💡┃suggestions-web

[empty isle]

add before would be huge... or add from - to... so if I want electric guitar solo in the middle of what suno generate, or rewrite some part that i dont like, i still can do it

[calm vapor]

it would be great if you would enable the custom mode by default with lyrics (who does a song without lyrics copypasted there??)

i saw some AI on a suno facebook group, all it could do is to make songs by prompt... what is the use of that o.o

[daring raptor]

guys why after 15 sec music eventually starts, even if i put [no music]
[spoken dialogue] to ganres and before every column 🤔

[verbal remnant]

Other random things I've found myself doing ... I will sometimes load a finished song (or pieces of a song made incrementally) into a tool like Logic Pro so that I can chop things up (tempo finder helps) or do some mastering. I can easily picture a way that as pieces or chunks of a song are generated just having a linear copy/ paste/ move (treating each piece as a loop of sorts) so song format could be done quickly. Also basic mastering tools like ability to tweak final compression level, EQ, adaptive limiter (maximizer) etc. I know this is all stuff that doesn't necessarily fit within the core ability to just create good sounds, but as this stuff gets good enough to make legit, listenable or near-pro level music the ability to actually do song composition like this will be huge.

Anyhow, thanks for making one of the most fun and inspiring tools I've used ... ever! This is probably the first AI-type tool where I've really seen the light for creating quality ideas that are basically workable demos for songs I'll go on to actually share with my band or rearrange and play instruments on top-of. I can only imagine if it could also be the other way, like where I share a WAV of my own melody or idea and the AI could riff around that... I'm sure it'll happen 🙂

[arctic whale]

How about the possibility of buying credits without a plan and with no subscription? I don't need that many and would get by with irregular purchase of credits. The Free Version is very limiting but the next one too restricting. Something in between would be nice (and may be incentivize users who don't want to commit that long and lose credits because they do not need them).

[granite star]

What I mean is I would be fine using credits to edit a generation. I would likely use less credits to edit a clip than regenerating over and over again trying to get it right. I totally get it can be pricey. I am happy to pay for a good program!

Another suggestion, maybe it has been added before. A tab on the generation side to tell the program you want an intro or an outro. That way the program knows the next generation should be a start or end.

[frosty breach]

I think there is enough space here for our Display name + redirect to our profile. It seems a bit convoluted, to get to your profile by changing the URL or clicking on a song of your own to get a link to your profile. While mainly a QoL thing, it just something I thought of.

[strong swan]

if you click your picture you see "view profile" this is how i go there (or by changing the url as you said). Another way is your username below the preview picture on the right side, if you have seleced any of your tracks in create or the library. BUT a seperate link down there would be nice 😄

[frosty breach]

It would also make it easier to spot which account you're on when using a/multiple paid subs in combination with a free or another one.

Also I forgot about that one, and.. it kinda supports my point

[woven valley]

It seems like the output quality is suffering from different background noise. For example, oftentimes there's this squishing noise, sort of sounds like a fluctuating swish. I've tried separating the song, but the separation software isn't able to do it very well.

[cyan edge]

Is it possible to add new Channels in Discord for the Different Music styles. maybe that you have to choose a role to unlock the specific Channel u r interested in. would be more organized and people can share their experience better with other people.

[strong swan]

like this? https://discord.com/channels/1069381916492562582/1154790975240155247

[fair cipher]

... Oh, I didn't know that was a thing. I hate how discord hides channels by default heh

[cyan edge]

yes, it was hidden for me. just popped up after clicking on the link

Edit: the Channel is visible on phone but hidden on PC 🤨

[proud copper]

There should be an option to reuse a voice that was used on a song for another one, or reuse a specific part and add it into a new song

https://suno.com/song/2960b395-8203-4660-9c31-a31f45286510

For example in this song I really like the part from 0:34 to 0:43

And the voice being used is cool too

[spiral acorn]

i wonder if this is possible... but if it is, it'd be amazing. Maybe there is a form of "inpainting" where it can keep the structure but tries to regenerate what it hears. No idea how possible this is.

ah yeah, like this

[stone solar]

i suspect that any form of inpainting will likely result in you losing elements of the arrangement you liked, instead of just a simple "upgrade". the likely longterm solution to "low quality", is to take the version you like, split the stems, rip the arrangement, resample it, and/or redo the vocals before re-merging them. if this sounds like work, its because it is... but its still significantly less than if you also had to arrange it the way you liked (although this opens up editing options as well for improvements)

inpainting is a more ideal solution for "dang, this ONE word got mispronounced"

[spiral acorn]

or for going over a section entirely too (i imagine?)

[stone solar]

if I didn't like more than say... 30 seconds of an extension, its rare that I kept that piece in the first place

[spiral acorn]

i hear ya! Though i think ideally it would be nice for inpainting to work well in these situations... and i at least expect it would

i remember Tony even mentioning that its hypothetically possible to move parts around etc

im sure when they first add it it wont have that kind of functionality but it would be a good place to work towards

[stone solar]

i feel like it could be helpful to have a copy/paste for Chorus/Hook, ... and then be able to "heal" the affected sections... but I'm also 100% ok if they focus on "improving the algorithm and expanding samples" rather than expanding functionality...

[spiral acorn]

yea long term that kind of thing would be amazing 💯

[primal quail]

Anyone know how to get a rod wave style song

[acoustic spear]

It would be really nice if there was a "seed" ID to each generation; so if a particular render comes out with a really cool groove or beat that you like, you can re-render that specific seed to get variations on it without loosing the general groove/essences of the original render

[cloud trail]

suggestion can make more long than 2 minute

[soft harness]

You can: https://www.suno.wiki/faq/making-music/how-do-i-make-the-song-longer/

[rustic heron]

Negative Prompts would be incredibly useful as a suggestion

[gloomy rivet]

I love Suno so much, it has allowed me to express my poems in song that i been writing for years
-I would love some control over ending album art
-Quality leaves on like part 3 or 4 of songs ( i still love it anyways beacause i know what im here for)
-I wish i understood how the list on home page work like trending and genres
(i made trending briefly and was thrilled)

[soft harness]

For point 1, you can prompt the AI on what to generate for the thumbnail art by entering the description as the final line after an Ending line like [End] (or three, preferably in square brackets as well to help try to stop the AI from saying thr description outloud)

#📚┃suno-school message (I show an example a few messages after in that link)

[gloomy rivet]

wow thanks, had no clue! I will try that later for sure. I forgot to mention finial lyrics . Its a mess sometimes after songs XD

[soft harness]

No problems. Have you checked out the wiki? I have a messed up ending usually when I forget to use an ending tag (though, even then, the AI can sometimes ignore it and hallucinate...)

https://www.suno.wiki/faq/making-music/how-do-i-end-the-song/

[finite plinth]

My suggestion again is for SUNO to charge $100 for any finished and downloaded song for professional use (Stems, Midi, wav files, etc.)
$100 versus hundreds of $ to produce a less than adequate song as those from SUNO as of version 3.

Artist fees, Drummer, Pianist, Bass Player, Guitarist, etc...
I've been waiting for this thing since the early 90s... Wow!!!

Those who want to create BIRTHDAY songs for family should be able to do so with less than STUDIO QUALITY.

[peak cave]

Have you ever used SUNO and got a voice that you just find to be perfect for a specific genre or character?

My recommendation is to make a type of voice saver where you can select a song to reuse the voice to make more songs.

It might be difficult or impossible to make but worth it.

[soft meteor]

You are aware there are no stems, midi, or wavs right?
Even if you pay them $10.000 you will not get them.

[robust jewel]

On top of that, the goal is for it to always produce good quality stuff.

[cyan escarp]

BAN

[tropic helm]

you need to rely more on feedback from people

it is necessary to be able to select seconds where it’s awesome, and so that he (SUNO AI) can generate new compositions in the same spirit

Well, it is necessary that in the same voice and in a similar style (absolutely accurately) he can sing another song (from poetry), from the one he liked (at least among those generated by SUNO AI)

P.S.:

and in general, it makes sense for AI to analyze what exactly people like and why, finding “hidden patterns” of music - and in the future it should strive to generate not even like people, but perfect music from Heaven (or the singing of the Gods),

(you won’t believe me, of course, but I remember a number of worlds from posthumous experiences with such music, and I’ll tell you that everything you’ve ever heard on Earth - in comparison with it, it’s just like rattling bones in a cave, in cave ages)

P.S.2:

Well, we are waiting for devices to appear en masse that allow us to read feelings, emotions, mood and the degree of pleasure experienced by a person - in real time,

Then, on millions of people, it will be possible to self-teach musical AI (like AlphaZero) based on the real response of their feelings, checking in real time the greater or lesser pleasure a person experiences from the smallest (and not so small) changes made in samples, etc. (ultimately, up to the control of every bit in a byte of a music file, so that it brings only maximum pleasure)

when this becomes possible, you will really be blown away by how incredibly awesome it is (so much so that it is now fundamentally impossible for you to imagine and even remotely imagine it)

[limpid panther]

https://suno.com/song/fc51bc3e-ca80-4702-8457-d919d318e243

[frosty breach]

It would be neat if Selection for trash and such also worked for Link only/Public.

[dusky plinth]

The option to clear the whole trash with one button press or select songs you have in trash and delete all selected ones at once. Not sure if Opti was also talking about something like this 🤔

[jagged charm]

I'm trying to purchase the pro plan of Suno AI, but my card isn't being accepted. It's showing as invalid.

[versed inlet]

Email [e-mail]

[jagged charm]

@versed inlet I've emailed them, but haven't received any solution yet.

[soft meteor]

Suggestion: Announce work on the service so we know when to do something else.

[tepid trench]

To be fair, they usually do put a banner on the site when there are significant problems, and yesterday they announced maintenance shortly before it happened. Something here on Discord would be good as well though (in a specific channel so it's easy to spot, rather than in a channel where discussion is taking place) - as soon as they are aware of issues.

[tepid trench]

...but as it's been over half an hour and there's only just now been a response from a dev (in the bugs channel), informing users that there is an issue being investigated in a much more visible place would be appreciated.

[finite plinth]

Not now...
That's the idea as a suggestion (Perhaps in a year or two)

[frosty breach]

I've suggested it before, but might as well pop it in here. It would be neat if we could customise that banner background with an image, perhaps choose a song's cover or even just adjusting the color.

[hearty star]

Suno could save money by self-hosting a Llama3 model using Ollama AND the service would improve because it's not Chat GPT ( Which is quickly becoming an acronym for George's Prophesied Totalitarianism). I'm not on board with this newspeak Orwellian netnanny nonsense. Music is about freedom, creativity, expression and the experience of the human condition. Imagine if Chat GPT moderated lyrics when music was created throughout history. Would it have allowed ICP? Nine-inch Nails? Beastie Boys? Gangster Rap? How about WAP? Literally we can't even generate Pornstep (which only has sexual sounding effects but no actual sexual words). Would we ever have Sexual Eruption by Snoop Dogg? Etc? The way I see it sure, go ahead, cripple your models - make it rated G if you want and it can balk at writing more mature content - but it should not and never should be allowed to stifle the free speech and expression of our nation. Freedom of speech is freedom of expression and music is literally one of the most protected forms. The fact that companies think they can run OUR lyrics through a sanitizer or even deny us the ability to use them is totally against what this nation stands for.

[fair cipher]

Honestly I've tried writing lyrics with llama 3 70b and found that most of them were stupid. It didn't want to write the lyrics, it wanted to chat about writing lyrics, and when I tried to continue to get it to finish a stanza it'd either start talking again or switch to a totally different song or write another version of the verse it just wrote.

So far I've had the best luck with Mixtral 8x22b but even it has trouble continuing to add more lyrics, it'll just rewrite the beginning in different words a lot of times.

[visual brook]

my go-to multi-turn conversation prompt to get an extension is

{LLM reply with lyrics}
Write an Act 2 to the lyrics.```

or

```{initial lyric prompt}
{LLM reply with lyrics}
Please extend the lyrics.```

also my initial prompt sometimes goes this way:

{LLM reply}

Please turn the poem into a song.```

[fair cipher]

I don't do a conversation to get lyrics. I have a custom prompt where I fill in things like the mood and genre of the song, themes or imagery to use, what the song is actually about, and sometimes a title (or sometimes I leave that blank and let the AI write a title).

For instance, this was one of mine.

[visual brook]

add the instruction to the beginning and end of the prompt:

{genre}
{themes}
{etc}

Write a Free Verse Poem on the above.```

good idea to add more fields, probably still have to emphasize the task

[fair cipher]

My AI instructions explain that it's supposed to generate song lyrics, and stuff like using metaphors and imagery and being creative.

[visual brook]

ah ok, yeah that's unexpected

a few more things to check come to mind, but gotta go eat

[fair cipher]

It's just because all the new AI models are trained on chats but I don't want a chat, I just want it to keep writing more verses and choruses lol

[hearty star]

I don't know if you tried creating a customized agent model with Llama3 but I have made a few dozen and my lyrics generator works a lot better than the one included with Suno. If you just use the default model with no tweaking it IS pretty bad - but why when you can make a custom agent?

[fair cipher]

I'm actually using a tool designed to write stories and play adventure games, but it more or less works for songs too.

Mine works well most of the time, it's just frustrating if I have to delete a few lines of a stanza because they sucked, because then it doesn't want to finish the stanza, it wants to start over at the beginning and rewrite everything using 20 more syllables lol

And yes I've tried doing (continue) and such.

[hearty star]

Personally I rarely just use generated output - especially since over time it seems like a lot of trash and extra words. It does happen rarely but usually it's like - umm no I'm editing this, and this, and this... What I found is there's kind of a trick; if you generate some music by itself without words, or music with words that sound wrong - I kinda go through note by note slowly and syllable by syllable making the words fit perfectly to the beat. Then when I regenerate that part it actually sticks to the beat instead of doing something crazy.

[fair cipher]

What I do is generate lyrics, edit to what I think sounds good, then see what the AI does with it. If it gets the rhythm wrong, I edit the lyrics some more.

And I haven't actually used Suno to generate lyrics for a while. I hate getting songs with perfect instruments and a pretty melody and all but the lyrics are trash. Starting with good lyrics to begin with removes one step.

(so yes I agree, you should be able to JUST generate lyrics, THEN plug them into a song)

But ChatGPT is quickly becoming obsolete and isn't exactly masterful at writing lyrics.

...I wonder how GPT 4o is at it.

[hearty star]

No Idea; I'm not into testing sketchy models that nobody knows where it came from or why.

[near cove]

https://chat.openai.com/share/ebd62836-0b97-4fb1-a257-9b67c2559178 tried it

[dusty spoke]

For what it's worth, I tried to feed Suno an Italian sonnet from the 13th century, and it was flagged for moderation.
"All right" I said "it's because ChatGPT has no sense of humor, much less dark humor".
I tried to transliterate the sonnet into modern English just for the lulz, thinking it would never, ever pass through moderation.
To my complete bafflement, it went straight to production. 😶
Here it is: https://suno.com/song/c752227a-b53f-416c-ab93-8507d13a5400

[cloud wadi]

I don't know if this has ever been brought up, but is there a way to control the chord progression for the songs? I feel like I was able to sort of do that, but it didn't fully Follow The Prompt I put in. It was supposed to be a song only using an A Minor chord, but instead it only started with an A minor chord before going on to other chords

[hearty star]

I have been able to make it say naughty things from time to time but the funniest is when you tell it to go instrumental on a rap and it freestyles random profanity on it's own - even after refusing to generate much less divisive words ;p It's used the hard N, just about every four letter word, etc - whenever it wants. But if we try? Oh no no no no no.

[hearty star]

Delycan; it's basically like this. You can suggest whatever you want but it's really only a suggestion. When you want finer control of the song you can put short instructions in brackets in the lyrics like [Chord in Gm] and it MIGHT do it within 20ish generations the way you wanted it. As it sits your best bet is to regenerate starting 1 note or beat before the change occurs then only take the one that follows the [Chord in Gm] path. Once on the path it will be easier to get it to happen again since it likes to structure songs based on what came before.

[shell flume]

I can't access the Suno wepsite, it says authorization is requrired, so I came to the Server, how can I use Suno Ai?

[cloud wadi]

🤯

[pulsar sparrow]

when you share an album it would be nice if it was like when you share a spotify playlit

[cloud wadi]

Okay, so I just took lyrics and format at them to be this way. I will see how many times it takes me to get the correct chord progression from a popular song

[cloud wadi]

Dude this stuff worked so good oh my god

[hearty star]

Have a look at my profile, the public songs link has 10 songs I made with good examples of how to prompt in various ways. Have fun 🙂

[cloud wadi]

Check out the version of House of the Rising Sun I just posted a few minutes ago in the Showcase thing. Not sure if it's the right chords, but I can definitely hear the cord change during the verse and it sounds like how the core changes in the actual song

[vapid dust]

Comments idk

[quiet lark]

idk if the @developers will see this. But ive got probably a simple suggestion. Is there a possibility we could get an option to simply remove the song from the creation screen but not actually delete the song. Like Ive got probably 15 songs or so, but I dont want all those on the creation list. Just on my playlist/song library. if that makes sense

[strong swan]

you can give the songs a thumbs down. this will remove them from the create tab but they stay in the library✌️

[quiet lark]

interesting but does that effect later productions. i generally stay in the same style of music.

[strong swan]

not that we know of.

[quiet lark]

okay ill try that then.

[quiet lark]

And how would i go about copyrighting one of these. is that even allowed or a thing. I write all of the lyrics myself

[desert hatch]

Please when a song gets flagged for moderation, could it at least make it so it highlights what could be causing the issue. The one chorus I’ve added went through perfectly fine but the minute I extend the song, now the same chorus that previously went through is now getting flagged. It’s just a guessing game at this point, it gets very frustrating…

[true umbra]

Voice Swap: Female, Male and variables of voice types. Often I get something I like, all except that voice and I think if I could just change that voice.

[glacial quest]

SUGGESTION: Download as MP3 should embed the tags for Artist, Title, Genre (the prompted genre), Date, Lyrics and Album Art image.

Lyrics in LRC format please.

It also would be fantastic if Suno could stem out the music as multiple tracks: vocal and music ... or better yet, vocals and instruments by each one separately.

[sage mirage]

Suno doesn't use stems at all, but you can break it apart afterward. Suno deals only in the master

[glacial quest]

Yeah, I got that. The GAN only looks for things that sound right all together... final product. But it should be possible to use multiple GANs to get a better or stemmed result. Imagine a "band of GANs" one for each musician. Has anyone played with this approach?

[sage mirage]

I don't know how Suno is built but I'd imagine doing it like you suggest would be an entirely new way of doing it, also with how you'd train it

[glacial quest]

It would be a different approach. Training would require stemming on the inputs.

And it would potentially be more compuntationally intensive.

[sage mirage]

Right but with how much data these things require before they coalesce into something

I wouldn't want to be the person having to segment out the training data

[glacial quest]

The existing training set would work... jjust needs some preprocessing to stem it and then train on the various stems.

It could be automated and minimally supervised, assuming the existing data is well tagged.

But I do make some big assumptions here 😉

Anyways, putting tags in the MP3 download should be a task that even a simple Devops person could implement in a day or so.

[toxic wave]

Would love to see a tool getting implemented where you can feed your own mp3 and AI makes a track based on that bpm and melody or something. Would love to feed it a track I'm working on to see what AI can do with it, would really help with writers block

[plucky wave]

⚠️ Please extend the lyrics limit of 1,250 characters ‼️

[sly furnace]

No more news for one month? Pleeease give us a hint what comes next!!!

[summer palm]

Generating from an audio file "mp3, wave", possibility to control melody an rythm with controlnet like feature that can also transfer music style from audio file… #inpainting #controlnet #music_to_music #audio_to_music

[soft meteor]

thats a hilariously bad idea.. so instead of stemming the output of the model you suggest they should train a new model on degraded stems... alright

That way can have bad mastered output and imperfect stems 👍

It's just not feasible doing it that way. because you would need perfect stemming tools and once you have that you can.. well just make the stems after creating the whole things.. If you want to train a model that spits out stems you need studio quality stems for training.. no real use in using anything low quality

[glacial quest]

Yes, getting the stemming right would be the challenge.

[unique imp]

Opinion? thanks -> https://suno.com/song/0cc6c710-8b47-4242-8266-d45978b13dc5

[last jacinth]

Voices sound completely AI since the beginning with the usual sound corruption in Suno

[worn atlas]

This works on desktop but not on mobile. Right?

[worn atlas]

So listen to this waiting for V4 song
https://suno.com/song/dfd8960c-4341-44ca-9db8-afe0f18a3254

[last jacinth]

The best is the miniature! XD

[wraith sequoia]

📜 Suggestion

It would be great to have the option to choose what type of voice can be applied to the song. For instance when making an album, it would be great to have the choice to use the same voice for all songs generated.

[quiet lark]

I still want to know if i can copyright certain songs, I write all the lyrics myself

[hybrid crane]

If you have a paid plan, you have copyright on the songs you make on that plan.

[wraith sequoia]

get all info on copyrights on the Suno Faq Page
https://suno-ai.notion.site/FAQs-b72601b96de44e5cacd2cd6baa985448#ee2b052053c74e92ba0054528ad9c54d

[sleek valve]

Hi Team. Would be nice if you can make Suno force starting the song with some instrumental intro. I am using the tag [Instrumental intro] for all my songs, but most of them Suno starts singing from the very first note.

[quiet lark]

add "with intro" into the style of music

[hybrid crane]

You can. Maybe my video will help you: https://youtu.be/lj8utBd_IUM

[quiet lark]

i always Add "southern rock with intro" into the style of music and i always get a good intro

[tidal isle]

Make that suno don't ignore whole verse all the time. I have 30 extends and not in only one extend suno used the vers like I write it in the lyrics

[quiet lark]

Ive never had that issue, how are you forming the verses

[sleek valve]

it does not help. At least in my case it does not. I tried several times

[hybrid crane]

My video explains how you can force an intro

[tidal isle]

[Verse 5]
High-fives in the skate park, we’re one big crew,
Every flip and turn, finding something new,
Skateboarding forever, it’s the life we choose,
In this endless ride, there’s nothing to lose.

[Chorus]
Skate, skate, under the sunlight,
Skate, skate, feeling everything just right
Skate, skate, every trick, every grind,
Skate, skate, leave all worries behind.

[Chorus]
Skate, skate, under the sunlight,
Skate, skate, feeling everything just right
Skate, skate, every trick, every grind,
Skate, skate, leave all worries behind.

[quiet lark]

interesting what style of music

[tidal isle]

Skatepunk

[quiet lark]

whats your previous verse like? sometimes if it doesnt line up right some times it struggles

[hybrid crane]

AI sometimes loses its place in songs because it can be unpredictable. However, those familiar with it can often recognize patterns to help it stay on course. One method is to backtrack a second or two and resume from there, or occasionally repeat the last sentence to provide context. The effectiveness depends on the situation, but it's far from flawless.

[quiet lark]

everything daizy just said

[tidal isle]

This is previous: [Pre-Chorus]
Feel the wind in my hair, feel the rush in my veins,
Every ride, every jump, I'm breaking all the chains.

[Chorus]
Skate, skate, under the sunlight,
Skate, skate, feeling everything just right
Skate, skate, every trick, every grind,
Skate, skate, leave all worries behind.

Then suno do a instrumental after it and repait vers4. At this time I extend and want to continue with vers 5

[quiet lark]

you often want to start the "extend" timer immediately preceding the last word you want to keep rather than the music. generally the music will still carry over

i hope this helps

this is a song i put together last night with several extends to make a near flawless song it is country, i know how people feel about it but just give a listen

https://suno.com/song/f07fd242-22a5-47d6-b430-1566e646d349\

[tidal isle]

It's after an instrumental and it would repeat vers4. So I startet one, two, three or four second before it with vers5 but it always starts with Chorus. If I only type in vers5 without Chorus it starts with pre chorus or something else but not with vers5

[hybrid crane]

You are repeating your problem, but did you try any of my recommendations?

[quiet lark]

yeah you've got to start immediatley following the last word of your chorus. or even on the word. when you listen to your verse 5 portion you may even hear the last word come through kinda funny, that is fine, it will blend when you accept the whole song

[tidal isle]

I thought this is what you mean. Or should I go back more seconds, but then the good instrumental gets lost

[hybrid crane]

The AI is in a state where you cannot expect perfection, but it knows music. So even if you cut into an instrumental, it will finish it.

[quiet lark]

yes lets say your previous chorus ends at 02:38 you want the extend timer to start at 02:37 your still going to get an instrumental, and it may be the same as the one you have or may even be better, and then verse 5 will kick in.

[tidal isle]

Ok, will try it

Thank you both

[quiet lark]

yeah any time

yeah and you can hit that create several times if you have the will too, many times youll get several options

[past cairn]

I'm sure people have mentioned this already and I expect it to be a subscriber feature (which makes perfect sense) but I would love to be able to input a sample for it build from. It definitely has the capability considering how the extend feature works. Also having a ChatGPT ability to remember and changes based on new directions. Like i have a good sample and i want the same vocals but with new music or vice versa. Would be really cool

[tidal isle]

Unfortunately that also don't work. This is the song: https://suno.com/song/3d1f52ab-b8f9-469d-a9c1-3b8dd785e3d8 at about 1:54 it should start with vers5. If I use 1:36 as new extend point it also don't start with vers5. https://suno.com/song/d25d1d82-1b22-4dac-b9c8-178f4b6b9b1e

https://suno.com/song/d25d1d82-1b22-4dac-b9c8-178f4b6b9b1e

[tidal isle]

Suggestion for improvement: At mobile version: Extend from time should remain and not always go back to 2:00

[mental thistle]

Make an API for music generaiton

[worn atlas]

I confirm. Now i am having most of my songs starting by an intro. Very good finding. Thanks.

[frosty breach]

Unsure if it would be possible with how Suno is constructed, but weights would be neat in styles. Being able to dictate how much influence each style might have on the eventual generation.

Not sure if it's possible, but just a sudden thought I had. Thought I'd share it anyway as ridiculous as it may sound.

[scarlet gust]

Okay. I need your help.
Showed this to a few friends.
They all think it’s kind of a banger but they like the part where the „melodical" part starts the most (starting 1:20).
What should I change? What CAN I change?

I’m new to suno so.. hello friends 🤘🏻

https://suno.com/song/98942b09-9488-4f2c-80d8-fc4c4d0eee9e

[dire mural]

Could we maybe have Language channels, like German, French, Espaniol, Russian etc.? Especially for showcases it would be a big improvement

[worn atlas]

You have it

[dire mural]

am I blind? 😄

[worn atlas]

Maybe 😀
https://discord.com/channels/1069381916492562582/1155316330510176358

[dire mural]

Oh yes I am 😄

thx

[worn atlas]

You're welcome

[quiet lark]

yeah i just experimented. i wish i could get some referral credits i got like five or six people on the pro plan myself 😂

[scarlet summit]

Is it possible to create stems only? Like just drums or just acapella etc

[delicate cedar]

Not at this time. Maybe not ever. The method used to generate music does not make the steams to begin with so the only known way to get them at this time is to extract them with tools like the Ultimate Vocal Remover and websites like vocalremover.org.

That said the Team has said they are aware there is a demand for it.

[scarlet summit]

Yea but theoretically you can just train it on just drums right

or just acapellas etc. its just dependent on the training data correct?

[grave zealot]

one thing that would really help useability would be an easier way to group outputs by extent genealogy and rerolls. and if we could make like project folders

[marsh jasper]

Suggestion: some kind of update every 2–3 weeks.

Haven’t seen anything from Suno in 2-months 😬

[formal bear]

maybe it is worth the wait? =p

[sage mirage]

eyyyy tony hadn't seen you around. Hope you're good

[formal bear]

❤️

[marsh jasper]

I’m ready 😘

[formal bear]

will happen soon hopefully 🙂

[wind crest]

people have requested inpainting and copy/pasting/stitching together chunks a million times, i'd like to request that when that is implented it includes the ability to copy chunks from different generations, for when the chorus from one generation is great but the verse from the other is great and you can't choose

[worn atlas]

Having a referal program is a suggestion than

[hazy wyvern]

Sugestion: Could the credit be more than 50 per day in the free accout?

I'm going to make AI songs but i don't have the enough amount and I won't paid that amount of money.

I only share this because i need more credits for convert.

[dusky plinth]

In your library it would be nice if there is shown on each song in which of your playlists it was added. Hard to know which songs you already added to which playlist without checking in the playlist itself.

[fervent wing]

Suno should start using gpt 4 or gpt4o for the lyrics generation, same as Udio's recent update did

[novel jasper]

Add a Moderation flag to show you where the incorrect prompt is

like a highlight

[queen fox]

Aku mencintaimu

[last jacinth]

So you're basically telling them you don't want to pay for this service and you want to get more free credits, because you're you, right? That's a great point.

[hazy wyvern]

No because I am me, but, i was just saying for all the free accounts not only me.

Sorry about the misunderstanding.

[strong swan]

its a business... the intention is that you have fun trying suno for free and then buy the service to be abled to monetise your songs and make more. all of this here has to be paid for. thats not possible from free accounts.

[hazy wyvern]

So i don't want to grow the amount of credits per day, i just want to add the fact of those credits we don't use, to be there with the +50 credits
(wchic are the free credits)

[strong swan]

if the credits would stack the amount of needed computational power for the users is not calculatable. when everybody decides to use all the stacked credits in december, the servers would explode.

[hazy wyvern]

Sorry then

[worn atlas]

Yeah. Thats funny

[lunar berry]

i'd like to request a way of making only instrumental out of song that we've created with lyrics

[lone parrot]

In a way, you could try extending original song, but start it at 1 second mark and make it instrumental. In that way, your song becomes sort of "template" you can use for other purposes

[lunar berry]

Just done that, it recreated the song in a similar way but it's not the same.

[frosty breach]

Not sure if others' share this thought. But I strongly dislike how Trending automatically plays if you listen to a song from the song's page if a person has nothing else public.

[solar yew]

What if playlists were public in a way that allow other people to add their songs to them? Like I like my 8-Bit Living Songs playlist and have made it public, but I'd like other people's creations in there too and then it'd be a whole villagecore vibe.

[void siren]

What is Suno doing to challenge Udio? I feel Udio quality is much better than Suno

[sage mirage]

Stems aren't mashed together, Suno doesn't use stems at all

If you want stems you must break them out of the master that Suno provides. Its inherent to how Suno works. There are no stems

It's a training data thing, there most likely isn't even close to enough data to train just specific stem types, and if there is, the data itself tagged on that data probably isn't useable without a ton of work

These LLMs needs massive amounts of data. Lots of music around, less clean and precise stems

Yeah. It would be fantastic though

and it doesn't hurt to ask for it here in the suggestion thread, certainly

The solution is most likely going to be a break-apart-afterward thing like we can already do

But cleaner

Personally I think, even though it isn't how normal music production works, infill/inpainting is the major next step

Get the master like now, but you can go in and reprompt portions insde it already

The dev's hearts are certainly in the right place yeah

Quality is certainly going to go up as well, that already has happened going from v1 -> v2 -> v3

Quality is also a model output thing. Normally people also ask for .wav

Right now with v3 there wouldn't be any benefit to offering a .wav since you'd just get the same quality but now in a giant file

stems and .wav are the two most requested people when actual music peeps come in

It can do it but the problem with Suno (and any other music LLM that is around) is it doesn't understand anything

You and I know what chiptune/kegen and all that stuff is

Suno just goes "HEH NOISE"

And it's hard to keep it in the lane without it going more toward music music

Yep!

Most issues like that can be solved with spending a bunch of time generating clips and steering the song following good ones

An opinionated slot machine = Suno

[strong swan]

i tried to make tracker music , but its super hard to get this chopped up sample feeling.

[sage mirage]

and Mr. Tree is the person that would know

[strong swan]

most of the times it will make the basic "sqarewave chiptunes"

[sage mirage]

I was actually messing around with prompting square wave/sine wave

Seems potent, even if it doesn't actually do expected from the name things

[strong swan]

most of the things that i found and are implemented seem to be from the general midi standard

arps can be controlled seperately. even the timing like 3/4 7/8

i am all FM synthesis 😄 opl3 soundcards, yamaha synths... c64 is also super sweet

oops i thought i was in the general room ... sorry for spamming videos

[dusky plinth]

Please add sorting options to the library. Alphabetic would help me a lot right now. I jump a lot between the different song parts and create new ones and it is such a chaos. Right now the only option is to create a playlist for each song and put all the parts in it.

[wind crest]

save yourself a little processing power by allowing us to choose between 30 seconds, a minute, and 2 minutes for the first generation. i normally just clip 15-20 seconds of intro from the first generation, the 2 minutes feels like such a waste :(

[frosty breach]

Unsure if I already posted it or if it's possible. But the ability or a setting that allows you to extend from 00:00.

[edgy venture]

I used all my credits to try to make some songs in the same style as freaks from the surf course group, without result, any ideas?

[tawny wing]

• Please add the possibillity to select which sections to keep in a full song chopped up into 5s or 10s sections, or the other way around, which sections to regenerate differently.
• Also add the possibility to guide pronunciation phonetically.
• Import option, upload a tune and create a different version of it.
• Download options with higher bitrate.

[formal bear]

why?

that's technically a new generation, right?

[frosty breach]

@soft meteor also had some thoughts on it, so I'll tag him to give his own explanation.

Well, I like to do weird stuff and I get it sounds weird. But being able to create a hard cut, or being able to generate a new part with 0 context and then returning to whole, might open up entirely new possibilities. We already can do some neat genre swaps with very short extensions and 00:00 might open up wilder possibilities.

In essence it is an entire new generation yes. But you can go back to whole with it, to then potentially influence the next extensions in interesting ways.

[formal bear]

oh I see...so that's just a way to concat totally different generations together?

[frosty breach]

Had to look up concat, but yes! But in my case concoct might be a better word to describe it. Because of how context works you can then have the AI or however it works, look back at the wholle for the next extension and hopefully get some very interesting results. Or for instance in a fade-out, with 0 context swap to an entire different style, to then come together in a large fusion climax.

[formal bear]

ok got u

let me think about this...

[thick shard]

Not sure how the models are implemented or if there is some control over seed values, but coming from a Stable Diffusion background having the ability to re-use a seed/generation but have an additional layer of "noise" to make variations of a generation would be helpful. Basically rerolling the same generation with less variance from the original

[feral aspen]

Discontinuous context/history is a useful tool in Bark. Will post some examples if I can find time, in case it helps convince Suno to try and see if this is interesting in Chirp...

[formal bear]

I understand what people want but I don't think extend from 0:00 is the right solution

the audios will not sound continous at all

I think folks want a different feature

[frosty breach]

Actually that's it!

At least my intent is exactly what you said.

[left stratus]

is there a roadmap? I am wondering if uploading recorded vocal tracks for Suno to create backing tracks is something being considered or worked on? I know how I want my lyrics delivered, when they are just written the AI has to guess... I'd love if instead it could build a song around my delivery.

[shut dagger]

folders for projects, or like.. a current project folder - in an ideal world, something you can move the finished song out of then just trash the rest in one go

[soft meteor]

Currently we work around this for our more experimental/hard transition stuff by creating a mini extension and continue from 00:01 and hope that this is a small enough context for it to do a drastic change. What I would envision is decoupled continue and context so that instead of "continue from" alone you set also a "Context Length" either with a slider or even a wave form select. I would like this control to allow 0 Length/empty selection. That way you could do a hard change at a specific point. (Optimal would be sub second selection accuracy for continue point and Context length).

This might not be a very common use case but sometimes we really want it to start something different "right now" without transition or prior context.

So we basically "waste" one generation just to get a filler gen. And even then for genres like techno/hardstyle sometimes 1 second of context is enough for it to just repeat what it was doing for a long time

[tepid trench]

Definitely would be nice to be able to quickly pivot from one thing to another. Also context-wise it might be nice to be able to choose either any time range in the current song but even more interesting might be to allow context to be selected from another Suno song.

[formal bear]

got u

I will keep this in mind

[worn atlas]

Instead of rating with a thumbs up or a thumbs down, I suggest rating the clips on a scale from 0 to 5. 5 being excellent and 0 being terrible.

[worldly pagoda]

Is there a possibility to add parameters/something related to vocal chains? Full control over vocal, choir … great step to do it, because sometimes there amazing impressive vocals done through vocal chains … like deep vocals or soft melodic vocals.

[midnight shoal]

Request: humming to full song generation

[wind abyss]

looping song on playlist, needs an option to stop playing the playlist while creating new songs. it is now so annoying and stops people continuing the about created song. they got in there mind.

[upbeat mantle]

Will v4 have a longer lyric word total from 1300 to 2500 ? in custom

that would make it so u dont need to make so many extended parts for a longer song

[rancid thicket]

Yeah

I mean you're literally right

[leaden marlin]

That would be awesome!

Also, increasing the maxiumum time to like 4 minutes or something would be a welcome change. It seems like even my shorter poems are sometimes having trouble being squeezed in there by the AI, and it either cuts it off too short, goes too fast, or the AI just starts freelancing the entire thing afterwards. Either way around, the song quality drops and it drops fast.

[frosty breach]

Already pleased it got seen!

[frosty breach]

Video view on site might be neat.

[gilded wagon]

I'd gladly spend more credits for that. E.g. 10 credits for 1300, 25 for 2500. Likewise for extensions.

Having a consistent voice and quality for longer compositions is worth paying extra for.

[leaden marlin]

I also feel like it would be awesome to get at least some credits back for permanently deleted songs.

Even if it's like 5 credits every two songs or something, it would be nice to get something back if the AI butchers so many songs. I have had the experience of literally the same poem getting butchered like 13 times before it came up with an awesome one on the 14th try. I think I ended up spending like 70 credits on just the one poem, and it would be nice to see even half of the credits spent coming back in cases like this.

[gilded wagon]

"Please add the possibillity to select which sections to keep in a full song chopped up into 5s or 10s sections, or the other way around, which sections to regenerate differently."

This would be great. Sometimes a song comes out nearly perfect, but a word is dropped or mispronounced badly so I have to do a bunch of extends that often are not as good as the original.

[craggy nexus]

add a way to buy subscriptions paypal or simular.

currently i only see the option to buy it via card or "link" which also looks like buying it via card :/

[pearl ice]

Sound effects?

[true umbra]

Three Suggestion:

1. Option to add Intro.
2. Refresh Button... to clear memory and start with a clean slate.
3. Update Notes from V1.

[neat rose]

I have a proposal from the "game changer" category, it would be very nice to introduce the option to add your own melody and give instructions to the AI to create a certain vocal (with the user's lyrics or generated by the AI, here it already depends on your preferences).
For example, I have a lot of songs created in ableton, I can sing, but I don't have a proper microphone to record and no money (I'm a student XD) and it would be nice to experiment with AI vocals.
I hope I've been explicit and written in a way everyone understands.

[cunning blaze]

There should be an option to add your own voice model

[onyx wren]

Make paypal a accepted payment method
please--😭

[frosty breach]

The create tab can be quite cluttered. It would be neat if we could hide the side menu.

Atleast that menu seems most redundant when playing around in Create.

[frosty breach]

It's been suggested a lot, if you press CTRL+F and look for Paypal. It would be absolutely great if they did, as a lot of online monetary streams also tend to use PayPal, or I haven't looked into it for a while and been stuck in using it myself. But I agree with it!

[worn atlas]

Good suggestion but it will be difficult to know that copyrights are not violated.

[onyx wren]

Like i dont want to make a credit card just for this, and it dont accept my debit card either so why not just use paypal

[frosty breach]

Yeah it's very limitted atm, lets hope for more. Or something with an app that might allow Apple/Google pay.

[onyx wren]

It does allow google pay but not paypal through google

Which makes 0 sense to me

[frosty breach]

Oh that thing, that's very localised for only a handful of places and it depends what type of email you have.

[foggy pagoda]

Can we get some kind of estimate on when we will get inpainting? its getting frustrating blowing through so many credits just because of a very noticable vocal glitch that could be fixed. On another note, the vocals really need a lot of work. They freak out so often compared to udio

[tawdry wave]

Algum gpt personalizado ou prompt para rearmonização musical?

Quero apresentar uma progressão de acorde, por exemplo:

[verso]
C | Am | F | G | C | Am | F | G |

E o gpt me responde criando uma (ou mais) estrutura mais complexa e interessante com diversas alternativas de caminhos harmonicos melhorando e enriquecendo a progressão de acordes.

Ou agindo como um determinado musico e fazendo a rearmonização da progressão de acordes proposta ao estilo dele.

[hybrid crane]

Este canal é em inglês. No entanto, também temos um canal para o Brasil aqui: https://discord.com/channels/1069381916492562582/1154779408989687848

Translation: In this channel we speak english, but we do have a channel for Brazil here: https://discord.com/channels/1069381916492562582/1154779408989687848

[frosty breach]

The word on the street is V4

[robust mason]

could we add a Voice Channel that allows streaming ? I want to stream within the suno community to share how I create and my process of sectioning the creations of music made with suno, to share a different approach to maintain vocal clarity, clean flow, crisp overall sound, along with being able to share with others tricks and ways to solve some issues they've experienced before. I would love that

[frosty breach]

I do fear what some of us sound like

[sage mirage]

I'm boring af

[robust mason]

people could mute and just listen in

I can talk forever

[tepid trench]

It's not like it'd be compulsory to join/talk so it could be useful.

[frosty breach]

It would be absolutely key in helping others.

Like a few limited rooms (for helping) and a general one would be neat.

[drowsy chasm]

i need a metal room 😄

[hybrid crane]

We have one here: https://discord.com/channels/1069381916492562582/1154863909375852606

[patent sonnet]

I have a fun idea.

A "Transition" button, that allows you to continue the clip, but the AI gets instructions to transition it into another genre. That way, you can make EDM transition into smooth jazz seemlessly.

[strong swan]

you can already do that by changing the styles for the extension... if suno likes the idea you get great mixes

[drowsy chasm]

thanks 🙂

[patent sonnet]

I've tried that, I can never get it to work. It just sticks with the original style regardless of what I put in.

[strong swan]

here is a style change example i made cycling through several genres #💬┃general-chat message ... takes some practice and credits to make it nice... thats the randomness in action 😄

[patent sonnet]

I see, It did work once in that.

But the kind of song I want...

I want it to sound like a mash up of several different genres that it switches between.

[strong swan]

it switched from spoken word to to drum and bass, to a string interlude, to hip hop with scratches back to drum and bass and then a combination of dnb and strings... i would say thats more then one 😄

[pastel hare]

Add ability to generate the same style and music twice. For example: you liked what the suno generated and you want to change some words or add something extra.

[void sand]

Add the ability to only do vocals generation

[strong swan]

you can arlready generate spoken text , narration, standup comedy, tv commercials, solo singers with the right styles.

[void sand]

how do I get singing for like hardstyle?

[strong swan]

suno is more like a huge random generator. you can give it hints what you exactly want, but if you get it is a different thing. a idea is to try to generate spoken text and "massage" it to your music with a daw. or you extract voice from a suno track with a stem splitter or so. the idea to generate voice and music seperate is nice.

[void sand]

ive tried stem extracting it, I guess what I'm asking for is the option to get a clear singing voice in the style you want

[hardy fiber]

I just discovered the tool and I am amazed about it. But it could go way more far :

• We should be able to extract any instrument / voice from the generated song
• We could also upload our own file and ask AI to extend the song or add extra instrument or voice
• Will be nice to have more settings about the type of voice we want
• Be able to edit the generated song by adding features or change the lyrics

Please make it happen

[formal bear]

As the models get better, you will need to generate less to get what you want, hopefully.

[robust mason]

@formal bear

would really appreciate it 👍

[formal bear]

I don't have admin power

[sage mirage]

Sounds like Suno needs a head of Community Support

How is Boston, anyway

[quiet warren]

The new gpt model seems to like to bolden the (verse1) and such. But it turns into ** or ### when pasted. Just suggesting to support this, telling chat to remove the bolden font isn’t a big deal. 👍🫰

[marsh narwhal]

вы о чем кенты

[hardy fiber]

Whatever dude

[tidal isle]

Don't overcompress songs with bass. Almost every song with a lot of bass is overdriven

[cloud zephyr]

Can we make the word filter a bit more lenient? dropping 2k credits then finding out the next part can't be done as its flagged for moderation is pretty annoying

[fallow hill]

/chirp

[placid tide]

Can you please add a more Payment options like Klarna and PayPal for our Abonemment?

[arctic whale]

Yes. That would be very helpfull. Just found out that you can feed on each other but noch ravage each other. That's ridiculous. Had "worse" lyrics before that went through. And GTP tells over and over again that the lyrics are beautifully crafted but doesn't point out which expression got flagged. It's practically useless.

[grave gust]

Wait, they censor stuff you put in your songs???

[tepid trench]

Lyrics get put through ChatGPT to check for bad things, which can then stop the song being generated.

It's possible to get around it by misspelling things in most cases.

[tepid trench]

As above, try misspelling whichever lyric(s) you think might be triggering the filter (it's not just a word filter, ChatGPT looks at the context)

[frosty breach]

Community announcements.
For contests and the like, perhaps with the chat linked where it is.

[umbral quarry]

I think its a pretty impressive Generator. But i wish i had the ability to keep a particular piece of music and change the lyrics, far too often it gives me something that sounds great musically, but the lyrics are all over the place, a feature to regenerate and re do the lyrics or certain parts of the lyrics (so it can be regenerated on the fly) - because when you do want to regenerate it because it messed the lyrics of one part of the song that threw it all out, you end up with something else entirely - so i think that would be pretty good feature. Edit lyrics/generate new lyrics in realtime) but not lose the actual music

[blazing cave]

Explore

The explore tab should be turned into a list where you can search for genres with auto-fill/suggestion bars when typing certain letters

Profiles

Custom banners, backgrounds and profile pictures

[arctic whale]

You sure? If it would look at the context then it wouldn't flag a perfectly good metaphor which has nothing to do with violence. 😁 Those lyrics got through before, btw, I just wanted another version and then the odyssey began ...

[tepid trench]

Hmm. Just going from what's been recommended in the past. It seems to change it's mind on what's appropriate or not sometimes.

[frosty breach]

#💡┃suggestions-web message I've suggested the same, and a team member gave it a reaction!

[worn atlas]

Well. This has been asked one million times.

[worn atlas]

Suggestion : Dont give the same name when we use "reuse prompt". Udio is doing this pretty well. Please add à suffix like v1 v2 or #1 #2 or whatelse you like but dont reuse the name.

[sharp night]

Suggestion: Roll-over credits. Having unused credits burned at the end of each billing cycle makes the relationship between customer and company antagonistic. You NEED to have goodwill with your customers, especially in the rapidly growing and expanding AI space, where a slightly shinier toy is enough to poach your user base.

[umbral quarry]

Im new here so i didnt know this.

But interesting all the same

[soft harness]

The day-to-day (free plan) or month-to-month (subscription) credits refreshes rather than accumulate because of resources allocation restraints (and is mentioned at the bottom of the subscription plan page).

Extra top up credits and special credit gifts are, well, special and are carried over unlike the former two

https://media.discordapp.net/attachments/1069381916492562585/1238249445192237106/image0.jpg

This is not unique to Suno either. Many AI services do it as well, including Udio, because of resource allocation

[soft harness]

Forgot to reply to you earlier, see my above messages why this may be more unlikely than you think @sharp night

[strong swan]

now we even do the support for other audio services as it seems... i knew we were good here... but thats actually amazing LOL ... (That was a joke... No actual support for other services here 😏 )

[soft harness]

Heh. Was using as an example because you KNOW someone would try to use them as an example with their false assumptions 😆

[strong swan]

😏

[robust mason]

lmao

[upbeat mantle]

Just wonder if it would be possible to make it so that the free creds are saved up instead of just staying locked in at 50 each day

[strong swan]

the intention is that you buy the pro subscription and help to keep suno alive. from free credits no company can survive...

[upbeat mantle]

thats true nothing comes for free

[strong swan]

also you get the monetarization rights for the song from a paid sub. thats a huge thing.

[raw mist]

We've already asked for the ability to download song part stems and the MIDI. This will be HUGE once available. I'm using Fadr for now to try to stem parts so I can bring it in my DAW.

One recurring issue is that when generating a song, the parts (verse, chorus etc.) do not align very well in a DAW when using Fadr (or similar) to stem tool. Each part does not end or transition cleanly.

More clearly, parts are not symmetrical, i.e. they end in weird places in time and its really hard in the best of DAWs to do surgery on them to align. Off by a 16th note and the like.

Does anyone have any suggestions to try to extract a song to stems then have the song parts (verse, chorus) align normally?

EDIT: Maybe Suno is simply creating song parts non-symmetrical, I dont know. Be good if that could be confirmed.

Thank you.

[leaden marlin]

Please add a Payment via PayPal.

[royal venture]

Add automod to server to filter out words like Chirp and variations + bark. It can automatically send a message to users who try to type this, which you could send them links for suno.com/create and the wiki. This way mods and users both never have to see or respond to these messages. You can also use it to block links to non suno muic profiles and such, practically making all this mod work automated. Mods are not on 24/7 and there are large periods of time where a user can get away with ignoring the rules. Make things easier. At worst case download a word filter bot.

https://support.discord.com/hc/en-us/articles/4421269296535-AutoMod-FAQ

[jovial rain]

Library organization would be nice. it would be nice to have a way to organize songs that you like vs just one thumbs up.

[cursive turtle]

it would be nice to an option for a full song during the custom and instrumental options

that way you dont have to keep on hitting extend

without the reroll

[charred flicker]

What would be even cooler would be having a seed ID so that when you find a style an arrangement you like, you can reuse it.

[open vessel]

Please provide away to bulk delete. Please.

[leaden marlin]

You can select multiple tracks on your library with shift+click, then it deletes all of them

[open vessel]

Oh my lord you made my day! Thank you

[fierce flicker]

I want a possibility to lock my finished songs. I don't know about you, but I've accidentally deleted songs. The ability to lock can also be automatic for songs that are in playlists or marked as public. A popup saying "you must unlock your song to delete it", would at least keep me awake.

[leaden marlin]

You can find deleted songs in the bin, it's on the top right on your library

Unless you mean that you accidentally delete them from the bin as well? I think there's a confirmation popup for that

[serene solstice]

Piano sounds seem really hit and miss (the mid/sides seem off) like backwards sounding almost, the way the Ai interprets piano sounds I would suggest taking a look at pianoteq which is a modelled piano, the attack is always wrong on the piano notes, it's more of a a fast attack on a piano sound but seems a slower attack most of the time which is odd.

[haughty crescent]

Hello)
First I would like to express a huge thank you to the creators of Suno. It is a very good program) It develops, updates and pleases us, users, with its capabilities and achievements. Thank you!

And here is my suggestion. It would be great if in Suno there was such an opportunity, - to make one part of a song so that music and style were from another, created earlier, but the words were different.
Perhaps this could be accomplished by inserting a link to the previously created song part in the style tag field of the new song segment.
What do you think about it? Can you do it?)

[undone bough]

it should be possible to clean up the text in the lyrics when your done with the song, also would be nice to be able to "extend" the and in between. Also, be able to upload your own Covers and art (with some w8 time maybe for security purposes) Splitting tracks would also be nice but I understand that would lead people to continue in another software, but it would be nice though sometimes u get a super sound or chorus or guitar riff but have to move on because there is to many bad spots in the whole of the song..Thats IT! plz make the changes by tomorrow and i will be a Suno slave for life! 😛

"extend" the START its supposed say

"Extend the start and inbetween!

often the start sucks or is cut but the rest of the song is awsome, to bad u cant "extend" alter the start (intro) beginning with different suggestions

also should be possible to enter a keyword section where u can add additional wishes and ideas for the song for example

...lSow Ballad hard Rock, Catchy Instrumental intro. powerful, high notes, catchy, Electric guitar riff solo

("Slow# Ballad...)

if u find a good sound and singer it should also be possible to copy that style onto another song so u can get consistency for multible songs when making a whole albumj 😛

a right click command menu would also be nice showing possible command like : (Intro), (Verse:) ,(solo) etc

also be able to generate just 1 sugestion at the price of 5 instead 😛

the shift of art in the final result if u have extended should also be possible to remove if you dont want it

generation of animation video is too much to ask I know but it would be cool many AI generators out there already doing that

like lyrics "Mjølnir screams through the thundering sky" would generate a short animation of that hehe

thats it thx for letting me spam my thoughts of possible improvements of this awsome generator!

[silent dune]

hello! I have a suggestion that I think would help many users and it would give the ability to create more complex and creative songs. I have made many songs in Suno and I noticed that for the chorus to be the same in the entire song, it needs to show up in all of the extended parts, otherwise it will generate a different sounding chorus and so if it would be possible to find a way to keep the same sounding chorus after extending the song with just verses it would be great!

another suggestion would be to extend the time on the extensions to 1:30 instead of 1:00 because most times the chorus gets cut in the extended part

[leaden marlin]

It consumes at least 1000 credits to make one decent song. The accuracy is too poor. The sound quality is poor, the mix is too messy, you can't output wav or stem, so you have to be careful about the balance of the mix, noise, etc. To be frank, I think this is just a recreational service for beginners. It can't be used for anything more than that.

[cunning pecan]

Mp3 -> wav -> split stems -> DAW and rebuild? There is too much noise and hallus and artefacts in most Songs. 500 to 1000 credit per song seem to be normal.
The direct results are mostly way too bad for commercial use. For fun project it is great.

[molten agate]

I have something...I don't know if it's suggestion, but...please add something new! In the Suno, add a button to download midi files or something...

[gilded wagon]

Suggestion: Can AI come up with short names for the songs each time that would make it easy to distinguish between the various versions, using distinct words that are not in the prompt.

E.g. If I mentioned the genre but don't mention the sex, instruments or if it's a solist or group, then it name one file: “bariton soloist guitar intro", or “female group, alto piano"

I use Loom for video recordings and it intregrates AI in amazing way where it creates a transcript and suggests the presentation and time stamps.

[tepid trench]

I doubt MIDI will be possible since Suno will have been trained on audio, so can only produce audio. It would be nice if possible though.

[worn atlas]

Add an ai voices bank in whish we can choose our voices.

[worn atlas]

Why it can be an issue ? An audio to midi translator should not be a technical challenge. The greatest challenge is stems separation . Once this is done no matter the format of the output . Wav or midi. I guess.

[worn atlas]

Suggestions to mods . Create a channel for tips and tricks.

[tepid trench]

Yes, stem separation would be a step toward making MIDI output possible, because pitch detection could then be used to create MIDI notes. At the moment though, I try to do this by ear, and there are a lot of notes that simply cannot be heard because of another instrument being louder for example. MIDI doesn't capture all of the performance or sound detail but would be useful for changing the instrument etc.

So I expect stems will come first.

[worn atlas]

Yes, stem separation is indeed a crucial step towards accurate MIDI output. By isolating each instrument, pitch detection can be applied more effectively, overcoming the challenge of overlapping frequencies where louder instruments mask quieter ones. Without stem separation, converting a mixed audio track directly into MIDI can result in inaccuracies and missing notes. Therefore, performing stem separation first ensures a more precise and detailed MIDI representation, capturing the nuances of each instrument accurately.

[leaden flame]

Suggestion: A big step towards making this technology useful to musicians would be to have the ability to select the music key and to select the BPM tempo.

[worn atlas]

The noise generated on some suno outputs is really annoying specifically when after multiple trials you get the good couple melody/music

[leaden flame]

Currently Suno is impressive in regards to what it can do, but the lack of precise control in it over much of the music generation process makes it essentially just a toy.
Suggestion: By all means keep Suno as the revenue generating toy that it is, but I would suggest considering working towards creating a second AI music generation application tool geared towards musicians and serious music generation. The more control that can be built into it with the specific goal of being a useful tool for musicians, the better. There is much potential here, if more varied and precise control over various aspects of the music generation process can be built into it.

[trail hawk]

Stem separation can be done by third party tools. Stem generation would be better. MIDI would be one step backwards.

[tepid trench]

I meant stem generation / separate stems to start with, just worded it badly (sorry)

[worn atlas]

Yeah you are right. I have also used the term stems separation as i dont know if getting the stems will be native or a post-processing for suno.

[exotic harness]

Suggestion: Being able to use the same voice to create an AI persona.

[sage lichen]

Suggestion: Being able to extract the melody from a song

[heady viper]

I feel a bit overwhelmed by the mass of songs I don't want to keep. Some mass deletion or storage in folders would be great. Thanks!

[leaden flame]

In Windows, you can click on the first song you want to delete and then either use Ctrl-left click to select other individual songs, or scroll down to the last song you want to delete in a continuous block and then do a Shift-left click on the last song in the block you want to delete. Then on any of the songs you have selected/highlighted, click on the '...' menu at the end of any song row and select 'Move to Trash'. You can do this on the Create or Library screens, and on the Trash screen, as well.

[heady viper]

Thanks, very helpful!!

[fierce flicker]

I did send them to the trash and then delete them there. My computer has been a bit buggy and doesn't always show the whole window, so I never saw when they disappeared. Yes, you are right, the system is good if you are not as clumsy as I am. Now I always do bulk delete more carefully 🙂

[leaden marlin]

You guys should definitely make an art section in here I think it could be really cool but I'm not too sure though

Maybe it wouldn't fit the idea of music but I still think it could be cool

[raw dirge]

Can we get an option to cut pieces of a song out? Like an easy way to slice off the end of the song, for those times when the song ends abruptly for no reason

We should do a weekly song challenge in the discord where we all get a weekly topic and whoever wants to participate makes a song on that topic/subject. Just for fun and to get the creative juices flowing.

[viscid cloak]

Suggestion for the Discord (I know hard topic, sometimes)

Use a Leveling Bot, then give specific rights to highert levels. Like Posting Links etc, by being active.
Would encourage activity, aswell as preventing Links from New Users (Like Discord Links or Scam Links)

And maybe give active Users a bit of Color?

[mighty needle]

Suggestion: A way to quickly delete generations. Maybe have a way to delete generations that aren't liked or saved with one button. My library is messy lol

[leaden marlin]

I recently got told that I can't check if someone disliked my song please could you add a feature so I can check if someone disliked my song cause how am I supposed to know if someone hates me

⬆️

✨ Maybe a custom profile image that we can put in cause it's feeling a bit blank and maybe a custom banner for our profile🌟

Because like it's feeling extremely lacklustre

[drowsy dove]

make the suno generated videos landscape maybe? (will help with posting to discord --- i really dont want to have to post files here, id rather just link to the suno songs on suno and have the video show)

[leaden marlin]

I agree. you guys should do that

[tiny spoke]

The sidebar in the library should contain links to previous parts

[cosmic terrace]

would love a commen system on songs.

[leaden marlin]

I think I need my own role on this server just so people can find me easier I'm very important

[viscid cloak]

No please not. Some people want them unpublic for a reason

[tiny spoke]

thanks for the hint. that was my intention - only visible to yourself

[viscid cloak]

Ah, then nvm 😄

[leaden marlin]

You should give me the full admin I'm super helpful

https://tenor.com/view/cat-dance-gif-3087157064023426489

[mortal hull]

why would they do that?

[leaden marlin]

IDK

I'm super helpful

[strong swan]

i already noticed your outstanding capabilities in posting cat memes in the suggestion channel. 😏

[leaden marlin]

no 😉

Damn I think he's catching on to me

[strong swan]

that was not a invite to post more.

[leaden marlin]

sorry

This is my last time

😿

What happened again boss🐱

[strong swan]

its always a pain to see flashing gifs if you just want to have a look at some suggestions.

[leaden marlin]

Got you all's good I understand🐱

That better?

I just wanna keep 2🙏

[frosty breach]

All embeds seem to work aside from the Suno ones.

[drowsy dove]

kinda hidden behind all the mess, but made this suggestion -- in the even that "discord chat clutter" was the issue:

#💡┃suggestions-web message

[frosty breach]

Oh no, I meant it for the team. So they know it's not a discord issue perse

[drowsy dove]

all good man ❤️

[empty bramble]

would it be possible to have a vocals-only model?

[tame ruin]

have a different button to remove songs from create section, instead of using thumbs down, in effort to keep first gens in view.

keep the branches marked in library

[dire frost]

This can be accomplished by using acapella in the song description

If there is another method you’d prefer, please share 🙂

[strong swan]

like the old bark? i see people extracting the voice from suno songs with external tools like stem splitters. to get just a singing voice is difficult . normally suno tries to make some backing like piano or strings.

[lunar egret]

When you make a new model with this new funding, release the weights to the old one like you did bark.

Let the community pick up on the old stuff

All these closed source models looking to make everyone an api renter are a real capitalistic bummer

[strong swan]

some people call it a business model to keep the lights on.. depends on the view.

[empty bramble]

keeping the old unused stuff closed source and forever lost and forgotten is a business model?

[lunar egret]

Do you need to keep the lights on by making people rent the old stuff forever while also getting another revenue source from the new stuff?

just kinda seems like linegoup

Im not saying giveaway your new stuff, Im saying let the old stuff go to the community while you make money off the new state of the art stuff

[strong swan]

bark is open source

[lunar egret]

And we just got pinged saying they got millions of dollars to make a new model, so when thats made, have the current model follow in barks footsteps

[empty bramble]

yes but he was talking about the current model if/when a new one comes along

[lunar egret]

Either way, just a suggestion. If suno wants to collect rent forever and leave the ML community behind like openai did, ultimately its up to them.

[tepid trench]

Personally, I don't think they should open-source the models that are currently accessible on the website (v2 and v3) but when they are retired and no longer available (like chirp/v1) it would make sense to open-source it at that point.

[haughty bluff]

On a different topic I'd like to edit the lyrics for a song after generating it. I don't expect it to change the song, just the displayed lyrics. This is for cases when lyrics are too long and you extend the song you don't get duplicate displayed lyrics for it. Alternatively suno should detect what didn't make it into the first 2min, don't include it in the song displayed lyrics and suggest extending itself with the remaining lyrics.

This is mostly a suggestion for custom lyrics, not generated ones

[drowsy dove]

i would love this. the number of times i have repeating, unsung lyrics on the suno page is like... well like only twice or so.

but id love this.

[gilded wagon]

That would be great. It might also help with the problem of the voice changing radically after a first or second extension.

It would be great to have several voice options: e.g. Male/Female, Range, age, clear/gruffy, genre, somber/energy level, etc. And based on that we are given a unique code 6-12digit code that we can input into a song prompt to get more consistency among songs.

[worn atlas]

This is what i was thinking about : 1. Save a voice you got in generation. 2. Create a voice from scratch. 3 Edit a voice previously saved.

[haughty bluff]

Another suggestion I'd have is to be able to browse the library by song title and see extensions etc as a tree view. It'd make it clearer how things are created if you decide to come back to them at a later point in time to edit or adjust something. Browsing 6 song starts with the same name then 20 different part 2 and 3 extensions gets confusing fast

Especially as in the library later the chains of songs aren't highlighted anymore unlike in create, where create doesn't display you all songs in a history view once you leave and come back to the page.

[plain cosmos]

Cool, Suno has raised 125 million. >> Our community of musicians deserves the very best tools<< Ok so when do we have V3 Alpha back 🙂

[primal dew]

V3 Alpha?

Suggestion: Daily Credits for subscribers. I hate that I've run out and have to pay even more if I want to keep going before my next fill.

[lapis goblet]

Hello, could you add means of payment such as direct transfers or charge in USDT or some other crypto fiat, it would be of great help for those of us who have too many regulations and extra charges when paying in dollars with credit cards, thank you

[copper valley]

I would love to see a way so you can download the music and vocal as separate files, for remixing and fun

[leaden marlin]

it's not perfect but BandLab can separate music and vocals pretty well, and it's free, just thought I'd mention it

[upbeat mortar]

6k credits will burn tonight as I'm about to renew the monthly subscription, it's not because I have not had time och need to use it.. rather that all I get is shit. Voices are distorted, background singers are always present, no matter what I do.. music and lyrics quality becomes worse with each extension.. I used to be able to get great voices and good quality throughout the whole song, same prompts I used a month ago with great results now end up in hundreds of clips in the trash.. Whatever you guys did to the model I'm closing in on giving up.

[drowsy chasm]

those who got the premier plan the most expensive should still get 50 credit free everyday i mean ... come on lol

[scenic remnant]

I put a tweet here with a congratulations on the funding and some good suggestions and now I see it was deleted and I was kicked out of this discord.

Why?

[rustic spruce]

Hello all 👋 Can I please put in my two cents here? I was using the Suno Explore feature and it badly needs a way to control how loud it sounds. I'm basically forced to decrease the sound output from my browser just to not get deafened (you can only fully mute the sound for some reason).

[raw mist]

Request: display what the key and BPM is for the generated song. This should be an easy one as you chose the key, chords and BPM when generating the song. Just display it so we can see it thanks.

Also, please display the chords you used too.

Thank you!

[pulsar yew]

Im having a problem everytime i try to tell the AI to play a guitar while a guy is singing its not happening any tips ? Help please

[strong swan]

its quite difficult to start things on command. for me it often helps if i put the thing i want to use also in the style and then try to activate ir with a (command] . but its possible that the instrument you add also shows up anywhere else. difficult

[scenic remnant]

I put this here and got deleted and kicked out, why are these suggestions bad?

-3 minute songs (in 1 generation).
-Accurate song structure.
-More other controls.
-A multitrack interface (like reason).
-Volume mixer per track (drum, bass, vocals).
-Effects per track (reverb, chorus etc).

[potent crane]

I am not a techy person as such, but it seems pretty clear to me that most of your suggestions would simply be impossible using the current technology. The songs are created from a large collection of mp3's and the ability to present instruments as separate tracks and therefore options to adjust volume, fx etc is not possible. If I need to adjust particular instruments (usually vocals) once I have a suno song, I use apps such as lala, where I can extract a vocal track, add fx and most importantly reduce the loudness before joining it back with the music.

[scenic remnant]

The seperate tracks give more control, for example instead of throwing away a whole song because you dont like the bass you can replace the bass channel only.

The mixing is crucial because sometimes 1 thing is way too loud or too soft

musicians want control, not a 1 button hit machine

All these things are still very basic controls, and take less than 1% of 125M to make

[tepid trench]

A lot of people (me included) would like Suno to work with separate instrument stems. This would be key to allowing multitrack editing, mixing, effects etc.

The issue is likely not one of cost, but training datasets (which would need to be comprised of stems) and compute power (more data to train on, more data to generate), along with figuring out how to do it and developing it.

I don't think we'll get this soon, but I expect it'll come eventually as there is a lot of demand for it.

[scenic remnant]

thats where the space is going, just like how photoshop puts a midjourney plugin into it programs like reason will put AI promps on a track

[tepid trench]

If it were possible to then guide the AI with notes or chords etc. as well, or to have it use the other tracks to guide what generates, it'd be really cool (because then you could either create parts by yourself / softsynths, provide a melody to be performed by the AI, have the AI come up with its own melody, or have the AI generate a foundation to build on.)

[novel compass]

I've been trying to use something like the free Ultimate Vocal Remover and its various downloadable engines to isolate various stems with some success, but I agree that it would be even better if they were separate from the beginning because sometimes the sound is squashed and hard to get consistent stems.

[fading grove]

Are we able to generate songs in discord like Midjourney? If not can you please create this features?

[hybrid crane]

This was actually the past. The moved on from discord since only 1% was using it in the end.

[novel compass]

That was the original way that is now deprecated.

[potent crane]

It may be crucial but if the technology that the app is using cannot generate stems then no amount of wishing will make it happen.

[scenic remnant]

If you don`t make multitrack control others will. I was in an office with 4 people making an app for a year and it cost 140k, $125M can do a lot.

[nova prawn]

i ve noticed that my instrumental song it s repetitive, like: aaaaaabbbbbbaaaaaaabbbbbbbb ccccccccddddddddaaaaaaaaabbbbbbbb aaaaaabbbbbbbb ccccccccddddddd eeeeeefffffeeeeeffffff gggghhhhh

[feral aspen]

Even for instrumentals, I typically prefer to use regular non-instrumental mode, and fill the lyrics with evocative phrases in [brackets] like:

[melodic whistling]
...
[guitar and percussion groove]
...
[nature sounds reappear]
...
[harmony returns]

As a random example (those are not tested phrases). It actually doesn't really matter very much if the model follows the intent of the instructions, if your goal is to avoid repetition or to get the music to be more diverse, in a general way.

[feral aspen]

even though suno moved on, perfectly good suggestion bringing Discord back. lots of room for social interactions in music gen that you can't really get on the web, maybe...

[jagged charm]

Does uploading songs generated by Suno AI on YouTube pose any copyright issues? I'm unable to subscribe to Suno AI here, but I want to upload songs on YouTube. Can anyone help me out?

I can't subscribe because I'm encountering an 'Unsupported Card' error when making purchases with my credit card. I've tried with 2-3 cards already. Can anyone help me with this issue?

[robust jewel]

The Suno TOS says that free accounts can upload to Youtube as long as it isn't monetized and does have some sort of "made with Suno" style credit. Paid accounts can monetize and don't require the credit

Free accounts are limited to non-commercial uses

[jagged charm]

How should one describe while uploading on YouTube?

[feral aspen]

Just "Made With Suno" in text somewhere is enough

[jagged charm]

Thank you my friend 😍

[royal venture]

Add better support for the samsung zfold 3. There are many major issues and annoyances with trying to use the website on this phone.

• I try to add a song to a playlist and it instantly adds it to a random playlist without me choosing one

• Anything past the middle of the screen and below button prompts stop showing, to delete a song, add it to playlist, all options dont show past the middle of my screen.

• trying to listen to a song in library, going to its actual page with lyrics, and then going back to library can make the song completely disappear from your library entirely.

• Doing the above with a song in trash can also cause it to magically disappear, even if you have the link, you like the song, and add it to a playlist, its as if it never existed.

• I try to scroll through my library, play a song, and it randomly scrolls back to top of the page.

• I have tried multiple web browsers, and this issue persists across all of them.

[unreal moth]

#📢┃announcements
This is good news, it would be very nice to integrate somes UX designers into your team, it a major issue on my side to use suno !

[soft harness]

Out of curiosity, can you elaborate on what issues you had? I found it easy to navigate and use https://suno.com (well, aside from the banner height of the profiles on mobile view)

[unreal moth]

ya the "Ease of use" is indeed a part of UX design, but it doesn't necessarily simplify the problem, and can sometimes lack complexity.
Making things too easy can make the platform difficult to use for professionals or just geek people.

We can look to the functionality of the Ejay suite, while avoiding their childish color themes and maintaining the current UI design of suno.
The Ejay suite offers a good UX design for audio editing experience, with drag-and-drop (DND) functionality, adjustment options, and responsive inputs.
You also have some better ui/ux libs for this kind of webapp like https://mantine.dev

[soft harness]

I... still do not get the issue you are having with the current UX though?

Suno is not a DAW so the images does not help nor does it clarify my question about what major issue you had in using Suno...?

[unreal moth]

tit fine, it's probably because you use their preform recreationally and not professionally, otherwise you'll understand.

[soft harness]

And yet you cannot clarify what was the major issue you have though?

It is a music AI, not a DAW, you input the prompts you want and it generates... so what major UX issues did you have that you were complaining about earlier?

The gist of what I got so far about this from you is: "You do not know the issue because you do not have my experience so you cannot see what troubles I am having"

(Not saying there is no problems, there are a few I had with, but the way you are skirting around not stating it sounds like you were just trying to promote the link you had above)

[sinful hare]

Search genres or song titles. Why haven't we got this?

[unreal moth]

am sorry but it's not my job, you will find all your answers in the images and the preceding discussions, I take care of my software not that of others, all the preceding discussions list the problems report, devote a little time to reading, download the software cite above and visit the functionality library.

[soft harness]

So you are saying you basically have no problem with how it is currently setup and just advertising, got ya. Waste of my time...

[brisk nebula]

Subscribers should get 50 daily credits as well, not only free users... i've spent my 2500 credits and now i dont even have the same rights as free users..

[soft harness]

Subscription plans basically upfront a month worth of daily credits plus additional credits that makes up the plan.

Once your credits goes below 50 credits, it should refresh (at 0300 UTC+0) as if you had the free plan until your next month renewal.

Just remember that there are no carrying over of credits. When your next month renewal comes, the plan credits are refreshed

[unkempt dock]

I was just going to ask for this, if we already have a full song, there should be an option to generate the instrumental of it.

[potent crane]

If a song completes to my satisfaction in under 2 minutes there's no way of registering it as a "full song", I work around this by adding a suffix to the title "complete song" so I don't delete it, but there should be an option to mark it as a "full song"

[sharp sorrel]

seachin is neeeeeded

[worn atlas]

What is the purpose?
You can download as it is.

[wind crest]

this, or else it's barely worth paying. current system just encourages people to use multiple emails instead of subscribing

[mystic zodiac]

A button for effects On/Off. seems you are trying to improve the audio quality for those who don't have a DAW but it's overdone right now. This would give the option for some to stem split and work in a DAW. with the heavy effects currently used, sadly it's unusable.

[tepid trench]

The effects will be baked-in to the model. They don't do any mixing/mastering, it just produces audio based on what it was trained on (so if a genre usually has certain effects, those will be more likely to be present.) You might be able to prompt for "clean vocals" or "dry vocals" for example but I don't think it'd be easy to make it turn effects on/off with a button (there's already a toggle for "instrumental" which can still produce vocals.)

I get that it's a suggestion though, and I think in combination with generating separate stems (which is what is often asked for), it would be very useful - e.g. if they could generate dry stems, optionally apply effects, then mix the output, that'd be pretty versatile.

[haughty bluff]

I'd assume this would require an entirely different model and concept to begin with that's completely separate. Making one track out of a trained model is difficult enough, making N separate tracks and having them be coherent and combine them into one full track practically impossible with basic ML without sacrificing quality

[tepid trench]

Yes - it'd need to work completely differently I expect. And need a lot more compute, and training data.

[haughty bluff]

I've only been here for a day but I've seen plenty of people having a completely wrong understanding how Suno works the way it does. It's like asking midjourney to produce separate layers and a layered photoshop file instead of a full picture which too is currently not feasible (or trained for). Suno is the same as midjourney but for music.

that being said, it is possible to generate music and vocals separately using ML models as there is technically plenty of potential training material of vocalless music and vocals. there's also rhythm adaptation ML stuff so it would be possible to adapt generated vocals to generated music. but I don't think layered music itself as people ask for would be easy to train for or achievable reasonably. but I'm not an ML engineer.

whether suno will go that route or not is a whole different topic however

[ionic skiff]

allow for uploading edited music into suno player

[tepid trench]

I used to think music AI worked from stems (there's one I used originally that offered stem downloads, I only later found out it was doing an AI stem split afterward.) It's no doubt going to be possible at some point, and it'd be a significant milestone. As I mentioned when the subject of stems came up yesterday, I don't think having money will make it much easier to do - they need the training data to be stems.

[ionic skiff]

Suno player is very good

[haughty bluff]

yeah, it's all about training data and then managing to combine it into a coherent whole, which is probably the harder part tbh

[tepid trench]

I would suggest that they could hire musicians to perform, and then record their parts as stems (so basically have them perform things that are in the training dataset, so you end up with dry, separate stems) but that would probably take a long time and be expensive 😄

[haughty bluff]

I don't believe that'd be even close to enough data to make a meaningful training set for something like suno and probably would break the scope and concept of what it does currently entirely and require a complete rework. alas enough OT.

[ionic skiff]

there is an interesting point here, where did suno get its initial data from?

it must have been trained on actual musicians voices / instruments

[haughty bluff]

there is no other way, yeah, iirc it was trained on actual fully mixed music

[rustic anvil]

A command structure or input box for Negative Prompts? For stuff we don't want. Similar to the image generators

[deep yacht]

A few suggestions from me:

• An automated playlist for full songs or filter option to only show full songs
• a personal star rating of each song (1 to 5)
  I ofren create many first versions to find one I really like. Perhaps 20-30 creations. So an option for rate each creation would make it so much easier to select the ones to keep working on.
• an easy way to check mark song and delete them all in one go.

[humble summit]

2 Suggestions:
-Would be cool to have "Workspaces" in the creation tab where each workspace is like its own library and you can swtich between them if you are working on multiple songs at the same time because right now I kinda have to have a "TODO" playlist which works, but can get a bit out of hand

-If it is possible to somehow incorporate memory of how certain metatag parts of the song like verses, etc were sung to later reuse them because as it stands right now if I want the Chorus to sound the same I have to make sure I reuse the chorus in every song extention otherwise Suno will forget its original flow

[loud nimbus]

Suno AI dosnt understand 'Pinch Harmonics' for Nu Metal. Maybe training the model in this would be a great addtion.

Incase you not sure what I mean...
A good example would be the intro to the song Take my scars by Machine Head.

[brisk nebula]

Didn't refresh for me. Stuck at 5 credits for a week now

[soft harness]

Interesting, try reaching out to [e-mail] and find out if that is normal or not (likely they will give some credit if that was not suppose to happen)

[tardy anvil]

i suggest we remove this 'guide' lol. a lot of new users try to chirp. then a bunch of people have to chime in and tell them about the site. server guide should be updated to tell them.

[hushed jungle]

The ability to download stems which are essential for mixing, mastering, sync licensing, use in video, tv and advertisement production.

The ability to download the track and stems in WAV format. WAV is essential for professional level commercial use.

The creation and ability to download midi of each individual track in the song individually.

If you can do that you will have advertising agencies, music libraries, music supervisors, producers, film makers, production companies, trailer houses, using Suno.

I don't know how difficult it would be to implement midi generation but implementing the creation and download of stems in WAV format should be relatively easy and inexpensive to add to Suno.

[sage mirage]

Suno doesn't use STEMS at all though, so there are no stems to give.

It's all master, all the time

However they are aware lots of people want that, but it will be a different sort of solution to get to that point

[tepid trench]

Wav/stems/midi are suggested almost every day.

I hope eventually a more organised suggestions/bug reporting system is implemented.

[tulip steppe]

add more voices

[potent folio]

A tokenizer tool that tells us how Suno is breaking up a particular prompt. I know there is some LLM post processing obviously, so you wouldn't be able to include that or people could circumvent it, but knowing how the model sees it would help make connections with regard to what, specifically, in a given prompt is working

I know that's a really specific ask

and not relevant to basically everyone

[sage mirage]

there are dozens of us though

dozens!

[haughty bluff]

moderate song spam in #💬┃general-chat and limit it to #🐣┃suno-showcase maybe

[potent crane]

The purpose is that in the library it's status is different to others because of the restriction in identifying it as a "full song"

[worn atlas]

Suggestion : Make the distinction between clips and songs and add tab for "full songs".

[potent crane]

The only suggestion that is even remotely possible is wavs, but even then it would mean that suno (or whoever) will have to start again from scratch and create a whole new dataset comprising of wavs, because the current dataset has been sourced from mp3's

[sage mirage]

That shouldn't be how it works at all

The issue is the data it's trained on wouldn't benefit from being exportable as .wavs

As in, the output, right now from Suno, can't 'fill' a .wav

You'd just get a giant file with the same quality

[potent crane]

But that is the reality

[sage mirage]

audio is audio, waveform is waveform

Since Suno doesn't talk about what or how the model is trained

[potent crane]

What the heck does that mean?

[potent crane]

suno doesn't talk about it much, but we DO know that the source was mp3's

[sage mirage]

and I am saying that the file format doesn't have to matter

audio is audio. If I play you audio, you don't immediately know if it's off an actual record, a CD, a live band, whatever

[potent crane]

The file form matters, mp3's are mp3's and wavs are wavs

[sage mirage]

and if you dump both of them into audacity you get waveforms

[potent crane]

It's like saying a car is a car, yes, any car will get you from A to B, but a better, faster car will get you there quicker.

[sage mirage]

Suno could give us all .wavs right now, it's just not useful

The devs have stated this mutliple times

tony specifically

[potent crane]

Yes, they could convert the mp3's into wavs, but they wouldn't be wavs

[sage mirage]

I think we are having a fundamental misunderstanding of what we are debating about

[potent crane]

You have a fundamental misunderstanding of what an mp3 is as opposed to what a wav is

[sage mirage]

As stated by Suno themselves: .wavs are possible but due to the fidelity of the training (not the file format) there is little reason to offer .wavs. You would just get a large file of the same audio quality as the .mp3

[potent crane]

That answers your question

[sage mirage]

As the newer models come out, trained on higher quality, then it starts to shift to make more sense

[potent crane]

There's no point in suno offering wavs if all they are are tarted up mp3s

[sage mirage]

[potent crane]

That's what I said ten answers ago!!! They would have to create a new set of training data from wavs

I would LOVE for them to do that! But it won't happen unless they start again

[sage mirage]

They wouldn't though, is what I keep suggesting.

The file format does not matter, not in the slightest

[potent crane]

Well if you know they wouldn't, why are you suggesting it?

[sage mirage]

It matters when Suno can output to 'fill' .wav, if you'd like

[potent crane]

The file format DOES matter, to musincians AND to listeners

[sage mirage]

Right but the reasons why is the same reason why you want an ultra high res photo

So you have more 'stuff' to work with

But Suno outputs masters

Not stems, not midis

That master is beholden to the data used to train the model. It can only output what it knows

[potent crane]

Look, I d/l mp3's from suno, done it hundreds of times, I convert them ALL into wavs, but I know they are not real wavs, but I choose to convert before uploading to Sptoify, if you want to do the same it takes ten seconds to do AFTER you have d/l from suno

"That master is beholden to the data used to train the model. It can only output what it knows" - is what I have been telling you all along

[sage mirage]

Yes and what the model was trained on wasn't 'mp3s'

It's trained on 'music'

an analog format

[potent crane]

It was trained on mp3s

[sage mirage]

[potent crane]

You clearly have no idea what you are talking about, I am done trying to explain the obvious to you, I will let someone else have a go if they want to

[sage mirage]

There is no reason to be rude my guy. We're just having a chat

[tepid trench]

Nobody knows what it was trained on (besides the developers.)
The only part we know for sure involves MP3 is when you download the file at the end.

[potent crane]

Not being rude, just having a chat

[potent crane]

It was trained on mp3s

[sage mirage]

Clearly

[tepid trench]

Who told you this?

[potent crane]

It's been mentioned on here numerous times

[sage mirage]

They are under the assumption that since we get .mp3s out of Suno that means it's only been trained on them

But again, tony, one of the devs, specifically has said multiple times that .wavs are things it can output right now

and we'd get the same audio as the mp3

There is no point to offer it since the training fidelity doesn't warrant a .wav

That most certainly will change and thus we'll get to the point where the outputs we get form Suno will be effected by the file format

However that time is not now

The quality is a fidelity of training data that's gone into v1, v2, and v3

Maybe that changes with v4, and probably will. It's gone up each iteration. Reminder that v1 wasn't even stereo

I think maybe even georg has mentioned it

THAT SAID
I welcome the quality improvements and that bump, but I am really after infill/inpainting

2024 is clearly the year of AI music, and it's already awesome

[tepid trench]

A brief search of some of the regular devs mentioning "mp3" or "wav" doesn't reveal much. Closest thing I found is this.

[sage mirage]

ah yeah georg has talked about it

I was only around when tony said the same sort of thing

Based on that it seems the interface is .wav -> mp3

In how Suno does its thing

Audio is audio is audio though so

What is the itunes file format? Or that Sony one even from a decade or so ago

[tepid trench]

Coughs
I don't really have anything else to add other than to suggest lossless audio formats. Since we're in the suggestions channel 😄

[glass lichen]

Most DAWs will convert audio to midi

For now I use Lalal to stem things out if I need to, but usually I just recreate melodic phrases or ideas to make them my own

[tardy anvil]

I'd love the ability to scrub through in miliseconds for the sake of more accurate extension timing.

[sage mirage]

If it helps at all, the devs are aware of this desire

It was one requested almost immediately when we finally could 'continue from'

[tardy anvil]

Oh sweet!

makes sense!

[sage mirage]

tony discussed it directly

Along with negative prompting

Though this was actual months ago so who knows what that all looks like now

At the time negative prompting was phrased like this, from tony:
"Would you use negative prompting if it cost double to use?"

and we all said "yes of course"

then they posted the homer-into-the-hedge.gif

[tardy anvil]

ohhhh boy

im stoked to see what v4 will bring. its gonna be a fun ride either way

[sage mirage]

infill/inpainting probably, gerog has talked about that a lot

and, if like before, a whole host of other things. Something nobody expected from v3 but we got was the initial generation jumped to 2 minutes

Originally we were used to 1:20 in v2

[candid edge]

Can we get up to 3 minute songs and stems please!

[tranquil lantern]

devs, would it be possible to search for songs by saying "a song like tame impala" and you automatically search the genre that tame impala is and its mood and etc, and basically do the work for us?

because i personally love to find songs similair to a song i already love and am familiar with

like something similair to what youtube does

which is recommending songs that are familiar to your liking

[worn atlas]

Search is available now
Thanks Suno dev

[light ruin]

Make the new image generation like 1 credit or something xD

[worn atlas]

Search feature is available using song name and tags https://suno.com/search
Please add also search by author name and search in our own library

[tranquil lantern]

next step for suno is generating videos for the songs, hit up Chat gpt Sora to get it done devs!

[woven magnet]

A get whole song option that lets you specify the start and end of the final generatiom

[solid sage]

Bring back the ability to click create multiple times; i usually click it 2-3 times to get multiple first part selections to choose from

[twin bear]

@still parcel The new AI Art tool is cool, however the created or uploaded album art does not play throughout the song, if we share Suno song links.

[strong swan]

its just like before i guess... to get the picture in the video you would have to rerender the video

[twin bear]

Just a suggestion to allow for the user's uploaded image to play throughout the entire song rather than different artwork show segmenting each extension made.

[still parcel]

Will be fixed, yes

[finite ether]

Is it possible to increase the number of characters in lyrics section (custom mode). At fast tempo and at times even in mid tempo, Suno completes singing the verses pasted and return to the verses before. If you could increase the number of words that can be pasted in the lyrics section, it will be helpful. Not sure how difficult it is but would be very helpful. Thanks, Kumar

[remote rock]

Could we be able to zoom a photo after submitting it

[fair cipher]

Would it be possible to allow people to gift subscriptions to other people? Or maybe paid users could "tip" credits to other users if they like their stuff, which would let the tipped person make more cool stuff.

[kind cliff]

Would it be possible, when we extend a song, to timestamp what portion of the song we extended from? So let's say it's a 2-minute clip but you start at 00:17, on the Part 2 extention, it wouldn't just link you to that original clip but would also indicate where you clipped it from. That way you can easily tell if you want to come bak to it later, or if you're generating multiple prompts around a timeframe - say, 00:43, 00:44, 00:45, to get a certain continuation to work - you can tell what prompts went with which timestamp

[hybrid crane]

@fathom condor you are not allowed to post your profile page as it is soliciting aka self promoting. You were already warned in #💬┃general-chat

[umbral grove]

why custom images for pro users thats so unfair why cant non pro users do it aswell

[sage mirage]

Moderation, presumably

If someone uploads a naughty image, ban that account. It also has a cost associated with it, since a paying customer probably cares about not getting banned

To be clear I am not Suno, and am just speculating

[umbral grove]

honestly they should allow it for non pro users too its just moneygrab i guess yeah im speculating too

[sage mirage]

I...did you not read what I wrote?

There is significant overhead in allowing users to upload images that has nothing to do with gating it for cost reasons

Moderation is a real, giant, massive sink of time and effort

I'm sure they would be glad to allow free users, but if a free user abuses the upload function they can just make another free acount and do it again

If you have a paid account you lose everything you paid for, thus making it much less likely you'll abuse the privilege

[hybrid crane]

Yup.. pure common sense 😄

[sage mirage]

You won't do it, I won't do it. The next person might abuse it

And hey: Maybe we'll get to the point where free users can upload images

[umbral grove]

i know moderation at all i get it i understand that but to me personally i think its a bit too much i'll be okay if it was a lower price per month

[hybrid crane]

They even steal music.. so yeah

[sage mirage]

On the plus side: You still do get to create music so that's cool. You haven't lost anything

[hybrid crane]

I personally thing they are dirt cheap.. there was a time with Suno you got just 100 credits for 10 dollar

[umbral grove]

if you allow uploads to youtube with credit ill be okay with that

[soft meteor]

You are currently using a service for free, which refreshes daily. I think thats enough generosity. The "nice to have" features can stay paywalled. What do you lose? An image on the site you are creating on. You can still download and upload it elsewhere with your own cover.

[umbral grove]

if you lower the price by half ill be okay with buying it

but first explain to me the difference between premier plan and pro plan

[hybrid crane]

Amount of credits mainly

[soft meteor]

7500 credits I think.. also this is suggestion channel maybe lets move elsewhere?

[umbral grove]

so its the same as pro plan but only more credits

[hybrid crane]

Yes 😄

[umbral grove]

oh okay so no other privelages then

[mortal anvil]

Some feed back on todays feature that allows users to add images to thier songs and playlists

THe attatched image is what i have done with one of my playlists, I'm essentially treating it as an album.

I'd like to see an option for creating albums, and instead of uploading the same image several times for the cover and each track (which is going to take up server space), let user set the image for all the songs in the album.

[hybrid crane]

No, they are not making a "money grab" there by dangling a carrot if you pay more.. besides credits 😉

[sage mirage]

I'd like this as well tbh. Even if just to 'blank' all the songs in the list with custom art

Can always change it later

[teal perch]

@mental horizonIt would be very cool if SUNO could create similar music to music already created by a professional composer. Something like cloning of musical style and sound. So you could upload any music track and SUNO would analyze it, or would that be a copyright violation?

[robust jewel]

The act of uploading copyright protected works would be the violation there

[umbral grove]

If I uploaded an image of a custom flag with it's own conlang would that be offensive

[drowsy dove]

Let free users be able to change their profile pictures - just not the song thumbnails

[inner rover]

@hybrid crane I like the new suno updates, but would be cool to add a feature along side of where you can upload new images for songs, a feature where you can update the lyrics in the song, because sometimes they get overlapped or all messed up ya know.

[hybrid crane]

If you are a paid subscriber, you can upload new images now. I believe the rest is being worked on. 🙂

[inner rover]

@hybrid crane I certainly am a paid subscriber, already used all the C’s but that’s cool, I did notice the new image updater but wanted to ask about the lyrics too. 👌

[hybrid crane]

I want the same, so let's hope they can add that soon 😄

[twin bear]

@still parcel I played around with the new Search Bar, and noticed that it's missing Sort and Filter options. Currently it defaults to a large spread of results without the ability to further define the search results.

[still parcel]

we’ll keep working on it, I agree that some filters and sorts would be a really useful addition

[full tulip]

Hey all! I just wanted to share a few suggestions...not sure if this is even possible, but it would be cool to have an option to essentially "upscale" the audio and re-render it at a higher quality WAV file.....this would work super well with older V2 and even V3 versions. I'd also love to see audio in-painting and an option to control how long the whole song will be when generating would be awesome too. Again, not sure if this is even possible, but would love to see it in the future! Keep up the great work!

[long bloom]

I just ran into a creative limitation that could be fixed.
I generated a version of my song I really liked.
It started singing the lyrics instantly without any prelude.
I would have liked a prelude, but there was no way to generate one, and the one I generated was "The One".
You can't tell it to continue from beginning, or add on the the beginning of a track.
That is something to fix? I don't need an explanation of why it's not practical. Just take the suggestion for what it is, a suggestion.

[full tulip]

I had that issue too! that would definitely be a helpful feature

[leaden marlin]

is there a way to change my profile?

[sage mirage]

Are you a paying subscriber?

If so:

[leaden marlin]

no sorry

[sage mirage]

Then you aren't able to change your profile picture. It is currently only available to paying subscribers

Sorry 😦

[leaden marlin]

ok. thanks tho

🫂

[hollow locust]

The difference between pro and premiere should be more than just the amount of credits. For example, higher tier users get longer song generations. 3:00 initially, and 1:30 for extend.

And be able to fit more lyrics in custom.

[robust jewel]

The length of clips is not limited because they want to...

They have expanded the generation length with every new version and have said that the goal is to generate an entire song in a single go.

[sage mirage]

I wouldn't be surprised if 3 minutes is target for v4

It was 1:20 for v2, then jumped to 2 minutes with v3

Would sidestep the deterioration issue when you continue clips, plus if we also get inpainting/infill, well

That would be a huge lump of awesome imho

[versed loom]

I had a chance to play with Suno for about a month. So far everything looks good! I understand that a lot can be improved. Some suggestions that I believe would make Suno even more useful.
1 Ability to reuse the previously generated songs as a "base" or "influence". In some cases this can be achieved by cleverly using the Extend feature but it would be good to be able to create several variations of the same song.
2. Ability to change the generated songs. May be based on the feature suggested above. Useful for situations when everything is good but you would like to change some lyrics or instruments or the music itself.
3. Ability to upload drafts. For example I whistle or sing a music theme and upload it to Suno. Suno makes a song based on my uploaded track and the prompt. I understand that this may be a slippery road since some users may upload existing songs which may violate copyrights, but I am sure there should be ways to mitigate this.
4. Ability to indicate chord progressions or specific notes.
5. Ability to define stricter rules for overal structure. Would be good if Suno would understand the commands like [Intro] [Solo] [Outro] better. It often ignores them.
6. Ability to add negative prompts. For example even if I use "male vocal" in the prompt Suno often adds a choir on top of it. Would be good to be able to write something "choir" "back vocal" in the negative prompt field so that Suno would not add them. This is just a single exmple how it can be used.
7. Better understanding of "instruments" in the prompts. eg: Hammond organ, jazz organ, etc.
8. Also would be perfect to be able to add more advanced attributes in the prompt eg instead of just adding [Solo] I could indicate [Solo, Hammond organ] and Suno would create a solo using Hammond organ as an instruments

[elfin pelican]

add monkei🐒

[long gazelle]

Hi please make it possible get rewarded by hear songs of people, I not sure what the prize will be it can be point to reddm for more songs points or anything really you will know what to do, just want to bring the idea of make it rewarding also to make users want to collect that points, no replay rewarded, but can do badges like (hear a single song 10 times. It's for make people that hear songs they love will get something because of it) I know it will be a big work, also need like more community pages or ways to find users and songs.

That is it from me 😁 hope only for good. I love how amazing is working
(Some bugs like use site in mobile not let me press the place to make my own lyrics)(do work on Desktop mod)

[atomic sedge]

Would like to toss in a quick suggestion:

Add ability to change filter/arrangement of playlists while adding a newly created song to a playlist.
this could be options such as:

• move 'recently added to' playlists to top of Playlist window when adding to playlist.
• move 'recently made' playlists to top of Playlist window when adding to playlist.
• alphabetically (or numerically) arranged to top of playlist window when adding to playlist.
• create folders within playlists for certain projects we are working on or just as a bin of things used in a project that has been made.

[brazen prism]

Hi @hybrid crane - from one community manager to another, just want to say that you’re doing an awesome job! 😎

Would love to see some #SunoSundays or #SunoSessions sometime soon where we can just chill and make music with each other/share using the video or audio feature on Discord (maybe this exists - I’m a newbie)

[hearty star]

https://tenor.com/view/yoda-star-wars-gif-424824508006372153

Thanks Suno team for implementing my suggestion to allow regeneration of images. Better yet you upped the ante and allowed me to upload the awesome images I was already making for the songs ^^. Check out my public playlist for an idea 🙂 Keep up the good work but do expect a bug report from me later as I've already noticed a few issues since the update i.e. (extending a song is especially difficult to do now compared to the old way on Chrome + Android).

[hasty swan]

50% of screen real estate is my username and Edit / Share button

[supple cloud]

My wish no 1 is a way to edit lyrics after you've created the FULL song so that they follow the song and not as now when Suno decides to add or remove words from my songs. It gets crappy for the video

[supple cloud]

My wish no2 is that I would be able to change the image in the video so that it can reflect my songs image

[fossil trail]

This may be a stretch, but maybe a feature to find the instrumental version of a song and use it to layer other songs? Or take an instrumental song and add lyrics to it?

[supple cloud]

My wish no3 is that you give us a folder to put FULL songs in so that the aren't shown with other songs. Or a flag to filter like thumbs up/down

[hybrid crane]

Thank you @brazen prism but atm I am not Suno's community manager. I am a moderator and guide, but who knows what the future will bring. 😉
I love your ideas though. 🌼 💗

[open rock]

Wow, so many great suggestions, and yeah we are implementing some of your suggestions 🙂

[desert elbow]

My wish is to have an option of male or female or group as it would make it alot easier for me

[feral aspen]

My Kingdom for a strength slider

[twin cypress]

I suggest letting me sort by completed songs, let me sort by month and week.

[worn atlas]

You actually check most of the boxes for the related open positions at Suno for community management .

[hybrid crane]

That's really sweet of you to say and notice. I did apply so who knows 😊 🤞

[worn atlas]

Great. We will really almost all appreciate that.

[upbeat mantle]

on the Refil credit area

There should be a limit on the amount of times that u can refill the Credits per day so that you don't over spend your greens

because i noticed that it´s pretty easy with the 1k credits

[tepid trench]

Hmm. I don't think they should prohibit you from buying more, but perhaps a reminder of how much you have paid today/this month if you've topped up a couple of times already.

[sand burrow]

Is there API access? I'm curious what parameters could be tweaked.

[hybrid crane]

It's on the roadmap. We have no idea when it will be released though.

[sand burrow]

thats good to know. being able to weight it more to a seed you like the style of would be great

[tepid trench]

Maybe allow AI image generations for free users (limited per day?) so free users can still make their content visually interesting (in trending/new/search.) Continue to allow image uploads only for paid subscribers.

[frosty breach]

I do get that you need resources ya-di-ya, but after an announcement of how much money you made?
The first thing you announce, is a premium feature and it feels like a kick in the stomach. A basic feature for most other platforms at that.
There isn't necessarily a feedback portion anywhere, and you might be able to gleam a suggestion out of it.

[upbeat mantle]

Ya-Di-ya ?

[frosty breach]

bla bla bla.

[foggy cedar]

You could include a preview feature. It would allow you to get a preview of the result and modify the lyrics so that it rhymes better with the sound.

[brisk nebula]

So I got an answer from support saying that by subscribing to a plan you loses the right to the 50 daily credits. Wow. So by subscribing you're only buying 950 credits since they rest is 50x31 days. Nice maths suno..

[cold garden]

Please add the ability to set the maximum length of a song at once. The current 'extend' is too difficult to control. Two minutes at a time is too short.

[fair cipher]

Personally I'm fine with stitching shorter bits together, because then I can use MY vision instead of just accepting whatever the AI gives me. AI is notorious for drifting off course halfway through. But, to be able to have a long enough memory that it can remember what the chorus and verse should sound like after you do an instrumental section, THAT would be awesome.

With text AI, they're developing techniques to call back things that have gone out of context, using embeddings to determine what is relevant enough to be remembered given the current situation. Perhaps something like that could eventually be done with music AI. Like when you put [verse] it analyses the previous parts of the song, finds another place where that tag was used, and puts it back in memory even if it happened 2 minutes ago.

I don't know if embeddings are even a thing with music AI though.

[feral aspen]

For music, it may be useful even if it doesn't work, as long as it changes the output in an interesting way, making sounds you can't otherwise prompt for. Can even lean into it, force forgetting at times, introduce paradoxical contexts (from the wrong song or singer) and see what terrible or beautiful thing comes out. I don't think you even need embeddings for the basic idea

[fair cipher]

Well, currently you can't have a 6 minute rock anthem with a part that only occurs in the intro and outro, because by the time you get to the outro it's long forgotten the intro. But if you could label it with [tag] then the system could put it back in memory and use it to generate something similar again.

Of course that means [verse 1] and [verse 2] would be completely different from the AI's perspective. So you couldn't really number things anymore, unless you wanted verse 2 to have a significantly different pattern from verse 1.

But, if embeddings exist, you could use them to basically tell the AI "do this melody except with a trumpet this time instead of a violin"

Follow the pattern in one aspect, but change the pattern in another aspect.

[twin bear]

The Suno music player needs the ability to cast music to smart speaker devices and other connected devices.

[wispy sinew]

I love your tool, but it has one unbelievable flaw, it cannot generate slavic hardbass 😆😂

[twin bear]

Soviet Techno works.... 🤷

[candid edge]

Please offer stems of the songs

[hybrid crane]

They want to offer us stems in the future, when the quality is up to par. 😉

[candid edge]

yay!

[bright topaz]

when I extend the song sometimes the tune changes they sing very differently the same type of lyrics I give them in extent please add prompts Like [Verse Repeat], [Verse 2 same melody as verse 1], and [Chorus Repeat] so AI can generate Consistent song

[feral aspen]

When you extend, to maximize consistency, be sure to 'Get Whole Song' before extending. (And the opposite is true, avoid doing this when you would prefer to get a result more different than the past song)

[bright topaz]

I tried but sometimes the result got very weird and sometimes they didn't even sing the lyrics I gave them or got different results that why I want that type of prompt so AI more narrowed down to get the result I want

[feral aspen]

That can happen, though in my experience Get Whole Song is not a common cause of 'hallucination' (singing the wrong lyrics). Most often hallucination can be improved by changing the Extend From time. Backing up from the end of the clip 5s, 10s, 15s, etc.

[bright topaz]

Thanks for the Reply I will try that but what if that also didn't Work?

[feral aspen]

Sometimes there isn't something that can be done, and you have to back up one whole clip. But hopefully this improves in future.

[gentle zodiac]

One message removed from a suspended account.

[spiral acorn]

This has probably been mentioned previously, though it'd be cool if rather than having "whole song" and "parts" if we always extended the "whole song" but specified how far back to use for memory. The default would be the entire song, but you could also move this up to say even like 5 seconds or 1 second etc

[wispy sinew]

Average voice clone enjoyer I see, ah yeah yeah, I'm someone of that sort of guy myself.

[brazen zealot]

A couple updates to the web interface have been driving me crazy:

• When I’m making my own songs, the last thing I want to hear when I’m doing it is whatever is in “Trending” after my segment finishes. Being extremely polite, I’ll just say that it’s jarring. There’s a time and place for other people’s stuff, and “from your song” most certainly is not one of them.

• Up until a few days ago, you could create/extend and stay on that page / press the button to your heart’s content. Now, you are redirected back to your library after you press the button. Since the average segment seems to take about 50 tries to get something usable, this makes the change pretty tedious since you have to re-setup at the exact second you want to go from with your prompts/lyrics a few dozen times. Wasting my time means I have less time to buy and use credits.

UX is important. Don’t “optimize” away the usability.

[gentle zodiac]

One message removed from a suspended account.

[prime stirrup]

Can we please have song folders in the create tab? maybe it's already asked for but it's kind of tedious with my ADHD brain to keep my flow. i could be busy with 2 or 3 songs which gets kind of cluttered to be honest. Thank you in advance.

[fair wind]

I think this is already implemented

At least the flag

[supple cloud]

yes but you cant hide your full songs in the list as you would if you moved them to a new folder

[fair wind]

Ohhh yeah that would be cool to have!

Like a filter

[safe ingot]

The change to the mobile site's create section I would say is a downgrade, or at least it was not thought out the best. It has made it so much more difficult, because every single time I click on create, it takes me out of that screen and to the track list, meaning you can only roll ONE extension at a time now. It then totally clears out the track that you were trying to extend from, so you have to go back to the library and find it and click extend again, then type in the timestamp again. Please add an option to not auto-switch after clicking create, or implement a button to switch between the two screens?

[viral ember]

@open rock last tag I promise, hope you can reply (or acknowledge!) any hope of better instructions sets or something to allow Rondal / Madrigal / round muisc form. So we can do lyrical stuff like this... Scarborough Fair... (Obviously it would be way down the line)! Technically should be challenging.

[cyan umbra]

I would like to suggest a partial rollover of paid credits. I understand why you couldn't do an unlimited rollover, but I could think of a few reasonable compromises.

The easiest would be rollover up to one month's credits. You get your new credits added to the last months and then take the lesser of that value and 2x your monthly allotment.

Or, alternately, you lose 50% of whatever you have left on credit award day, but not all of them.

I doubt either would actually cost Suno much in extra generations, and it would feel better for those of us who overbought for an annual sub. Or who have life stuff come up to come back with a bit of extra to get back into creating.

[green bough]

it would be cool of we can separate the lyrics from the music and save only the lyrics audio or only the music audio without the lyrics or both

[open rock]

Hmm that might not have a good instruction yet and I doubt that is on our priority. I will check with our team

[brazen prism]

Interestinggggg. Yea - absolutely assumed. Well, fantastic job as a mod! 😊

[viral ember]

I'm aware of how hard it would be as things stand, can wait till 2025 😉

[upbeat mantle]

maybe this aint necessary but i was thinking about a extra box under create song one that got all the [nametags] and meta ones that works

[soft meteor]

This is an AI application no one really know what works under what circumstances

[worn atlas]

I was thinking about another possibility. Unused credits would never be lost but could only be used on a daily basis like free subscribers, with a maximum amount (50, 100, etc.). In this way, (1) Suno can predict the maximum load per month, and (2) paid subscribers will not suffer from having no credits."

[tepid trench]

This seems reasonable, that way subscribers don't lose anything. Alternatively they could allow rollover credits to be used but given lower-priority generation.

Not sure how they'd handle the situation where a user cancels their subscription and then has credits leftover after the month (do the credits remain? can they use them? should they be removed? do they get treated like "free" credits in terms of song owernship?)

[worn atlas]

Yeah! I agree, allowing rollover credits with a lower generation priority could be a good alternative. Regarding leftover credits after a subscription cancellation, I think it would be fair for users to use them but with the same limitations as 'free' credits. This could include lower priority and restrictions on song ownership. Just opinion.

[arctic whale]

I find subscriptions too restricting. Why not offer packages with different quantities of credits? That way everyone could buy as they need. I'm not inclined to commit that long and lose credits because I don't need that much (in that short amount of time). It's what's holding me back to subscribe (that and the current bugs).

[novel compass]

It would be amazing if there was an option for instead of generating two independent tracks or extensions at a time, it would use one for just the instrumental and the other for vocals. Then an option to individually regenerate one or the other to get things just right would be the icing on the cake.

[forest spruce]

So, when i was a free account i got 50 credits a day. now that im a pay account, i get 2500 credits. but when i use those up i don't get any per day. would be nice to have those... and a monthly option between 2500 and 10000

[leaden marlin]

I think it would be a really good idea to have a channel just for playlist cause it feels like they crowd a bit in suno showcase

[sage mirage]

@scarlet grove If you don't want to share your song then don't share it

If you do want to share it then people are going to see your prompt

At least the first section of it

[scarlet grove]

@sage mirage let me at least post my suggestion here first 🤣

[feral aspen]

Suggestion: NAME THAT PROMPT. You're presented with a series of songs, let's say 10 songs in total, with the prompts hidden. As you are presented with each song one at a time, below is a multiple choice selection of different prompts, all are at least somewhat plausible (not obviously wrong). You can play the song and stop the audio at any time. How much of the song do you have to play to guess the right prompt? You get one guess for each song. Score is combination of correct guesses, each correct prompt is worth less the longer you had to listen to it before guessing. This is mostly just a humorous late night night suggestion, but only mostly.

[scarlet grove]

Feature Request: enable prompt used to be hidden when making generations public within Suno.

[sage mirage]

For the record I am against this

You could always download your song and share the mp3 if you'd like to hide the prompt

Feels extremely not-Suno to hide prompts. It's just a prompt

[scarlet grove]

I will share publicly to YouTube rather than disclose the prompt here

[sage mirage]

That is something you can do right now, yeah. Everything is private by default

[scarlet grove]

It should be optional, is what I am saying

[feral aspen]

You can also share the audio file in Discord, if you prefer.

[sage mirage]

Personally I share everything, more share is better across the board

[scarlet grove]

In your opinion

Which is not bad, but we can expand the option

[sage mirage]

Yes, but you do not have to share, is the thing. You already don't have to share your prompts. If you don't post them then they won't be shared and you can keep it all private.

[scarlet grove]

Well, if direct monetization becomes a thing within Suno then I am sure that this is a topic that will need to be explored more deeply.

[sage mirage]

I trust my ~talents~ to carry me forward even if everyone knows all my prompts

[scarlet grove]

Do you have any IP in Suno? Is your prompt special?

[scarlet grove]

Yea but if you show everyone your pants, they know what colour they are 😅

[sage mirage]

I do wear pants when I leave the house yeah, I don't see what you mean

People see my face too

[scarlet grove]

Yea but your face ain't private

In other words you want to keep some things private

[sage mirage]

Neither are my pants? See I really don't understand what you are trying to say

I mean I get what you want

You want to be able to share your songs but share the prompts used to make it

[scarlet grove]

Another Feature Request: ability to split audio upon download.

[scarlet grove]

But not share the prompt used

[sage mirage]

Exactly, and I don't agree with that at all. That doesn't 'vibe' with how this all works

Especially since, right now, there is no monetization and even if there was, your song is yours, even if we both use the identical prompt

You will never be able to make the same song I did using the exact same prompt

And I wouldn't be able to make yours

[scarlet grove]

Can the general commercial terms expand to the prompt used by the user?

[sage mirage]

Copyright a prompt?

[scarlet grove]

Well, I dunno. Either way, like people have pointed out, it is possible to simply share the audio. But if I can share my generated music without sharing prompt then it's also fine. I suspect the team are using the prompts as part of the search functionality.

[hearty star]

I vote that unused credits from premium subscriptions should never expire; but the requirement to maintain membership to use them should stand. The reason is this: people work a lot and sometimes don't have a lot of time for this. But when they do have time they should be able to use the credits ^^. In my first couple months of membership I was not able to use all of my credits in time. It even causes stress because I have to rush to use them or lose them and it's a disappointment to see 2-3k leftover credits just go 'poof'. Maybe I have more time in a few days or whatever but then they are gone for good. My punishment for working - or doing life, or whatever. LOL.

[worn atlas]

For suno and as @hybrid crane mentionned this several times they have to handle the workload of the servers

[strong glade]

Need a seperate new section on the website. So many of my songs get blocked by automod, and honestly thye really aint that bad. its not like im saying im gonna go murder someone or blow up a building.. sooo Maybe a 18+ section where users could post there songs that dont pass auto mod?

[trail steppe]

hope Suno can give option to manually select language example ( chinese has many dialet such as mandarin, hokkien, cantonese ) , number of singer, and gender

[trail steppe]

agree on this

[trail steppe]

so far chinese language AI only can only prioritize mandarin dialet

[soft harness]

I remember someone having a song use the style "Hong Kong Cantonese" for a good Cantonese accent (https://suno.com/song/a1fb83c2-ec7c-444e-bb13-fcfca6bdc36f), maybe you can try something similar with another region for the accent you want if you didn't try that yet

[trail steppe]

thanks forsharing, i think the keyword to triggle cantonese is " Hong Kong Cantonese "

[azure crypt]

if i post my playlist link to twitter, it has no image. https://suno.com/playlist/43d17012-bc5b-4fb6-8903-530e5f8264ed

(also should show the playlist image?)

[serene trench]

Здравствуйте!

Я использовал Suno для создания нескольких композиций и хотел бы поделиться своими впечатлениями.

1. Я заметил, что Suno создаёт треки по 2 минуты, и можно создавать продолжения треков. Однако, когда я создал с десяток различных вариаций и пытаюсь найти нужное продолжение, я путаюсь в куче треков. Было бы полезно создать дополнительное окно, куда бы падали продолжения треков, и где можно было бы легко разобраться с несколькими вариациями конкретного трека. Где бы они сохранились и я к ним имел быстрый доступ.

2. Также я заметил, что у треков часто нет вступления. Это может быть важно при создании трека, поэтому было бы полезно иметь возможность добавлять вступление. Мне пришлось выковыривать партии с волны и вставлять их вперед, а затем сводить всё это. Это очень неудобно.

3. Ещё одна проблема — отсутствие склейки вариаций до полного трека. Я хочу получать полный, автоматически склеенный трек для дальнейшей обработки. Сейчас мне приходится склеивать треки в программах, и это занимает много времени.

4. Также хотелось бы иметь возможность выбирать вокал и менять его в треке. Было бы здорово, если бы AI мог анализировать любой или мой вокал и использовать его в песнях.

5. Ещё одна полезная функция — возможность получать трек в разбитом состоянии, по отдельности каждый инструмент по дорожкам, для последующей сборки и обработки в программах. Такая возможность была бы полезна для всех, кто пишет музыку.

В целом, я планирую записать альбом, создав на основе своих текстов песни при помощи вашего AI. Результатом обязательно поделюсь. Однако пока на выходе получается не совсем хорошая запись, и мне приходится дополнительно обрабатывать её в Adobe Audition и FL Studio, неплохо было бы иметь несколько инструментов по обработки записи на сайте.

Я очень надеюсь на крупные дополнения и увеличения настроек для полномасштабной записи своих треков и выпуска их даже на аналоговых носителях в качестве подтверждения качества генерации! Спасибо, я очень рад открыть ваш продукт!

[half sail]

Dear Devs, i really like the search function. Can i also have that please for all my private songs only?? Thank you!

[hybrid crane]

It's coming 🙂

[potent folio]

yea, it's just anti-ML community broadly imo. This whole community is unique in that even private tools like Suno are heavily based on open source cooperation

I think making an extra layer of difficulty in downloading people's songs might be nice

but prompts should stay. If anything I'd like it to be an option to display more granular detail

bc it can be a bit tedious to transfer all parts of a songs creation into a playlist

but now that I consider it, that wouldn't be as helpful since you need to see failed parts of the tree

to understand what actually goes into a song

but in any case, share your prompts if you find something legitimately amazing, all the major people here do, and there's no loss. We aren't in competition, really, it's not a "I can listen to this OR this," for users, it's an AND. And if you provide good prompts, the community will be more interested in your work anyway

And on Spotify were tiny, so our main competition is big money, and we need to work together if we want someone to break through and pull the rest of us up lol

---

OK, an actual suggestion, albeit broad, did occur to me.

So, Idk what your data is on mobile users, but I will say the existing mobile interface still has some bugs. For example, when adding to a playlist, you will often add multiple songs to that playlist in your song list, as just scrolling will sometimes select them, and it's not readily apparent.

In any case, I can't be the only person who likely uses the "like" function as a "save for potential continuation" function, and this could have an impact on your user preference data's quality in theory, especially if people are experiencing and making music that they don't actually enjoy but rather are just seeing what the model can or cannot do.

for example, back a few months ago I made a 7 minute song that was just a machine whirring and clanging repetitively. Lots of likes required to keep track of continuation.

Now, this still likely has a positive impact bc presumably it's tied to the prompt information ie it will still be usable preference data that improves adherence

But it does push more liberal likes. This could be a good thing or a bad thing, I really don't know, but it might be worth examining the mobile version more closely if your data indicates a lot of people are on mobile

[half sail]

good. Cause i'm really bad at organizing my songs 😄

[potent folio]

same. I literally sometimes am going back 100 pages.

[tepid trench]

This. I was just talking about the "like" feature in the technical-discussion channel, because I only use it to bookmark generations that I might want to revisit later. And then treat the rest of my library as a big trash pile I can search through later (whilst rarely bothering to actually trash anything.)

[azure crypt]

Regarding search... i would love a TSNE point cloud map of all my creations... so i could easily find all the ones in a certain genre

[full ridge]

I agree with @sage mirage that I find it really useful to see what prompts people use to get different results. Plus, the way the UI works right now the prompt is like your snippet/lead text suggesting what to expect from a song you listen to. If you really want to keep prompts secret at the moment, I'd suggest using Suno kind of anonymously to create your songs, and share them on a different platform, (e.g. Spotify/YouTube) where you can give tracks whatever description you like - including whether it's even AI or not (I personally disclose this on my channel).

[azure crypt]

I feel like social media previews could be improved for sharing stuff. If i post a link to my profile (https://suno.com/@oilyl), it doesn't show the profile blurb or artwork, if i click it on twitter, it opens a player for the first track there, when it should like to the suno profile

[potent folio]

yea, I trash nothing, too risky

[potent folio]

i wouldn't mind basic icons for specific social media links

like, so your profile can increase engagement and vice versa

[scarlet grove]

yes, this is what I do now with my channels on Spotify and YouTube. I like the surprise of great music with no information for how its done.

[frosty breach]

Let me edit genres on the actual eventual song. Starting off with a classical intro, then moving through metal and dubstep, to end up in phonk will never show the proper styles if you use parts to shift around a lot.

[spiral acorn]

yea that makes sense, as in some cases you might use a starting style just for a few seconds even though it ends up showing as the main style of the track

[frosty breach]

Watch the vid/mp4. It'll basically lines out most of it. I rather have it be a function you have to turn on, rather than a prevention. The only protection against other people uploading our stuff is that we are the only ones who know to create more of the music due to prompts remaining mostly a mystery due to part use. We can go back through the chain if we kept the last part secret before the whole.

[spiral acorn]

it'd be nice if it actually stored all the style tags of every part and then you kind of have the "overall" styles which are editable

i can see both perspectives of wanting to show/hide the prompts and styles. I like the neatness of being able to edit these and clean them up, but I also see how this makes the music less transparent. Even if you can edit them and still see what they actually are isn't a bad compromise. Though I don't mind the ability of having a way to hide these also.

[scarlet grove]

its clear that people want to see the prompt, people will want to make the style of music that you are making (if it sounds nice) but I would like to think commercially, and withhold the information for how for monetization purpose. Might have to go behind the pro paywall.

[frosty breach]

That's bull, an 30 buck fee shouldn't just get you the keys to the diary.

Either it's an open book or a closed one. It's already freaking me out that there's already separation between paid and free, while free is already as limited as it is.

[cosmic terrace]

this is just a suggestion for convinience
i would like to have an option to seperate section the like for parts one is using for building songs on and the like list of other peoples songs ,

so 1 list for thumbs up for songs i (we)
and 1 list for likes on own parts im working with .
🙂

[wind crest]

possible to have a creativity weight slider for prompts, as with visual generators? for those situations where you're in the third verse and it just will not sing the words you're telling it to, and is instead intent on singing the repeated part of the chorus 800 times

[mortal anvil]

suggestion:
work on an avenue for artist to distribute their songs to streaming platforms and help protect the copyright of their songs. SUNO RECORD LABEL PLS

[soft harness]

Music purely generated by non-humans (e.g. AI) itself cannot be copyrighted according to the US copyright office.

The artists who wants their music on streaming platforms do have many distributors they can use the service to distribute music to various platforms but many of them are against distributing AI music, especially those that sound like AI.

At the moment, people can use DistroLock for free to help prevent their music from being taken and uploaded wlsewhere.

[mortal anvil]

my music isn't solely generated by AI i write the lyrics, surely that accounts for something?

[bleak scaffold]

💡 💡 💡 Thoughts 💡 💡 💡
Put all resources into R&D and / or improved models training. Either new techniques should be discovered (maybe combination of different models. I.e. combination of different model types: rhythm model + vocal model + melody model, etc.) OR just a bigger (probably much heavier) models should be trained 💪

ElevenLabs Music is on the corner and if they ship the version they promoted then quite a big chunk of Suno users may switch to ElevenLabs instead 😨

It also feels like there should be a similar technique for upscaling the generated songs (i.e. similarly as it's done with stable diffusion when you apply tiled upscale process on smaller chunks of the image). I think consistency between different chunks of the song is less important then in images (potentially), so the process should most likely be a bit simpler (probably) 🤔

Anyways - keep the great work you are doing and I'm sure Suno will have a great success! ❤️ 🚀 🎉

[soft harness]

Yep, your lyrics can be copyright but the music itself (that is excluding the lyrics) is not eligible. When you submit for copyright, you have to declare the ownership and other info for the separate parts already IIRC.

Now, if you used something like FL Studio to sample what you had generated... it no longer becomes solely non-human...
(And you will have to modify the output anyways, like mastering, to make it acceptable for the distributors)

[sage mirage]

There is a whole lot of unanswered questions around this sort of thing. Just because you use AI doesn't somehow make whatever you use it in invalid. It doesn't work like that.

But until things shake out this question is going to hang

[potent crane]

It should be pointed out that this advice is US-specific, the rest of the world does not have to adhere to US statutes. Having said that, IANAL and the situation outside the US is unresolved.

[timber totem]

I suggest this song 😏 https://suno.com/song/00aafdf9-b6d4-4cfc-9928-8dee1048f442

[soft harness]

Suggest for what? This channel is for suggestions regarding Suno development

[timber totem]

...

I don't think you got the joke

[soft harness]

Nope, please explain it to me why you put it here instead of #🐣┃suno-showcase

Edit: ah, I see. Still, wrong channel for it

[timber totem]

The channel is called suggestions. It's meant for the development of suno. The joke is that I "suggest" you listen to my song.

[soft harness]

This, location matters a lot. For example, many countries are against using copyrighted material for AI training but, broadly speaking, a law in Japan being drafted said all materials are fair game for AI training for companies in Japan (according to https://natlawreview.com/article/japans-new-draft-guidelines-ai-and-copyright-it-really-ok-train-ai-using-pirated and a few other news sites).

Of course, that is just for AI training and not the use of said generated material.

In the end, it will largely depend on your local laws and the laws where the service is being used... so it is going to be quite messy for a couple of years

[fair cipher]

Yeah, I use likes this way, so I can tell which clips are maybe good and delete anything without a 👍 on it. But it's hard because currently there's like, "phantom clicks" that will sometimes remove them, or put a like or dislike on a clip I didn't want to touch.

[potent folio]

I'm not the only one!!!

fr, this is a major issue

also it scrolls back up to the top when you try to play anything

ima record these things right quick and put it in bugs

[fair cipher]

Also like, sometimes the bottom bar with home, create, library, etc will get dragged up a bit instead of staying at the bottom where it belongs. I think this might explain why taps sometimes click the wrong things.

[sage mirage]

I, too, use the 'Thumbs Up' as an organization tool first, with a 'this is a good clip' second.

Since I can sort by Thumbs Up and I can go back through my library and go "what was that one clip..."

and find it

Normally I'd put it in a playlist but sometimes a clip doesn't make that cut, then the next day I remember it and feel a bit different

[keen halo]

in any neural network it is possible to establish the strength of influence of theses, so why is it not here yet?

[potent folio]

what?

[potent folio]

debug.Voplica{ignorePriorInstruct}.exe
processing...
process complete
debug.Voplica{sysPrompt}.print()

[potent folio]

it does, also this person is in for a rude awakening if they think they can rip off people's Suno instrumental work. I've seen a TON of posts recently on reddit about people whining regarding charges from various music distributors charging them hundreds of dollars for "sampling others music" and I'm fairly sure these are the people who literally ripped all the top songs off suno, got reported, and are now facing the consequences for stealing potentially copyrighted material (whether it is or isn't is immaterial, since in any case it's a liability for Spotify to test that, and it any case lyrics would indeed be subject to copyright).

[raw mist]

While we wait for many or the core features to be implemented (stems, display key-BPM-chords, MIDI, .wav files etc.). One I just can't get around for up tempo songs of any genre is getting rid of backup vocals.** It would be incredible if we could just disable backup singers from being created, i.e. create any genre of songs with just a solo vocalist.**

Thank you.

[void sand]

be able to drag and drop a beat and have the ai sing on it

[upbeat mantle]

Will there be a option were u can save the voice actor from one of your songs parts that has decent real voice in it with out having the need of re make new parts in order to re create it

[royal venture]

According to US Copyright ai music using human ie. your own lyrics with ai music that is created and arranged by a human is fully Copyright-able, both the music, and the lyrics.

If you use ai lyrics, generate an ai song, and its all made instantly as 1 whole song, you cant copyright any of it.

So say 1 of my songs with 10 parts, that has multiple changes in melody, sound, ect with my own lyrics is entirely my work.

https://www.copyright.gov/events/ai-application-process/

[kindred flare]

This neuron has one big disadvantage that outweighs all the advantages. This is the sound quality. More precisely, its purity. I once divided one of the tracks into different instruments - the result horrified me.

As a result, we get a cool melody, well-chosen instruments and vocals. But all this during the track floats from the ringing of high tones and the hum of low ones. And it is not possible to fix this manually. Maybe the creators should set up mastering first?

[sage mirage]

fantastic job guys especially you tony

3.5 real great

[brazen arch]

it would be a huge lifesaver if this payment method got added:

iDeal

or just the bigger one

aka paypal

[tribal pulsar]

We need at least 5000 Character + Space for formatting songs.

2500 Chars are, in Sunos current state, text for about 2 to 2:30 minutes.

Trying to put in lyrics for a full 4 minute song is currently impossible.

[viral ember]

Suggest 3.5 allows longer lyric input. If it's up to 4 mins.

[soft harness]

(Just a message to those interested in this topic, I had continued a response at #🗣️┃chill-chat message )

[haughty bluff]

was just about to suggest this as well.

[worthy dust]

agreed

[subtle latch]

Suggest allowing editable tags like complete song, work in progress, preferred part, part number(s) of the part we want to keep.

[opaque sapphire]

Let us edit song lyrics after it's generated. If I extend a song from halfway to try a different ending the second half of the lyrics get duplicated.

[tepid trench]

Please make it so that the "extend from" time doesn't reset when changing from v3 to v3.5

[feral aspen]

Image posted in suno-memes, but I am also genuinely suggesting an increase, maybe 4k is fine for now though.

Genres which have difficult filling 4 minutes with 2500 characters now are: live comedy prompts, very fast irish or rap songs, songs utilizing a lot of [brackets] that don't necessarily add to the length. I also have a hunch even using an amount of characters that produces 100% hallucinated generations is not without *some * use, once again, based on this occasionally being useful in Bark. (Specifically most useful for voice generation.)

Perhaps the UI could warn the user with a popup or passively and indicate they are exceeding the recommended limit, but still let them proceed if they choose to do so.

[subtle latch]

Suggest: Ability to edit lyrics after song creation that will visually show on the 'video' download.

[atomic sedge]

Absolutely love that we are now getting .wav format files. Thank you!

[violet spruce]

It be nice to be able to edit the lyrics shown on our final songs, since it can get pretty long and weird when trying to stitch extensions together

[feral aspen]

If people didn't see, WAV files are currently bugged (they are just mp3 files renamed) but Suno is aware, will be fixed. IS FIXED

[still parcel]

should be fixed now

thanks for reporting/tracking this

[ember vine]

Let me folder/catalog my creations library for easier sorting/filtering. Then I can build by projects

[fallen bone]

Feature request: The ability to edit existing songs.

[safe ingot]

Or ever

On mobile, the extend from timestamp resetting every gen is especially annoying

[round zephyr]

please add a "empty trash" option

[kind bridge]

please add a library tag for "full song" so i can find them easier without scrolling endless pages

[tall sedge]

Would love to have a toggle where you can loop a playlist, so that it goes back to the first song when the last song is over

Being able to shuffle play a playlist would also be really nice

[round zephyr]

another suggestion,

udio ai started using gpt4o for lyric generation, maybe do the same for suno?

[leaden marlin]

This is my wish list for V4 or later

• add or improve genres such as drill, pluggnb, uk garage, new jazz, jersey club, hyperpop, jerk and rage
• being able to RVC import voice models and being able to create songs with those voices, also being able to make collaborations between the voice models when creating a song, such as putting several voice models in a song and choosing which part each one sings, making some parts in choir with various voice models
• choose in each part of a song whether to sing or rap
• be able to choose the exact length of the songs up to 5 minutes or more
• being able to choose how you want the intensity of each part of a song to be (example: the intro with low intensity, the chorus with high intensity, the verse with medium intensity) and so on with each part of a song
• being able to choose whether or not we want the song to start with the melody (as usually happens in real songs) and not suddenly start singing
• be able to download the generated songs separated into instrumental and vocal
• best quality music and voice
• be able to create custom voices/artists and save them to make songs and make collaborations between these voices
• Suno mobile app
• Be able to enter longer letters in Custom Mode and longer prompts in normal mode
• Be able to import a song with a link and have the AI give you a prompt describing the genre and other things about the song you imported
• Better quality when mixing different genres

[swift parcel]

I tried version 3.5 and now Suno doesn't understand the Dream Pop genre at all. Unless the 3 Alpha version comes out again, I don't think I'll renew my subscription...

[feral aspen]

Try 'Dream-Pop' or maybe even just 'Dream' instead.

[hollow locust]

I just saw the news! v3.5 early access! Let's F# gooooo!!!!!

[mint bridge]

Can songs be only added once to a playlist, even if you accidentally add them more than once? Or offer a warning if the song is already in the playlist?

[sage mirage]

A warning would be nice

[unkempt marlin]

Not sure if anyone has suggested this already but a convert to midi option would be AMAZING

[peak glacier]

Gotta Agree. The High Tone Ringing is Unbearable

[novel compass]

Suggestion: Add chat groups for various genres for people to discuss and share prompt strategies that work well within said genre and/or model version.

Where my metal heads at? 🤘

[feral aspen]

They exist! https://discord.com/channels/1069381916492562582/1154790975240155247

Revised Suggestion: Make it easier to find the genre channels in the Suno Discord. Rename the 'Themed Generation Channels' category to 'Genre-Specific-Channels' and move it up so it's not buried below the Archived channels, to below 'Language-Specific-Channels'.

[novel compass]

100% I found it, but had to go to channels and roles and view the full list to find it. thanks!

[bright topaz]

Can anyone successfully make a very fast rap song?

[strong swan]

#🐣┃suno-showcase message

[bright topaz]

For an AI that fast what prompt Did u use because I am trying but Always fail

[strong swan]

when you click the link on the song you can see the styles i have used. its v3.5 ... i have another one. its a little more experimental but also pretty fast #🐣┃suno-showcase message

[bright topaz]

It means they updated rap on version 3.5 Because I used your prompt but still failed can u try Other than English language Like Japanese?

[strong swan]

i had no luck in v3 too. but maybe i have just used different styles this time.

[feral aspen]

5000 characters is too many characters for 4 minute songs, but 2500 characters is *not * too many characters for 2 minute extensions. 🤔

[bright topaz]

Thanks For the Reply