/female where asked mostly

[past phoenix]

well, it works as a song
but isn't really what you prompted lol

[quiet solar]

As it stands, the art of doing duets is a mix of getting reasonably close in the original generation and then using creative fixes if you want it to perfectly fit a particular vision.

[restive pebble]

it takes several gens to get it. At least it's how I get a duet wirthout fiddlin with stitching it together

[restive pebble]

unless you don't care how many iterations or credits needed to have Suno get it exactly as you envision it

[past phoenix]

yeah then we agree
problem is that when you'll get good vocals you'll wish you got another track's instruments

[restive pebble]

? not sure I follow

[quiet solar]

...and that's where I reach for the editing tools and splice them together. But totally understand why most sane people don't want to do that. xD

[restive pebble]

hence SANE PEOPLE... I am not one of those!

[past phoenix]

once you will get the vocals sung by who was supposed to
then maybe you won't be happy with the music part

[restive pebble]

I just don't know editing well

[restive pebble]

save stem for vocals and add instrumental.

works/sometimes

nothing in suno is 100% but I'll take 85% all day

[past phoenix]

yeah but then back to the problem that stemming doesn't work/sounds really bad atm

I'd have no problem if stemming was good

[restive pebble]

sometimes its good others not. I sometimes make four or five stem passes before I get agood enough one

[quiet solar]

Vocal/instrument stem splits are fine, provided you marry them back up with a vocal/instrument split that's similar to the original.

[restive pebble]

Kanthric undertands the stems much more clearly than I as he has mucial and midi knowledge

[past phoenix]

I don't even find the detection of instruments too bad, all tracks are different, sometimes one has 2 instruments (or vocals) but if they're separated enough in time it's ok
it's just that they have never summed up to the original to me, it has always sounded from worse to awful

not sure if it's because it also messes with phases

[restive pebble]

I do find the separation bad imo. guitars on synth stems, background vocals on the main vocal stem etc..

no idea why

[past phoenix]

yeah but if they don't overlap it's ok
you can still see them as separate tracks

[quiet solar]

FWIW, this was the Audacity breakdown for a three vocalist song created using vocal/instrument stems from a mix of covers and extends from the original song. Two different instrument profiles, about 6-8 different vocal profiles. End result (to my ears at least) sounds absolutely fine apart from one line that had artefacts I didn't notice at the time.

[restive pebble]

what do you charge for lessons!

whe n I mess around like that every splice is noticable

[restive pebble]

I have credits to give lol

[past phoenix]

there are really different ways of splitting a track in the freq domain
normally you'd use an FFT for things like this, and you should be ok with phases
but if you use IIR filtering, which is also a valid possibility, you're gonna mess up phases that the tracks would sound ok individually but would sound terrible when added back together
but I don't know what it's doing

[restive pebble]

out of my league now

[past phoenix]

also depends on the genre I suppose

[quiet solar]

I actually streamed the process live several times about this time last year. Idiot that I was, I didn't save any of the videos to put up on YouTube.

Maybe I should get around to doing that. No charge if I do, there's no particularly deep wizardry to the process. But it can be a little finnicky at times.

[restive pebble]

my issue is the split and splice. I cannot seem to find the proper place (downbeat) and when I do the splice always has a noice or something noticeable.
Take the turkey time song I did for the thanksgiving challenge. The intro is spliced and is noticeable

Not the topic so let me query ride

[past phoenix]

it's easier if you set the proper tempo in the sequencer/editor
now.. suno songs can have a fluctuating tempo (in theory, but in practice only subtly)

[restive pebble]

Kanthric knows all about fluctuating tempos 🤣

didn't you post something about that?

[quiet solar]

If it helps, the main factors are:

1. Pick your splice points carefully - you want to disguise them when the audio is at its most similar between the two versions as much as is possible, and when the song is making a natural transition of it's own (e.g. Verse to Chorus). Sometimes you have to grudgingly accept a sample just won't work invisibly and either make a new one or compromise and keep the original.

2. Try to ensure the tracks overlap for at least a second. You'll have to manually line up the audio, so the sample you start with will probably be longer to facilitate this

3. Adjust the volume of the tracks to be as similar as possible, if necessary

4. Crossfade the tracks

[past phoenix]

the timestretcher that they use in suno's editor by default, for that reason that suno songs don't have a perfect tempo, should really be the first thing to disable if you use it. It sounds terrible (phases problems very audible on vocals)
(lesson learnt from someone here)

[quiet solar]

Also helps if you're only splicing in either vocals or instruments, rather than both. A consistent instrument track does wonders for hiding vocal splicing going on behind it.

[restive pebble]

I'm out of my scope with all this technical jargon. timestretching and such.. I need to stop focusing on lyrical skills and dive into editing

[restive pebble]

thanks for that tidbit

[quiet solar]

Yeah, just to echo @past phoenix - you really don't want to have to deal with tempo or key variations on top as well, if at all possible.

That's why my workflows mostly use extend rather than cover, since those will tend to bring across the same tempo/key far more often than not. If you use covers, you really want to specify the original BPM and key signature (and/or use a high audio influence) if you're planning to create samples for editing.

[past phoenix]

if you ever use suno's editor (even just to adjust the gain of imported tracks), switch this on asap

[quiet solar]

^ This. 100% this.

[restive pebble]

I use that as it seems always enabled

[past phoenix]

is the tempo tag even reliable btw
I never use it but maybe I should

[restive pebble]

But I use it in studio not the editor. Editor I find always makes it worse. Must be me

[restive pebble]

Yes what he asks

Lol

[quiet solar]

As a style prompt, yes - more so in v5 than v4.5. Not perfectly reliable, but what is? Audio influence also helps a lot.

[past phoenix]

it's generally on manual by default
which isn't normally a bad thing, it's just that suno's timestretcher is really poor phasing-wise (it's a true technical issue with timestretching but it has been solved by many)

[quiet solar]

I find the Editor to be more of a headache than a help for this type of editing. Studio is a different ballgame.

[restive pebble]

Many other than myself.
Googling time stretching for music now lol

[past phoenix]

no you don't need that
that's for devs

just saying that the one in the editor is on by default while it should be off, it's not even generally useful there

[restive pebble]

OH

Now I understand why changing tempo bpm messes with the sound or pitch

I'm being schooled in Suno School

[past phoenix]

well it won't when the timestretching is on
but trust me you don't want to touch tempo or pitch afterwards

[restive pebble]

I don't want to touch it all all lol

I want a button that says create and it reads my mind and creates what I see in there

neural implant anyone?

[quiet solar]

To add to my not especially clear last post:

If you prompt for a general tempo in the style prompt (e.g. "fast-paced", or "allegro") then so long as it doesn't conflict with the genre too much that tends to work pretty reliably. You can also specify things like BPM and key signature in the style prompt - slightly less reliable, but definitely an improvement over not specifying.

Putting it in lyrics tags to try to switch tempo during the song, or to set the song tempo up front, is likely to be less reliable than either of the above, as is the general rule of all lyrics tags.

Once you have an existing song, higher levels of audio influence on Cover/Extend helps a lot to ensure what you create matches back to the original.

AFAIK, Suno may still introduce accidental slight tempo variations all the same. But if you need to correct that in editing, you can usually adjust tempo by around 5% - 10% with a decent tool that corrects for pitch without it being noticeable, so it's not a huge issue.

[restive pebble]

Just wow man

[past phoenix]

good to know, thanks

[quiet solar]

Maybe v.178.5? 😂

[restive pebble]

I'll give a shout out to Elon and see if he'll take me as a test subject lol

[past phoenix]

you're probably too old

[restive pebble]

So from what I got is tempo in Suno isn't spot on, sty;les prompts help is specified using musical terminology over bpm but they both help keep Suno in line. Slider/audio makes it stick closer to the original.
Yes?

[restive pebble]

old maybe but never grown

[past phoenix]

I was only mentioning the tempo because you said you had trouble knowing where to splice, and if you wanna splice on beats it helps setting the editor to the tempo of the song
but you can also visually (& audibly) know where to splice

[restive pebble]

wow, I have got a steep curve to round

this'll be fun

[quiet solar]

Yeah, Suno has a tendency to sometimes make minor tempo variations during the song. It's not normally a problem, but it can sometimes be in editing if one version is slightly out of kilter with the other (e.g. your vocalist is at 120bpm for half a line in the version you're trying to add, where the instruments behind are at 110bpm all the way through).

Separate to that, if you want to prompt Suno for a general pace and don't care about a specific BPM, then I'd suggest using the musical terminology for the ballpark range you're after.

If you need a specific BPM and key signature, which is usually because you're creating something specifically with editing in mind and want it to match the original, then to be honest I'd focus primarily on the audio slider. Specifying BPM and key signature will help on top, but if it's not easy then don't worry about it.

BPM you can get fairly easily by opening the original song in Editor/Studio. I tend to confirm key signature with a keyboard while listening, but I'd be surprised if there isn't an AI tool somewhere that can work it out for you.

[restive pebble]

drown me instead... a simpler death

that's confusing even though it's straigh forward

Appreciate the time brothers.. Today I will read up on this and play around in studio/audacity/cakewalk and see what I come up with

[quiet solar]

Most of the time you won't need to worry about tempo at all, don't let it put you off. xD

[past phoenix]

I haven't read "cakewalk" for ages
when it was the name of the app & not the company

(well it's not even the company anymore)

[restive pebble]

think it's bandlab now but don't care as long as it does what I want lol

Not ever going to let soemthing put me off. I face it boldly and accept the beatings until I an beat it back

First, I think I'll bake an apple pie. That;ll motivate me or atleast give me a sugar rush

[brisk zinc]

Hi how to make voice different than most people make ?

[restive pebble]

Vocal styles are tired to genre in a way. Metal won't have a vocalist singing a ballad style timbre from what I know.
You can influence vocals using descriptors in the style prompt which shifts them closer to what you might be trying to get.

If you can be more specific maybe I can help more

[brisk zinc]

I mean voice consistence tool. For hybrid genre songs.

[sleek stump]

For consistency, a Persona can help with that. Split STEMS and use the vocal split for the base of a Persona.

[polar spade]

Voice different and voice consistent are two different things. So let's start with voice different.

1. Among many ways, one is to just sing as close possible to the style you want. The cover your voice choose male.or female with prompting some genres or styles continuing in the direction youre looking for.

Your singing can give a more human characteristics than just trying to prompt Suno to give you something unique.

2. The second, use "persona" to have a consistent voice across your new creations and covers. With Covers, the idea is to crank audio influence up high so not much changes but the voicing and occassional styling of the music is pulled from the persona into the Cover. .

[polar spade]

Just reminding that 5.0 Persona now stems the vocals already before making a persona. So it cuts out that middle stemming step.

Unless you feel theres a distiguishable difference.

[brisk zinc]

Is any helpful tips regarding prompting for suno AI ? Like a website with tutorials or prompts examples with expected results.

[unreal narwhal]

Put ur song into Chat gpt and tell it to give you a song description for Suno thats 50 words or less. You can edit it if you need to after that.

[polar spade]

Honestly, there are dozens of posts online with lists of prompts. Just searching "suno prompts" i see two pop up. Good or bad, most will get you started in experimenting with prompts.

So its past the point of a " definitive" book or list. That idea is pretty much out the window.

Best to work it backwards. Start with your concept of what you want to hear. Then as @unreal narwhal mentions just as any LLM AI to come up with some simple word prompts and style to head you in that direction. Try them out. Hear the results and start tweaking with a prompt or two. Dont spend so much time over 4,000 prompts to find a "perfect one".

When thats a bit comfortable , start learning how to use metatags in your lyrics as part of your prompting.

You can also peruse the workflow https://discord.com/channels/1069381916492562582/1252267444773326848 channel for LOTS of more ideas from "advanced" styles to just some prompt lists.

[tawdry scarab]

Some time ago I made this post on prompting, it still works in 5 though 5 has different abilities on top of it: https://discord.com/channels/1069381916492562582/1365192916687126650
And today I made a follow up specific to extending, when, where, and why:
https://discord.com/channels/1069381916492562582/1468637403647447062

[brisk zinc]

Is letting Suno Ai auto generate and later learn from that good way ?

[tawdry scarab]

anything where you see the inputs, the process, examine the outputs is suitable for learning

but you only practice to do what you do when you practice

which just means that you can learn that way but what you'll learn that way is doing suno auto stuff

[unreal narwhal]

How can I get two male or female artist, in one song, to sound like two different people?

[tawdry scarab]

meticulous attention to all advice on the topic and building in extends one piece at a time or just getting very lucky, very clearly and carefully describe their identities with meta tags that include identifier tags then used before any time they speak

have a block of tags specifically dealing with vocal performance, in the lyrics and with some keywords making it to style

it helps if you can make them contrasting

such as a tenor and a baritone

[urban forum]

sorry if i sound rude, but this is the most ass way of doing it

you're better off regenerating 20 times of a singular sample than working in Studio

being able to make use of a context window and even adjust its size is a godsent, but the output quality is beyond doodoo to call it proper

you'd need to have infinite credits to be able to play around with it for that long in Studio, otherwise it's simply not worth it

[wet minnow]

its what the tools allow to do. or you do it externally.

i usually dont use that many credits in studio because i already have a pretty large library of stemmed stuff where i can get clean sounds and samples from.

[urban forum]

you do it externally
imho that's not an alright thing to say if people have to pay to make use of the studio.

[urban forum]

ye that definitely helps

i suppose for long term users it would indeed be quite nice

but for new users it's rough

[wet minnow]

some dont even have studio. so mixing and matching tools is always an option. i also have my daw besides studio. each one has its benefits (my daw is more "precise" but doesnt have the "generation magic" that makes studio fun)

[urban forum]

yea but i'm stricly talking about Suno atm, not anything external

the only interesting thing about the Premier plan is access to Studio, and it's beyond underwhelming

even the upgrade text by the way makes 0 sense, but that's for the feedback channel tbh

[wet minnow]

before studio the benefit of prem was just to have 10k credits 😄

[urban forum]

😶

[wet minnow]

i see it as a basically free tool on top of my 10k credits. For that its quite neat. AND its still beta. If you find things you miss... suggest them.

[queen linden]

another way to acquire a style ( it does not help with female voices) is to google or grok a simple describe fill in the style or voice, type of genre. keep it simple when you are using them

[tawdry scarab]

Studio is worth it to me to be able to take a flat track and overlay the stems on top of it, unmuting the original, and using eq on each

That makes most things usable

[polar spade]

By that idea, all the generative outputs also would not be , well generative, they would be exactly the same.

Im just saying we have to keep the perspective that generative means we can NOT create a vocal stem and a drum stem that will sync perfectly against each other.

The only conceptual means is to create maybe a click track stem, then generate a drum stem against that then create a "cover" track of the drums to generate a rhythm track, but then weve never determined the pockets for establishing a melody and vocals so we keep hitting these walls. These ideas are normally avoided by a writer by creating all the ideas at once and massage things, chart them down , write lyrics , then basically record all at once.

Generative AI does basically the same but backwards. It starts with a block of noise and chisels away at it, leaving the sound we want. There are never just one track of drums or vocals. And if there were, they wouldnt sync up.

[polar spade]

Well, to bring it in from the extremes, maybe an expense of $50 total a month is about my estimate of what it presently costs in credits to do any version. Work it out in Studio, run 50 generations until one pans out, or Frankenstein it between Suno and outside resources.

At some point, not everyone can afford to enjoy front row seats, but still most can get some enjoyment, even from the nose-bleeds.

There are workarounds that can get even the least costing tiers to produce great results, but I think experience lends a hand there as well. I still use 3.5 a few prompts and extends with much joy and good results.

That said, things get cheaper and faster the more skills we learn. Even knowing when to alter their focus and try something else is a very good skill that seems to frustrate many others who keep repeating a task and hoping things change.

[polar spade]

Kind of an aside, but the cover version in Sudio is a good place to run several versions of tracks of drums, vocals, guitar, etc. To similarly choose something that might work better and fit the vibe. So it similar to just a musician trying several takes in a studio setting.

About every month theres a new tool to integrate into Studio to help conclude tha Studio is nice tool to have in your pocket.

[trim pike]

what kind of song should i make for a retro 80s skate rink like dirty dancing video but roller skates

[solemn sierra]

Style / Genre (EN):

1980s retro disco, synth-pop, roller disco vibe, funk bass, analog synths, warm pads, melodic arpeggios, nostalgic pop

Tempo (EN):

110 BPM, steady groove, smooth dance rhythm

Mood / Atmosphere (EN):

romantic, joyful, nostalgic, energetic, cinematic, youthful, neon-lit night

Instruments (EN):

analog synthesizers, disco bassline, electric guitar funk riffs, classic drum machine, hand claps, reverb-heavy leads

Vocal Style (EN):

expressive pop vocal, warm tone, emotional but playful, light harmonies, 80s pop delivery

Lyrics Theme (EN):

falling in love at a roller skating rink, dancing and skating through the night, freedom, movement, young romance, music and motion

Song Structure (EN):

intro synth → verse → pre-chorus build → big catchy chorus → verse → chorus → emotional bridge → final chorus

[strong sandal]

Is there a way to make it so the music stops and it goes spoken word? I've tried every which way I can think of and the most I can get is a very quiet single instrument .

[tawdry scarab]

it's possible but it is hard to make sudden changes

techniques that make it more possible will burn more credits than getting lucky, but, they improve the ability to suddenly change the track

the one I always end up using is micro-extends, picking only silence or spoken word, and then with a few tiny little extends and no get whole your tagging all over for just the spoken word, using "" marks, using [spoken word], putting spoken word in the style, not loading the style with a bunch of "no instruments" just maybe "instrumentation" in the avoid keywords

[polar spade]

Please mention if youre Pro or Premium.

[ocean jetty]

I need help on the Suno Studio, downloading stems an ensuring they are on the right BPM. I've tried both downloading "Follow Track" and "Manual BPM" and than setting the correct BPM in my DAW however throughout the song - the BPM goes off time.

[glossy rock]

If you are in pro, i found that:
[Break]
[Spoken]: "Your speech here"
[Resume Music]
works quite well

[hollow slate]

this one song refuses to be used by the studio just sits there black sscreen

[glossy rock]

I have question here. I have been trying to make Deep House trance music or just also basic trance. I can't get Suno to space out words and repeat them as they would in a normal song. How would I structure and prompt it to make it work? I am Pro btw

[neat nexus]

Have you been able to locate anyone on Suno whose outputs match your goal sound?

[waxen laurel]

On the free version, I've had several songs that just cut off . I'm assuming 7:59 is a max time length? Is that different for pro accounts?

[hollow slate]

not sure saw one guy with a 14 minute song

[neat nexus]

It's one of those 'up to' deals. Some songs come out longer than others for various reasons. If you extend you can make them go even longer.

[glossy rock]

This can be done through repeated extending.

[hollow slate]

oh thanks

[glossy rock]

Just a warning. The music can start to loose touch if you make it too long. I was once shown a song with 20 extensions where the last 5 mins was just noise

[hollow slate]

Yeah whats up with that horrible noise..usually on the end of the good stuff

[neat nexus]

There are upper limits of course, but one way to work within that is to periodically get full song after doing a number of extends, extend from before the noise and go again.

If you don't allow for enough time between get full songs though it can cause loops.

The AI will just grab something and sing it over and over and refuse to move on.

[glossy rock]

I have not.

[neat nexus]

You'll be the trailblazer then? That makes some things harder. There are two options there:

• Have to keep trying and coming up with new ideas and methods until you hit on something that works.

• It might not technically be possible to create the exact sound you're looking for (yet?)

[neat nexus]

Have you tried formatting the text in any way to see if it drags out the vocals or spaces out the words?

Have you looked in to phonetics as a way to draw out the syllables of words to see if you can get the spaced out vocals that way?

May be able to get a LLM to assist you with that as long as you don't take their words as gospel

[glossy rock]

...

Ban em

@neat nexus

[glossy rock]

I have tried this and still trying to figure it out. Unfortunatly SUno does take the structure to heart

[neat nexus]

You mean you investigated the style with LLM or you mean you tried formatting the speech with the LLM's assistance?

[glossy rock]

I tried formatting the style. I know the style of trance:
it’s like music with words repeated. Then it goes blank and there is just a spoken quote and then the music resumes. Before I brought it to a LLM, I tried this in Suno but it couldn’t get it

[neat nexus]

Do you have any examples yourself of what you tried?

Or a track of the exact style example you're trying for?

Sorry...tired of working this blind.

Give me some idea of the attempts made and the goal.

[glossy rock]

Yeah I got something lemme send you a link

https://suno.com/s/HYJEBEHuaUXtDSCG
this is the right structure kinda. But this is dubstep so not the right music

https://suno.com/s/GlPuY66AYf2wQS0k
and this is the right type of music, but the vocals are too much. The music could also be slightly toned down

https://open.spotify.com/track/2FlMWBfcjQwSo28rv4DUck?si=qNMkKjuiRlekZYTK01sTLg
finally a real trance song. What I am going for

Although this one has a lot of vocals. I would do less than this

[quiet solar]

@glossy rock I may be wrong, but it seems like you've got some of the main ingredients sort-of working, just not all together. What's the specific problem (or problems) that you're running into?

[glossy rock]

Well when I bring it together, that’s where it falls apart. The midnight one I sent was as good as I got. But either the timing or the spacing wouldn’t work. Maybe it is just me trying to get a very specific sound instead of being open to diversity?

[quiet solar]

I'm still not entirely sure what you're trying to do, but if I'm grasping in the right direction then you're trying to impact the spacing and repetition of the words in the song? If so, that's a little bit tricky to do, but the biggest impact you can have is how you're structuring that in the lyrics themselves.

For example, your structure in midnight is essentially asking for a very short verse followed by a very short chorus followed by a very short bridge with no reptition. You could alternatively try structuring it more like you did in the first, but making the repetitions yourself rather than relying on Suno's contextual interpretation. For example:

[Verse]
Stars...
Align...
Stars...
Align...
Stars...
Align...

[Verse]
Whispers...
...in the dark
Whispers...
...in the dark
Whispers...
...in the dark
Whispers...
...in the dark

[Chorus - sustained vocals]
Dream on
Dream on
Dream on
Midnight's spark.

[Bridge - building to drop]
Lost in time,
Floating free.
Lost in time,
Floating free.
Lost in time,
Floating free.
Lost in time...
(Free!)
[Drop — heavy 808s, cowbells, hi-hat triplets]

[neat nexus]

OK....take a look at these....

Just your lyrics: https://suno.com/s/v4A82Ism2JAumQ5V

This one I added some speech in SPRECHGESANG to give it something at the beginning: https://suno.com/s/vvXvqbrDpd30cVbE

This is what I meant by formatting the text. Oh I'm sure my word salad in the style prompt helped too, but the way I tweaked the syllables bent it towards more of a dragged out vocal. Which was enough to influence even the beginning narration.

I'd also definitely look at what @quiet solar posted above tow ork with those too.

That's some nice lyric rewriting there.

[quiet solar]

Artificially stretching the words by repeating letters as Hepcat suggested is also a good shout. Suno is often pretty good at working out what you're trying to achieve with that.

[glossy rock]

Okay I’ll give all of this a try!!

[neat nexus]

An attemmpt combining the two things. My word stretching and Kanthric's structure with your words.

https://suno.com/s/XMYcezmyG17OxWad
https://suno.com/s/6COI1OpnKUWBC53V

Now unfortunately I have to get ready for some stuff tonight so I'll be taking off for a bit. I hope this was at least inspirational to you?

[glossy rock]

It was!!

Tysm

[dire jacinth]

hiiiiii

im new

[heady fiber]

Hey

[sleek stump]

Hi there, welcome to Suno Discord! For general chat and community there are a couple channels for that, https://discord.com/channels/1069381916492562582/1069381916492562585 and https://discord.com/channels/1069381916492562582/1159929934270181437 are where you can find most of the chat happening. This channel is for discussion about Suno prompting and what not.

[scarlet elbow]

Howdy!

**What's the best way to update the song with a few lyrical changes? **

Want to keep everything pretty much the same otherwise

[quiet solar]

If you're happy with 'pretty much the same' then Cover the song and make your lyrical changes as needed. If you're on Pro/Premier, set the audio influence fairly high (e.g. 65%-85%) and weirdness low (e.g. 15%) and use the same style prompt as the original, which should help ensure fairly decent consistency with the original track.

If you'd prefer just to make changes to those small sections, you can try that in the Editor (Pro) or Studio (Premier) by selecting the section you need to rewrite and using Replace Lyrics.

If you're looking to do a very surgical change that only switches out the lyrics and leaves everything else unchanged, you'll need Pro as a minimum to split out the original instrumental/vocal stems, and then Studio or an external DAW/audio editor. Use Cover or Extend to recreate that section of the song with your lyrics changes, then get the vocal stem from that song and use it to replace the original vocals in Studio or the DAW.

[nocturne flower]

Can someone explain why so many credits are being used in the Studio version? I rarely do any covers or replacements there, yet I end up burning through 70-100 credits every time.
do I lose credits even if I cut some parts?

[sleek stump]

Almost everything you do in Studio has a credit cost.

[scarlet elbow]

Thanks @quiet solar !

I am stuck on "Loading DSP Engine" -- I have Pro

[heady panther]

on 4.5-all, what to write in lyrics section so that a phrase is looped or repeated at certain intervals instead of directly treated as a single block to be spoken in one go? say for techno/edm “vocal bites”

[stiff sentinel]

oh i am getting this too on another random song of mine i tried to load up

[scarlet elbow]

woohoo! 🙃

[frozen dagger]

The genre of the audio and the style prompt matter. Assuming it's already in the EDM ballpark you would want to increase the influence of the slider of the one, or both, audio or style influence sliders (if u can, said 4.5all, so maybe not available with free tier).
Regardless, you can try prompting in the lyrics section where the repeated looped verse you want. Let's say I want for example a repetitive backing vocal chant saying "blueprint". The lyrics prompt box area might look something like:

[Delay-loop vocal stab]
(blueprint, blueprint, blueprint, blueprint)

You can also use quotes for different "spins" on the vocal delivery, but sometimes it changes vocalist. (sometimes that's good)

[Vocal loop]
"blueprint... blueprint... blueprint... blueprint..."

Use terms that most closely describes effect you wanting, but things like:
Tempo-synced delay, feedback delay, ping-pong delay, slapback delay, tape delay, dub delay, vocal loop, looped vocal sample, vocal chop, vocal stab, ostinato, mantra, canon, round, reprise, iterative repetition, stutter edit, glitch repeat, phrase looping, sample looping, delay throw

[heady panther]

no sliders on free 4.5-all, but gonna try the rest. many thanks

[hoary mauve]

hi, hopefully someone can help me out a bit. I want to create a song with a lead singer and one single background singer, not a choir. I tried every prompt I can imagine, even asked AI for help but all I get is either a layered standard choir or the echo of the lead-singer for whatever I put in (). any suggestions?

[sinful tartan]

Hi everyone, I'm having trouble creating a jazz piano trio instrumental. I want a slow tempo for the bass and drums but a fast tempo for the piano. I hope someone can give me some prompt help.

[quiet solar]

Anything with multiple vocalists tends to get tricky quite quickly.

What I'd suggest in the first instance is to try adding something like "The song is performed by two vocalists; a smooth male tenor lead and a breathy female support" near the start of the Style prompt to try to get Suno to generate two separate vocalists in the first place.

Suno is more likely to do this if the two vocalists are distinctively different - this will be a lot tougher if you're after two male tenors, for example.

Then in the lyrics, tag the start of each line where the singers change over, e.g:

[Verse]
[male vocalist] The male voice sings this line
And this one too
[female vocalist] The female sings this line
[male vocalist] And we're back to the male again

You can also put the support vocalist lines in () as well, if they're fully separate lines, as that may help to reinforce that you're calling for the backing singer. Just be wary of doing that if they're switching mid-line, as it'll typically be read as an echo of the lead's line - you'd be better off splitting the line into two lines, one for each singer.

If you have a lot of switchovers throughout the song, and it's important that the correct voice sings the correct line, you might realistically be looking at a mix of prompt and editing work to get that to happen. There's a more comprehensive guide for that here: https://discord.com/channels/1069381916492562582/1466598465969590465

PS. Should also add that to expect that this will take multiple attempts to get something approaching what you're looking for. It's not really a question of the 'right' or 'wrong' prompting, but that what you're trying to do is recognised as one of the more difficult things to get out of Suno's current models.

[languid junco]

Why is suno?

[waxen jackal]

Any prompt idea for a banger powerful but groovy sultry dance track slow to medium except rap? Which genre? I want to cover one of my old demo.

[quiet solar]

Making personas from audio uploads, or anything derived from an audio upload, has been disabled.

Basically to prevent creative attempts at voice cloning existing artists

[untold lava]

it still works though... for some tracks

if you extend

stopped working for tracks uploaded with no instrumentation/././

[quiet solar]

That's likely a bug.

[untold lava]

looks like gonna have to do the new method i made up... if u want your own voice

just use add instruments to the uploaded track instead of extend

why have they disabled this does anyone know?

they should integrate a voice verification system like elevenlabs has, so we can use our own voices lol

im assuming its so you dont make ronald grump sing WAP

WOW WONT LET ME SAY HIS REAL NAME

[naive quail]

They don't want people uploading recordings of other people and making personas from them. they have no way to tell whether it's your voice or not.

There was a period shortly after they introduced personas where we could make personas from an instrumental recording, such as a guitar, but only if no vocals were detected. At some point, they started blocking even those. 'Sounds like there have been a few times since then that creating personas from recordings has worked again, but it probably hasn't been intentional - 'maybe not enough regression testing with updates.

[untold lava]

"They don't want people uploading recordings of other people and making personas from them. they have no way to tell whether it's your voice or not" -- sure they do, it's not free, for them to put such a thing in place, but sure they can tell, just have to create a database of audio same way they already do with music library's. or CHEAPER would be what elevenlabs does. if you try to create a pro or instant voice clone it can be flagged and you are asked to verify with a custom recording done in real time on their platform

to match the voice

to your own

can make AI agent to analyze and perform the task autonomously

suno needs to do this. Because some of us use the platform for serious work, former musicians/vocalists looking to use their own new or pre-existing recordings in outputs was the appeal

im just happy managed to make quite a few voice personas before this funny business

thanks for your answer

[naive quail]

How will they be able to distinguish my voice from my coworker's? She sings, but doesn't have anything on the internet to be scraped for training data. I could secretly record her singing on her lunch break, make a persona of that and start having her sing songs about cats. 😄

I get what you're saying wit hthe elevenlabs route and maybe they'll go that way at some point, but I suspect that it won't be any time soon. 🙁

[quiet solar]

You can still do that, it just involves removing the vocal track and taking the song out of Suno to work with your own voice in a DAW/audio editor.

[untold lava]

my voice is damaged

[brisk zinc]

Can you extract voice channel only from track to download and tweak it ?

[untold lava]

this allowed/allows me to sing again, well as close as possible

i have uniquely studio recordings from abck then...

used them

for this

[zealous solar]

Are there constant changes to Suno V5 Pro? I have lost four songs to trash again. This was not recognisable. The beat was wrong, the music was bad, and then two of the tracks became a terrible screeching interference, with no accountability for such a strange sound. I am completely perplexed by this. One minute it works just fine, and the next it goes bonkers. Has anyone a blinker of an idea how to override such terrible results?

[quiet solar]

Ah. Well firstly, really sorry to hear that mate.

Secondly, you could team up Suno with a separate AI vocal replacer trained on your previous recordings and use that on the vocal stem. I genuinely don't know if Suno has any plans to find a way to reintroduce allowing vocal personas, or where that would sit on their list of priorities since it's not something that (AFAIK) the vast majority of users do.

[quiet solar]

Not on a minute to minute level, no. Could you share a link to one or two of the songs that went haywire?

[untold lava]

yeah i would have to do a ton of work in POST to give it the same sort of polish that you get with one click on SUNO

but i have a work around...

if i share it too much word gets out they will take that away too 🙂

but My voice personas i have already, ssound identical to me, people who are familiar with my music cannot discern a difference

people from 20- yrs ago lol

[brisk zinc]

I think vocal voice generator should introduce with specific trait impacting voice and it's quality of sound produce. Because now, almost all songs contains same voices.

[opal orbit]

Anyone got tips on the best way to instruct suno to product only that backing bass or synthline when regenning in studio? So it doesn't keep replicating the melody?

I'm particularily stuck on this one

[neat nexus]

I'm confused. You have voice personas you say sound identical to your own voice that work. What's the issue?

Why aren't you able to just use those?

[nocturne flower]

can you be more specific?

[sleek stump]

You dont lose credits when using studio, when you make an edit, that has a credit cost. When you regen an instrument or vocal track, that has a credit cost. Anything you do that requires compute time has a credit cost.

[nocturne flower]

It was saying 10 credits per cover within Studio so I did like 2-3 changes. When I checked my credits I had over 100 less

[clear zephyr]

I cannot sign into my Suno account. I have a paid-for pro subscription that I use intermittently. I just tried to sign in as normal to continue working on my songs and instead I am directed to provide my either a telephone number or some other contact details. Once this is completed I am then asked for a new username and why I want to use Suno. Then a new account is created. I don't want a new account, I want access to the account I already have and have paid for. I try to email support and the email bounces. I try to email billing and the email bounces. I have now spent hours researching what to do which has led me here. Can anyone help me deal with these RIP OFF MERCHANTS?

[sleek stump]

Firstly, there is no tech support here in the Discord, you need to do that through email, [e-mail] or [e-mail] (for billing issues).

That said, are you trying to log in with a phone number? There is a known issue with phones adding or deleting a 0 at the start of the login number. If that is the case for you, try deleting that leading 0 and try logging in again.

[clear zephyr]

There is no tech support ANYWHERE for suno, that is the problem. I have just signed up to discord specifically because I cannot contact Suno by any other means, and they suggest on their own site to contact people here. The emails you listed have bounced my emails

[polar spade]

If you are using the add track feature to add an solo instrument, then you may be having issues with the outcomes. Im experimenting a bit with that to see whats going on.

But if you havn't see it, it's below the waveforms/tracks. You can choose any instrument and add style prompts with the idea of creating a bassline, riffs, drops and so on to your composition.

[clear zephyr]

I am trying to log in on my laptop, as I always do. I don't know what else to do. The only other suggestion users with similar problems have suggested is to go to my bank and provide proof that I am being ignored and not provided with the service I have paid and get my money back. Surely there is an easier way?

[opal orbit]

Yes I'm very familliar with it, I'm just having trouble with this particular gen in getting a clean backing synth that doesn't follow melody. I'm looking for help with prompting

I'm guessing things changed a bit with this new studio update, the prompts I usually use aren't working atm.

[polar spade]

What you likely got was an auto-response to say you are in queue where it may also suggest you try Discord to self-help with the FAQs and knowledgebase info.

That is a standard auto-response.

As for your issue: " I emailed support and sent a ticket about 3 weeks ago, about not having access to the "Explore" "Radio" or "Hooks" sidebars despite being having a premium account. Is there any chance I could get an update on this?
I had access previously, and it stopped working about 8 weeks ago."

Have you cleared both cache and LocalStorage , both. Log out, Close all Browsers. Just open one. See what you get.

Done this already?

[opal orbit]

Yep, done all that. It's an issue that persists across multiple browsers, computers, and incognito. Hooks doesn't load on the mobile app either.

[clear zephyr]

Not an auto-response, no response at all. I get an email from mail delivery subsystem saying email failed to deliver

I can't even give feedback on the suno site. It seems suno really does not want to engage with their customers once payment is received

[polar spade]

Most dont know how to clear LocalStorage, (they just clear their cache only) BTW, so good job that that you know that "developer" move.

[quiet solar]

If you originally signed up using your phone number, it's worth trying Ginger's suggestion a few posts up. It's a known issue that's hit a few people recently, and a very common fix for it.

[opal orbit]

This is what my sidebar looks like. I've reached out in #🪲┃bugs-web before, and sent a message to the mods when they requested, but no one ever responded.

To add to this, I've never had the hooks sidebar, but I used to be able to access it by clicking on someone's hook in the homepage and scrolling or clicking "Create Hook" there. Now it gives a "Not Found" webpage.

[opal orbit]

I'm gonna do it again in any case. I probably would need to for any fixes on this to work anyway. But yeah, since its across multiple devices it's certainly seems to be something linked to my account.

[polar spade]

You've run through all options. That helps.

However, then that leaves us not being able to go further than to see if there is an explanation directly from Support.

Reporting in bug channel wont get us a reason, but sometimes it might get a validation that others are having the same issue.

As I had shown in my previous screenshot, I dont have Explore or Radio either. It makes more sense then that its a developer decision -- perhaps working on those features -- than it being your subscription.

[opal orbit]

I figured, though everyone's response so far has been confusion. If it was a developer thing then I'd assume the Suno team would know about it happening? In any case, I hadn't really seen anyone bring up this issue. There's certainly plenty of hooks I've had waiting off to the side I can't upload yet, lol.

[polar spade]

Yep. If others though also dont show, then not confusing at all. They [Suno] are revamping the UI.

Logging out of your account and back in is all I can think of to try to get the most updated features to show. Certainly can be a combination of their working on it and your account needed to upload the lastest bundle from a new log in.

So last suggestion, though im sure you've done, is to log out everywhere, and log back in on only one app or browser.

[opal orbit]

I think I've been trouble shooting on isolated devices, one at a time. Not sure I've done everywhere at once. I'll try that + clearing date/caches and see if that changes anything.

[polar spade]

And Sunos team certainly would know. Youre right.

Many UI reconfigured are not privy to the public. So asking may not get a response for just routine changes in formatting.

But youve got your email out to Support as well. You seem to have done everything needed. So the only two expectations are

1. to get an eventual response from Support
2. observe a change in the layout

Again, I dont think its personally to do with your account as I see i dont have them either.

Maybe a couple other users in here might check their layouts this morning.

Otherwise, glad we could get a bit further in the resolve and rule out a few things.

Good luck.

[opal orbit]

Well, tried it across all devices, still nothing. Thanks for the suggestions though.

[polar spade]

A negative could point to a positive. No change may simply mean its Server side.

[opal orbit]

Hopefully, haha.

[neon pewter]

can someone help me with the studio?

[vestal beacon]

Need some help. At 2min 51.712 (ish) mark, there is a crack in the track. Tried repair in audacity, but it's too large. Tried replacing but everything sounds a bit off. 🙁 https://suno.com/studio?project_version_id=3f5f4442-cbb5-4a2c-b998-ab7fe92e5f4c

area right here:

[polar spade]

A couple of thoughts:

1. edit the crack by lowering the volume only.
   Then
2. go to New Edit UI or Legacy, Select slightly left and right ( about 6 seconds minimum) , click "keep duration" in the bottom cog icon in Legacy edit, or Fixed (below exclude box) in New edit, and regenerate?

Or

Remaster the whole... or...song section crop out that bit and remaster just the segment. Then lay that new segment onto a new track and adjust fidelity as needed?

[vestal beacon]

Thank you. I'll give it a try

[stiff sentinel]

i'm taking a look rn too, i'll share my thoughts as well once i'm done

[polar spade]

An issue or how to? Just post what you need help with, mate.

[stiff sentinel]

these stems are the ones from suno? did you check the Version 2 as well?

well i thought something like this would work since that section is similar to the few bars before it, but i can't see waveforms so i can't adjust it to be seamless 🤷

my next route would be to try with remasters as already suggested

[neon pewter]

how too

[polar spade]

How to use it? As in .... how to get started?

[neon pewter]

yeah i have no clue

[azure crane]

What happens? When I'm logged in, I don't see any songs at all. When I log out, now I can't log in...

[polar spade]

First, just drag one of the songs from your library list seen on theleft, drag into the larger workspace panel in studio and you'll see the songs waveform.

One of the features we use studio for is to layer different t bits and segments of the song to give the best final product.

So we might use the first half of one song. Then drag another similar generation and use the second half of another song.

That can be a first time project you can try.

[neon pewter]

oo i see

[polar spade]

The playbutton is at the bottom or tap spacebat to start and stop.

You have controls to set the speed, thats tempo. And a few other tools.

Studio is also called DAW , or digital audio workstation. That just means a place to splice and edit all sorts of segments and tracks together to create one song.

[neon pewter]

oo so i can make a nightcore style songs or?

[polar spade]

Its a leaning process for sure and a learning curve involved eith just doodling around a bit.

To be honest, you may want to go to the New Edit UI and see how that functions first, then apply that to studio as well. Just to keep things minimal in Edit UI until youre used to all of the tools.

[neon pewter]

oki i can try that :D

[polar spade]

Well, yes. Sure. But the Create panel is just fine for creating the nightcore style first-- fro the style prompts and lyrics. Then drag that into Studio for more mixing together anything you wish. Adding extra tracks of drums, or synth or even sound FX.*

Have you just purchased and started using Suno?

• sound effects can be made under "Sound" found top of Create panel , drop down menu from the Cover button.

[neon pewter]

i just subscriped

:D

[grave heron]

Hi I have Pro plan and I'm trying to figure out how to iterate my songs and get the best versions.
Last night I made a song and was making various cover versions and got one where I really liked the style of the music, but not the vocals.
I made an instrumental version of it and I love it. But when I try to add the lyrics back, it changes too much of the song too. Is there a way to keep it the same or at least "close enough" that I'm not aware of.
Also it's a hip-hop song and at first the vocals sounded like I wanted, which is more of a "conscious" hip-hop style, but as it progressed it sounded more and more like Eminem just shouting the lyrics, and I don't really like that style for this.

[polar spade]

Do a few baby steps first and play with just the Create panel. Click Create on the left icons to open Create panel. You put in lyrics, type in a style like "nightcore, dark, edgy" and click Create button. Thats all you need for just the first try.

[polar spade]

Sorry, im may also be using terms that dont make much sense at the moment. Without describing each button and step in small detail. But its a quick learn even by just noodling around and clicking things to see how they work. Youll get used to it pretty quickly.

[vestal beacon]

When you ask/want Suno to generate a few bars of a song, whats the best way to ensure it sounds coherent? No prompt? Same prompt as when you remndered originally? I'm struggling to get something that sounds similar to the song itself.

Maybe thats because I downloaded the song, tweaked it, and re-uploaded?

[brisk zinc]

How to make voice non-generic with pro plan subscribtion ?

[vestal beacon]

All help greatly appreciated

[ocean jetty]

Also curious.

[polar spade]

I would lean towards putting in the same prompt rather than leave blank. Otherwise, it will use the sounds seconds to the left and right to interpolation what it should do.

[polar spade]

Did you "insert a selection"? Maybe not do that.... but....

You may simply try to drag and highlight the selection from about 3 seconds left to 3 seconds tight over the click area.

[ocean jetty]

I find sometimes when you just want make a vocal change, just a few lyrics for examples. When the Vocal is Set, sometimes it has instruments.

[vestal beacon]

Not sure what you mean here. What I did outside of Suno was cleanup some slight issues with loudness and looping using ffmeg/claude.

[vestal beacon]

thanks. Yeah.. using no prompt a few times now and the results are very strange -- so at least need something in

Sort of wish it could suggest a prompt. Have a feeling that the original prompt is not ideal for rendering a section.

[polar spade]

As for what I was saying:

I was asking you to only drag a highlight across the small section -- my mind is in Edit mode.

So this would done in Legacy or New Edit UI for me. Not Studio. Though, it can be done in Studio.

OK. I would do best to be sitting at my desktop while making these suggestions.. . sorry.

Ill run in there and see whats available in Studio.

[polar spade]

So. in Studio. I selected the track and made 4 substitute or replacement tracks to choose from. All sound great and blend in perfectly. I'll now try to make a click or glitch in a track and see what it does to "fix" it with a replacement.

[stiff sentinel]

i've found the pro editor (not studio) does better at that

though i haven't been able to test things thoroughly on 1.20

[polar spade]

Seems quite easy in studio. Heres a video where I made a glitch then used Studio and Replace to make a couple of samples to replace. I also attempted replace the Lyrics as well, though that wasnt as clean so ignore that -- i should have just used an instrumental but was curious about lyrics as well -- but you can hear the audio of the guitar was blended nicely and the glitch removed with just two outputs. So in 4 or 6 outputs samples one would likely sound very good and solve the problems .

You would choose SONG and add styles or not, and no lyrics for it to blend the instrumental... and hope this removes your click.

Whats this?

[half dragon]

No details = spam

[grizzled finch]

It's a bollywood cover song a lil try just wanted to let u guys hear it
Small disclaimer = listen it at low volume.

[polar spade]

I see. Yea, we are all proud of when we create something new.

Heres the thing. We should just use suno links in here to keep the subject focused there. Having just a song and saying we made it with. suno doesnt help as well as to go to the song page to see the evidence and skill.

And being a Suno-School channel , its not often used as just a place to post songs or even critique songs.

but also, just noting we cannot use Suno to cover an existing piece of work, especially by another artist.

Anyway, I hope you are having fun with Suno.

[vestal beacon]

Is anyone else having UI issues?

I can't even find the extend button here..

[half dragon]

Under that Cover drop down... you'll see it - nvm that's already shown lol

[polar spade]

i see the word Extend there to the right up top

[half dragon]

I think he meant the "create" button

[vestal beacon]

yeah..the bright button that normally shows up and I press >< Think it's create

[half dragon]

The Extend option is there but no button to create it

Mine still shows.. so maybe some cache issue on your browser?

[polar spade]

oh yea... well.. ummmm... its not there in your screenshot... true. I see it in my panel.

[polar spade]

Im noticing some odd things... yea .

[grizzled finch]

If it's really that way then I apologize for my act of the same. Didn't had the intention to disrupt the harmony of SUNO

[polar spade]

No worries. Just bringing you up to what the channel is all about. Enjoy!

[grizzled finch]

Thanks bud

[polar spade]

Is the task bar supposed to be up here in the corner hidden behind Studio?

[stiff sentinel]

mine doesn't have that, but they're doing some UI stuff today, for a few minutes nobody had a "create" button lol

[vestal beacon]

update - to resolve, had to clear suno.com in browser cache. thanks all.

[half dragon]

Yeah it seems messed up for me too... the side menu bar appears on top, then hit the arrow on Suno Studio to close it.. it's disappeared altogether

[polar spade]

OK... seems a few things are going wonky. That means an early dinner for me perhaps. Just wait it all out.

[frozen dagger]

I was at work earllier using the web browser on my android phone to listen to music and the UI shrunk where I could brealy see anthing listed in my library. Suspecting more UI changes are being made and it's causing some trouble.

[polar spade]

oh.. did you see the video.. any of it make sense?

[polar spade]

we can click the layout button top right to bring it back, in order to get the Library if needed.

But also , if I click Layout and go to the "other" layout then placing my pointer over the Create or Library to minimize gives me a block icon. Oh well.

[half dragon]

It'd be fixed soon... 🤞

[polar spade]

I didnt notice this but Radio and Explore have been move down below to Labs

[opal orbit]

That's good to know, though I think I'm still missing some access, haha. Clicking on labs brings me to "Genre Wheel" and "Live Radio". Unless Genre Wheel is the new name for explore

[lime snow]

Yes, mine won't bring up any of my songs at all, all day.

[stiff sentinel]

anyone know of a tool that can make my voice more female-like without changing the pitch? to help suno cover a duet i'm working on

[frozen dagger]

use the male persona but in advanced options select female and prompt for female vocals.
This in theory sounds like a good bet. Similar to yesterday when we made a song singing in spanish, turned that into a persona, and then used the persona to generate a song with english lyrics but have a spanish "accent". That worked pretty well. Would think similar method to achieve the similar results. Not tested that to see how closely it resembles the previous voice. Just did it to switch vocal genders.

[gritty mulch]

there are voice model/swap programs like audimee and controlla but i'll be honest i think they're kinda butt. i just sing into studio and tell it to cover me as female vocal and it does nice work. you don't even have to be a great singer (source: me, an incredibly mid singer)

[stiff sentinel]

i meant my actual irl voice

i know what i want the melody to be like so i'll sing it manually

but suno will definitely get confused between who sings what when covering if both male and female parts are the same voice

if i could turn my own voice into a female, i could record the duet with the correct roles and suno's cover will do better at keeping it correct

[gritty mulch]

if you have studio can you put them on different tracks? you can cover the parts meant to be female on their own

[quiet solar]

Sort-of. Just to set expectations, I roped in a friend to try out something along these lines - it was possibly a little bit more accurate than a pure prompting approach, but Suno merrily did it's own thing with who was singing what.

If you have a melody worked out in advance, what you might find interesting to try is singing the entire thing yourself to get the benefits of audio input steering what you want from the vocalist when you cover it with a male singer. Then cover that song to make a female version with a high-ish audio influence and splice together the vocal stems as needed (assuming you don't need the two to harmonise).

[stiff sentinel]

ah, okay. I was going to try the role of your friend myself in this case, if i could find a way to female my own voice. (and i'll still try if somone knows of a tool that can do so)

[stiff sentinel]

this is exactly my plan B should the myself-female-voice thing not be plausible

[gaunt raven]

How do you make sure it's 2 diferent voices who sings cirtain lines?

[stiff sentinel]

the plan b is the guaranteed way, i will have a solo male version of the song and a solo female version of the song (yes this means they will sing parts that they don't need to)

and i will choose the lines myself and put it together in Studio

[gaunt raven]

ahhh got it ! smart.. but i need 2 males diferent voices.. lol

[stiff sentinel]

you can do the exact same thing

have a Male A version and a Male B version and then pick and choose yourself

[grim glade]

I'm having trouble with prompting / negative prompts. I'm trying to prevent my female singer from belting/yelling lyrics, but she keeps doing it. I've included "yelling, belting, shouting" in the negative prompts, and have included in the main style prompt that the voice should be subdued, whispering, etc, but it just keeps starting to get super powerful, especially in second choruses. Any ideas?

[quiet solar]

If it's useful, there's more info on prompting for two male voices at the same time, and workarounds for when they inevitably don't sing the right lines, here: https://discord.com/channels/1069381916492562582/1466598465969590465

[quiet solar]

In most genres, Suno will tend to follow the format of typical songs that follow a softer section with a louder one. Negative prompting isn't usually enough to overwhelm genre influences for something like that, so what you want to try is putting more of what you want in the Style prompt - not necessarily just about the singer, but for the song as a whole.

You might want to try adding words like this to the style prompt: "introspective, static dynamics, pianissimo, soft, relaxed" and if there are problematic sections of the song where it typically goes more energetic, reinforce those by tagging them in the lyrics as well, e.g. [Chorus - soft, introspective]

Unless you're doing a very soft, acoustic piece in the first place I've always found volume dynamics quite tricky to control though. So you may need to try something like creating the song using multiple Extends - basically, every time it goes loud, extending the song from just before it and getting Suno to try again with more focused prompting to be softer. Repeating as many times as needed until you get to the end of the song, then Get Full Song and do a Cover or Remaster to smooth out any oddities caused by the process.

[grim glade]

thanks for the tips. i've also found the negative prompts tend to not really do much, so i've stopped really even bothering with them. good call on the lyrics tags, i think that could help a lot, as they tend to get followed pretty well

[quiet solar]

I find lyrics tags tend to be a little unreliable, but they're still the most effective way to attempt to give a steer to a specific part of the song.

I don't use negative prompts all that often either. Like most AIs, Suno is much better at responding to positive prompting, so it's usually easier to explore a fix there in the first instance.

[rapid bobcat]

What is the deal with suno and inline prompts? Is there not a way to get it done reliably? I generate a song..it's like 95% of the way there, so I want to do a few more generations without changing a thing, and it just goes off the rails, like I generate a song that has (lead) in it with 10 generations. No issue, then from 11-12 onwards... it sings the word (lead) every single time, generation after generationd espite doing fine for a 10 generations first.

[hoary mauve]

Thank you. I will give it a try

[quiet solar]

It's the nature of a model based on probability. The problem is that while you know what you want to keep and what you want to change, Suno doesn't. And although you can influence what it does, it can't be precisely instructed, so it will inevitably make variations you don't like as well as those you do.

If you want to make really surgical changes, you need to start bringing in aspects of editing as well as prompting and regeneration into your workflows, as that can be both a faster route to the same end result and one which ultimately involves fewer changes - albeit in exchange for taking more time.

Where you run into heteronyms that Suno has to try to infer by context (e.g. live, lead, wind, close) it's a good habit to express them phonetically (e.g. lyve vs liv, led vs leed, winned vs whined) or to use punctuation to steer the pronunciation (e.g. l'ive vs li've, l'ead vs lea'd). The more you use Suno, you'll also start to spot other words that it finds difficult and learn to head them off at the pass.

[wintry saddle]

or short extend sections, so suno doesnt have too much time to deviate

that said - if you ask for something suno barely understands / cant do the chance of it doing it will always be low to zero.

like you can get a cannon sfx, or thunder, or rain - but you will burn credits waiting for suno to actually do it.

realistically everything in the input (even the lyrics themselves) interact to form the final result -> so the more you have the wider the final potential cast net will be... a fun example of this back in V3.5 I tried to write an ironic BardCore 'rap song' and Suno could clearly tell the lyrics where 'rap' and refused to use the bardcore instruments (which where easy to attain in a none rap song!) and switched the whole result.... basically overriding the whole genre and even any meta suggestions!

[slender glacier]

So my Playlist is missing a song, should be 20 but theres only 19

Check out my playlist made on Suno! https://suno.com/playlist/17a03a19-671d-48c5-88e3-2bad52627c61

[quiet solar]

Looks like there's one stuck permanently loading. Is it possible it's been unpublished by the creator?

[vestal beacon]

When you break apart a Suno generated song into stems using Studio, and then you sort of edit it and put it back together and export it. Does anybody notice a loss in quality? If so, how do you get this back? Do you have to use remastering? I haven't been successful with that yet.

[fickle heart]

Hello everyone, I need some quick help: How can I set the Suno interface to English? Currently, everything is in German, which is confusing me.

Okay, it's already been taken care of. It was the automatic translation in Google Chrome. 😉

[ocean jetty]

When creating a new track, and choosing guitar, with a simple style prompt - it also adds other instruments? I.e Kick. Which really clouds up the mix with having two kick drums.

[vestal beacon]

Noob Studio question: In my bass track, there's a section that was really poorly done, so I cut it. It's better, but it would be great if there was bass in this section. My problem is, whenever I generate it, I have to provide a prompt, and no matter what prompt I provide, it sounds like it's steering it very far away from the rest of the bass track. Any help here? I tried guessing the instruments, that doesn't seem to help. All help greatly appreciated.

[zealous solar]

What is thedifference between V5 Pro and Studio?

[sleek stump]

@keen crescent As Dasvon noted in Ticket Help, this is not a model issue. There have been a few discussions going back to the middle of last year about this, Here is one back in June that you can review.. #📚┃suno-school message

[zealous solar]

Are there constant changes to Suno V5 Pro? I have lost four songs to trash again. This was not recognisable. The beat was wrong, the music was bad, and then two of the tracks became a terrible screeching interference, with no accountability for such a strange sound. I am completely perplexed by this. One minute it works just fine, and the next it goes bonkers. Has anyone a blinker of an idea how to override such terrible results? It looks like the Studio has very similar problems, too!

[sleek stump]

This is something that happens with generative AI platforms, not just Suno, but all of them. You should not expect to get "your" perfect songs every time you push the Generate button. Some folks spin the wheel dozens of times before they get something they will work with. Additionally, there are issues when using extreme slider settings, or certain settings with Cover and Inspo.
Another thing to consider, the v5 model is still just a beta model, and is likely getting actively worked on and updated in the back end, but for the most part it should not deviate massively unless there is a bug.

[quiet solar]

I'll repeat my response to you from yesterday - if you can share links to a couple of the songs, it would be a lot easier to help if it was clearer what you've done in terms of prompting and settings and what is going wrong.

It's definitely the case that sometimes the Suno models churn out something off the rails, much as ChatGPT and Gemini will sometimes throw out an image of a person with too many fingers and a disembodied head leering over their shoulder. But those types of issues are typically occasional rather than constant.

[neat nexus]

V5 Pro allows you generate stems, can edit whole track in editor.

Studio allows you to generate full stems, edit them in a DAW like system in the browser, can generate within that DAW.

Since I honestly am still only scratching on the surface of Studio myself, others who use it more will need to explain its capabilites more.

[dusty coyote]

Hi guys. Does anybody aware of a way to download multiple songs at once? Like a whole workspace or selected files?

[polar spade]

One obvious difference it that Studio allows for sequencing different/multiple outputs in one project.

Pro can only work or edit within one output.

[polar spade]

No. Its not been foreseen to happen. This helps, among other things, to avoid the ability to pirate or farm songs.

[blazing crystal]

Is there a way to get two Personas to sing a duet?

[half dragon]

Possible.

• Generate one with one Persona - then Cover with 2nd Persona - then cut and mix them in Studio or DAW
• Generte the song using Extend and generate each section using a different Persona - then get whole song.

Either way, it will take many tries to get it the way you want, like anything else.

[blazing crystal]

Okay, thanks! I don't have access to studio, so I may have to go with the second option.

[polar spade]

Pro version: separately extend every verse three time with the prompts and persona added of male or female. Splice out the verses you dont want.

For alternating lines, in one verse, at least first try to just prompt with those tags which lean towards a male/female duet and hope the stanza you want with alternating lines is produced in one of the outputs. Usually that tends to be the third verse, bridge or penultimate chorus.

[wintry saddle]

an older school crazier 3rd option is to generate the same song with the 2 personas, stem the 2 vocals, and blend them together -> you can do that offline and upload it again... you can actually get near 4 part harmony going with the uploaded sample vocalists -> WHICH you can persona / extend 😉

well ~4 channels of vocals (as suno will sometimes put many vocalists together as a group on one channel)

pre v5, you could have 2+1 generalist channel for vocalists.

but that all changed with there big updates back at the start of 2025ish

[blazing crystal]

Working on a DnD themed album, so I have personas built for each character. It's going to take some learning to get them in songs together...

[fluid wren]

Are you using cover or replace

[cyan jewel]

hey i've got a track that I basically just need to slot an interlude in it between a chorus and a bridge. what would be the best way to make this change?

[honest hinge]

Hello, can anyone advice, with the update the persona creation for my own voice is disabled as it counts as uploaded audio, or even recorded audio

[half dragon]

Maybe in Editor, make a cut then insert ? same in Studio if you have access

[cyan jewel]

so im in studio rn. i'm not sure what im doing though

in the editor i originally tried cutting the song to where I want to extend

and then extended from there

but I wasn't getting the results I wanted exactly.

[half dragon]

Same thing , cut where you want to insert, then add a new track, highlight the section, then type in your prompt for your interlude [Instrumental Interlude] then generate
Listen if you like or not, if not, keep generating...

[cyan jewel]

so like im highlighting it

but it feels like its not getting context

it feels like im doing something wrong

so like this is what im doing rn.

I want it to have context for the instrumentation/cadence of this particular section so that I can blend the chorus, spoken word interlude, and bridge together

[slender glacier]

Ok, I like how a song turned out, but.... I want to change a cpl lyrics and take an instrument out...whats the best way to do that all the while keeping the same feel, tempo and sound of the song?

[half dragon]

To take an instrument out, extract all the Stems... and locate the stem for the specific instrument and hope it doesn't bleed into other stems , then remove it 🙂

[slender glacier]

Stems? I dont know anything about that, and I guess what I meant was replace an instrument, like take out the cello and instead have lead guitar

[half dragon]

Stem extractor is a tool Suno offers to let you seperate out the tracks from the generated song. You'd have several instruments tracks and a vocal track... or you could just seperate into 2 tracks, one vocal and one instrumental

[fresh knoll]

You have experience in studio? If not there's some videos. I generally spend 6-12 hours on a major project like that

[slender glacier]

Ooph

Lol

[half dragon]

Right. Replacing an instrument is not going to be a simple work.. it'd probably take a while but with Studio, same concept, locate the instrument stem you want to replace, hightlight it and Cover or replace with another instrument... and hope for the best 🙂 might take several tries... to get what you want.

[fresh knoll]

To even explain it they need to be familiar with the tools.

I think this was the one that is that https://discord.com/channels/1069381916492562582/1468681865648144434

[slender glacier]

Ok but cant I just try remix or cover or something and just change the lyrics and the song description while maintaining the style?

I'm still learning

[half dragon]

Yes you can try to Cover and keep Audio Influence slider as high as 85-90+ to keep the melody - don't change the style, prompt for that specific instrument.
See if that gives you anything better. You won't get the exact same arrangement though so keep that in mind.

[slender glacier]

Right on , i justdo t see what the difference is between remix and re-use style (keep in mind i mainly use the app, and desk top when I want to keep a persona)

[nocturne flower]

Hey guys could someone help me change just the outro of a song? I like everything but the outro, how can I change just that

[fresh knoll]

It would be a different song still. If you want the same song, you have to studio it. If you smash the create button it's new. Otherwise everyone would complain the gen the same song over and over

[half dragon]

Reuse style & Lyric is just that: it'd copy the style and lyric to your Create panel
Remix is just a generic term for Cover, Add Instrumental, Add Vocals...etc..

[fresh knoll]

Crop it out. Hit the splice where you want it. Then drop in the new outro onto the timeline

[half dragon]

Either in Editor, Studio by cut and regen
or Extend, by choosing the extension point a bit before the existing outro, then generate that outtro section again... regen until you like, then get whole song.

[nocturne flower]

Editor worked just fine, thank you sir

[nocturne flower]

Do we spend credits even if we crop in studio?

[acoustic egret]

whats the best settings to use consistantly?

[half dragon]

It depends - there's no one settings work for all songs.
Audio Influence - higher to keep the audio more like the source audio, lower will give Suno chance to change a bit
Style Influence - same higher will stick to your style prompt.. lower will let model have more freedom
Weridness - is weird 🙂 it'd affect the performance.. or even the arrangement... you'd just have to play with it.. more weird? crank it up. lol

[acoustic egret]

ok thanks so basically just tweaking it and pressing "create"? lol s

[half dragon]

Yep... you got it.

[nocturne flower]

Remaster doesnt use credits?

[half dragon]

It does.

[nocturne flower]

should be very little then

but Studio eats credits and I dont know why

I just went to studio to crop 1 part lost 80 credits

[half dragon]

I've never paid attention as I have so many.. but someone did say almost everything you do in Studio costs credits!
With Premiere, I have plenty to work with - and still have plenty left over at the end of the month ...lol

[fresh knoll]

50/50/20 unless you want something unexpected or different like more creative

[acoustic egret]

so basically the default setting

im after something exciting with amazing arrangements and banger riffs. i find when i put it on high expermental/chaotic it glitches and you peopole having conversations in the background lol its weird af

[fresh knoll]

Yes, but best is a perspective, if it came out bad try default. If it's still bad, then check prompt style and lyrics for possible cause

[fresh knoll]

Add [guitar riff] in lyrics where you want it. Sliders won't do that

[polar spade]

If you cover and add a persona, the audio slider affects both the audio covers and the persona used.
The style slider also covers the effect of the lyrics, which control , in part the style and feel of the piece.
Suno, also tries to vary outputs as a matter of giving us the best of options.
So, tweaking and pressing "create" is the plan. 🙂

[polar spade]

The closest to "enhancing" in Pro is trying Remaster and High. This tries to vary the instrumentation along the way while giving a good mastering to the fidelity. It's often not my choice as I just want some warmth or clarity. But certainly try it a couple of times to see your thoughts.

[slender glacier]

No i didn't unpublish anything...bummer

[acoustic egret]

So your talking more about the "mixing" rather than the arrangement right?

Btw, is there a way to add multiple audio samples and tell SUNO to write me a song that incorporates all the audio uploaded?

[sleek stump]

Unfortunately, nope. Its one file only currently.

[sleek stump]

It may actually have slight arrangement variations.. They can be really really close or you may get an outlier thats just bonkers different.

[midnight mason]

so, Suno generated this. This is a REALLY long set of instructions and, invariably, the vocalist starts SINGING these instructions. What do I replace this with that will just make this section a long instrumental? Without messing up the entire song? I tried remix but I can't ever get the same song.

also for some reason the beta editor duplicates my song's bridge???

[midnight mason]

I CANNOT make heads nor tails of this beta editor. It is incomprehensible 🙁

is is an issue with trying to use beta editor on v4.5 song? So for example it will bleed sections into other sections, duplicate sections on the LYRICS page.... feels impossible to make headway into fixing anything

is there an easier way to just

take a song I already have

and generate a nearly identical song but with the lyric changes I made in the song info/displayed lyrics?

I am trying remix now with audio influence set to 100% and all other influence set to 0%

doesn't work, generates an entirely different song....

[polar spade]

Making a Cover in 5.0 has been doing well for minor lyrics changes only. You can add "live performance" in styles promt box to give it that stage sound and audience response.

[midnight mason]

every time I try to cover it will like, completely change tone, vocalist, instrumental, or just make an entirely new song, I don't know if this is a slider issue or what

[sleek stump]

You can get a 90-95% cover by setting the sliders to Weirdness: 20 Style: 50 Audio: 85... and delete the content of the styles window. This is how I do my lyric edits using Cover.

[midnight mason]

I'll give that a try

nope it's still making completely new songs

completely different vocalist, completely different instrumental albeit with the same melody

[polar spade]

So. Here is the original👉 https://suno.com/s/aJGsuDnFRmFzow7S 30 seconds

👇 Here is clicking "Cover" , leaving the Sliders at default and adding "Live Performance" in the styles.

I change two lines in the pre-chorus to "This is the line I replaced" and This is all I wrote"

So, is set up to get a close iteration with vocal character similar, instrumental, not change the genre or tone.

Here is the live Cover version.
https://suno.com/s/Ng7YZu5fH9KQhF6C 41 seconds

The live performance changed Suno to also change the key. Not that it should or shouldn't but there was nothing to tell it NOT to. Same might happen if we change any part of the style. But if we only use Cover and only change some lyrics the idea is to keep it all the same and just swap out the lyrics, provided they are almost dead exactly the same syllables and meter as the original. Otherwise, suno has to re-tokenize or re-assess the lyrics in that swap and may need to change the melody to accommodate the new lyrics edits.

[leaden lake]

Lowk what I imagined suno V6 to
Sound

[raw inlet]

The other day I had this error message and tracked it down to my style prompt simply containing the word “stab” as in synthesizer stabbed chords, lol. So many precautions these days to make things safe from cyber bullying I suppose

[fresh knoll]

Yeah that happens with a few styles not just that one

[midnight mason]

why does "replace section" just straight up change the vocalist every time??

I'm trying to make a simpler edit where I change some lyrics but now "replace section", while cleanly making the edits to the section and fixing the lyrics, generates a second, completely different vocalist from the rest of the song

[stiff sentinel]

use the lyric replacer to replace lyrics, in the normal editor (not studio)

[midnight mason]

I don't have studio

I have been using the lyric replacer

so I'm trying to replace that highlighted section

I select it and click "replace" and one of two things happen

it gives me garbled nothing / replaces the vocalist / skips the lyrics entirely

so I tried to just replace entire sectionb it always replaces the vocalist

[midnight mason]

so, now it's mostly working. except it can't generate a section that pronounces the word "ancient" correctly.

"Through a cracked screen AN-INT code" ???????

I'm completely perplexed. Every word I try to replace on the second line the AI will just garble

"Red alert
Through a cracked screen
An-CHRENT code

Red alert
Through a cracked screen
Dat-CHRED code

???????

What I'm trying to do is basically just de-sloppify my opening verse. it was a quick song I made with basic walking to town that was rad except very plainly generated an uninspired opening, rest of the song is great. remixing isn't doing what I want it's just making a new song completely

on god it's doing it on purpose. it just suddenly cannot pronounce any word, ANY word I slot into where "ancient" goes

[stiff sentinel]

huh ok weird, i tested it myself right before i mentioned it here, it took me two tries but I was able to change the lyrics on a recording of my own singing (changed "square" to "star")

[midnight mason]

from what i've read everywhere I'm doing it right but for whatever reason it's completely resistant to pronouncing anything right for me

I'll get "ancrient" "an-sea-ant", just.... everything except the right word even if I try to spell it phonetically or something

[polar spade]

Its a less than three seconds section. That will be quite difficult for it to change any word placed into it. If you can highlight at least the three lines "Red alert.... Code", but in any case, 6 seconds is easier for Suno to work with. Or, try using FIXED instead of SMART. Fixed is for short selections.

Secondly, its doing its best to squeeze "Ayshn'tkoad within that 1 second interval and likely just cant no matter what you try. That can happen. One way is to shorten the words before, so type "threwa kraktskreen" as the previous line, to see if it "lends" some extra "tokens" or space to the next line Ancient Code.

[midnight mason]

tried all of that, it will pronounce everything perfect until I get to "ancient" and just mock me. "anwehn, "ankchsen"

just anything but "ancient"

even with the replaced section bumped to 7 seconds, even on fixed, even doing phonetic right before it

https://suno.com/song/d2d35cac-26e3-4eda-b711-c6e4665e1f57

here's what i just got

like no matter what I do it will not budge on that word, even if I try a synonym or something

i've gone through like 80 credits now probably

the original line was
"Every cracked screen
Every code"

which is borderline nonsense so trying to unslop it a bit but just

can't figure it out

[brisk zinc]

Can you add vocal later if managed get nice music with pro plan ?

[midnight gust]

Has anyone had any issues with tempo change when downloading stems? I have a song that is 109bpm. About two-thirds through, the song goes to cut-time for a few measures and then returns to standard 4/4. When listening to the stems before download, everything is perfect. Once I download the stems, the tempo becomes much slower in the begining, then when it hits the cut-time portion, it becomes a runaway train, getting faster and faster, all the way to the end.
I've wasted 300 credits trying to fix this and am lost. I've locked tempo to track and it does not help.
Anyone ever experience this?
Were you able to fix it somehow?

[polar spade]

Did you use this step before downloading stems? https://discord.com/channels/1069381916492562582/1468681542913101865

[midnight gust]

I have not.

I'm using StudioOne but the tempo change is even prior to putting them into my DAW.
Literally, playing the wav files thru any media player shows the tempo changes before putting them into my DAW.
But I will check that out and see if it gives me any pointers

[eternal sedge]

Hi all, still being a Suno beginner haven´t really edited something to improve it yet - i do have a line with a "vocal glitch" i like to regenerate WITHOUT changing the music behind ... is this something i´d only be able to do in Studio? Edit feature doesn´t seem to work here?

[midnight gust]

It's actually not tempo drift. It's literally changing from 149bpm to 74bpm during cut-time in "follow track mode. If I set it to manual, the first part is fine, but the cut-time portion is then double-timed and continues to the end. So, once again, I'm at a loss here

[keen crescent]

Hi Team, can you use more than one Persona (Male & Female) in a Audio track or can you only add 1?

[polar spade]

Starting with a free test, What's the full song do on download?

[midnight gust]

The full song plays fine upon download. It is only once the stems are extracted and downloaded

[polar spade]

Yea. I wouldnt know without just experimenting.

But rather than extracting stems in Studio, do the extractions from the output menu. Unfortunately, downloading in Studio, we cant seem to pull up the stems in the Get Stems module, but we can Get Stems and then use Studio and find our stems already available in Studio.

By the using Get Stems module, we have an additional choice to download stems as "WAV- Tempo locked".

Tried that before? . Its a possible solution. Im not to where I can test. Or frankly, create the issue to test as a solution.

[jagged matrix]

How do you guys manipulate and edit the stems outside of suno? I use Logic pro stem splitter and some other midi tools but the hardest thing ive come across is getting sounds similar to the once you get from suno.

[stiff temple]

Suno Master needed. I created a cover that I like a lot. However I need to change only the lyrics in the last verse. I have tried the extend remix and cover options, but the original sound of the cover is never preserved.I'm wasting a lot of credits...advice?

[jagged matrix]

Have you tried something like "Preserve everything" and in the lyric editor just changing the last sentences?

[polar spade]

You're Pro and not Premier, then?

[stiff temple]

Never heard of that trick. I put that in the style box (which is empty), adjusted the lyrics. Didn't work.

[polar spade]

Not to be a wet rag, but the Cover feature is really good at doing just that. Even with sliders in default position of 50/50/20, if i Cover a song and do little else but change some lyrics on the end, I get , generally, the same thing, with only slight variations.

So , as always, im going to test one to prove me wrong. 😁 Hold on.

[stiff temple]

That's what I thought too. But maybe because its a cover of a cover?

[stiff temple]

Maybe I should stop covering the cover and go back to the original.

[jagged matrix]

Would you then leave style box completely empty and just change the lyrics or have the same prompt that you started with?

[stiff temple]

I tried the style box empty or with same style as original...neither worked

[jagged matrix]

Have you tried the extend function?

[stiff temple]

On the cover yes. Style didn't stick...I guess I'm going to have to revert to the original and keep spinning it until I get something close to the cover I liked.

Cover of a cover problem was not something I anticipated

[jagged matrix]

Do you have a daw? maybe you could stem split, remove the vocal where you wanna change, put it back into suno and add all the text.

[stiff temple]

I get an orange suno screen when clicking your link. No song. Am on android.

[polar spade]

Ill make it public. Maybe that helps.

[stiff temple]

DAW? Hadn't thought of exporting stems....🤔

[polar spade]

Link work now? Not much to show but a cover sounding like the original.

If you really like the cover-of-the-cover and feel the chain of outputs is affecting somehow, you could download and upload the wav file to have a freshly uploaded audio as well.

[jagged matrix]

It worked for me. So what did you do? You chose cover the track you wanna change and wrote the new lyrics?

[stiff temple]

No, there seems to be some sort of authentication problem. It doesn't know that I'm already logged in. Hey, when I
Try to use my google account to log in it.Just spins... Can you describe in simple steps?What you did that worked for you?I'll try again.I'm up for anything

Thanks for even taking the time to respond

[polar spade]

Yes. No magic. I just changed the last line of lyrics. COVERED with Audio slider at 20% (which is default) but actually I think I had it at 18% in the rush of putting together. You hear the results. It does start with a female voice, on the first line. Thats about the only difference, then sings the new lyric line at the end.

[jagged matrix]

I see. Do you only use suno or do you export tracks into other programs anything?

[stiff temple]

Never exported yet

Tad frustrating, because i'm so close I can almost feel it. Lol

So you made a cover of a cover?