#💡┃suggestions-web

1 messages · Page 10 of 1

tepid trench
kindred yew
#

@tepid trench thanks for the advice'

real echo
#

Please bring back v3 alpha. why? Because it allowed you to direct your instrumental pieces using [brackets], instead of lyrics, an awesome feature, which is now seems almost completely lost in the v3 version! At best it interprets the [brackets] as choir, which suck if that's not what you want in every song....

raw cave
pseudo crane
#

How does it affect your usage of the website if some people use chirp in here?

#

Oh I didnt know they used more power than the website

worn atlas
pseudo crane
#

Oh cool I did not know it caused more strain on the ai

tepid trench
#

Where did you find out that the use of /chirp on Discord is causing overload?

torpid walrus
#

🎵 Suno AI Music Feedback 🎵

Hey everyone! 👋 I recently tried out Suno's AI music generation, and while it's impressive, I have a couple of suggestions for improvement:

  1. Fine-Tuning Control: It would be great if Suno allowed users to add control words directly into the lyrics. For instance, we could specify the desired style or mood, influencing how the AI performs. Imagine being able to add "tenor" or "bass" right in the lyrics! 🎶

  2. Chinese Pronunciation: Suno's Chinese pronunciation could use some enhancement. Clear enunciation and accurate handling of homophones (those tricky multiple-meaning words) would make the AI even more versatile. Let's aim for crystal-clear Mandarin! 🇨🇳

Feel free to share your thoughts! Let's make Suno's music generation even better! 🎹🌟

formal fractal
#

is a 4 minute song out of the question? continuation seems to never work for me as it changes the style

real echo
#

The option to pick v3 alpha as an alternative model

fair turret
#

I would like to have the option to keep the musical score and regen vocals or visa versa

#

Would also like to be able to keep certain parts of songs, Implement licks and riffs

#

I got long songs Leoma. Careful where you continue song. Make sure the next part picks up the feel. Took me a minute to figure that out

sonic laurel
#

you guys never actually explained why you're not allowing access to v3 alpha. you just said "it's a stepbackwards". well if that's true, you can still let us access that model. why not let it be up to the user? it's this kind of attitude that makes people want to go elsewhere

native mortar
#

That is, probably, the most common suggestion.

turbid slate
#

can I give such an order to Suno for example for the first verse? if so, how should I write it so that he understands it?

Verse 1: Soulful R&B
Verse starts with a piano riff, setting a melancholic tone

real echo
meager forum
#

What would be an awesome feature is if you could auto-combine a continued song, rather than trying to manually stitch them together

#

Like once you generate the 2nd part of the song, you can select which one and it will auto-merge them together

turbid slate
real echo
lethal cypress
# real echo You can simply use brackets []

I attempted to create a song structure with it, but it only works sometimes. For example, when I try to add another structure after the initial drop, it doesn't always work.

tepid abyss
tepid trench
real echo
meager forum
meager forum
real echo
tepid trench
lethal cypress
lethal cypress
thick glen
tidal wharf
#

whenever there's "heavy usage", and the service is left for paying subscribers only, it becomes so fast. I understand the need to still serve free subscribers, but can the resources maybe allocated in some different way? I.e. some resources that are dedicated to paying subscribers only and don't depend on free usage?

delicate cedar
magic jewel
#

I'm sure most if not all of these things have already been suggested, but those would be useful features:
• Allowing users to see and control the generation seed, so results can be more easily repeated - brownie points if there were two different seeds, for the voice and instruments
• Support for multiple voices in the same song (like duets) by specifying it in the lyrics commands (i.e.: [Verse 1 - Voice A], [Verse 2 - Voice B], [Verse 3 - Voice A])
• Downloading the instruments tracks and voice tracks separately
• Specifying weights in the "Style of music" box, like "80s (hard rock:1.3) epic"
• More comprehensive and up-to-date documentation (current one is from V1 and is really short, I had to figure out a lot of things through trial and error)

Are there any plans to implement those features? Are there public changelogs or roadmaps?

Thank you and keep up the great work, I love this tool!

fringe agate
#

Have you tried using the tag “no effect”, “one voice”?

delicate cedar
#

ИИ не может не думать о том, о чем вы ему говорите, чтобы он перестал думать. Отрицательных подсказок пока нет??) С Суно работает. (Для общения используйте Google Translate, приносим извинения за возможные ошибки.)

fringe agate
#

I would like a referral program integrated into the track link, but not too intrusive. I would like to create, but I am very poor. Is there any way to make the service accessible to poor people? Or is this discrimination based on wealth?

tepid trench
#

You can usually create for free and get 50 credits each day to do that, but sometimes this is not available when service usage is high.

It would be nice if credits could be traded between users somehow but I'm not sure how that would work best.

urban granite
#

Can i change to a different sounding vocal? on a track i really like that made with Suno

mellow laurel
#

It says my song cannot be generated as it was flagged for moderation, but i have no idea what caused the flag. How do I find out?

hybrid crane
urban granite
#

what other ai can i use to change the voice of the vocal in song i made in SUno??

native mortar
pine pendant
#

Suggestion: a bit of refactoring of the discord would be nice, maybe a dedicated place to discuss the new suno website and the generations on it, maybe some billing channel, maybe a q&a/faq place, most discussions are currently going on in one channel - "general" since there is no other obvious place for it

plain cosmos
sudden aurora
#

Suggestion: Let us search for genres and such so i can listen to more then just the suggested song

tough merlin
#

Suggestion : make the songs browsable by genre and language

tepid trench
fringe agate
tropic prawn
#

Suggestion: Ability to edit the lyrics of the song, if a song is continued from a point before a verse there's duplicate verses in the lyrics, or sometimes words have to be phonetically spelled for the AI to enunciate it correctly, but it would be nice to have the lyrics show the correct spelling of the word

languid gust
#

please let me X the Banner

image.png
#

or make it black

thorn tide
#

Can anyone here tell me who knows about partnerships? Because we have an upcoming Discord server and would like to enter into a partnership

worn atlas
hybrid crane
delicate cedar
# worn atlas Yes i am aware of the unofficial community wiki. Thanks for pointing that. But i...

That would be nice...but personally I'd love to see them focus on keeping the model running, fixing the issues we've been enduring and making improvements.

Since Suno works differently we discover as much as they do most of the time. So what does it matter if we get a handful of musings scrawled on a napkin during lunch or if the community pools together the things they've discovered as long as the knowledge gets out there?

Somehow I suspect not too many of us would be in favor of pausing those other things to write up documentation....

leaden marlin
onyx torrent
worn atlas
worn atlas
leaden marlin
#

suggestions:
• be able to make songs with RVC models
• be able to collaborate between the RVC voices in the same song and choose which part each one sings and make duets
• be able to choose the exact length of the songs
• choose the intensity of the songs, example: intro: low; pre-chorus: medium; chorus: high
• that songs can be downloaded separately by vocal and instrumental
• be able to make songs with bad words
• be able to upload or link a song to copy the style

sand bobcat
#

2 Things I found woring with the vocals in FL studio:
Short reverb helps with the artificial grit of the vocals
Adding a slight reverb with high-pass filter smears together the grit that comes from the generated vocal/noise reduction on the vocals.
Filtering everything below 100hz (50-100hz taper) helps make it feel less muddy with lower vocals.

sullen valve
#

I’ve been trying Ai mastering software, it makes the songs sound so much better. It be cool if you offered ai mastering to your online program

native mortar
sullen valve
#

It’s called kits ai

opaque gull
#

Hello, if I "continue from this song" the next part does not sing the whole next lyrics and therefore the merger does not contain all lyrics right. Will there be a solution where a song can be repaired or where suno can even produce songs of desired length with desired lyrics?

analog spire
#

Could you suggest a "style reference" inspired by Midjourney? and input a musical proposal to better guide SUNO's AI towards what we might enjoy, without directly copying the original

magic jewel
magic jewel
sharp lily
#

it's probably suggested before but: it would be nice to edit the lyrics of a generated song, without changing anything significant about the style of the song (same vocals an everything, just different lyrics)

magic jewel
tribal lion
#

Suggestion: Please no self-promotion in #general, it gets way too spammy sometimes and makes it hard to help people.

warped orbit
#

While i'm here, i'm sure this was suggested before, but vocal/instrumental separation. I've seen issues with the voices being too quiet in some songs, so being able to have that vocal stem would really help for fine-tuning the song if the generation wasn't the best balanced.

delicate jolt
#

Add an "English-only" alert on the general chat "General" description
If I got it right, it's preferable not to talk any other languages besides English there

image.png
echo lotus
#

You would be better off generating one song (instead of two) and setting up queues for free subscribers. This would greatly relieve the servers.

tepid trench
echo lotus
delicate jolt
native mortar
echo lotus
tepid trench
hoary jasper
#

Suno excels at creating music that follows genre conventions, but struggles to blend diverse genres and instruments.
Most of the generations are indistinguishable from ordinary creations if you don't examine the quality of the audio close enough. However, I encountered difficulties when attempting to blend different genres and instruments. For instance, I attempted to combine liquid funk or drum'n'bass with saxophone to create a DnB track featuring jazz instruments, akin to Feint's Clockwork Hearts:https://www.youtube.com/watch?v=32pHvXsxa9o.
The majority of the generated output is suboptimal: https://app.suno.ai/playlist/e9020748-814f-40a0-9cac-06f786b9bb31
either straight-up jazz, or liquid funk with muddled instrumentation.
I tried to make drum'n'bass with classical instruments like violin, and piano, with similar success: https://app.suno.ai/playlist/432d34b2-cd9b-4daa-811a-64ce4daa9165
Except this banger:https://app.suno.ai/song/a4a47cac-71cc-4a95-a4d7-981f3d60b8f8/, but it's more like Dance meets classical instruments.
I guess this is because of the limitations of the transformer model? It's good at generating what A and B is, and seemed to have a good 'understanding' of the "essence" of A and B is. But when you want something that is a combination of a part of A and a part of B, the transformer is less good at it. Or it could be because I'm shit at prompting.

ornate mulch
bold prairie
#

Hi, would be great to have ios and android dedicated app for suno ai, much easier for users and intuitive user experience to become

formal bear
#

working on it

remote igloo
echo lotus
formal bear
sudden valve
#

No doubt this has been requested by now but we could really use access to the seed parameter for generations.

hollow locust
#

I think a 'cheat sheet' providing a list of things we can do in [] would be great. I only found out from here that I can use [End], previously using [Outro]. [End] is so much better! Also, I learnt [Epic (instrument-of-choice) solo] here... If there is such a sheet/page on the site already, please point me it its direction. Does [Repeat Chorus] do anything? (sure would save a lot of space if it does).

swift tartan
#

id really like to be able to do disable 2 songs per generation, like sometimes you just want to explore different styles to see how it sounds and basically im wasting half my credits because i dont need 2 iterations for that

smoky current
feral aspen
hollow locust
#

Thanks guys!

tidal mountain
#

I want to upload my music data for suno to learn my preferences

#

You guys are amazing

calm wind
#

Two suggestions 💡

  • Edit Lyrics (sometimes they do not fit right with what Suno generates, especially in Custom Mode and with several song parts, so an option to modify them so you can share the song with the appropriate lyrics would be fantastic)
  • Specify Song Duration (would be ideal when creating parts to cut song hallucinations or parts with few words)
tame dust
calm wind
# tame dust how to specify desired output time?

I've reworded that. I mean for instance to specify if you want a part to be 10 seconds (does not need to be exact though, could be a range). With that it would be easier to edit song splits that you don't like

marsh pawn
#

Can you add the time you used in the previous clip in the 'Continued From...' section? Like if I chose the time to change sections at 00:25, can you add that somewhere?

void crater
#

There should be a choice on the amount of songs generated so that if we do 1 generated we have more options to change and fix things easier compared to using credits up quicker which allows us to save more credits for one song compared to multiple if we choose to only generate 1 type.

frozen bronze
#

I need more than 2 minutes. What can I do?"

tepid trench
wind crest
swift blade
#

It would be great to add a searchbar to find known songs generated by other users

worn pelican
#

Suggestion: On the library, let users type in a page number to navigate to. Currently if I want to go to page 80 of my library, I have to click the "next page" button 80 times.

Also, this might be a bug, but if you delete an item in the library, the next page button becomes disabled. You have to press the previous page button, then next page button again to be able to move forward in this situation.

tepid trench
worn pelican
#

Its basically unmanageable for me but I am trying my best within the current constraints

lament carbon
#

Suggestion: that Suno.ai should do a collaboration with Ace Studio to create a plugin that you could use within their software.

Maybe this would also need a stemsplitter tool - if you are not only working with instrumentals.

But Ace Studio is really wild. You can change voices of the vocals and change single words, pitches, add vibrations to the vocals as well.

This would make it possible to edit single words in the vocals that got glitched, mudded or pronounced wrong. Or change voice type and all the lyrics, all while still having the same melody.

hidden vault
#

ANYONE would like to share knowledge on ADVANCED prompting or workflows to make your songs with Suno sound godlike would be much appreciated. I'ma put all together to create an in depth GUIDE 😄

rapid needle
#

Suggestions:

  1. TRACKS as output. This would be amazing to get tracks that can be edited in GarageBand or similar. Would allow very advanced editing and take songs to production. Would allow for very powerful remixing capabilities. Each track is an object and can be moved between songs.
  2. Specify song length
  3. Visually place lyric blocks, chorus blocks, instrumental/solo blocks
  4. Specify outro: fade, end hard
  5. Be able to name artist for styles or see generic labels for their styles or iconic songs
  6. Seed and prompt disclosure to be consistent; Have voice seeds and track seeds?
  7. Able to create Albums
  8. Better ranking system for sorting
  9. Able to select song structure from a template ex. 3.5 min Intro/Lyric/Chorus/Lyric/Chorus/Bridge/Chorus/Lyric/FadeOut
peak pasture
#

Suggestion: Plz add Paypal (or another one) for Europeans, so the most dont use/have credit cards! ❤️

narrow talon
#

Suggestion: An option to specify when the song should end. If you are going to the option "get whole Song" you should be able to set a timestamp for the end, like you can do when you continue from a Song. You could prevent all the weird endings and unnecessary added sounds at the end that you´ll get these days. And we should be able to edit the Lyrics after we finished the song. The shown text and the Vocals are not always identical.

tidal wharf
tidal wharf
hybrid crane
tidal wharf
narrow talon
peak pasture
tidal wharf
tidal wharf
hybrid crane
hybrid crane
narrow talon
#

I just just got a prepaid Visa to pay for Suno, but real Creditcards like in the US are still not common

tidal wharf
hybrid crane
hybrid crane
peak pasture
#

A credit card cost 23%

#

you can use prepaid credit cards, but it cost 39€ year

#

Paypa cost 0%

#

Paypal i meaan

peak pasture
hybrid crane
peak pasture
peak pasture
soft meteor
hybrid crane
peak pasture
hybrid crane
peak pasture
hybrid crane
hidden vault
#

PLEASE, give us a list of ALL the genres and instruments that sonu have been trained on, so we can explore to the max the app and its posibilities!! 🔥

hybrid crane
peak pasture
hybrid crane
hidden vault
peak pasture
hybrid crane
#

I am not sure what you expect.. If you want to keep your own bank 🤷‍♀️
Suno is not able to offer more than it does now.. but is working on paypal.

peak pasture
pseudo crane
#

Add the option to hide the orange banner in the website, I only need to read it once

hybrid crane
peak pasture
soft meteor
#

look just get N26 or Revolut. If you don't request a physical card it is free.
You can use paypal to transfer money to the "credit card" and pay with that.
There are no hidden charges with Revolut i use it myself. If you don't want to do that because reasons then thats on you.
Sure it might be more convenient if they add paypal but that can only be changed by the payment provider or by changing the payment provider which probably is a hassle to do just because. So I wouldn't expect a high priority on that. Because if people want to pay they already can.

hybrid crane
soft meteor
#

If you only can have one bank and your bank doesn't offer free credit cards i would switch banks

peak pasture
#

Not for the card, but for the credit on it

#

Thats why it called Credit.

hybrid crane
#

That's why they invented "debit cards"..

soft meteor
hybrid crane
soft meteor
#

educate yourself

pseudo crane
#

How long ago did this go offtopic 😅

peak pasture
soft meteor
peak pasture
round maple
#

if you can reach suno please suggest that they introduce rvc or any tech that we can train our own voice .There are many option to convert ones voice to celebs but from others voice owns voice options are limited.Such improvement will be very useful

opaque cipher
#

Hi Team, just like to confirm whether the following song structure tags, when used in custom mode, will all have an effect on the final music produced? (Intro,Verse,Chorus,Pre Chorus,Interlude,,Bridge,Outro)

tidal wharf
#

why were the lyrics removed in downloaded videos?

pseudo crane
hybrid crane
hybrid crane
opaque cipher
warm fable
#

I would like to suggest the following features:

[Reference From This Song]: Create tracks by referencing a previously completed whole song. (Select from the library.)
[Custom Key and Custom BPM]: Provide options for customizing the key and BPM, similar to the instrumental option in Custom Mode.
[Convert to Instrumental]: Convert the entire song to an instrumental version.

Additionally, the options below could be offered through a plan that incurs extra charges:

  • 320kbps sound quality
  • Separation of stems
  • Convert to Instrumental
  • Convert to MIDI
delicate jolt
cinder berry
#

Please allow us to upload our own music clips - being proprietary I don't know how possible this is but I have self-generated portions of songs I'd love to collaborate with

delicate cedar
cinder berry
pearl mesa
#

Please make pages on explore tab and implement search for songs

delicate cedar
grim cedar
quiet pewter
#

I searched this server for the word 'seed' and got 3947 results, but think it needs another mention because it's so important. If this app were to tell you the seed of the song you just generated, and allowed you to input a custom seed, it would make the app 10/10

shut cypress
visual brook
#

i am again pleading with Suno.ai to release more open source repositories around their brilliant tech - open source community could help with scaling/bugs/CVE patches and features

worn atlas
quiet pewter
grim cedar
#

I've noticed that sometimes if you use continue from song, the next verse/chorus/line/vocals will start too close to the previous end, and there's not really an effective way to put in a gap, the ones we have are usually too long and also break genre a lot or break the current melody by creating a new one. There should be a way to have like a short natural break between separate segments of the song.

obtuse tree
visual brook
delicate cedar
visual brook
#

440mm revenue, i guess - maybe not my finest internet moment, admittedly

delicate cedar
#

Mozilla is an example of how to run a company into the ground while wasting money. And how to completely eat a market advantage and turn from industry leader to industry follower. Just about everything they do these days is rearranging deck chairs on the Titanic.

visual brook
#

edited based on feedback

potent rain
#

in short, to succeed at open source: write a kernel

visual brook
#

honestly i would even be ok with v3 being open sourced when suno chirp method/model v6 comes out - commitment to delayed open source would really help the larger AGI space, adding music as a domain to datasets could get us to a robot folding my laundry faster, maybe

delicate cedar
potent rain
#

i was being sarcastic

delicate cedar
#

I'm a bit of a nerd when it comes to this stuff...can't code but I'm fascinated by what people make and have followed stuff for years.

soft meteor
#

Sure I would also like if Suno would go open source for the chirp model but. I can completely understand why not and I don't expect them to do so. A nice middle ground would be if you could run the generation locally/custom cloud GPUs (still having to pay a subscription but no usage limitations) but then theres the risk of the architecture being stolen and additionally they would have to pack everything up in a usable package. Maybe if there was a way to keep the model encrypted while inferring (probably not realistic) but one can dream.

small pike
#

Or both

#

App for on-the-go funzies, and plug-in for sitting down in the studio/office

#

And while I’m throwing ideas out there, a “make a song out of this” mode where you feed it a riff, a chord progression, whatever unfinished voice memo thing you’ve got, and it gives you ideas for finishing the song.

winged python
#

put the lyrics back in, in version 3 videos, please

tidal wharf
native mortar
obtuse tree
worn pelican
#

open source is tricky with AI because money is needed to train large models.

#

There is no way around that.

soft meteor
wind abyss
soft meteor
#

I really don't think that is the bottleneck

wind abyss
#

it took me 6 hours yesterday to get video clips 4 in total, and another 2.5 hours removing the audio from it, because it came down has clips. then i had to find buy another program to remove the glichy/jumpy audio from the video.

#

to so i could upload it to utube and othersites.

spring dune
#

Hello, I have a question. Could you please bring back generations for non subscribers and make an offline app for Windows, please?

delicate cedar
#

An offline application is very unlikely. The AI model requires very expensive hardware to function.

tepid trench
#

I'll repeat my previous thought on open sourcing chirp: I think open sourcing older versions could be good, just so others can learn from and build upon it.

spring dune
delicate cedar
#

Signs unclear. Ask again later.

Fortune-teller.png
worn yarrow
#

And yet, is it possible to buy, for example, a bonus code from you and enter it in the application by paying for a subscription or some other option?

#

Because it is impossible to buy a subscription from Russia

delicate cedar
earnest leaf
#

Please relax the filter for lyrics. It's blocking words like idiot, stupid, dumb, etc.

hybrid crane
earnest leaf
#

Often the context is just
"You're an idiot ha ha"
Something like that isn't something we should have to try to get around in my opinion. I really just think it should be tweaked.
It seems to block anything that is even mildly insulting

leaden marlin
#

Please, please, please add a feature where we can unlink later parts of a song from previous 'continue parts.' I'm generating some great genre mash-ups, but I'd like to continue and then generate a whole song 1/2 way through rather from the beginning. To clarify: I may only want parts 4-8 to build my "whole song," and disregard parts 1-3. So, some way to specify "start at part 4." Or simply some way to 'unlink' from previous parts. Thank you.

#

(I could easily do this in a DAW, but I'm thinking for the benefit of the community.)

#

Alternatively, a method by which to generate a full song starting from a specific clip and time, would be great as well -- but that is probably a lot to ask.

tidal wharf
warm stag
#

Vocal only button just like how we have instrumental only button.

#

Like with the vocals only, even if the vocals are treated as a "instrument" by suno, just make sure that the other instruments are tagged, labeled, and muted on a vocal only option

#

Concurrently, you could also have volumes tagged to other instruments and anything that comes up as a vocal instrument could be the "loudest" to ensure music never overshadows the vocals which can often happen in these generations

pliant canyon
#

perfect would be stems in general without any processing.. so we can process them in our DAW 🙂 - okay we can only dream

soft meteor
tepid trench
# soft meteor That would be soo cool but that would require stems as training data i suppose w...

It'd need to learn more, probably keep track of which stems are related, and generations would be more complex. I'd guess it'd be necessary to also have some mixing/mastering applied afterward (that seems to be already baked-in to what Suno generates at the moment.) I'd like to see it, and no doubt it'll come in time, but I don't think it's going to be very soon (happy to be surprised though!)

soft meteor
split tinsel
#

I would love to have a search Bar for specific Songs Im looking for

upbeat rock
#

as a subscriber, I dont need to see this

image.png
still wedge
#

I would reallly love to generate songs using an API 😁 any plans on releasing an API?

thick glen
warm stag
upbeat rock
hybrid crane
thick glen
serene solstice
#

SUGGESTION, instead of a song cutting off if you have too much lyrics, a popup comes up with a quote of credits allowing you to generate the legnth of song fully equal to that token value based on the amount of MM:SS that it takes, we dont mind waiting longer for it to generate and it takes out the equation of having to rejoin songs with extra regernations. thoughts?

#

like you wanna do a quick one verse jingle or something or a short video and match the seconds, you hit generate, it asks how long you want the song with the value of tokens changing with a slider.

#

much better user experience.

narrow seal
#

If I can trow my note: non-subscribers get 50 daily credits, why paying subscribers don't get any of that?

Also I'm doubling on the idea for feeding the AI something and it would continue the song.

Also, I found that to be great tool for sampling. So I hope that you could get someway the part that you are interested in (isolated drums, guitar, etc)

soft meteor
thick glen
soft meteor
#

Also it's pro6more complicated than that cause you can generate more in one day with pro

warm stag
wild leaf
#

I’d like to request the ability to input chords so we can have a deeper level of control of the feel of the song(s)

warm stag
wild leaf
warm stag
wild leaf
hybrid crane
warm stag
steel quiver
#

I think song generation numbers for each plan are misleading since it generates two variations of each song. There should be clarification of number of songs/variations

patent heath
#

I have tested v3 extensively (15k+ credits used) and it is very good. I am currently working on building and implementing a stem extractor, audio to midi converter, and mix/master compiler. The idea will be to take a suno output and make the pieces usable in music production, while also maintaining simplicity for all users to click a single button and have their track upscaled to professional audio quality. I would expect your team may already be doing this, I plan to open source my work after completion. Thank you for building an exciting tool to play with!

Also I wanted to let you know I found a way to include artist names. Can't put Avicii -- but you can put "Avicii's" (i.e "Avicii's signature style) unsure if this useful but wanted to share so this can be fixed.

brisk delta
#

Re-use a song allready generated to continue the song or generate another with the same rythm

frail sonnet
#

Would love to control sections of songs. ie start of sad and slow than gradually ramp up into a hopeful uplifting place. Also define the length of the song. "I need a song 30 seconds long"

wild leaf
worldly galleon
#

Sometimes my songs on suno.ai get cut off early and I wish we could rerun the generation with the same seed to try and get a complete clip

brisk delta
brisk delta
tepid trench
worldly galleon
#

awesome, i'll try that

brisk delta
tepid trench
#

Strange. I had quite a nice song generated and I've used this to force it to generate different parts continuing on from the first 20 seconds or so and it's kept the same rhythm, instrumentation etc.

brisk delta
tepid trench
#

Ah yeah I think the closer to 00:00 you get, the less similar it'll be but it should at least be of a similar style.

hybrid crane
#

As far as I understood from the Devs:
The strongest aspect of the model is that it creates incredibly unique music, partly because the seed isn't fixed. That's what creates the magic. Holding onto fixed seeds would hinder this.

fair turret
#

Could we get a feature request section that is done like ideogram it would be very helpful? Only posts allowed are for suggestions not already posted then people vote and this will keep everything clearn and concise which would be helpful in showing devs what community is looking for

lethal sphinx
#

when can i finally pay you guys with paypal

#

really annoying the only options are credit cards

formal bear
#

We tried to fix a lot of things recently, but still one bug -- once in a while if your job restarts, you may get two unrelated clips concated together (so maybe even longer than 2mins). If this happens, sorry -- we are working on fixing it.

fair turret
#

Also I am begging for some folders tree structure to organize music. I have nueroulogical problems in my hands and cycling through 83 pages of music really blows for me:O Love the tool that much I have already cycled through a Free account, a 10 account, a 30 account and now going for the second 30 account in about two weeks. Again Please give us some organizational tools> I can burn 1000 credits on a single song and i get lots of variations I am not always ready to ditch

soft meteor
fair turret
#

@formal bear I have got a few good long clips 😛

formal bear
formal bear
fair turret
#

I am getting a lot of songs where it ignores the verse and i have to trash it. Usually about 2 in about 15

#

maybe they were flukes

#

^regarding concatenation

soft meteor
# formal bear no it should be billed once...

It happened to me yesterday one press on create 2 song created 15 credits taken. And again today. I won't press a refund as it's a measly 10 credits that are missing. But i want you to be aware that there might be something causing this.

formal bear
#

there were so many issues for the past 5 days

fair turret
#

Where is there a refund option

#

Im on the web dont see an edit option either 😦

formal bear
#

I will ignore all the issues in the past 5 days, if you see new errors, let us know

formal bear
formal bear
fair turret
#

I sent int bug reports when my songs were jacked because it was happening so much> Still love the tool its awesome. I know its in alpha. A refund option would be nice on the web though

formal bear
#

sorry

#

we will make up to y'all at some point

fair turret
#

Ok Do you want us to keep sending them?

formal bear
fair turret
#

I dont want to overwhelm you but as as a programmer myself though you should know.

#

Ok

soft meteor
fair turret
#

Yeah if it is this friggin good in Alpa release should be phenomenal

brisk delta
#

A search bar for explore songs, we can search songs by name song or key words to find a similar prompt o a better explore experience

fair turret
#

Yes that would be really helpful

formal bear
#

now that fires out I can finally get back to improving models... 😄

fair turret
#

Check my post above I think it would be helpful for feature requests

#

would be aquick reference for you guys

formal bear
#

I think we are going to do sth with discord

#

all these feedbacks -- huge amount of duplicates

fair turret
#

yes

formal bear
#

not saying that you shouldn't post but we know

#

just so many things on our end to solve

fair turret
#

that is why making it unique feature requests would solve it

formal bear
#

and we are only a dozen ppl

fair turret
#

well you can see what we desire and you can decide whats best for your vision

formal bear
#

we appreciate every feedback

#

just not enough hands to get through

fair turret
#

of course understood, sometimes though you may have reasons fro not going with what we want or not prioritizing what we are asking for, Backend issues etc etc

formal bear
#

one day 🙂

tepid trench
# formal bear all these feedbacks -- huge amount of duplicates

Yes the number of times I see different people suggesting the same thing, e.g. stems 😅 which is something I'd also like but some way to see "this was suggested - it's not possible right now because ... but we've taken this idea on board" would be good.

formal bear
#

we will get there

formal bear
#

we are going to try sth someone told me to resolve the duplicates / group them

fair turret
#

no definitely not. Feedback section would be just a list of desires and then you guys decide whats best

formal bear
#

let us know what to prioritize

fair turret
#

we agree on that 😛

formal bear
#

(if we can't make up our mind)

fair turret
#

Phenomenal work though I am very pleased with what I have been able to accomplish with this tool

#

people are blown away

plain cosmos
soft meteor
#

I would suggest a real bug tracker of some sort even for public feedback and bugs way easier to sort through and probably easier to see if ones error is already submitted and just upvote it. But setting that up correctly is also time consuming I suppose so... hmm i don't know. I for one don't really like discord for feedback and bugs it's so chaotic. Talking and discussion yes.. feedback... no

fair turret
#

Could we also have a section for requesting metatags?

#

@soft meteor yes definitely

formal bear
fair turret
#

Ppl could search and avoid duplicates

#

Prolly work better than discord

#

can assign ticket Numbers

#

prolly have a open source solution already out there

soft meteor
plain cosmos
fair turret
#

I wrote a proprietary one from scratch for a company I worked for

formal bear
#

but I mean these are the decisions like -- we can have one eng working on this for like 2/3 days, but we can also try to fix a major bug / feature / speed things up

#

only if you are at Google / Meta scale you have the luxury of having eng working on things like these

#

but we will figure sth out

fair turret
#

Longterm you may save yourself time and headaches though if you can find a point to implement it

formal bear
#

tech debts...feature debts..

small pike
#

Maybe try a discord forum for suggestions? I think those might allow for upvoting…

fair turret
#

you could also consider hiring a freelancer to do it

formal bear
#

we don't even want to give u a shitty demuc off shelves...

fair turret
#

well yeah it has to be reputable one for sure

#

growth headaches amirite?

formal bear
#

but fun right

#

what a time to be alive

fair turret
#

these are good thijngs tho

#

yes it is for sure

#

I wasn't really thrilled about some of the implications of AI but after seeing the benefits if we can just keep the dark forces in check

plain cosmos
formal bear
#

you are right

fair turret
#

yeah meetings get old lol

small pike
#

I really like Procreate’s suggestion forum as a format. Seems to be working well

fair turret
#

been there done taht. 2 hrs to move the ball an inch

formal bear
fair turret
#

I dont know discord well unfortunately

#

not my area in the past

#

prolly not hard but I am sure a time sink

small pike
formal bear
#

k I passed on the info

fair turret
#

then you could go with my idea just make a channel and refuse dupe subject posts for feature suggestions

formal bear
#

man look at bugs

#

I posted a definition of what's bugs before

fair turret
#

yeah i did

formal bear
#

🤷‍♂️

fair turret
#

need to restrict dupes there, Can you run a script in discord to assign a unique id number?

formal bear
hybrid crane
#

A lot of people simply do no read. Even if somebody just explained something they wil just repeat the question..

fair turret
#

could do the same with feature requests if so

soft meteor
#

Ah clear flask and fider where the two feed back tools i didn't remember

fair turret
#

There are so many posts in these channels you cant even get to the rules

hybrid crane
# formal bear 🤯

I tried something like that with a lot of data in chatGPT but still makes doubles since sometimes it is asked different.

fair turret
#

maybe put the channel rules under a special section so ppl can access them

#

@hybrid crane how often does it make doubles could this be easily moderated?

#

getting rid of dupes may be lesser evil

#

maybe be able to get a community vounteer to keep it clean

hybrid crane
fair turret
#

@hybrid crane In ideogram you get prompted for the topic. then it searches if it cant find lets you post

#

some dupes still slip through but it works pretty good

#

not assigning id num tho

soft meteor
hybrid crane
fair turret
#

@hybrid crane maybe they would share their script lol 😛

plain cosmos
#

Must be some good prompts out there. Is not the only discord with this problem

hybrid crane
#

A dedicated QA would help to gather.. hopefull Suno with get that in the future...

formal bear
#

this feels like sth chatgpt would solve

fair turret
#

quite possibly

#

Ill see if it can do it

soft meteor
#

"As a large language model i can't disregard the opinion of people just because they are similar.." No but on a serious note this should work. Ar least to reduce the duplicates a bit. But i would still suggest a dedicated feedback tool. Or at least discord forum if tool is out of the question for the foreseeable future

fair turret
#

@Tony Ok it gave me something this is on GPT 3.5 . Needs code review from discord scripter though. Lemme know

hybrid crane
#

Best is to have access to the full data then.. I tried by copying the whole deal.. because I was curious what the bugs were and what was on the roadmap (first month I came here). But if you copy the data like a user, things will fall out of context. You don't see who responds to who.. and ChatGPT doesn't understand then either.

fair turret
#

I can share my prompt if you want or i can send what it spat out

steady scarab
hybrid crane
#

I personally worked with Productboard (and ClickUp).. There you can gather everything and also show the user what is worked on etc.. also a roadmap can be shared publicly.. Great tool.

steady scarab
hybrid crane
steady scarab
#

Known Bugs/Suggestions List section maybe

fair turret
#

I glanced over the code seems pretty straight forward. @hybrid crane would you like the prompt I used?

hybrid crane
fair turret
#

@hybrid crane Here you go hope it is a starting point or solves it : I need a script to create a channel in discord where we are tracking feature suggestions. We want the user to only post unique feature suggestions so we want to search what we already and posts of and ensure we don't duplicate. If the user is not duplicating want to allow the post which is assigned a unique id number if this is a possibility in discord.

dire harness
#

I would like a feature to be able to download the voice and the music separately

fair turret
#

@hybrid crane noticed i typoed when i edited but it knew what i meant

hybrid crane
#

I cannot run it though.. I have no admin rights.. 🤷‍♀️

fair turret
#

@hybrid crane maybe if you verify it they will allow you to

#

anyhow back to music. Hopefully that will solve that issue for them

#

would actually solve 2 feature request channel and unique bug channel

tepid trench
# steady scarab Known Bugs/Suggestions List section maybe

If possible it might be good to have some kind of bot in the existing suggestions/bugs channels that can try to respond to people to point them toward a known bug/suggestion it thinks is relevant to what they are suggesting so they know it's already been acknowledged.

fair turret
#

@tepid trench this is what the script is 3Daizy and I were talking solves more or less. Could modify to display the issue already posted, So if it is different user can reword. This could be slippery slope tho.

soft meteor
#

Hmm to be honest I would probably look into using something already established. Not that it becomes another thing that needs maintenance

fair turret
#

@soft meteor good working code doesn't need maintenance. IT just runs. Unless parameters change

soft meteor
fair turret
#

@soft meteor I suggested looking for opensource earlier. Tony asked for ChatGpt so I went that route

#

@hybrid crane I cleaned up that prompt it was bugging me I had typos lol : User
I need a script to create a channel in discord where we are tracking feature suggestions. We want the user to only post unique feature suggestions, so we want to search what posts we already have and ensure we don't duplicate. If the user is not duplicating a post want to allow it which then in turn is assigned a unique id number if this is a possibility in discord.

soft meteor
fair turret
#

@soft meteor I understood he wasn't familar with discord at that level which I am not either. It would not be hard to do this I examined the code it is very straight forward but it is a time issue and Chat GPT is producing decent code that can easily be used as a major starting point

soft meteor
#

Guess only tony knows. Both valid interpretations i guess

lofty sonnet
#

scrolling on page 100 something trying to find a track I made a few weeks ago, please add search function in library 😵‍💫 😭

leaden marlin
#

@formal bear Take a song you've already done in suno ai, and then change the lyrics, but with the rhythm of voice, instrumental remaining the same.

red tide
#

I'd love to be able to find other people's music that I liked on the explore section of the website, I just found out I can't and it's terrible

reef marten
#

Five second fix:

#

We need a prompt craft channel like MidJourney

fair turret
#

@hybrid crane the programmer in me wasnt satisfied. I incorporated @tepid trench suggestion. Reran the prompt with some error checking. Would still need to manually update a post in the log with a note saying resolved and maybe later add code to look for that but that is fairly easy should it be needed. Anyhow this ran in python I glanced over it but you probably want to look it over and hopefully they will say ok or give feedback if they want a modification. I used a log because this eliminates the necessity of a database and this shouldn't be a big file to read so shouldn't introduce much over head by storing ithe unique feature requests there with a unique assigned message id - CHAT GPT 3.5 Prompt: need a script to create a channel in discord where we are tracking feature suggestions. We want the user to only post completely unique feature suggestions, these features suggestions will be stored in a log-file that can only be accessed by those authorized to run the script. To do this we want to search what posts we already have and ensure we have a completely unique post feature request and don't duplicate posts for feature requests we already have. If the user is not duplicating a post want to allow it which then in turn is assigned a unique id number, if it is a duplicate we will display the topic for the duplicated request that was already posted and tell the user that that topic for a feature post already exists.

limpid chasm
#

Could we get a pop up with a basic overview of what’s allow with songs on the site? I’ve had quiet a few non-explicit songs get flagged for moderation.

hoary jasper
queen lichen
#

after generate a music that we like how can we keep the same instrumentals and voice but change just the lyrics but keep everything like the original version that we like?

wind crest
#

would love a timer that tells you when your credits reset, so i'm not just refreshing every 10 minutes hoping the 24 hours has rolled over by now haha

tawdry crescent
#

would like to see limit max length of a creation, can be useful in various of tasks. loops, fragmenting creations and such.

shell mirage
#

Please show what Time, Clips are being Continued from in their Details. So like it would say, "Continued from 00:45."

leaden marlin
#

Suggestion: be able to train suno with your musical preferences as if it were a LoRA

shell mirage
#

Make Suno pronounce Brutality correctly

rotund orbit
#

For as poor as these v3 renders are, the two generation concurrent limit is more than annoying

native mortar
shell mirage
void urchin
#

It'd be good to open new song links or reload page and not have it be 100% by default, but make it remember what I set
it's a minor thing, but rather annoying when you are clicking every suno link that pops up on discord to hear other people's work...

image.png
alpine stratus
#

China's Tiangong Music APP has a very important function for creating songs based on lyrics. Users can upload reference songs, and AI will imitate the style and timbre of the reference songs to create, and the effect is very good. It is recommended that Suno also add such a function

vague harbor
#

I would love if we could format the songs and when I say outro fade out it would FADE OUT and end without getting cut off lol

#

So [Fade in, Intro, Melody, Verse, Chorus, Verse, Chorus, Bridge, Chorus, Outro, Fade Out, End]

#

Also the ability to remake intros with the same patterns or notes…(Im totally dreaming I know….) I have been using suno then adding my own notes on top to make like a music collaboration instead of just letting Ai do it.

#

Also a way to master the tracks and export to wav or midi. Pleaaaase.

worldly galleon
#

be cool if it got trained on boy's choirs and boy soprano/treble solos

tight beacon
#

I think it would be nice to add environmental effects like Thunder or Rain to our tracks.

shadow dawn
#

would be great if we can import songs from copilot using the Suno plugin so we can continue it and so on.

timid heart
thick glen
tight beacon
worn pelican
#

It doesn't do well with nature sounds FYI

thick glen
#

I managed to generate a gentle rain in the background by just writing "ambient background"

worn pelican
#

They typically sound distorted. Things like wind, rain.

#

They need a model dedicated to sound effects.

sudden valve
#

when generating many songs with the same title it would be helpful to have them automatically numbered like "song 001" "song 002" I am doing this manually right now and it is quite time consuming

spiral acorn
#

Since so many people are using multiple accounts with the Pro and Premier plans it would be nice if there was a more official way to do this. Something like a custom plan where you select the amount of credits (with an input or slider). I know you can email Suno already about custom amounts, but that would seem to be more for fixed plans that don't change. If there was a user specified custom amount you could always cancel your plan and resubscribe with a new amount that makes sense for that month.

Also, it would be cool if we able to see the roadmap, or at least a general one? So that way if our suggestion has already been asked we know it more clearly.

Though I do get you all are very busy, so even setting this up might not be worth the time at the moment 😆

narrow seal
#

it looks like that

image.png
fair turret
#

@narrow seal I think that is in Polish and I don't speak it

#

@narrow seal that doesnt look like what I Tom and I were discussing

#

@narrow seal if it does have what we were discussing tho let him know how to turn it on

narrow seal
#

i believe is the "forum" function of server, and the idea behind that so the users could vote for best feature.

#

hence the "+1" button

fair turret
#

yeah but this makes sure it is unique and can't post if not

#

the problem is channel clutter

#

they have this implemented on Ideogram AI discord

#

it works pretty well

#

if there are settings already in discord would need specifics so it can be implemented

narrow seal
#

the tags above are feedback, idea, suggestion, new function... and the icon for this channel is the same as in your language suggestion channel.

fair turret
#

yeah this isn't quite what is needed unless I am missing something

narrow seal
fair turret
#

I gave him a prompt which just does that

#

it creates a python script to make a channel record unique posts verify new posts are unique and makes sure log file is only seen with permission

sturdy vapor
#

Is Suno going to fix V3 or did they lure a bunch of users in with V3 Alpha and this is the plan? Anyone who says V3 is better than V3 Alpha, name one way.

gentle lantern
#

Please add the ability to use the same seed generation when creating songs, this way we can retain the exact same style of voice, instraments etc.. whilst changing the lyrics

elder dragon
#

Is there any way use the same song then edit the Lyrics

soft meteor
hollow locust
#

Really looking forward to any sort of 'in painting' feature... just rolled a GREAT song, and it ignored [intro] and [instrumental], Literally starts singing at 00:00.

Speaking of 00:00... sure would be nice select where to "Continue from this song..." within the second 00:00:99.

Where does it pick up from anyway? The beginning or end of the second?

sturdy vapor
#

Stop using a poll on discord to gauge user sentiment. For all we know it could be 100% fake. Either way yall need to fix V3 or do us all the favor and go back to V3 Alpha... Why not give us an option to use V3 Alpha instead of V3?

wooden fable
#

the v3 vs. v3alpha poll might be used by people who haven't even tried both versions – meaning those results are not very useful

#

people tend to vote for whatever they know

#

"Anyone who says V3 is better than V3 Alpha, name one way."
Actually there is one thing: The new version is much better at following clues like [Outro] and such.
I still prefer v3alpha though.

pseudo crane
wooden fable
#

right when v3 dropped, i was impressed by how well it followed the [instructions] compared to v3alpha, but after some more playing around, i realized that 90% of my generations are poor. and i don't like the choir effect. so i moved my vote from v3 to v3alpha after careful thought.

pseudo crane
wooden fable
#

this is the suggestion channel though, so here's a suggestion btw.:

  • More transparency about what the developers are actually working on / aiming to fix.
#

maybe some sort of status page?

pseudo crane
#

Timeline would be nice, but best is if they can focus on working and not updating us 😛 It sucks sometimes to be "in the dark" but for sure they are working on the next version and knowing that they are wont really change much for us

wooden fable
#

true

pseudo crane
#

"It is what it is"

#

But yeah, a poll on a discord which doesnt seem to even have any built in or extra bot protection activated should not in any way shape or form decide where the development is going imo

quiet scroll
#

Is there a way to covert free generated songs to pro ones so cover the owners rights? Like buy it for a credit or something?

wooden fable
#

i'm pretty sure that's not possible

hybrid crane
loud canyon
#

I'd like some way to group songs together while I'm working such as a project or session. Sometimes I'll be working on a song and make multiple generations of each part to pick the one I want. It would be nice to be able to group all those fragments together to only show them while working on the song... or the opposite to hide them while working on another song.

For example, sometimes I'll be working on a song and get a decent amount into it and want to wait till the next day to listen with fresh ears. I might have 20 generations for that song so it'd be nice to group them and treat them as a project.

hard schooner
#

UI Web

  • much larger list of the trending/popular tracks with the search function
  • implement top song charts for different genres, weekly, monthly , yearly, all-time
  • make artist profile visible so that people can build their identity based on specific style, let users follow favourite artist
  • ability to upload song thumbnails or/and integrate image generators
  • over time implement “radio” section where people can play specific mix of genre types with curated list of songs, in the future perhaps users can define type of music they like and all of it can be generated while listening to mix
  • current lyrics generator doesn’t work well. Improve it with some open LLM models, offer various song structure templates on top of lyrics

Model/generation

  • make songs longer if possible, generating additional chunks to getting 2<mins is bit of a hassle.
  • ability to upload existing songs and improve on it, music equivalent of “inpainting”
  • being able to split different audio channels would be awesome but probably hard atp

Big picture stuff

  • perhaps think of Suno phone app. much nicer to use than web app for listening experience
  • ability to generate specific audio samples and MIDI files. Would be amazing for music producers, suno DAW plugin?
wooden fable
#

the "big picture stuff" makes me wonder what the long term goal of Suno is: Normal consumers and musicians looking for random easy inspiration, or if being an actual tool for production professionals is also in the scope? Those scenarios have rather different requirements.

hard schooner
#

why not both/all?

wooden fable
#

that's probably a lot of work, and difficult to combine

hard schooner
#

they should think big, they can be central place for entire music experience for humanity

#

perhaps they can split it in 2. Suno music/creator/editor and Suno music app super player.

wooden fable
#

yeah

hard schooner
#

“you think spotify is cool cause it has AI that can choose cool songs. We have AI that can create Fu….g cool songs from scratch while u are listening to music.”

#

I think they should think big.

wooden fable
#

catering to the masses is probably where the money is though.
pro is a tiny segment.

hard schooner
#

so many bullshit AI apps getting infinite money for useless product these days, Suno is among 1% who actually are doing something incredible and in tangible way. I mean thats why servers are crashing lately I presume

#

If I would be Daniel Ek I would try acquire them like… yesterday, or live in mortal fear if that fails 😆

wooden fable
#

yes

#

and .flac or .wav export

#

of course, they need to fix the broken vocals first 😅

#

unless we're all aiming to make choir music now

#

right now the vocals aren't worth exporting IMHO.

hard schooner
#

you can in theory target ability to convert entire songs into format that could be interpreted by standard DAW and recreate entire music project with appropriate instruments, plugins

#

but that’s still bit far though

wooden fable
#

we don't know how the generation works. maybe it's not possible to split it up that way.

#

i have a gut feeling that the vocal part is separate, but i'm not even sure

#

so far, my best approach has been to use UVC to split music from vocals, and then RVC to regenerate the vocal track (repairing the damages done by Suno's mp3 compression and UVC).

#

but this only works on single voices, not choir

hard schooner
#

another suggestion: hire me -> former music producer turned senior webdev & UI guy. Would love to help 😁

#

but more seriously would be willing to volunteer to help out on ui quirks if the team is busy on model infra stuff. 😆

tepid trench
# wooden fable i have a gut feeling that the vocal part is separate, but i'm not even sure

From previous discussions I've seen about this, vocals are treated basically like an instrument which is why you get hallucinations/nonsense words with some genres if you don't provide lyrics. As I understand it, the AI is trained on fully mixed audio, so it can only produce fully mixed audio. To generate genuinely separate stems, you'd need to train using separate stems as the AI otherwise wouldn't treat them as separate parts. I expect this would be computationally multiple times more expensive.

Someone who just wants to crank out a song with minimal effort isn't going to care about that. People like me who'd like to be able to drag stuff into a DAW and do further editing would benefit from separate stems. I suspect there's a relatively small amount of people who want to do this, so generating separate stems for all songs would likely be a waste of resources in most cases. Hopefully this will change as hardware improves and costs come down (ha...)

wooden fable
#

oh okay!

#

then i wonder what happened from v3alpha to v3 ?

hard schooner
#

thats interesting. Thanks for explaining

wooden fable
#

different weights / training data set?

hard schooner
#

can you in theory train another model to just split vocals from the songs?

hybrid crane
hard schooner
#

or that would be inefficient/dumb

wooden fable
#

@hard schooner look for "Ultimate Vocal Remover" which is exactly that.

tepid trench
#

During alpha we were encouraged to use the thumbs up/thumbs down on our songs so this was probably used to reinforce what traits are good and reduce what's bad. To be honest I rarely did this except to just mark which songs I planned to revisit later. Maybe more people were liking the stuff with reverb/chorus. I mean, it does sound good when that's what you're actually wanting!

wind crest
#

ability to sort by language on the explore page would be fantastic, would love to be able to scout out spanish songs specifically

wooden fable
#

@tepid trench that's exactly what i did too. you don't care about rating all the bloopers, so i only upvoted the good ones.

#

but how all this ended up into vocals being turned into choir all the time... hmm... that still puzzles me.

#

if that's based on up/down votes, then it shows quite well that a simple thumbs up/down is not good information in itself.

hard schooner
#

RLHF is probably not ideal for music as there is a lot of subjectivity involved

#

perhaps only to mark something obv wrong

tepid trench
# hard schooner can you in theory train another model to just split vocals from the songs?

Yeah there are vocal removers, vocal isolators and stem splitters which have been trained on how different things sound. e.g. you can separate vocal, percussion, bass, melody (where melody is essentially "everything else".) It's not perfect and you can hear dips e.g. where vocals or drums have been removed. But it's something I never thought would be possible, years ago...

wooden fable
#

v3alpha + UVC + RVC worked pretty well.
v3 + UVC + RVC ... not really. (because of the choir effect)

hard schooner
#

I know the feeling. I am usually AI optimist who tries to be up to speed and I wouldnt have predicted Suno AI quality for the next few years

#

I remember sometime in early 2022 playing with Image generator models in EleutherAI discord thinking it will take 3-4 years for it to become good. And then Dalle2 gets released a month later, then SD and MJ arrive in the summer … GG

#

lesson: people are bad at exponential thinking 😅

quiet scroll
hybrid crane
quiet scroll
reef marten
final rivet
# tepid trench During alpha we were encouraged to use the thumbs up/thumbs down on our songs so...

Funny enough I tried to raise the concern about thumbs up/thumbs down feedback Could/will indeed introduce bias into the model. If a vast majority of users favor a particular style (like Celine Dion), the model might become skewed toward that preference. Thats why V2 for my end was a mush superior model for what I try to make. Tried I should say since V2 wont listen to old promts any longer, haven't been able to achive the sound I favored since the introduction of V3.

reef marten
#

yeah this is a question for dev. the feedback worries me also, but i wasn't using it because it would skew my own results. as i make stuff of a lot of different styles. If there's a global skew that will be much worse. No offense, but most people generate rubbish because they were raised on rubbish. If this is controlling feedback into the system then suno will die a rubbish death.

weak elk
reef marten
# weak elk This is a valid concern. I also feel that Suno is skewing towards genericism, ra...

yeah, that's what i mean by rubbish. I didn't mean a genre, i mean like the four chord song. The four chord song is a lot of major pop hits (see the axis of awesome video) but it's also the first thing i wrote when i was a music major back in college. It's not innovative and doesn't take a lot of skill, and there's already lots and lots of it. No one is going to need a million more iterations of it. And that's the sort of thing that can happen if the feedback is controlling AI results globally. It won't get better. it's already better than humans, but it'll get worse

This isn't a radical suggestion, it's what's happened to all major AIs. Chat GPT results are on average far less factually accurate now than they were a year ago.

weak elk
#

A creative AI such as Suno should be used for innovating music, yes.

hard schooner
reef marten
hard schooner
#

so music models will have intrinsically “better” taste in music than average human trained on rubbish and wont depend on humans clicking likes

#

like future system will be more like that record store guy from “HI-Fi” movie who will tell you why your music taste sucks and give you suggestions on how to improve 😆

reef marten
#

I mean, it's kind of perfect the way it is. I think the point is don't fuck it up.

#

If I ask for 1890s circus music, I get 1890s circus music. I don't want it automatically turning into celine dion or the lion king. no offense to celine and elton.

hard schooner
#

the more general systems in the future my deeply understanding human psychology and mechanics of how certain sounds and harmonies affect a person depending on their brain patterns and psychological to a degree that might be even uncomfortable

reef marten
#

it already does that, if you let it go off the rails.

#

especially when it glitches out

#

My best songs so far have come from errors. The time constraints are preventing it in most pop songs i think

visual brook
#

the robots are going to weaponize the brown note

hard schooner
#

like yesterday I generated series of synthpop songs and they all sounded soo good that it was almost uncanny. usually I would need to play 100 songs of various artists in a certain genre to find the good one. And suno was constantly producing bangers

#

I got the feeling that might brain got hacked lol

visual brook
#

maybe you had that 1:1000000 lucky gen

reef marten
#

on pop songs i find 5-10% are good enough to add to my playlist, which is still an amazing result.

reef marten
hard schooner
visual brook
#

but several in a row.... the odds....

hard schooner
#

after that almost ever second song was incredible

visual brook
#

if i could run suno locally, i would already have produced and listened to 24 hours of literal music i like

reef marten
reef marten
reef marten
hard schooner
#

but it depends if u like the style. I am synthwave fan and this type of songs just broke through my pleasure centers like a freight train

reef marten
#

suggestion: (saying it again) somene take the 5 seconds and make a prompt craft channel on this discord

hard schooner
#

I kinda got the vibes of that famous thought experiment of future AI being able to totally wirehead humans with infinite fun

reef marten
#

suggestion: explicity chord progressions, maybe even an editor where the ai can generate chord progressions and then we can edit

#

suggestion: editing tools for basically everything.

#

suggestion: vocalist generator, editor. Would be nice to be able to craft and reuse a vocalist

granite yacht
#

few days before my friend shared me a song from Suno by filming the computer screen, and now both of us only know the title of the song, thus, can't find & listen to it again. Could you please add a feature like "Search by Title" or something, so to let users search the genereated song database?

hard schooner
#

yeah searching for song name and filtering by various criteria (like genre) would be awesome

#

thre are sooo many great songs that were probably generated by now but you cant find them. There are just couple of dozens in the new/trending tabs

small pike
#

Suno might have a legitimate hitmaker on their hands, if not now, then within like a year or two

#

Suggestion: slow mode for the suggestions channel, with maybe a separate suggestions chat

indigo cypress
#

PLEASE ADD LANGUAGE PROMPT, SUNO AI ALWAYS SINGS THE FIRST 1-2 LINES OF HUNGARIAN LYRICS IN ENGLISH ACCENT. 🙏 ❤️ And thus drops the gems into the mud. 🥲

(Or just make it detect the language first, because starting the generation.)

wooden fable
#

If you specify the genre name(s) in hungarian, and generally stay away from using any english words, you might get rid of that problem.

#

i think that might help at least. it has seemed to work for me.

tulip wasp
#

Suggestion: Add more searchable items in the trend section of the exploration section. It is best to divide multiple sub sections according to the main style, and add a search bar. This will build a synchronized community in our application. I think this part can refer to Civitai.
Although I am not an AI related technician, customizing the sound of a song should be technically feasible. If users can choose their own singer for the song, even if it only supports uploading trained voice files, it can greatly improve the user experience.

#

I am surprised by the significant limitations of user and product communication in this software

#

I can't even search for existing songs on the platform based on relevant keywords

#

😫

dire scroll
#

After you buy the pro plan and run out of credits, you loose the option to use your account as a basic Plan account. Not getting the option to temporarily turn the account into a "basic plan" is very frustrating.
Thanks for doing this tool in general, I've had a lot of fun with it!

trail hawk
#

If you have stems, you can do it in your DAW/VST. We just need WAV/FLAC and stems.

hard schooner
desert moss
#

Wen suno v3 on discord??

spiral dome
#

Are there any plans to ever introduce an option to edit (regenerate) a specific piece of a finished composition (like inpaint in image generation)? It can be very disappointing when you receive a rare successful option, but it contains some small but unpleasant problem that cannot be fixed in any way

delicate cedar
desert moss
delicate cedar
trail hawk
#

You are right. For subscribers it could be slightly longer. And the textbox itself could be larger. It would be more ergonomic.

gusty silo
#

We need a way to buy back tracks we created in free mode. Even if it costs 20 credits.

trail hawk
#

It is not so complicated to obtain a license for a cover song.

magic jewel
#

I have another suggestion, to update the pricing for additional credits.
Currently, free users can't buy credits, which is fair.
However, if you are a Pro customer, you get 2500 credits for $8. Then, if you need to refill your credits, $8 buys you 1000 credits. It's cheaper to create a new account and sign up for Pro again than to buy additional credits as an already existing Pro customer.
Buying extra credits should at least give you the same 2500 for $8, arguably even more, since pretty much every service gets cheaper the more you buy.

The current pricing model for extra credits makes no sense to me. This model is even more punishing for a Premier user who wants to buy more credits.

delicate cedar
magic jewel
thick glen
desert moss
trail hawk
#

I would not use an editor with MP3 in its name...
Audacity is a free audio editor.
And Tracktion Waveform FREE a free DAW that can also edit audio files.

desert moss
jolly elk
#

I'd lik eto know is it me or is there only one female voice? I keep trying to get Suno to do a soulful gospel type vocal and I don't know if it's me, but I seem to get only the same sounding voice.

delicate cedar
regal ingot
#

Does anyone know if there are special text characters we can use within prompts to better direct what comes up or happens when suno generates the audio? i.e. "+", "-", "#" etc

trail hawk
#

Can the identical song be generated with and without vocal?

hybrid crane
regal ingot
#

musiq - I use RipX to remove the vocal(s)

jolly elk
regal ingot
#

However, places in the track where vocals are removed (using RipX in my case) lack legibility and definition. There is also a lot of harmonic activity in the suno tracks that make it difficult for third-party applications to separate elements within the track(s). Some work is necessary. AI can make me lazy.

delicate cedar
# jolly elk Female Gospel Vocal

Try giving it some emotion. Try mournful or joyous or something like that. You may also consider trying to split the gospel part with a comma to allow more focus on the female type.

#

Also is that the whole prompt for the style field?

slender plank
#

Does anyone else find that "continue from this son" does NOT work when stitching the song back together?

warped root
#

Hello, I am thinking 30 second "exploration mode" because 2 minutes kind of is a waste when I am testing a new flavor of music. I'd rather get more song starts when trying to find something I want to continue on. At the same time, generating more smaller segments there needs to be a way to organize them for a folder or project level. In general I find I have the following: 1) not going to use clips (but I rarely trash stuff), 2) ones I mark "like" so I know it generated a decent/acceptable outcome, 3) ones I really like - would use as a song start//base and continue with it. But I come back on these later and it is hard to find them with limited list in the Create section (have to go find in Library by chronology). Also when working on continuation there is a link to segments related, but you can't get them in the primary view if they are older. So basically need a better way to handle a lot of segments when working on a song.

jolly elk
trail hawk
#

Can I modify the lyrics in a song that was already generated and change nothing else?

thick glen
heady sluice
#

does anyone have created a gpt that create songs using suno special prompts like laughter] [laughs] [sighs] [music] [gasps] [clears throat]

clever hinge
#

I just wanted to say that all of the problems in that new poll are important, please don't make us just choose one of them 😄

final rivet
#

Indeed.

heady cove
#

My wishlist for V4 and forwards; 1) Record your own voice (for cloning) 2) Make it possible to view vocals and music lines separately for both generations 3) Allow re-generating vocals and music separately with tick a box. 4) Allow your own recorded voice to be used as the default mask for vocals

regal ingot
#

I'm assuming that syllables in a song's lyric have to indicate to the ai some sort of setup for execution in order to sound reasonable. If you change even 1 word, it has to "run again" with a completely re-mapped lyric layout. If the Lyric DICTATES any "suggestions" that the ai accepts, it will alter/change/create anew BOTH "elements" (Music/Lyric) and that's why we can't re-run a prompt keeping EVERYTHING the same EXCEPT a word (or 2). It starts with a new slate of players and a new "producer" overriding your preference(s). I find it difficult to complain about any of this only because the current state of the suno platform is a marvel in-and-of-itself.

fair turret
#

@soft meteor I ran the variations of the prompt and got code every time I will add the files if the forum allows it and it isnt working for @hybrid crane

hybrid dragon
#

The ability to export Midi & Generated Acapellas. Due to the tracks being heavily distorted and mostly in Mono, you can't even stem split them but I can see why. ^.^

fair turret
#

I know this isn't probably high on the list but probably would be a quick change. Could we get the graphics with the album in 1024 x 1024 so we can use them on broader range of things?

#

that is scaleable so it can be put on a tshirt etc

visual brook
#

suggestion: borrow UI ideas from these guys - multiple lines for the same song in an almost jupyter notebook - maybe call it "advanced mode" or something
https://arxiv.org/abs/2310.12404

arXiv.org
Loop Copilot: Conducting AI Ensembles for Music Generation and Iter...

Creating music is iterative, requiring varied methods at each stage. However, existing AI music systems fall short in orchestrating multiple subsystems for diverse needs. To address this gap, we introduce Loop Copilot, a novel system that enables users to generate and iteratively refine music through an interactive, multi-round dialogue interfac...

fair turret
#

I tried scaling what we have now but too pixelated for my purposes

hybrid dragon
#

Just so people are aware, if you plan on "Releasing" ai generated music, it's not mixed or mastered and it sits mostly in Mono channels. It's not "release ready" as the tool was designed mostly to give musicians and songwriters ideas 🙂

fair turret
#

Suggestion - can we have it release ready 🙂 ?

#

soundspretty good when Im blasting it tho just sayin 😛

#

Im sure that is worth a buck

delicate cedar
fair turret
#

GIMP

delicate cedar
#

I had good luck with the free Upscayl application.

fair turret
#

Ill check it out mine was all pixelated. I need 2500 by 2500

delicate cedar
#

I blew it up huge and then reduced it to the size I needed. Looked good.

fair turret
#

if I have a 1024x1024 that blowes up well

#

Ill try upscayl thanks never liked gimp anyhow lol

delicate cedar
#

(Note that I resized it to be smaller in my final image.)

#

(The file was 26mbs.)

fair turret
#

bummer i cant run it

#

trying to avoid upgrading pc but it is becoming more difficult

delicate cedar
#

Huh. I'm on a Thinkpad e14 gen2 Ryzen running Linux and it works for me. What kind of hardware are you on?

hybrid dragon
#

It's suppose to enhance our workflows within the industry, not for the public to abuse it for things it wasn't made for

fair turret
#

@hybrid dragon Abuse??

#

Ok

hybrid dragon
#

It can be deemed Financial Fraud depending on which country you are from. Be law aware before releasing AI generated stuff 🙂

native mortar
delicate cedar
sullen elbow
#

Can we have a better way to organise generations? i.e group generations into "sessions" or something when I'm making a bunch with a theme or for a single full song.

native mortar
# delicate cedar Yes...

This upscale is a bit better than yours, if you'll look at details. Colours and details are better preserved.

delicate cedar
delicate cedar
#

How's that?

fair turret
#

@hybrid dragon if you say so. It isn't fraud in the US. How do you know who is a musician if only musicians are allowed? Sounds to me you are nervous of the quality being produced here and want to scare off customers. Just saying. Anyway this isn't the place for this discussion general is a better place for this sort of thing.

native mortar
delicate cedar
#

Granted I was just making a simple cover image.

#

For my purposes it was good enough.

Farewell_Song-album.png
#

But I'm curious now.

native mortar
delicate cedar
#

Topaz. $200!?!

Upscayl is good enough for me (for now?) 😳

#

But that does definitely explain the quality differences.

#

Thanks. I'll have to remember Topaz for when I actually have the money to spend and the hardware to take it on a ride.

thick glen
frail sonnet
#

Would be great if you could seperate vocals from music tracks. So you have the ability to download vocals only, music only, or combined.

potent rain
delicate cedar
delicate cedar
potent rain
#

hey im just talking about somali pirates

wide adder
#

Hi! Would be great to have an ability to regenerate song with the same music and voice and singing style. I was trying to generate song but it pronounce words incorrectly, and I want to save everything and just regenerate it with correct emphasis in words and pronunciation.

leaden marlin
wise cape
#

Struggled with that as well. Worked it out with very odd spelling.

leaden marlin
#

don't know i never use those words in songs i use softer words

wise cape
#

no, not the point.

#

For example, Diver.

#

It sounds alot more proper, half the time, if you literally spell it Daj-ver

leaden marlin
#

ah.. thank u for the hint i'll try it

hard schooner
#

Something like this will be soo nice eventually🤞 🤞

new-suno-1.png new-suno-2.png
misty blade
#

Suno is a very powerful model, but when I aim to extensively utilize it for producing high-quality music projects, it seems the interface doesn't quite fit the bill. I generate dozens of sketches, with each segment of a song receiving several different sketches. I would appreciate a way to organize all this neatly, perhaps with a timeline, and have everything bundled into a specific project. I would be grateful if you could address this need, taking into account that it's about producing a music song project.

wide adder
#

Another suggestions:

  • regenerate a part of a song. For instance from 00:00 - 00:10 or verse/chorus. (same style, voice etc.)
  • ability add something in the any place of a song. For example I have Verse and Chorus and i want a piano solo between them. I know we can use metatags like [piano solo], [Intro] etc., but it works randomly and it works not often. Need ability to insert such things with 100% chance)) Without spending credits on regeneration to get appropriate variant). If insert at the beginning the rest of the song should be shifted. Maybe modifying a full song, if I what to add something extra.
  • Better documentation with working examples. With description of all metatags and what they do, I'm not musician and I need to understand how different tags will affect a song.
  • Generation that based on user's taste. If I liked some song, genres etc., I want to generate similar songs with music style. It should be toggle to turn this feature or off. (I know about Remix button but it is random)
  • As some one mentioned need to ability to download instrumental and song(words) separately. Or generate instrumental by provided text.
  • If possible, more convenient method to generate full song.
  • I know it is might violate someone's rights, but it would be super-nice to provide an sample of music and singing style to generate music.
  • Explicitly define language of a song. Because for example Belarus, Ukraine and some other languages has the same Cyrillic letters, but it has to be pronouns differently. Maybe it could help with pronunciations. Without making a crutches=))
  • Better music quality 😋
  • Regenerate song with the same music but different voice and style of singing
  • Take a piece of song from generated one as part for a new song, or use some part of generated one and add the same instruments and playing in a new song
leaden marlin
#

"Add intro" before the start of an already generated song. So like "continue" but to be added before the original song when you merge them. I have a 3+ minute song that could really do with an instrumental solo before the actual song starts. But currently I see no way to do that without external tools or generating the entire song again with "continue" in many steps.

surreal current
#

Hi, I have some suggestions and some interesting ideas

  1. Reward mechanism, the number of likes can be exchanged for new Credits, of course, there can be a reward ceiling, the reason is that a lot of credits are spent on the songs, not all of them are satisfactory songs and a lot of credits are wasted, and the points can be returned after the release of good and satisfactory songs, so it is easier to activate and motivate users.
  2. Songs can be viewed by others or others can view their own lists or song comments, the reason is that you can communicate ideas with a good person, and you can set the songs that are allowed to be viewed and private songs.
  3. The new Top Song Column page, the songs with more likes can be put into a new fixed page and will not be rolled down, and the top songs of the week can be added for classified search and scrolling, the reason is to keep popular songs outstanding, and there are more good songs that can continue to increase the popularity of the website.
  4. The style of the song generation is satisfactory, you can keep it with other words or songs, and you can also change the lyrics again, the reason is that sometimes there is a very satisfactory tune, but if you want to change the tune or redo a certain part, you may not be able to do it again, you can only continue to the next song like opening a blind box, wasting a lot of good songs and credits, including some other details control operations. Or simply create a template to annotate the hierarchical tags of the song, and the user can directly click on the tag to change the content, for example, like daw, there is a timeline to turn a generated music into a daw, but it is a simplified tag, where you are not satisfied, regenerate, or add. Of course, there are many interesting ideas, but I don't bother to elaborate on them one by one
queen lichen
#

BIG FEATURE SUGGESTION:

Let we upload a song made by Suno and identify automatic the vocals and instrumentals to let we continue the composition or change lyrics 🙂

surreal current
vapid geyser
#

anyone know is Suno accept payment via paypal ?

hybrid crane
#

not atm.. they are working on it.. I don't know when it will be there

leaden marlin
vapid geyser
#

@leaden marlin any guide to do that ?

leaden marlin
#

Google

vapid geyser
#

ok thanks

unique heron
#

Is there an API to build on top of?

stuck steppe
unique heron
stuck steppe
novel grove
#

my suggestion would be to have a section on here to help each other with lyrics or discussion for lyrics 😄

stuck steppe
novel grove
reef marten
#

the most needed thing, and this is not an easy one: inpainting

ornate mulch
#

@stuck steppe I second this, or just a channel to use to share our secrets for getting Suno to work for us better.

young ice
#

I'm not sure what's been recommended, but something I'd love to see is a way where we can either "clip" certain segments of songs we like. Alternatively, a way to "edit" clips would be nice. It's just the tiniest but frustrating getting the perfect thing get ruined because of one tiny part that completely breaks the flow.

late plaza
#

stems and/or midi

blazing lark
#

I have a recommendation (and it also would fix a problem with making a song based on an RPG character): Ask if a name is an artist name, or is it something else. I ran into an issue because one of the characters is names Mika.

jade crystal
#

A suggestion: Being able to cut the song parts so you can create full songs out of selected ones, like lets say i create a song with 12 parts, but then go back and create individual songs out of 1-4, 5-8, 9-12... so that i now have 3 songs but they are made in a way that if played after each other it sounds like one whole continues song. That way, we can create very large pieces that all fit together but are still individual pieces in their own right.

blazing lark
#

Another idea: Make continue also generated based on suggestion. The metal ones are amazing, and would like the AI to arrange the lyrics into a full song, and add another set of lyrics based on suggestions I make.

late plaza
#

Why does Free Users gets topped off with 50 new credits daily, but Premier and Pro users doesn't?

light egret
#

I absolutely second the wish for a high quality download / WAV. I would also love to have more control over the song dynamics, build ups, bridges... I'm trying with meta information as much as possible, but it is quite often ignore once it gets a bit more complicated, like spoken words, changes in dynamics etc

frail fulcrum
#

Suggestion, Generate (and allow us to download) TTML, SRT, and LRC files (closed caption timing)
TTML is for YouTube, SRT is Facebook and LRC is most karaoke systems.

jolly crest
#

highlight moderated content for easy lyric debugging

feral aspen
lament crow
#

A search function would be handy.

abstract lion
#

i have a few

vocal basin
#

Inserting my voice to use with the song

worn pelican
#

Others have mentioned that the trending tracks are often low quality. A better ranking algorithm can help increase the quality of the tracks on the trending page.

One signal that can be measured and factored in is average listen time of a track. Then, even if the listener does not upvote or downvote a track, there is still a signal that is captured which can be used for ranking.

deep imp
#

Currently, I'm working on Thai songs, and I've found it quite challenging to create a perfect one. This is because Thai is a tonal language with five distinct tones for each word. While Suno often generates beautiful melodies, the lyrics sometimes have incorrect pronunciations due to tonal mistakes.

To address this, I've resorted to regenerating the song repeatedly, but achieving both accurate tones and a fitting melody proves to be a rare occurrence, happening only around 1-3% of the time.

A feature that would be incredibly helpful is the ability to regenerate a song with the same lyrics and music. Additionally, I've noticed that Suno sometimes skips sentences within verses, which can affect the overall meaning of the song. To improve this, an option to set a "Lyric Strength" ratio during generation would be fantastic. With a setting of 100%, Suno would prioritize including all defined lyrics in the song.

Overall, I enjoy using Suno and am excited to see it develop further. I truly believe these improvements would make it an even more powerful tool. 🙂

rotund spade
#

I wonder if the training data labels all Chinese dialects as simply Chinese. Currently there are no way to specify the intended dialect, say Mandarin or Cantonese, the two major dialects in popular Chinese music. The most problematic part, is that the model will randomly pronounce parts of the song with an inconsistent dialect. And FYI, in Chinese, the dialects are usually incomprehensible with each other., which makes it extremely weird.

hybrid crane
feral aspen
# rotund spade I wonder if the training data labels all Chinese dialects as simply Chinese. Cur...

This is a shot in the dark as I don't know anything about Chinese, but have an idea for a workaround. If you can construct some lyrics that are very clearly and obviously recognizable as either Mandarin or Cantonese dialect, then maybe you could use those lyrics purely as an extra signal to the Suno model so that it interprets the rest in the proper dialect without changing. I would guess these lyrics would have to go in front of real lyrics, though it's possible they could help even at the end (after all your real lyris) as long as they in the prompt. This would also mean you have to edit the song in post production and remove the extra verse.

trail lichen
#

Hello all, is it possible on the web app to add a field that lets you know how long your full song is going to be, based on the Parts(timestamps) that you have already created/selected?

gray sail
#

I am assuming someone has suggested this but it would be great to either be able to specify BPM and Key, or at least have them tagged in the generated song/track. Also an append option to append extended parts to the original automatically.

severe crown
#

Please add one of the ancient languages in the world and the only African language with a script — Amharic(Ethiopia). 🙏🤠

leaden marlin
#

My wishlist for Suno V4:
• add or improve genres such as pluggnb, new jazz, jersey club and hyperpop
• be able to make songs with RVC voice models
• be able to collaborate between the RVC voices in the same song and choose which part each one sings and make duets
• be able to choose the exact length of the songs
• choose the intensity of the songs, example: intro: low; pre-chorus: medium; chorus: high
• that songs can be downloaded separately by vocal and instrumental
• be able to make songs with explicit lyrics
• be able to upload or link a song to copy the style
• be able to train suno with your musical preferences as if it were a LoRA
• best quality music and voice

sturdy vapor
#

My wishlist: Go back to V3 Alpha! AI is unusable now, loses quality even with just a part 2.

wicked pike
#

If you guys are training off "liked" songs, I'd recommend you add a "silly button". Because I've liked alot of songs that sounded terrible but made me laugh so much. They should not be used for training.

sturdy vapor
formal bear
wicked pike
# formal bear are you serious ? 🤦

What did you expect? Half the people on here use it for poop jokes and cringe memes, I for one find the hallucinations hilarious and most of the times it has better beats or flow. Maybe have a second ai pass over the training data to insure the lyrics match the transcript.

feral aspen
#

I understand the spirit of the suggestion: more granularity in reactions. Thumbs up and down is adequate for YouTube, but music might benefit from a more interesting rating system with more axes. I know I rarely use just a thumbs up on people's posted Discord songs, I like to mix it up. You know you want to see a list of the most 🤦'd songs...

formal bear
sharp sundial
formal bear
feral aspen
#

Because liking is also one of the only two filters currently, liking my own clips is heavily overloaded. It could be any clip I want to find again later and benefits from a filter, which is often because I like the clip, but also could be a clip I want to follow up on 'why did it come out like that', and even music styles I couldn't get to work and want to remind myself to come back to later and try again.

I think I also most heavily use likes while working on multi-part songs to mark candidate clips while choosing which to continue from, and use it much less often when doing other kinds of generation. I almost never use dislikes except in this specific situation, where removing it from the create screen reduces clutter. (And if I remember I usually go back and remove the dislikes later since they usually are not especially bad clips and I don't want to signal that.)

worn pelican
#

Should just be able to tag clips. Then you can add a "silly" tag and filter by it in your library.

#

I am an advocate for being able to tag our music as that is highly useful.

lament verge
#

I am curious if suno or bark will implement speech to speech conversion

formal bear
worn pelican
formal bear
#

organizing them is a bit hard...if pre-defined, multi-lingo, etc

worn pelican
#

Yeah tags get out of hand as they scale up

#

It can become tag hell

#

But tags + semantic search is super nice. But high GPU use on the semantic search probably.

#

Then if someone searches for "sad" it will also find a track tagged, "melancholic"

formal bear
worn pelican
#

Yeah there are probably ways to optimize it well

cosmic dock
#

My wishlist for V4; 1) Download each part of the song as separate tracks. Vocals, guitar, hi hats, drums etc... so fine changes can be made manually. 2) Add lyrics to existing generated songs. 3) Save a vocalist so the same voice can be used in future songs.

formal bear
worn pelican
#

I think they mean if you have an instrumental that is good, you can overlay vocals/lyrics.

#

In case they don't write back.

#

Also I will +1 that request because there are a few good isntrumentals I generated that I would like to add some vocals on as well.

formal bear
#

that's just infilling

worn pelican
#

Yeah pretty much.

sturdy vapor
#

@formal bear Will there ever be a way for us to use the V3 Alpha model again? Kind of like we have the option to use V2 currently.

formal bear
sturdy vapor
# formal bear see pinned messages

I have seen that. It is not going backwards I can assure you. I have yet to complete a single song without massive quality loss since the "Upgrade". I wanna see Suno progress, but we have taken a major step back

feral aspen
sturdy vapor
#

Also with that logic, whats the point of having option to use V2 lol

worn pelican
sharp sundial
cosmic dock
latent falcon
#

Allow us to make models of our own voices, so we can make our own music!

sturdy vapor
wicked pike
sturdy vapor
#

V3 Alpha was so good it was rare I had to bring these into a DAW to fix it up. These are unfixable without the Stems thats how big of a quality loss I am having since Alpha

feral aspen
latent falcon
sharp sundial
wicked pike
sturdy vapor
sturdy vapor
sudden valve
#

it would be really nice if suno had regional pricing. currency adjusted the monthly subscription feels more like $$60 a month in places like Japan and Canada

sturdy vapor
#

I know with the company I work for, If we were to release a new version like this I would be out a job. Wish we could just select V3 Alpha and I would have no complaints

potent current
#

Are you guys going to release an api as well?

feral aspen
sharp sundial
# feral aspen Can you link (or DM me if you prefer) a few of the better V3 Alpha instrumental ...

Here are some of mine for you to compare to. Those are made by V3 Alpha.
https://youtu.be/3bRyY2i6Mns?si=DoylYSQCoRf-1RHk
https://youtu.be/TedOpYyYx_4?si=k-yprR92qx4UUUDa

It's a mix of both for the current V3. The composition can be too noisy, and if you are lucky enough to get a crisp one, it's easy to tell the difference that it degraded over time.
It's bad enough that I couldn't continue making songs to upload to the YouTube channel anymore. The difference is outstanding.

sturdy vapor
feral aspen
sharp sundial
karmic snow
#

Rework the "Remix" button and add a "copy lyrics" button

The Remix button should allow you to either choose the vocals or the instrumentals of a song and allows you to create a new song with that same vocals/instrumentals and change the opposite to what you want

whole falcon
#

Is there an API? That'd be my suggestion.

oblique elk
#

Suggestion: Rename #🐣┃suno-showcase New users don't understand the chirp reference and the webai doesn't even refer to chirp either. Maybe suno-showcase?

odd glacier
sturdy vapor
odd glacier
#

But as far as vocal clarity, it's no contest against V3:

sturdy vapor
odd glacier
sturdy vapor
odd glacier
# sturdy vapor Inexcusable for V2 outdoing or even coming close to V3 imo

The main problem I'm seeing with V3 which seems to be contributing to it's poor vocals is that no matter what genre the song, V3 will add new vocal layers with each subsequent bar. Notice in this version, the first & second bars are fairly clear, and match the clarity of the intruments, but then it just keeps piling on more and more layers until the vocals sound like a choir consisting of people who just returned from the dentist...

https://app.suno.ai/song/774cc73a-a304-4fe7-8210-a691a801087c

One Step Away From A Smile | Suno

alternative, rock, upbeat, electric guitar, harmonica song. Listen and make your own with Suno.

▶ Play video
sturdy vapor
odd glacier
sturdy vapor
odd glacier
sturdy vapor
lofty sonnet
#

would love for CREDITS to ROLL OVER. burning 2.5k credits rn before I get home from work tomorrow and I've lost them. not cool.

#

all major sample libraries let your credits roll over...

odd glacier
lofty sonnet
#

suggestions to suno not me plz

sturdy vapor
#

Common denominaor is V3 = credit burn

lofty sonnet
odd glacier
lofty sonnet
sturdy vapor
#

lol

sturdy vapor
wicked pike
sturdy vapor
wicked pike
#

V2 changed?

silk owl
#

a few suggestions:
continuing songs is great, what would make this even better would be to stitch songs together so one doesn't have to edit them. It's not practical to keep generating all the time, as often you want to repeat sections exactly, i.e. for choruses
for example in https://www.youtube.com/watch?v=8Tlm6DFtKAk
the AI generation is going great, until i have to stitch in the chorus again and again (and bridge and outro). then it requires manual editing - which i did really clumsily after the stich in, first at about 3:05

The other big thing, which im sure you are aware of and will be coming sooner or later, is it would be great to generate each layer of a track separately (i.e. vocals, drums, bass, etc). As often an amazing AI section is generated but it has a little thing wrong, and it sucks to have to throw the whole thing out, because of some little things.

YouTube
Jules
My Design

"The priest, who holds the secrets of the gods, shall shape the destiny of the land." - Babylonian Inscription

What lyrics should be, for future AI:
0:26
Verse 1:
I forge the lion's grip, the compass in my hand
Weaving wisdom's web, a cosmic master plan
Kindling ancient flames, crafting squares of might
Whispers weave a path, where secrets come...

▶ Play video
flint void
#

If I dislike my own songs, I should never have to see them again. ESPECIALLY NOT IN THE LIBRARY.

flint void
ruby lantern
random creek
#

I hope that the time of the songs will soon be increased to 3 minutes, because the program often cuts off the text 🙂

flint void
fierce sky
#

Hello, I don't know if this is the right channel to ask this question. But when you have the PRO paln subscription on Suno IA
Is the generation of music longer like 3 to 4 minutes instead of 2 minutes on the free version? Thanks in advance

feral aspen
fierce sky
#

Do you have any tips for making combinations because being a composer I can manage but if I want a perfect result I don't say no to tips

#

thank you for your reply

feral aspen
fierce sky
spiral acorn
feral aspen
# spiral acorn Are there any disadvantages in sound quality when continuing from the individual...

When you continue from a single clip, the music AI only has that single clip in memory.

This is an advantage in some ways:

  1. if there is a problem, there is less of the problem in memory, so it is less likely to make problem worse. (In general LLM terms: garbage is more likely after garbage)
  2. if you want to change the music style, there is less of the existing music style in memory, so it's more willing to change things.

The disadvantage is that you have less long term consistency (can't see earlier part of the song) and less long term music coherence or quality (the new music doesn't mash as well with earlier stuff). Though you can "get whole clip' after doing a single-clip continue for a specific reason like changing-genre, and that can bring some long term stuff back for the next clip.

TLDR: 🤷‍♀️

spiral acorn
grim cedar
feral aspen
grim cedar
#

Regarding the recent mini-poll, and from being fairly active in general over the last month, I don't think anyone really knows whether upvotes 👍 and downvotes 👎 are actually helpful for the model or not. The minipoll also only refers to upvotes. Can we get some specific user advice on how upvotes and downvotes should be used and how they affect the model, if at all? I think it would help to get everyone on the same page.

vast stream
#

BATCH SELECT functionality in the Library! So we don’t have to select unwanted clips one by one to Trash! PLEASE.

wicked pike
light egret
wicked pike
#

@feral aspen thanks

feral aspen
wicked pike
vague oak
#

Does the guys from Suno really pay attention here to submited sugessions?

wooden fable
#

i think so, yes, but they're a small team as far as i've heard, and they're probably struggling to keep up with the popularity of this thing

#

just the server (over)load alone is probably a tough issue to deal with

dawn fjord
#

Bulk Download

wicked pike
#

They need an ai bot that only directs to FAQ or creates priority tickets for questions unanswered.

dawn fjord
#

Search

teal horizon
#

Can I control the female or male voice for the verse?
If not, think about whether it's really necessary.

Please make a dock page that describes what kind of markup we can use.

little venture
#

Some tool for auto clip maker? 😉

light egret
teal horizon
teal horizon
light egret
#

ahh, yes, to be fair, I've never tried switching between both

frail fulcrum
#

after a song is generated, it would be nice if we could upload a custom image, write a separate prompt for a new image, or simply just regenerate the image.

teal horizon
#

It'll be great if the music ends smoothly instead of abruptly

worn atlas
frail fulcrum
#

also somewhat related to adding ends.. I have seen sometimes that the song ends.. and then adds a bit at the end that gets cut. It would be nice to manually crop things too.

wind abyss
#

but i don't know how Chatgpt is at moment it might not work anymore

violet crest
#

💡 : Adding a "Voice Record" to use our own voice instead of generic (random) one. Because I created a Nü Metal song and it really sound like Ivan Moody from Five Finger Death Punch.

violet crest
strong drift
#

Are there plans for an API? I could envision a future pass through applicaiton that takes the results of a different AI tool for say branding and generates a prompt for Suno and then returns it to a user. For me personally the ability to push results out from Suno to another application would help justify the cost of the tool when say I don't need 500 songs in a month but several clients might want one each.

wind crest
wooden fable
#

[instrumental outro] has worked for me many times

jolly crest
worn atlas
umbral silo
#

add option to actually empty trash

#

I don't care if the songs still remain on suno servers after that, I just want some way to completely clear some experiments away from my account

earnest axle
#

Please give paying subscribers the option to use V3A. I paid for the service to use V3A, not this current V3.

prisma aurora
grand musk
#

hello! i would really love to see a mode that lets you upload instrumentals in mp3 format or even better wav and add vocals to it. i have a dozens of self made songs but as i have no vocalist, it would really be amazing to let suno sing for me

solemn heron
#

I would like to have an option to download individual stems tacks

crisp veldt
hybrid crane
plain veldt
#

One quick question guys! Is there any kind of searching engine on the app for especific songs!?

violet crest
violet crest
sterile meadow
#

Two suggestions:

  • Have a "More Like This" or a influence slider when remixing. Choosing a musical genre can still give you a very wide arrangement of song options. If you get one that is almost right, you don't get the option to change the lyrics and have the same style.
  • Have a "describe this song" option where it can take an existing song and create a prompt for it. Even if it was internal only, it'd help to get a more specific sound, similar to the first point.
violet crest
blissful wolf
#

Good songs have lyrics that rhyme. 2 lines of a verse or chorus ( either contiguous or alternate) will rhyme and also have the exact same number of syllables.
This maximises the effect of the rhyme when each of a pair of rhyming lines are sung with the same TEMPO/SPEED/SHAPE.
I have uploaded to Suno 6 song's lyrics. All of them rhyme and the pairs of lines that rhyme have the exact same number of syllables.
In my experience Suno does not recognise ( and not act on these rhyming lines). Invariably the 2 lines are played with so much difference between their TEMPO/SPEED/SHAPE that the listener misses the rhyme effect altogether.
( This despite the fact that on occasion the number of syllables is made by splitting a one-syllable word like "bright" into "bri - ight" or "no" into "no - o". Suno , miraculously recognises this and sings the 2 syllables just like a human would! )

charred minnow
#

Will suno have a telegram bot?

worldly falcon
#

A NEON emoji

unique surge
#

Ability to search creations by prompt keyword/style keyword

prisma aurora
sly furnace
#

Suggestion: It's currently about a 10%(?) chance for a song to appear in the new section of the Explore Area, which is quite disheartening after investing hours into creating a quality track only for it not to be featured. Therefore, it would be beneficial to prioritize paying members in the submission process to ensure their efforts are more likely to be showcased.

unborn basalt
#

Please please add a function for MIDI export for each song. I will pay a monthly subscription for this function.

haughty nymph
#

How do I change to my premier account? I signed up by error in my cellphone with another account and I can’t do nothing to delete it or change to the other one…

hybrid crane
granite egret
#

Is there any way to get the v3.1 upgraded to include a input for time and allow it to range from 30 seconds to 5 minutes in 30 second increments for lenghts? The 2 minutes is nice, but it keeps cutting off the end of the lyrics, and most songs these days go around 3-5 minutes, so it would be great as a paying customer for the 30/m plan if i could have access to such a feature. 🙂 Is this possible?

#

or is this something that would be too difficult without a v4? Let me know thanks!

#

I am a developer and i would be happy to donate time to this project if coding time is needed to implement my suggestion? 🙂

acoustic pike
#

i have a problem. i can't continue with my old generated songs/musics with new lyrics. part-2 is empty. please can you fix it?

reef marten
#

I'm not sure about this, but I'm having trouble continuing songs from before the last server maintenance
If I try, I get a completely different musical style.

I might be wrong about this, but if my hunch is correct, here's a solution:

If there's going to be a small and subtle change to the AI, it would be nice if that were still to be called a new version number
like v3.1 or sometihng, if it's not going to be compatible with things created with v3.0, so that we could select v3.0 to allow us to continue working on songs generated on v3.0

If there was a change, it might have been really minor, but enough that I just get radical changes when I try to continue old songs

pale ocean
#

Dear devs, I want v3 alpha back. It was perfect. But V3 has so much noise and glitches. Just get v3 alpha back as an option!

#

Dear Tony, V3 is not an improved version of v3 alpha. It is really terrible. v3 makes so much noise background and vocals are waving in the most genres. Please get v3 alpha back as an option till you get a new version of app.

woven pilot
#

am i the only one who keeps getting the same song over and over and over????

hybrid crane
nimble wharf
#

Does anyone know what happens to unused credits? Will they be reset to 0 the next month or stay until empty? I don’t want to waste them all with the current noise sounds.

woven pilot
#

yes they will be reset

woven pilot
hybrid crane
woven pilot
reef marten
#

there's no legal IP associated with chord progressions. being able to specify them would be awesome

#

also, being able to gen an instrumental and then have an overlay-continue of vocals would be epic

woven pilot
unreal moth
#

i feel like the web app need improvement for the ui/ux , it pretty not intuitive for continue song, remix!
need more pro design.

hybrid crane
woven pilot
hybrid crane
woven pilot
hybrid crane
tepid trench
#

I'd try varying the style prompt a bit, add things to it etc.

karmic ocean
#

guys its possible choose the music tune? like A minor for example?

hybrid crane
woven pilot
tardy bloom
#

Would it be possible in V4 where after finishing a song you can add effects like (sped up, slowed, slowed and reverb and others). I feel like this would help to get different versions as Suno doesn't respond well to those commands

woven pilot
#

all music is supposed to be unique

hybrid crane
woven pilot
hybrid crane
#

Afro-pop?

#

Afro-fusion?

woven pilot
#

will try

long monolith
#

too many problems on new parts

#

most of the time my credits goes for the bad or incomplited generations

#

half the credits blows for nothing

reef marten
#

here's another: negative prompts. like seriously needed.

long monolith
woven pilot
#

negative prompts ?? meaning?

long monolith
#

is there a full sorce of promps? a refrence to learn the promps i mean

reef marten
# woven pilot negative prompts ?? meaning?

meaning i have a song here about a girl going to war from the first person. She is singing, or would be, with a female voice. just got 8 generations of all male singers, with my prompt that now says "girl singing 1940s style, fast female singer, women, sisters song about war" and still all male singers

#

so something like -male vocals

woven pilot
#

ohhh ok

long monolith
reef marten
#

it's needed for a lot of other things, like subgenres. If you have a really large genre, like for instance, swing, and you want a particular style, like USO, you can't get it because you're going to end up in new orleans

reef marten
#

where did i get the negative prompt idea? all the art AIs use it

#

on midjourney it's --no stuff if you want not to have stuff

long monolith
#

any body in the suno team can hels?

reef marten
#

there's like millions of posts here, i assume they'll compile it later or an AI will

#

I'm sure they have their hands full

#

I'm just posting it here so it ends up int he suggestion box

dry robin
#

get ready for dislikes

#

i have 4 sub par songs compared to what i have previously generated, studio quality my bottom, i can see what you can do, but priorty should be to reslove quality issues rather than new customers. keep the loyal ones first, i will only wate 30 dollars a amonth for so long, hears hoping. how is going back to something that actually works goiing backwards ?

light egret
#

Is there a chance to improve the sound quality? Especially with many layers (like with synthwave), there is so much crunchy noise :/

hybrid crane
dry robin
#

THEY CURRENTLY CANT DESPUTE COMMENTS

hybrid crane
#

You can try to clean it with other software. @light egret

light egret
#

Do you have any suggestions? I'm new to this 🤓

hybrid crane
neat relic
#

I created an instrumental song in custom mode with the only prompt as "ambient harp"

Now I'm trying to continue the song, and generate a 2nd part with drums, using prompts like "drum and bass, jungle, drums" or "glitch hop aggressive", and after 8 continues, I'm still only getting harp.

Any suggestions?

wise moat
#

I think a great feature to add would be uploading an audio sample of what you want the song to sound like. Maybe a voice recording of you humbing or even singing the song yourself and then having the AI make something around it

pastel lynx
#

Ok I have the best suggestion ever:
When a name isn't specifiec for a continuation of an earlier generation, default the name to the timestamp it was continued from.

So it can be named 0:43 or whatever and you can know what point it was continued from.

wise moat
#

I think another good feature to add would be if you want to close out a song with a Chorus line then timestamping in the previous part with a prompt like "Repeat line from 1:12" would help.

#

Also having continued versions be longer than 1 minute would really help as well

edgy temple
#

current credit pricing model rewards having multiple non paying accounts instead of one paying in some scenarios. fix would be to give those 50 extra credits to paying accounts everyday when remaining credits are below 10

formal bear
#

suprised that hasn't been rolled out yet...X.x

edgy temple
#

oh its rolling out in the near future?

formal bear
#

yeah

formal bear
edgy temple
#

perfect ❤️ love the product so much btw

formal bear
#

again, sth should be rolled out...

astral copper
#

Music Granting Documents Needed:

Hi Suno team, I'm a paid member and your FAQ has mentioned that I own the music generated for commercial purpose. However, when I try to claim the music on some distribution platforms, they ask for proof of license grant. I'm wondering whether you can provide that? Manual processing is unlikely to be scalable, so I'm wondering whether you can add some feature to allow downloading granting document for music generated during paid subscriptions? Thanks!

languid ether
#

Suno team,I highly recommend you to add alipay,weixinpay or some chinese payment method.Many cn have a strong need for it

worldly galleon
#

This could all be in my head, but I am noticing when generation is faster (virtually every time the site goes subscriber-only), my perception is generation quality is much higher

hollow locust
#

it'd be great if Suno understood musical terms like [rit.], [crescendo], [decredscendo], etc.