💡┃suggestions-web

[sage mirage]

i.e. I put in 00:26, it would play from that point

[potent crane]

It's very frustrating when you hear it get through a first verse correctly then it goes wrong in the next verse when it sings two lines quickly, squeezing it into one, then that puts the rest of the song out of whack.

but it does, you can select the second

[marble vessel]

It will be awesome if these features are implemented as soon as possible -

1. Search for my song by the song name.
2. Ability to add multiple songs to a playlist and create a new one if it doesn't exist already.
3. Export all tracks from the playlist either as a single file or individual output. This will be handy to create an output of curated tracks that I create. Right now its individual downloads which is painful for a big list.

[vestal sinew]

yeah i second that.. being able to search by song name or lyric would be great i've made a ton of stuff that ive forgotten about and struggle to find again because i have to click back through the pages. it would also be good if when you generate full songs it showed you which parts were combined. as i have a habit of continuing a song and then realise that its got the wrong part so i waste credits 😦

[sage mirage]

Yes I mean if the I could constantly get it to go back to the time I've chosen

Rather than having to drag the play bar back to make sure I've got it in the right spot

Minor convenience thing. I'm more bothered that the 'Continue From' time is reset if you change off from the 'Create' tab

[green patrol]

I use 300 creditis per music

[stoic glade]

Would be nice if this project went open source so everyone can enjoy it on their own hardware if that is feasible

Since the free tier is bleeding money

[untold bramble]

great work on the recent 'trending' and 'new' expansions for the explore page. I would suggest also having an 'old favorites' tab as I think there are many tracks that might have been forgotten that were superb on the explore page, and can still be discovered by newcomers

[formal bear]

top month, top each month?

[sage mirage]

Sounds good yeah

People like recognition

A rolling 'Top for the month' page

Go back and listen to the previous month's hits

[stoic glade]

When text gets rejected, a highlight of the offending sentence/word would be nice so I don't have to play guess the phrase that is not allowed. I like to listen to more violent songs that are fantasy oriented and its a little difficult to use custom mode when some words or flagged

Kobolds, dragons etc that kind of fantasy

[hot flicker]

Please give us ability to download stems! That would be a game changer.

[scenic acorn]

Make a feature that implies to make music like an artist.

E.g you pick Travis Scott and Suno makes music in his style(this would mean that Suno needs to analyze past songs and recognize what elements make the track sound like Travis' tracks)

Also: make some forum/thread where people can give tips for every genre (bass, what pattern to follow, how to make it less monotone, etc)

I'm good on dnb(not an expert in producing but I've got some ideas about how it should sound)

[indigo marsh]

is that possible to generate MIDI？

[worldly falcon]

Not really a suggestion for the app but for the Discord server: Add some custom roles for regular people, so we can have colored usernames and the roles we identify within the app, like sound mixer, listener etc etc...

[hexed wharf]

i want the same. please

[frank anvil]

Sorry, it's not a public one

[tidal wharf]

I use 800-1000 (averaging 50 created songs for 40,000 credits)

[dim hearth]

not sure how this could be implemented over discort, but some kind of "known bug list" would be very helpful for the bug channel. currently its more a complain area, which I can imagine is less helpful for the devs as they need to filter out all the complaining. also it would reduce the unnecessary discussions there.

[safe latch]

I would like a slow mode that uses less credits

[dusky charm]

a ''finished song" tag and associated filter button in the create and library tab would be nice.

[frank anvil]

That's very true! And a list (with poll) with the most desired features. So we don't have to talk about the same topics again and again.

[hidden vault]

----------------------------------------------------------------- 💡 IDEAS to add in the WEB💡 ---------------------------------------------------------------------------

1. In Pc when you delete a song, the MESSAGE TO UNDO is over the music player, making you wait to move or change songs. Move the undo message up and that's solved!
2. The **ABILITY TO ADD INTROS **to clips that start directly in verse. Shouldn´t be impossible cause you already have the ability to continue a song using the style of other clip and you can stick 2 clips together (to make the full songs) keeping style and making flawless transitions. You only have to figure how to do it backwards, but it would be the same as creating a continuation clip calling for a complete silence [silence/mute] and then call for an [intro] and the rest of the song, stick all together and then cut everything before the silence.
3. Saying this: the ability to TRIM CUT clips, and to try different clips sticked together like a puzzle making the AI do the transition between them would be CRAZY good. So a simple editor would be awesome.
4. You could add to every track generation info like BPM or key. You could use or partner with free software like https://tunebat.com/Analyzer. It´s been a great thing for me to learn about music and have more control off what i want. It´s probably not difficult to do and would be awesome!
5. Add a VOLUME SLIDER! for the web page.

(Respond this with more ideas for the developers and i´ll add them!, Let´s make Suno better together 🔥 )

[worldly falcon]

Although a minor thing, I'd like to suggest a separate "tab" for full songs in the library. That way you can swiftly get to them instead of having to sift through all of the clips

[green patrol]

wow, I use this when the music is so dense to me, can you share your music ?

[wise helm]

Maybe more space for lyrics in costum mode

[pliant canyon]

Suggestion - a switch to deactivate the noise reduction in generations

[hybrid crane]

Suno tries to offer the highest quality possible with the current techniques and never stops investigating to improve that. If there was such a thing of just adding a switch, they would have made the audio in general better...

[junior surge]

Suggestion: The ability to choose between male and female vocals

[pliant canyon]

ok thanks for the info.. I just have wondered why it lacks of details in intimate parts and this parts really sound for me like the details are filtered with some kind of a noise reduction filter ...

[hybrid crane]

The perceived lack of detail in more intimate parts of a track isn't due to a noise reduction filter. Instead, it's a result of how the AI processes complex sound layers. AI-generated music sometimes struggles with clarity and distinctness, especially in subtler, quieter sections, leading to what you might perceive as distortion or filtering effects.

[coral charm]

1. most useful thing would be to add a switch "vocals only" (as there's the "instrumental" one - or better, a generate music+vocals in 2 streams);
2. a tool to align the vocals to the beats, as it is it's crazily random;:
3. instead of "continue from time" expressed in seconds, it would be better a "continue from measure";
4. generated time, both from fresh or from "continue this song" is too erratic. Some times it generates 2 minutes, some other 30 seconds, totally random;

[hybrid crane]

You can influence the gender of the vocal by specifying it in your style or genre prompt. For example, use terms like "female vocals" or "male singer" in your request. However, this is not always completely reliable due to the AI's interpretation.

[pliant canyon]

Thank you for the insight

[coral charm]

5. also, it would be extra useful to expose the seed used for generation and be able to input/reuse a seed

[hybrid crane]

1. V3 (for paying customers) offers more control over vocals and instrumentals.
2. Suno's model inherently integrates vocals with music, and their alignment is AI-driven and not manually controllable.
3. The "continue from measure" feature is part of V3 for paying customers, offering more precise continuation points.
4. The variability in generated time is due to the AI's interpretation of your input and the complexity of the music being generated.
5. Suno doesn't use or expose seeds for generation as each creation is unique and based on the input prompt, style, and other parameters.

[cloud python]

Is it possible already to get just vocals?

[hybrid crane]

Generating vocals only in Suno is possible. You can use terms like "a cappella" in your request to prompt the AI to focus on vocals without accompanying instruments . But is not always consistent.

[cloud python]

Prompting guide would be great 🙂

[hybrid crane]

I know they are working on something like that.. So stay tuned!

[coral charm]

1. I use V3 and can't find any way to have generated vocals only (unless i missed something); Beside being able to ask for female or male (often ignored), that's it.
2. ok.
3. As said, i use V3 (paying) and there's only "continue from time", again, unless i missed something;
4. yes, maybe. So i wonder why with the same input, same lyrics sometime it generates 2 minutes and some other way less (?);
5. mmm... ok.
6. i'll try (again) acapella, but as for now, it only generated (some times) "acapella songs", not isolated vocals (there's a quite big difference).

[delicate cedar]

Until they get a prompting guide out, you might find some good insights at the unofficial community wiki? suno.wiki

[hybrid crane]

1. you have a toggle for instrumental only (voice you can do with adding a cappella in the style of music prompt)
2. see my little video
3. this can be due to several reasons, the pace of the song, the lenght of the lyrics, the style of music, the progress of the song that the AI feels like it's ending.. a mix of all and sometimes just one
4. depends on style genre.. I personally have managed to get pure a cappella with minimalistic genres

[coral charm]

Thanks!

[coral charm]

Ok. As for 3) , as said, that's continue from time, not from measure 😉

[hybrid crane]

You mean beats.. yeah sorry.. I though it was a language barrier

[coral charm]

🙂

[hybrid crane]

This is new tech.. but I can assure you the Devs are musicans theirselves and what to give us as much control as they can achieve. It it is possible but not there yet, it is just a matter of time

[cloud python]

Would be amazing to be able to download by track (or instruments vs vocals at a minimum)
And would also be awesome if we could 'continue from' uploaded audio

[steady scarab]

Is there a method to assign a generated voice to continue producing more content and use it as a template for building a library?

[sharp venture]

Might be a niche suggestion:

But allow us to upscale the quality of our music with AI. The AI will try to replicate the music as best as it can, but simply make it sound much more crisp. Obviously this is something that is more advanced and is down the road.

[indigo cypress]

Specifying language. 😊 Suno sang my Hungarian lyrics as an American singer. It was beautiful btw.

And seed extraction and specification. The song at my first try (using my lyrics, on web, not Discord) was beautiful, we just need some refinement functionality.

[frank anvil]

Great Idea! This would also help separating into Voc/Intr tracks, Cause it would help "reparing" the instrumental track.

[quasi forge]

Greetings, all! I have been thoroughly enjoying Suno ever since V3 was rolled out. I honestly enjoyed it with V2, but V3 has been a game-changer. I wanted to throw my suggestions out into the void, to see if anyone else had the same thoughts.

1. I would love the ability to edit or regenerate over smaller sections of a completed clip. Sometimes Suno will generate a lackluster verse followed by a mind-blowing chorus. It would be nice to try regenerating the verse without losing the amazing chorus.
2. It would be great if we could set specific musical parameters like time signatures, key signatures, tempo, instrumentation, etc.
3. The ability to remove vocals and create a karaoke-style track would be cool. I've seen lots of suggestions for this already, so I'm guessing it is a common request.

[livid ruin]

Suggestion:
The ability to save a Chorus for a song so we can veer off the path but always come back to the Chorus that made the song great.

[hybrid crane]

1. Suno is working on a 'paint in' feature to regenerate specific sections of a song, addressing your concern about redoing certain parts without losing others.
2. Setting specific musical parameters like time signatures and tempo isn't feasible yet in Suno, as it requires extensive research, control that would increase computational demand and cost. But anything is possible in the future.
3. For removing vocals, external tools like https://ultimatevocalremover.com/ exist. Suno is exploring stems splitting for more track control, but it's a complex, ongoing development.

[lofty sonnet]

Need to be able to Create a new playlist from the Add to Playlist interface

[marsh jasper]

A Consistent voice feature would be awesome. So that you can use a particular voice you like over and over in varying genres of songs, with the ability to slightly vary the pitch, raspy, accent, etc depending on the song

[tidal wharf]

what makes you think there is noise reduction at all?

[lofty sonnet]

as "investors" or "subscribers" it would be nice to know how much progress they've made or when the expect to be able to implement it, if known.

Also who annonced in paint research and plan to implent?

[pliant canyon]

I got generations with intimate parts wich sounded like there is noise reduction applied ..

[hybrid crane]

See

[tepid trench]

Would be nice to be able to "continue from" a more precise moment in time, possibly by having a "continue from now" button that uses the exact spot in the playback timeline. Sometimes I want to take a song in a different direction but do so at the sacrifice of a note or two just before the next beat (where the part I want to change begins.)

[sage mirage]

I have noticed this cropping up more and more for myself

Where am entire second is too long

Cutting back the the previous second usually works since Suno is continuing

But introduces the possibility of a dramatic change

[formal bear]

you want 0.1 sec?

[sage mirage]

00:00:00

Or

If we could just input that

idk it's a specific request I know, most users won't bother

[formal bear]

is it really necessary...

[sage mirage]

I mean

I'm a user

Of course it is, every request I have is the most important thing ever

[formal bear]

ok I am out

[sage mirage]

It's just sometimes there is an instrument or singer that starts at the tail end of a second

00:00:00 is a bit much yeah

.1 second should be plenty

Minor stuff though

[gentle cypress]

Ooh I actually strongly agree with this. Full seconds is not only a little confusing (is it right at the start of the second? the middle? the end?) but also having more control over it would be awesome. I've thought about that a lot as well. Right now I usually end up going back a couple of seconds and hoping it continues without me missing out on stuff I liked

[lofty sonnet]

or like a set of recommended points to generate from???

[opal pulsar]

• In the library to navigate between page, arrow are cool but for UX it can be cool to add feature, when you click on page number (between the arrow) you have a dropdown menu to select the exact page you want

• Another feature usefull, when adding clip to a playlist let people add it to a new playlist directly without need to go to playlist page. It's common UX behviour

[formal bear]

ok I will keep this in mind

[opal pulsar]

When continuing a clip, by default expected UX behaviour should be to use by default the actual paused timecode instead of the end timecode of the clip

[sage mirage]

Thanks tony, you're the best. Sorry about everything you have to listen to

Can't you offload listening duty?

Probably not since you have to hear the things Suno is doing in order to like, work with what the model does

[opal pulsar]

Another is when continuing reset the clip name field with the continued clip name

[lofty sonnet]

I would LOVE to see the waveform. Like soundcloud, maybe above or under the continue from part in the ui when continuing, or a minimizable window above the timeline on the bottom. After seeing music finished and mastered for a decade it's just a whole different medium of ART when you think of energy and dynamics, colored ones like the ones from TRAKTOR DJ are so awesome to look at and understand the sonic dimension with color too

[feral aspen]

This is one I've been pulling for so nice to see it's popular, thank you KakerMix

[lofty sonnet]

Def full color waveforms

[lofty sonnet]

Thank you 💯

[tepid trench]

I'm on the fence about this, as I would also like it to behave this way however there'll inevitably be users who don't realise and continue from the wrong point. I guess "continue from end" and "continue from: [time]" could be a toggle.

[formal bear]

yeah I know

[sage mirage]

It was Furrybeans

[formal bear]

sth better is coming

I think

[lofty sonnet]

and ai recommended continue on from points

[feral aspen]

Nobody really used it so I haven't updated it (not sure if still works) but I did this one silly evening where I was just going to add a volume control but things got out of hand https://github.com/JonathanFly/chirp-studio

[formal bear]

yeah I could but I love talking to you folks

got to stay close to the users

[lofty sonnet]

This is amazing, can SUNO just pay you???

Much more interactivity with clips, for now!

[opal pulsar]

The ai have somme issue when failing to finish the asked lyrics

You add your lyrics (verse 1, pre chorus, chorus) the ai generate the whole chunk (it's perfect) , but when generating a second chunk (verse 2, pre chorus, chorus) the ai quite always fail in creating a new complete chunk and cut in my chorus. When continuing just after the cut (or when asking to regenerate the whole chorus) most of the time the AI is lost and dosent keep the previous work and get totally different mood (it cost me 100 credit sometimes just to retrieve the same chorus)```

```Expected Behaviour :

Add the whole lyrics with a bigger length limit, generate the first part and when continuing the AI "remember" where she stop (and what she does previously) to generate another chunk and avoid forgetting here previous work```

[sage mirage]

Well I apprecaite that you guys are still so close to the users, so thank you for your sacrifice

[steady scarab]

Wow, where is that!

[opal pulsar]

sound like a good UX choice

[feral aspen]

I bet it doesn't work any more, Chirp has had a lot of UI updates since then. But I can fix it if there's demand. I woudl also like a wavform not just a real time visualizer, which the same library should be able to provide client side I think.

[opal pulsar]

the native HTML audio api can give you wave form information, i think it can be coded in vanilla without lib

[feral aspen]

Suggestion: Show precise continue times in the "Continued From..." display for each time. "@ 32s" etc

[sage mirage]

32.1 seconds

[lofty sonnet]

I think some kind of phrase sectioning could be interesting, like it shows you labeled markers with a concept of a part of a song, like intro, verse, turnaround, break, chorus, build, pre-chorus, Chorus B Section, etc.. Could be a fun way to somehow apply the prompts to different sections before, after setting like a rough tempo slider it would suggest point to enter a lyrical prompt and a style prompt if you wish for 2 minute gens that have multiple guided and more precisely engineered moments

some music is so complicated and some so simple and droning and monotonous, this sort of rough sketching a framework of styles and shifts would be amazingly powerful

[sage mirage]

They've mentioned before that there is a lot of things just not customer facing

[lofty sonnet]

Like an advanced mode

[sage mirage]

I wonder how much of what we're asking is 'just' UI

[lofty sonnet]

necessary

for the people of the world!

the musicians in all situations

[sage mirage]

There is a divergence with users that's become more clear as time moves on. There are the people that are upset that Simple Mode changed and now they get 'the same' tracks, wheeras there is us all going "if I could precisely pluck the bass at the exact moment"

[warped canyon]

hello and thank you for Suno. After using the paid service, I have had the problem of finding a great sound to the lyrics that I wrote but want to go back and tweak the lyrics to the same music and do not see a way to do it. Also would like to lay out the whole song in part1 then just have to press continue and it follows the original script without having type anything in the part 2, etc. boxes

[lofty sonnet]

the sections could be so open ended but somehow influence to section, "Guitar solo" "techy perc fill'. with inpainting feature a different MODE, like Composition Mode, could be selected where you see the left to right layout of the track and it's current stemmed layers, that can be auto inpainted or trimmed, namable points also of development for the given length of the track generated and modified in one waveform or channel style layer thing where you can recombine elements as a get whole song after edited etc, perhaps showing you the layers or potential combinations or development of different layers

It already imagines so much, it would just be an absolute mind blowing moment where it could do that but still have a section be broadly regenerated with a sample crossfade in (weight of regeneration) window of regen, and the crossfade out all draggable. then the ability in Composition Mode you could actually do that with whatever groups or individual categories of sounds. With a fractaling interface of groups or buses I mean you could go so deep and have the average 70-100ish channels some pro tracks require, like inpainting with a spectral microscope for these stems that can now be recognized and manipulated with ai by other developers

[opal pulsar]

Going to library > select a clip > Continue. It open the create page with an empty list. You have to click in the left panel to listen the first part```
```Expected behaviour

Going to library > select a clip > Continue. It open the create page with the continued clip in the list. It can be usefull to listen the first clip and check if it fit good with the newly generated clip without going to left panel (and see the lyrics in right panel)```

[tidal wharf]

"generations with intimate parts" is not a phrase I expected to read. In either case, I think what can sound like noise reduction applied could actually be the neural network mimicking a noise reducer, the same way it mimicks delays, reverbs, autotunes, and other effects

[hidden sparrow]

i have a suggestion for suno ai to have it where it continues the voice of the person you made the first song withj that way you dont have everysong sounding like different singers lol i am a paid customer btw

[lofty sonnet]

style, instrumentation, vocalist lock would be awesome. Like we could go in hear examples of different textures and tags and pick our own pallete of sounds, styles, rhythms, archetypes and music direction

[hidden sparrow]

like my post share it bro it would make it where we could make full albums and also i was thinking how about we make it where instead of it being an ai all the way it does the instruments you record your voice singing and it splices everything together

[steady scarab]

You'll agree with me that this technology is poised to completely revolutionize the industry in just a few months.

[lofty sonnet]

Yes it already has

but it has the potential to become SO powerful

like a style profile.. or like a defined psuedonym/style preset, almost like custom gpt for a artists/band/type of media generated

could select your customgpt style you've already programmed with negative prompts, etc.

I'm simultaneously working on countless genres and niches that it would help organize things automatically too if you could search by your style preset/band custom preset in libraries to and work on multiple musical projects

remix almost does it, but I think this is a distinctly different option we could explore

[steady scarab]

I like that!

The AI trend was lacking in music, but they changed that with this publication.

About time

[lofty sonnet]

Explore by tags, like instagram, and have them have there own trending charts, discoverability of types of content able to be produced with suno, like in explore you could type and search for tracks with "tuba" in its tags or "elevator music" or the combination of tags, so the more you add the more you get closer to a type of sound, and they have a charting aspect to them still even in combinations or less broad fewer tag niching which would be insane for potential to discover new tags and types of music through, follow new ai song creators and radio station builders getting so down into

[south meadow]

I'm excited with this music generation tool. I haven't done enough tests because of the limits, but I have suggestions to improve the system.

1. I don't know when I should click in like or dislike. The tool often generates good songs with the wrong genre. I think the evaluation should regard three aspects: whether the tool generated the correct musical description (genre, speed), whether it generated good lyrics and whether the music is good/pleasing.

2. The custom mode has a field to insert the Style of Music, but it requires existing lyrics instead of the description of the lyrics to be generated. If I keep the field blank, the generated song will have random voices without real words.

3. It would be good if it allowed to edit existing songs writing about details to change. I would ask to change an instrument, the harmony, the voice, the speed and other details.

[sage mirage]

1. be heavy with the dislikes, a bit more stingy with the likes. I also use them as an ad-hoc organization tool.

[sage mirage]

2. enter what you want thr song to be about, hit this button

[steady scarab]

I encourage the developers to create a portal for precise suggestions, utilizing specific selection menus. This would enable them to prioritize the most critical needs from both industry users and end customers. By doing so, they can swiftly focus on what truly matters. There are some great ideas within the comments on Discord, and I fear they might get lost in the flow without a structured system in place.

[lofty sonnet]

With the option of employing a customgpt element, ai producer agent, it's like you build your own project, or your own studio environment equipped with exactly the assets and elements or palette you would like to work from, converse with the gpt like a studio manager and get them to better understand the type of music you wish to create over time through a conversational approach, so you can hop back in, still play outside of the bounds, but have a foundation designed for sonic consistency and aesthetic purposes

[opal pulsar]

Discord improvment : Adding a role with the language of the user. It can be cool to have some discussion about a song. It's difficult to have advice on some clip choice when it's not in english

[haughty valve]

It would be great if we had Suno Gift Cards similar to Steam Gift Cards

[sage mirage]

more money = better than

[lofty sonnet]

THANK YOU FOR Continue From drop down on the right to track original gen!! Just saw it! I thought it was necessary.

[oak owl]

I would love to be able to upload a clip and continue it. For example I have a song I stitched together from several different generations (cleaned up some nonsense hallucinations and rearranged verses, etc.). Would be nice to continue from my "clean" clip.

[sage mirage]

An often requested feature but there is nothing stopping someone from uploading copyrighted music

[carmine igloo]

I have an instrumental piece of music id love ai to write some lyrics

[lofty sonnet]

plays and like data for tracks in our Library that are Public

[hidden sparrow]

that was my idea lol

[humble tendon]

@stark fulcrum I don't think this will happen, they're already having problems with people uploading copyrighted lyrics and getting them on trending
people will just upload full copyrighted songs then, I like the idea but I just don't see it happening unless they want a massive lawsuit on their hands.

[dusty dove]

Try to use REMIX is not 100% but sometime it help us. 😄

[untold bramble]

very sad that suno does not know what a cadenza is... wanted to see if its classical side was up to snuff 🥲

[opal pulsar]

It can be usefull to have negative prompt like -drum. In my actual song during a flute instrumental break a drum just pop from nowhere like a savage pokemon xD

[humble tendon]

That's actually a great idea, like a negative prompt for stuff you DON'T want in the song, +++ this

[sage mirage]

It's been talked about before. tony mentioned if we'd still use it if negative prompting cost twice the credits

On account of how prompting works

You'd prompt twice, once for your positve prompt, then for your negative. It isn't free.

We all said yes, at least the people here at the time

[opal pulsar]

not an issue tu pay

[sage mirage]

That's what we said, yeah.

I'm going to burn the credits anyway, either trying to chase what I'm after by brute force, or spending the credits to get closer right at the start.

[opal pulsar]

Yes working to remove this 🤬 drum cost me so much credit xD

[dusty dove]

Questions!

1. How many characters TextArea Box accept??? For now is 1000 characters max.
2. How many minutes we have directly without use "continue..."???
3. How many minutes we have to continue??? For now is 1 min (60 segs)

Thanks!

[red ember]

Be able to specify a min and max length
Be able to specify a key

[feral aspen]

Here I fixed this one tonight at least, simple wavform/volume/speed control. https://github.com/JonathanFly/chirp-studio It's slapped together so it only works on desktop full screen probably, I just positioned it on top of things for my display.

[lofty sonnet]

Are you able to make one like traktors? that changes colors based on frequency content? It's so nice to have a visual cue too nowadays

[feral aspen]

It's using this. https://wavesurfer.xyz/examples/ I am not very knowledgable about audio tech, maybe you can check if there is some option for that?

[lofty sonnet]

Bro still awesome, thanks so much!

[feral aspen]

Wait until you see the next bookmarklet...

Actual super useful. Maybe coming soon

[lofty sonnet]

O.o

[lofty sonnet]

I mean who doesn't think this would be sick, and helpful when inpainting/continuing from???

and those section markers could be snapped to what it knows the tracks beats/bars are for ease of continuation or fast editable section re-prompting

[split tree]

Voice cloning

[sinful mulch]

Can't seem to find a way to permanently delete playlist from the trash, should be added unless im just blind

Also, I think that the clips library should get rid of the pagination and load more songs on scroll like playlist do. Right now it is easier to find all the clips i've ever made by having a playlist of every single clip and scrolling through them rather than using the clips library

[radiant shoal]

After a song is generated with lyrics, it would be great to generate an instrumental of that song.
It would also be amazing to export stems!

[spiral summit]

• As JF said, it would be great to remove the vocals without losing quality.

• Also, it would be nice to control the timing of the song before and even after it has been generated.

• I’d like to see a better fusion of genres, so we could experiment more.

• We could also have the option to choose between male and female voices.

Thank youuu!

[strong cradle]

• We need a negative prompt

• Would be good to be able to add things to the beginning and not just the end of each clip. Sometimes Suno cuts with vocals right at the start.

[tranquil whale]

Hello, the app seems incredible to me. I am a music producer and it is a great help. I have several suggestions that you could implement.
1st Being able to extract the midi of the songs. (to be able to work them in the Daw)
2nd Being able to do Acapellas
3º Being able to edit the songs that have been created, for example, to add or remove elements.
Thanks in advance to the Suno team. You do a great job

[dire hollow]

• I would like to be able to generate a piece and then later generate the same item with isolated vocals or isolated instrumental.
• I would like to be able to stop a generation when it's not going into a direction I want it in the first place

[desert whale]

The ability to upload and swap our own vocals would be amazing… maybe too OP ha

[hybrid crane]

Thank you for noticing. I will let the Devs know.

[hybrid crane]

I see why you'd find these features valuable! Generating an instrumental version of a lyric-based song and exporting stems are highly requested features in Suno. The team is aware of this demand and is exploring ways to implement them. This investigation involves addressing the complexity of accurately separating song components and ensuring quality in the instrumentals.

[hybrid crane]

1. Removing vocals without losing quality is a feature under consideration.
2. Timing control both pre- and post-generation is being explored.
3. Fusion of genres for more experimental freedom is a noted suggestion.
4. The ability to choose between male and female voices is already available through style prompts and metatags in Suno.

[pliant canyon]

suggestion - maybe a bit more control for [] ... don't know if this is possible .. for example [Intro, 8 bars, 145 BPM, Instrumental, no drums] [Verse 1, 16 bars, 145 BPM, female vocals] [Instrumental Break, 145 BPM, 2 bars, riser] [Chorus, 16 bars, 145 BPM, catchy tune] - and the "continue from" should begin a few seconds before to avoid cutted vocals

[hybrid crane]

1. Negative Prompts: The idea of using negative prompts to specify what you don’t want in your music is indeed under exploration. However, this feature demands significant computational resources. As a result, it could potentially double the credit cost per song, reflecting the increased GPU usage necessary for more precise control.

2. Editing Song Beginnings: Your suggestion about adding elements to the beginning of clips to prevent abrupt starts is linked to the 'paint in' feature that Suno is researching. This feature would offer more detailed editing capabilities, allowing for adjustments not just at the end but also at the beginning of clips.

Both features reflect Suno’s commitment to enhancing user experience, balancing the need for more control with resource management.

[hushed oasis]

would be cool also to have more control on the start or the end of the song like: [intro, 10 seconds] or [outro, 10 seconds]; i would not expect exactly that time but more likely around that time

[true rover]

As a user,
First, when I find a soundtrack with the right style
I hope the system can tell me what the label of this soundtrack is.
This makes it easier for me to find the style I want in subsequent creations.

Secondly, I hope that I can systematically ensure that the songs created are complete.
(Including the complete ending of the lyrics and the ending music of the ending paragraph)

AND!
There are currently a large number of errors in Chinese pronunciation
Please make sure Cantonese and Mandarin are two languages

[hybrid crane]

1. MIDI extraction for working in DAWs is a notable suggestion, is asked before and could be considered.
2. Creating a cappellas can be done with adding "a cappella" in the style of music field, but as with all prompts, it can be unrealiable still
3. Editing created songs to add or remove elements is linked to the 'paint in' feature under research, aiming to offer more in-depth editing capabilities.

[hybrid crane]

1. Generating the same piece with isolated vocals or instrumentals after the initial creation (aka stems) is a feature currently being researched in Suno.
2. The ability to stop a generation mid-way is not available in Suno as of now, but it’s a valid suggestion for future updates.

[hybrid crane]

Uploading and swapping (your own) vocals in Suno is a feature that the team would be excited to implement. However, there are significant copyright (how to make sure it is your voice and not a celebrity) and technical challenges to consider, which currently make it difficult to offer this functionality.

[hybrid crane]

Your suggestion for more granular control in Suno, like specifying bars, BPM, and specific instrumental or vocal instructions, is an area the team is keenly exploring. These advancements in AI music generation, while exciting, come with their own set of challenges. More detailed prompts require more sophisticated AI processing, which can increase the cost of development and possibly the credit cost for using these advanced features. Suno balances the pursuit of these innovative features with the practical aspects of their implementation and user costs.

[languid gust]

Add Volume Control please!

[hybrid crane]

Using specific time-based metatags like [intro, 10 seconds] or [outro, 10 seconds] is an interesting suggestion for Suno. While exact timing control is currently not a feature, it's a valuable idea that could guide future updates to provide more precise control over song sections.

[stiff wagon]

I hope when playing the songs created, it can highlight the current lyrics in the right side, let say I am working with foreign language , it would be easier for me to continue and finish the song
Also allowing no login play song for visitor

[hybrid crane]

Highlighting current lyrics during playback is an interesting suggestion that could enhance user experience, especially when working with foreign languages.
Playing songs without logging in is already possible in Suno. If you share the URL of your song, anyone with the link can listen to it without needing to log in.

[stiff wagon]

1. Yes it is nice for UX
2. song link is public but playlist is not the case, playlist need to login to check it seems.

[hybrid crane]

which is logical (with playlist) because that is a feature for users.

The team could consider giving a public option to toggle 🙂

But that brings another problem, since eveyone can put songs in a playlist, also from other users. And some might not wanting their songs to be shared outside of Suno.

[stiff wagon]

playlist public option would be nice if possible, private in default

[mint thorn]

add in the music style the possibility of writing the name of the artist?

we may not have the same voice but we can get closer to the style of this artist? There are already typebeats on YouTube and voice models like RVC... why block the names of artists? 😢

[hybrid crane]

Voice models of RVC are often illegally created. Most do not have consent from the celebrities they (ab)use. Suno has to abide copyright laws.

[mint thorn]

I know well ! but that's not at all what I meant, sorry, I wanted to know the reason why Suno blocks artist names? For example I would like to make sounds as catchy as juice wrld but I can't describe them x)

This is just the vibe we'll get with Chirp and no copying or plagia?

[hybrid crane]

Same answer.. copyright

[mint thorn]

in these cases, the music can file a copyright complaint, I don't really understand, for me I call it inspiration

sorry for inconvenience

[hybrid crane]

If you don't understand, it's hard to explain. We cannot use their their name. Nothing I can do about it. Sorry

[mint thorn]

oh k, is just about "the name" ! 😮

I thought taking inspiration from their vibe was plagiarizing but I misunderstood, sorry

so sad for me eheh

[hybrid crane]

If you have more questions about that feel free to mail them [e-mail]

[mint thorn]

thanks

[worldly falcon]

Add a "Your year in Suno" thing like Spotify does, where it tells you which songs and genres you listened more throughout the year

[thorn comet]

would love to see a sample (Melody) only button without drums & also a super clean stem download option would be also cool to generate unlimited songs & stuff for a yearly fix price

[worldly falcon]

Do "song challenges" where people can compete in creating songs in a certain genre or theme decided by the mods, and the best song gets a price of credits

[silver peak]

In the app, it would be great to see the "dislike AND delete" button. Because it is far too often when Suno generates 15-sec endings or just some nonsense that we don't need in the library.

[dreamy arch]

it'd be cool if you could donate your credits to artists that you liked within suno,

[potent crane]

This is nuts...had a song (called "Journey", that's not important) - I have a playlist for completed songs. So...I added it to playlist. Fine, I am guessing the copy on my current workspace is no longer needed, so I move it to trash. Then I'm thinking, hang on, I hope there is still a copy of "Journey" that I have just added to my completed song playlist...noooo! It's gone! So if you trash a song from your current workspace it ALSO trashes the "copy" you have sent to your playlist. The solution? Have a look in trash and restore it. You would think so, but trash is not searchable and I strolled through ONE HUNDRED pages of trash, hoping against hope my eyes didn't glaze over and I would miss it, but, no, it's not in my first 100 pages of trash, and life's too short, so I no longer have a song called "Journey" - no biggie, but what if it had been a banger? Cue weeping and wailing. The solution - You should be told that sending a song to your playlist means that if you delete the version in your current workspace it deletes all versions.

[steady scarab]

Despite the gravity of the topic, the post brought a smile to my face.

[hybrid crane]

The playlist, is a "playlist". It's not a folder. There is a misconception here. 😉

[potent crane]

Well, I have been reminded of that in the most stark way. Folders should be implemented asap.

Thinking about the whole idea of a playlist within suno, I don't see the point, all completed songs I create in suno get's copied to my hard drive, and sometimes extensive work is applied to each song, editing, re-organising verses, adding intros, adding effects, mastering, meaning the raw copy sitting within suno is of no value whatsoever, and why would I want to listen to it within suno ever again?

[hybrid crane]

You will see.. there is a reason for playlists 😉

[sage mirage]

never trash anything

NEVER

[hybrid crane]

yup

[hushed oasis]

That would be great!

I think many noticed that people asking often for an intro before jumping on the lyrics or a less abrupt end to the song 😄

[tranquil whale]

I understand your frustration

[hushed oasis]

at the moment you have to have good luck or make an instrumental track first; it would surely reduce the generations attempts by users and maybe the load on suno in general aswell, while reducing the frustration that might occur to users trying many times to get a "specific" intro or outro lenght

[potent crane]

I think the number one problem at the moment during creation (disregarding layout/features/functionality) is the consistent lack of instrumental intros, and the lack of prompt suggestions that actually work. I try "intro" "guitar riff intro" etc but I would say 90% of the time suno ignores this - and yet suno regularly implements "ends" and "fades" without a problem, so it's not that it ignores ALL prompts.

[hushed oasis]

i feel you 👍 i think aswell that the biggest inconsistency is especially in the beginning, but maybe it's given by how suno generates the songs currently, but dunno

[lofty sonnet]

I'm excited for playlists to hopefully eventually be able to turn tracks into seamless mixes

[hearty goblet]

they really need an intuitive folder system... drag n drop...it should be a high priority

[grim flume]

Good evening, everyone. Is there any news about v3 development progress or v4 release? At the moment in my opinion the main problems are:

1. Does not generate repeated verses, pre-choruses, etc with the same musical and vocal line as the original (biggest problem!).
2. Does not generate a musical beginning, it is either missing or there are hallucinating vocals there
3. can't correct the selected fragment of the song, from the right second to the right second.
4. The hint is too short and does not allow you to define the desired result more precisely. Would like support for working with sids and samples.

[red ember]

[Changeup] tag for forcing a different direction of the song, while keeping with the same overall theme

[strong pollen]

Some very nice-to-haves that I hope are fairly low hanging fruit:

1. Start a new song from a particular position. Sometimes after its decided that a song is done and gone through an outro, it'll start a whole new sound. Right now, extending from there takes the rest of the old song with it, so it would be cool if we could start from there. Then, ideally, the model wouldn't be influenced by what came before and try to repeat elements.
2. Continue from playback position. I find myself scrubbing around in the player for the spot anyway, so might as well use the player's position as the starting point for "continue from this song."
3. Use the same info as 'Remix' for "Continue from...", minus the lyrics. Usually I don't want to change the style or title when continuing, so copying that into the dialog would be a nice quality of life improvement.

[inner coral]

Since not everybody was open to have a star rating system (For some reason). I though we could have something in between, that is more granular and easier to see what you liked and what you really liked. This would look something like this, with a pop-up when you mouse-over.

[silver peak]

Say 'hello' to colorblind people. You may change that easily with emojis: 🫤, 😊, 😀 or 👌🏻, 💖, 🌟 if to be a bit more creative.
And no need for more than 3.

[inner coral]

Settings --> Color Blind Colors --> Save

[echo flower]

Why not just let people add and sort on tags and turn this into a social media platform? </sarcasm>

[silver peak]

Tags actually a good idea, especially if there will be more complex compose options.

[echo flower]

Once there is an API for this thing, maybe we can just make our own myspace out of it, lol

[inner coral]

Yeah Mateo's idea look like social media and it's bad. Having a granular option to know quickly and easily what you like a bit, more or a lot is quite useful. Tony seemed quite open to the star rating option, but I think it would take up too much space.

[echo flower]

honest suggestion: They should just have an output that pushes outputs to an external (partner) audio editing webapp (something DAWesque) that they don't have to manage themselves.

[silver peak]

Something like Fadr? This may lead to a bit complex legal stuff. Besides, their current compose generates new little fillers to pieces - this you will not get on external DAW without API. So... kinda of not win-win.

[twilit flare]

may i ask and may well have missed this is the a way to keep the music or vocal voice typically the same i do an new song but keep the same voice and music cheers

[silver peak]

I guess this is a bit off-topic.
This may help, with respect: deepl.com/translator

If I understand you correctly - there is no way to get consistent music or voice. Kinda the way how AI models work in general

[echo flower]

Yes like that. I'm going to have to mess with fadr in a bit. I just use flsudio.

[nimble vortex]

BIG suggestion - ability to take a clip/song and extract just in the instrumentals from a fully run song with voice and lyrics. That way - if you like the output - you can immediately produce a karaoke version!

[echo flower]

you can do that in flstudio and it looks like maybe fadr does it too.

[silver peak]

Yep. So @nimble vortex may make his dream karaoke stuff right now.

[nimble vortex]

Cool thx

[faint plover]

Curious, can the songs be downloaded in midi? especially for the inst. would be super cool

[tepid trench]

Pretty sure there's no MIDI involved in this so that wouldn't be possible.

The closest you'd get is if Suno could do an audio-to-MIDI translation but there's likely services/applications out there that already do this

[echo flower]

it looks like fadr might do that. I haven't tried, I just now heard about it from @nimble vortex You can do it in flstudio with effort.

[echo flower]

fadr does it. I'll be testing more to see how good it is, but it's there.

[silver peak]

...this chat turned from suggestions into the fadr promo xD
But if to be serious, I guess that Suno generates music directly as mp3 or similar format, not MIDI.

[sage mirage]

It does

[echo flower]

Yeah, but there is nothing stopping from you from taking to the next step. And if there is a suggestion anywhere in there, it's that suno should partner with someone else for post processing at least in the short term so that they aren't wasting dev time creating stone wheels when tires already exist.

[sage mirage]

There is no point for wavs or anything else because the model itself doesn't contribute

It outputs as a master, full stop

It's been talked about a few times by the devs

Actually let me see if I can find it

#💡┃suggestions-web message

There is the link to tony specifically talking about it, it's in that general conversation

[formal bear]

hello

[sage mirage]

whatup tony, talkin' bout that hot quality issue

[formal bear]

wave file again?

[sage mirage]

ya

[formal bear]

😅

[sage mirage]

I take it Suno wants to keep as much in-house as possible, right?

I get that vibe

[tepid trench]

This is about MIDI at least, not wave

[formal bear]

Interesting.

Is this about generating the output audio + MIDI?

[tepid trench]

Getting the instrumental/melody part as MIDI I think - #💡┃suggestions-web message

[formal bear]

We can probably give you something low quality if that is REALLY necessary, but I am pretty sure that's not what we want to push for now.

I mean...v3 hasn't been released yet, right?

shouldn't I work on v3? 😅

[faint plover]

yeah, i think some of the songs are nice ideas, but with some good elbow greese they could be bangers

or at least stems

[formal bear]

for stems. there are multiple software around that can do that

if you want a cheap one, just use torchaudio.demucs...

[faint plover]

but midi for inst. would be perf

yeah thats step 1.2, 2.0 midi, so on

[formal bear]

honestly I don't love ppl uploading suno gens onto other platforms, but my boss made it pretty clear that if the demucs quality isn't SOTA we won't give demucs to you

so...got to wait

[sage mirage]

You guys getting a surge of users lately?

Seems like Suno is really starting to bleed out. When my father in law mentions it I'll know Suno has made it

[formal bear]

we still have a long way to go...

v4, v5...v6...

[silver peak]

By the way, Tony. Any insides about more controls over generation? Besides inpaint. Like weights, tags, etc?

[sage mirage]

I can't wait to hear v4 and dump all v3 gens like I did with v2

We might get negative prompting and have to pay for it

x2 baby!

[formal bear]

yeah...at some point these can be pro features

[silver peak]

well, if that will get my results in less than 500+ points...

[formal bear]

I'd love to see how you folks experiment with them

none of these controls are that magical, which is kinda why we don't want to scam you folks...

[sage mirage]

I want to break suno so bad

[formal bear]

lol you don't want to hear that I am sure

I have heard truly terrible things

[sage mirage]

lmao

[silver peak]

Yeah, that is understandable. But still, at least some sort of steering. At least my back from LLM's and SD tells me that.

[formal bear]

yep, true

[sage mirage]

Really though you can't just keep doing that though, right? I know you have to have your hands, er, ears to the grindstone and all that but surely you're going to forget what music is at this rate

[silver peak]

Great to hear

[formal bear]

which is why it is still on the table

[formal bear]

just wait yeah

but these controls will not come tmr, I am pretty sure

but hopefully also won't take another year...

you can make an educated guess of the time 🙂

[sage mirage]

v2 to v3 was pretty quick

[formal bear]

was it?

[sage mirage]

I mean yeah

[formal bear]

no

[sage mirage]

Seemed like it to me

[formal bear]

got to move fast

[sage mirage]

I also don't know what 'fast' is anymore because all this LLM stuff is a vertical line

[formal bear]

yeah it is hard to define...

[silver peak]

...oh, you got the same problems here?)

[sage mirage]

sure, it's discord

[silver peak]

There are tool to limit links

[formal bear]

any other quick question that I can answer?

[sage mirage]

I got nothing. Thanks for chattin', have a good night!

[silver peak]

Like the Meaning of life? Good night and thanks for a talk)

[formal bear]

42

[silver peak]

tried

[formal bear]

just enjoy it then

hope y all sleep well, tmr is going to be a big day...

[sage mirage]

Oh right, good luck on whatever the end of the countdown brings

[formal bear]

I tried

I hope you folks like it

[sage mirage]

Suno hasn't missed yet

[tepid trench]

except for lyrics sometimes, but maybe they fixed that 👀

[formal bear]

maybe... 😏

it's never going to be perfect

but hopefully it is except for lyrics few times...

[silver peak]

Hallucinations are a part of the current state of AI stuff :/

[sage mirage]

Slip over whatever that line and yeah, it's golden

Most of the time it works everytime

[formal bear]

we are all against the odds of the computer, but may the odds be with you when you generate 😄

[tepid trench]

Yeah I know. I'm not really that bothered to be honest, just interested in seeing what improvements have been made 🙂

[hybrid crane]

At least we live in interesting times 😉

[formal bear]

yeah me too honestly, very curious of the feedback

[sage mirage]

"wow its magic"

probably going to be most of it

[formal bear]

lol I am like a singer waiting for critics reviews

[sage mirage]

Wear your best suit tomorrow

[tepid trench]

In AI land we don't wear suits. Just hope that everyone hallucinates them on us.

[abstract wadi]

@formal bear free credits for the subscriber tomorrow? 😭

[sage mirage]

I'm dressed to the 9s right now what ever do you mean

[kindred carbon]

You guys can convert these mp3's to midi by doing the following: Get RIPXdaw - import suno mp3 file - tell it to strip/isolate layers and spit out midi for each layer. Now you have to import that MIDI file into a DAW, and re-assign each layer with all the sounds/drums/etc. So far this is the most accurate and easy one stop solution, but it is still hit/miss at times. You can also use UVR5 to stem out first (freetool), then follow below.

Midi data is just that, it's data, there is no AudioData in Midi files, they are simply just triggers :). SUNO would have to basically do the same thing, isolate/stem each layer, then covert those layers to Midi/PianoRollFormat, then save them as a multi-layered Midi file (and no vocals, they are replaced with midi notes too and treated as another instrument/layer of trigger notes).

There is also a free tool called NEURALNOTE that can deconstruct your audio to midi (or google wav to midi / mp3 to midi there are other wonkier solutions), even layered, and you can adjust sensitivity. This is best used with stems not a full song, use UVR5 to stem out first (free)

If you're in Logic, you can also use FLEX after you stem out/isolate a layer - then export from FLEX as midi :).

Bonus: If you're already in a DAW, the best realtime VST/AU isolator I've found is by ACON DIGITAL (very handy for real time quick checks or isolation of a layer to bounce etc).

[glacial rune]

I would be nice if Suno just offered the option to export as stems.

[cyan umbra]

v3 rarely hallucinates for me any more. I learned it's secrets LOL just in time for the upgrade which will change all the rules again I'm sure

[sage mirage]

The issue is that isn't how Suno operates, so there would be this sort of conversion that would have to be done

[sage mirage]

as it always is

[kindred carbon]

Yeah we've had this convo many times it's wayyyyy up there though, in great detail, and higher bitrate .wav for destructive editing/much better initial separation/more data/polyphony... For now these are great workarounds. I upsample to 96k/32bit often as a first step to preserve and better isolate/edit without more loss.

If you haven't seen it yet, get UVR5, it has the newest isolation/stemming AI models to play with not available in commercial solutions yet and you get a variety of results based on what you're feeding it. EDIT: and is free..

[sage mirage]

Yeah UVR is cool as heck

[glacial rune]

Very nice, I'll check that out as well. I wonder how long it will take Celemony to start adding features like this. 🤔

They seem behind the curve.

[faint plover]

My laptop blows up if I run uvr 🥹

[sage mirage]

[kindred carbon]

^ LOL, the best web-based isolation tool I've found is LALAL - it even does things post process to deverb, demud, normalize, etc. I think they have free options too.

[rapid oracle]

Where can i find all style of music models?

[languid gust]

please add more sites for Explore site. And Filter

[rotund orbit]

Okay, here's a quality of life update that's totally attainable.
In it's finished state, it could be a visual tool that lets me clip either the beginning or end of a song, so I can get a more coherent piece. Or, if for example, I don't get the result I want until I'm a verse in. That happens. Alway, you would be able to sort of pick out a range, click the button and get the song between those two timestamps. In it's basic state you could do it tomorrow the same way you do Continue from.

It would also be good for clipping unwanted fake out endings

[formal bear]

yeah

sure

not hard but not going to happen tmr...

[rotund orbit]

No worries. It doesn't need to happen tomorrow

[formal bear]

I am sure it will happen...

[rotund orbit]

cool

[grim cedar]

I don't know how feasible it is but one issue I run into is this. Let's say I have Song 1, I continue and get 2 clips: Clip A and Clip B. I'm trying to add a verse and a chorus, but clip A has a great verse, clip B has a great chorus. It would be nice to "Get whole song" with Clip A and then stitch Clip B onto the new song, as if I had continued from it. If that makes sense.

[formal bear]

I mean it is feasible...

but I think it adds more complexity than resolves it

[grim cedar]

Yeah I guess I can see that, maybe being able to reference another Suno generation? so you can take say the verse from Clip A and then when you generate again you can reference clip B so Suno knows what sort of sound you want

I dunno, I just feel I lose so many great generations because of an issue earlier in the clip which makes the later part of the clip unusable and difficult to recreate

[formal iron]

spill

[earnest pewter]

I'm using V3 and want to comment some issues: the voice generated have ghosting (an extra voice in the background) and this terrible and ugly electronic accent, the song generated has** low bias gain (V2 sound stronger and more transparent without artifacts) the quality of the music generated are very low and bad, you can hear chip midi electronic instruments I want super high quality musical curated instruments**, also want to promote the use of blockchain as NFT to better protect the copy rights , V3 understand better the music styles, also very important want to create the playlist file at the moment to label the song in the song name (wen save this automatically will create the folder playlist for all the music generated with the same song name) . last when SUNO is in default mode (not the Custom mode) please make available the possibility of Suno itself to continue from this clip to create the next part of lyrics. V3 mp3 generated are low bias V2 is better better clarity, music fidelity is gold and you need to improve quality of generated songs to do that you must separate the voice from the melody since the song generated add artifacts to the voice generated lowering the quality.

[covert schooner]

Do you guys have a public REST API available by subscription? We are thinking about an app...

[sage mirage]

Suno does not offer API access at the moment

Later they might

[covert schooner]

OK, thanks!

[queen plume]

I finished my first full song today, but now really want to strip the vocals out so I can record over the instrumental. I tried lalal and another tool with not great results

[sage mirage]

How powerful is your computer? I ask because UVR, Ultimate Vocal Remover, works well for me, but it's locally run

[earnest pewter]

for male or female I use this in the music style " rock [male] or latin pop [female]

[hushed oasis]

user discord roles based on paying subscribtion tier, pro and premier

[willow root]

Suno: Add a Name, creation date, length section to the top with the ability to click the sections to list according to name or length etc.
Suno: When creating a song continuation have a checkmark to tell Suno to end the song with this continuation. It's annoying to have to keep making lots of continuations with the hopes that one of them will have an ending.
Suno: Instead of having an obscure and really hard to find trashcan in the upper right corner of the Library, add the trash can to the left upper menu under since there is plenty of space anyhow
Suno: It would be incredibly nice to have a "Negative prompts" section. I was trying to make ballads and it kept making piano songs. If I could have put "piano" into negative prompts, that would have been awesome.
Suno: Ability to color code rows to symbolize a project. So for instance, maybe I need ballads for one project, I could color code those songs pink. Then maybe I need songs for a video intro, I could color code those songs blue.
Suno: Using the thumbs up option does what you expect, but the expectation for the thumbs down is to add a red thumbs down symbol INSTEAD of deleting the song. It's extremely bad usability for it to actually remove the song by pressing thumbs down!! And then when you consider how hidden the trash is, many people will go into a panic as to where the song went when all they simply wanted to do was to "mark the song as not so good".

[cyan umbra]

In English, at least, Suno v3alpha is very responsive to capital letters for emphasis. If you listen to how it's trying to deliver and compare in your head to what it should sound like, you can get it to match the rhythm very close with creative capitalization. And you can get even more with some apostrophes (vict'ry). If Suno can find a good delivery it rarely hallucinates. And honestly the times it does tend to improve the song.

[formal iron]

Awesome prompt engineering 👍 thanks

[formal iron]

Can you expand your point. using an apostrophe causes what exactly? Does it add a mute? Or an inflection?

[cyan umbra]

Here, one example. Even if you don't like the music, check the capitals versus how he delivers the lines. https://app.suno.ai/song/bc192375-4e5c-4273-9dc9-de10b8a2144d The other thing that prevents hallucinations is lots of ad libs. As they're like "controlled hallucinations". It'll put them in if it likes them, or skip them if it doesn't and not hallucinate where you don't want to.

But I tend to under-ad lib mot of my songs, so they're not really necessary most times.

[formal iron]

“The Storys truly Messed…” completely illustrates your technique - changing to shouting on the capitalization

Pretty good song 2

[worn pelican]

Hello. Please allow us to make simple edits to our songs, such as trimming off x seconds from the end. Also the ability to add simple fades at the end, sometimes a continuation is not needed to finish up a track, it just needs a simple fade or cut at the end

The ability to write notes for our tracks can also be useful for planning edits in post. As we review our generations we might write notes about the tracks as we are listening to it.

Future models may even be able to use these saved notes to possibly improve the generations for the user (at a credit cost), to better align it with their vision for the track. While a simple note like "add wind and wolf howling sounds to the intro" is useful for me to manually edit the track, eventually these saved notes could be very handy to have for the reason I mentioned. They would essentially serve as prompts to better align the generated tracks with the users vision.

I would also like the ability to further differentiate between a liked track (thumbs up) and a truly exception track, which could be denoted with a star icon. A favorite track. This will improve filtering if I am trying to only find my very best tracks.

I might have many tracks that are pretty good, but only a few that are truly my favorites.

While its true I could create a playlist, that creates more management for the user, as they now need to maintain this playlist. Truly looking forward to any improvements to the library and web app in general, I know these things are very hard to build...

[broken maple]

2 suggestions:

1. a text-string Search feature / filter for user library. If you have ever used 'Search' on Discord then you will know the value
2. maintain active V2 along side V3 on release. Even if no further support or updates are given to V2, it expands the abilities adding value to users, much like how previous stable diffusion / Midjourney models can be used ongoingly

[sick marsh]

Have you considered building this into a Desktop Audio Workstation plugin for existing professional musicians? Such software sells for a hefty price tag, & you'd essentially just need to get the model to do less work, & allow an engineer to craft it segment by segment, with the ability to tweak after generation. Just thought I'd suggest it. I think you'd make a billion. Cheer!

hm, I perhaps should've googled it before saying anything. this is a very competitive marketspace, actually.

https://www.productionmusiclive.com/blogs/news/top-14-ai-plugins-and-tools-for-music-producers-in-2023-for-mixing-mastering-composition-sequencing-more#3

[versed inlet]

Just generally for the audio quality to not degrade over the course of the song.

By the end its just squeals and white noise with V3

Is this the way it's compressed?

[worldly falcon]

Not a must because it doesn't really make sense I guess, but a "Creator" page you can customize and put your songs and the songs you like on would be cool.

[hybrid crane]

🤫 That's a great idea... And they are working on something like that. 😉

[hasty pagoda]

yeah 100% , to build of this , Making it clearer when your songs have been added to the explore page and just also have many views/likes yours songs have might be a good addition.

[topaz dagger]

Add MORE cowbell..

[hushed oasis]

would be cool also to search for songs in the website, like by genre, creator, by language etc...

[desert whale]

Thanks for replying @hybrid crane 🙂

I can see how it would present a lot of challenges… would still be absolutely amazing. In the meantime I did have some feedback, I’ve noticed that vocal quality of outputs seems to be slightly less ever since the 2minute song expansion. This could be purely coincidental but just wanted to note my findings. Would also still very much love an option to export as stems and higher res .wav (I also understand this could be challenging and would require a lot more gpu power but maybe as the company grows?) 😊😅

Thx again, having so much fun using suno.

[cyan umbra]

Ohhh. I got one. A legit suggestion. Some sort of an "eyedropper" tool for the voice extraction. I've noticed the voice can change over the course of the song, especially as you add more parts. If you could say "At 0:15, the voice was perfect. Use that for reference for every continue." I would be in love with that.

[dusk terrace]

I think you should add an option next to the get full song menu item. This option should be called produce, and it should only be available for a full song, and what it does is take a completed song and polish it, it should put on a professional hitmaker music producer’s hat who has the final say on the song and do what currently is missing to make a professional radio ready hit song with options to export the final mix on higher quality wave type files. I’d pay extra for that feature.

[upbeat glen]

Más medios de pago 🙏

[past tinsel]

Do you have non-profit pricing for qualified NGOs or 501(c)(3)?

[mint thorn]

I know this is a rejected suggestion in advance..
But have the possibility to put a link for Chirp inspiration.
For those who want to make experimental sounds where the existing prompts don't work. or even add an audio note
It's as if the writer of The Monkey King would file a copyright complaint against the author of Dragon Ball... that doesn't make sense
yes I like to complain sorry x)

[deft mirage]

does suno have an API / dev plan?

[deft mirage]

yah or judt extra for hq 320kbps wave or "mastered" audio would be reallu cool

[echo flower]

Oh man, either my luck is really good lately, or v3 is dealing with starts and endings a lot better. I like the 12 lines of lyrics rather than 8. My big suggestion now is: PLEASE COME UP WITH A GIFT CARD SOLUTION.

[split owl]

no, still

almost 1k

[spiral summit]

Sorry if it’s a repeated suggestion.

• We could be able to capture a song we made on v2 and try to improve it on v3, so we could have a better quality. I personally loved a song I created on v2, but it has no beginning or ending and I cant replicate the melody using v3. It would be nice if the AI could detect the instrumentals, voice, melody, and make an improved and complete version of it.

Thank you! 👐🏻

[glad rain]

In custom mode, can we have separate input for the track's image prompt? Sometimes it picks up words in the lyrics or style that are out of context.

[naive elm]

id love to have other payment options for ppl without a credit card >_>

[mint thorn]

An opensource i can buy x) To make a possibility to add a link to inspiration x) But only one version i can buy in private XD
Because is impossible to me to create song like this :
Rounhaa - 2b2t
Les grands enfants - Wall-E
Les grands enfants - Joshua wong
Jima - grotatou
Luther - Garçon
impossible to explain on Chirp 😢

[sick marsh]

Please make a mobile app, or at the very least, make a PWA.

Also, this suggestion box might be easier to manage if it was a forum channel?

[brittle hatch]

For Arabic, Need accent based on countries like , Kuwait, Saudi, Iraq and Egypt. because most of the creating either from Egypt or Syria accent.

[ashen dome]

API integrated with applications for a micro saas. Consuming API and generate the song for a client.

[vital quarry]

an API would be awesome, then i could integrate it in my twitch bot to let suno generate unique songs for each new twitch sub 😄

[fresh latch]

I was able to use my debit card just fine

[shell quest]

WHY DID YOU REMOVE THE LYRICS FROM THE VIDEO EXPORT 😠

[hybrid crane]

it's temp. They are doing some updates in that part, with better font and such 😉

[glacial spruce]

How to access v3

[worn pelican]

Would like to pay a lower credit cost for slower generations.

[sage mirage]

They've mentioned that the cost is more reserving the GPU at all

[jade nimbus]

Am i missing something? Currently the trash bin is not showing my trash at all

I really cannot beleive that we have no folders. It is getting to be a big mess in my library. I would like to separate files and keep each song in different folders. the playlist is nice but this does not help the mess. We need folders.

I would pay extra to have folders so i can keep things sorted

[tawdry gazelle]

The new V3 has a big drop in creativity compared to V2 and misinterprets prompts sometimes.
I would type something like "true evil lovecraftian organ, impending doom" and V2 would create something close to the feeling. Now V3 just abstracts these prompts into something like "heavy metal dark intense" and it'll be a completely different genre.

Edit: seems the issue was temporary as the same prompts now produce much better results.

[vivid tulip]

I would like to insert a language in the prompt that don't lock me in the rithms of the country, like, for instance, if I put "Brazil" in the prompt, or "brazilian", I´m stuck with samba and brazilian funk. And if I put portuguese, I get European Portuguese accent.

In the Remix, I can insert "(language Brazilian portuguese) and it's ok, but in the prompt, I can't make it around.

[mint thorn]

Is there a possibility of creating a special Chirp with a very expensive private subscription, but which would allow the possibility of adding a typebeat in an option so that Chirp raps on it.
But also the best in my eyes, the possibility of coding a URL zone where we could drop a YouTube link to give inspiration, AND ONLY INSPIRATION

I even think that I would be ready to pay 80€ to 100€ per month for these options... Nojoke...
And for copyright concerns, Suno will have absolutely no legal problems. The only one who will be called into question is the person who publishes something bad.

I really want to make it clear that I don't care about copyright, listen to my 4-5 sounds that I have dropped below, and you will understand that it is literally impossible to do something similar...

Songs : (go listening the song, you really fast understand my problem x) )
Rounhaa - 2B2T
Rounhaa - Kontakt
Jima - grotatou
Sokuu - Try to live
Les grands enfants - Wall-E
Les grand enfants - Joshua wong
P'tite soeur - Malkhut
Luther - Garçon

etc... is a good base to see you what i want to make

[cyan umbra]

Ok, probably not going to happen, but going to ask. Any chance of adding v3alpha back into the drop down to finish any songs started with it? Was the code nuked? Just for a little while? Please? 🙂

[leaden marlin]

Allow us to upload reference songs. Casette ai does something like this

[jade nimbus]

humm as far as i read the copyright, suno has rights to our lyrics once we put it in suno. they can do anything with the lyrics, we own the melody fully but not the lyrics. yes we can use the lyrics as ours but if suno chooses they too can use the lyrics without credit or payment to us. even the lyrics we add/our own lyrics.

[shell quest]

It takes like 1min to start the generation of 30 songs, but then it takes 10mins to download all of them. I would appreciate a batch download feature.

[silk turtle]

Re-add lyrics to the video... i just made my mom a birthday song and she's hard-of-hearing, and i discover now that lyrics were removed. Why?

[jade nimbus]

yes the rights remains to pro and premium members but not to the lyrics. they have the rights to use the lyrics too, however they want. even your own lyrics..

[silk cargo]

We need consistency in the voice

and also more clear vocal

a woman voice turned to a man voice in a course of 4 mins of song

[fiery linden]

Try to merge your clips together more often. So instead of keep using "continuation", you merge into "whole song" after 2-3 clips. And then continue from there. This way the AI will be able to keep consistent voice more often because it takes into account of the whole song instead of just the last clip.

That said, I would still like a feature that allows us to have consistent voice not just within a song but carry over to new songs.

[grizzled heath]

Bring the lyrics back, as you can see the first clip is v2 but when i continued it just dissapeared even if i entered the text there

it just says the genre :/

[sick marsh]

Allow a toggle to default all generations to Public.

[grizzled heath]

okay wait

https://app.suno.ai/song/c21404ba-b10b-4f61-b02d-f6e7bb8de437

[mint thorn]

They're a Suno Team admin, to ask one seriously question pls ? i dont distrub long time... 10 minutes MAXI

[sick marsh]

Also, do a Discord role for paid users maybe?

[sick marsh]

maintenance crisis at the moment, be patient.

[grizzled heath]

it didnt work i tried few times but yeah, it's broken

[hybrid crane]

Your understanding of the copyright situation in Suno is not entirely accurate. While Suno holds rights to the performance of the lyrics (i.e., the AI-generated rendition of the song), the lyrics themselves, including those you add, remain your intellectual property.

[sick marsh]

They are currently overloaded with traffic, so they aren't available, likely. I am just another user. But if you have questions about features or something, maybe ask in #general & we can make a thread?

copyright issues will put a stop to a lot of what you're asking for there, but switching genres can be done with continuation.

[mint thorn]

....

[solar holly]

My suggestion is doing a true remix. So taking an existing clip but just converting that specific clip with the same voice/lyrics into a house beat for example or reggae beat. There is a remix function but it doesn't do a true remix, it essentially generates a whole new song clip.

[kindred carbon]

The system would have to be TRAINED on said copyright material first/inRealTime, it's not just you who would be liable. I get the desire, however I'm glad you can't just type in my own artist name and take my whole style so easily.. (this isn't my artist name, but I'm 'somebody' out here with a big catalog and tons of streams, and am sensitive to how AI is trained.

[lofty sonnet]

videos and tracks should be labeled to show which suno user made the song, the video released on instagram showed the "artist"

[sick marsh]

This is a lot of conversation about this in what's supposed to be a suggestion box....

[mint thorn]

yep sorry bye bro

[half lotus]

id love to be able to add more lyrics to custom mode. Right now I get enough characters to cover 1 minute, maybe 1 minute and 10 seconds. Id like to do full 2 minutes.

Is that what the update is for today? A way to set a 2 min request?

[solar holly]

Another suggestion: add more genres to the reference library. I tried zydeco and new orleans bounce and DC go-go music and it must not have any reference for them. They are pretty easy beats to replicate.

[past wigeon]

Generate viseme .json data with the audio so animators can use the .json viseme data for accurate lipsync. Checkout aws poly viseme data for exact examples

[echo flower]

is that a post processing thing?

[sick marsh]

Dude. You don't understand what the underlying issue is. They can't do what you want.

[fiery linden]

Using a reference song is just too big of a risk for them. It doesn't matter if the lyrics are 100% original, if they use a reference track to create a new song, they can get sued by the original artist which the song is inspired from. All you can do is try to prompt so the song sounds like what you want.

I think a good idea would be to ask someone who is knowledgeable with music and try to break down the song for you. Tell you what instruments, genre, style are used so you can try to put those in the prompts - hoping it'll match the style you want.

[haughty valve]

Be able to see view count on a song even if it's not trending.

[red ember]

I didn't see Psy-Trance listed in genres on the wiki, but it seems to understand. May consider adding to the list?

[nova dawn]

Is there not away to implement a length option? Say I only want a clip to be 36 seconds we should be able to say we only want 36 seconds. Likewise if I wanted a 2 minute clip it will make a 2 minute clip. That way it will eliminate Wasing credits when we want a full 2 minute but only makes like a 30 to 43 second clip. If that all makes sense.

[sick marsh]

I hope you find a tool that works for you to express yourself. Cheers.

@delicate cedar action might be warranted here, please.

[sage mirage]

[sick marsh]

apologies for the mass mention but i'm pinging those online.

@solar sable @echo canopy @feral aspen

[hybrid crane]

There is no need to police for us. If people to vent their ideas that is fine. I don't see the problem? And pinging the staff is like saying the servers are on fire. Please keep your cool.

[sick marsh]

I apologize. I was overstepping. Thank you for your reply.

[hybrid crane]

You've made your point.. moving on

[echo flower]

I just want to say before you give up, you can totally get the sound you're after if you keep playing with the style prompts. You sounds fairly cost insensitive. just keep describing what you want till you get some winners.

[molten egret]

I have an idea for a feature. A Remaster function/button. This would apply to songs generated on older versions of Suno to bring them into the newer versions. Have a song you loved in Suno V2 or V1 and want to recreate that exact same song? And have it sound exactly the same but in improved fashion for the newer versions? Just click remaster, and choose the version of Suno you want to remaster it to and voila

[echo flower]

that sounds almost like the img2img thing.

[echo flower]

a remix with a denoising slider would be very cool.

[cyan umbra]

Could there be a "reuse this continue" button here somewhere? Or could remix auto-set the continue from field? Having the original continue listed and clickable here is much better than going paging looking for it, but if we're going to remix a segment, anything that makes it less clicks to get what we're continuing from would be helpful.

[worldly falcon]

Not sure how interesting this thing can be, but giving credits a more custom name would be cool

[formal bear]

n_gpus? 😄

[worldly falcon]

NEON

[tacit idol]

They the way it is now, Chirp-Showcase is kind of a free for all.
Having a channel with genre specific threads on the Discord where we can share our songs might be a nice idea.
If someone is only interested in a listening to a certain genre, they could find them more easily.
and on the other end, if there's a genre that you're not interested in listening to, you can avoid it more easily.

I realize that it would be impossible if we tried narrowing it down to every subgenre, but just the base genres shouldn't be too hard.

A thread for blues
A thread for country
A thread for hip-hop / rap
A thread for metal
A thread for rock
A thread for pop
etc..

Hopefully it wouldn't lead to arguments over crossover / "close enough" songs being posted in the wrong channel. 😅
(edit to change "channel" to "thread", for clarity.

[sage mirage]

Who would even head the genre zone?

Trying to shove songs into defined genres, oh man

I really do like the idea because Explore is a free for all and I just kinda go "..." when I look at it now

But practically

[short blaze]

Request to add "Song Description" to custom, so that I can choose Style of Music and Title, and have it still generate the lyrics for me.

[short blaze]

oh thats right, in the non custom, you can type in the style of music along with the title and description all in a single box, but it's just weird because i always forget they work different because the inputs are different
So I think the 2 options "custom" and not, should be revamped to 1 interface screen with options.

[delicate cedar]

Can't you just put the song prompt in the custom lyrics field and hit "Make Random Lyrics" to do that?

[short blaze]

no

[short blaze]

"generate lyrics" are irrelevent to the title you put in

okay, so you have to type in what you want the generated lyrics to be, in the lyrics box and then hit generate lyrics... just weird

so this can be fixed very easily

[delicate cedar]

So do it like this:

That might work better...

[short blaze]

so adding that to the gray text, would remove all confusion

but worded better

[delicate cedar]

Fair enough.

[short blaze]

my assumption, based on the wording and the button, is that if i type something and click "make random lyrics" that it would delete what i typed in and generate random lyrics

I'm sure there is a perfect solution that is very elegant and doesn't require a paragraph to be read.

that explains everything crystal clear, but I can't make a perfect suggestion (at least not yet).

[untold bramble]

i am not exactly a UK drill fan, but I tried to generate some songs and the voices and sounds are EXTREMELY muddled still. think it has to do with the voices being trained behind heavy 808s in drill but would be worth looking into 🙂

[sturdy rock]

Agree, I like it when it had the lyrics. This cover art-only is not as useful for sharing as many times subtleties of lyrics are non-decipherable to the listener given the still-developing audio quality.

[livid ruin]

Suggestion:
It appears that older songs never make it onto “new” in the Explore section even if you just made it public.

I just checked and it appears the “New" area is for brand new songs and not brand new to public.

What if you had a banger you made a month ago and want to dip it into the explore page it'll never be seen if it's made public.

[leaden marlin]

non lyrics?

[tacit idol]

Well, the moderators would be able to moderate and remove things that are definitely out of place.
But it would be up to users to post their generations in the thread they feel their song fits best.

[worn pelican]

It would be cool if Suno could provide timestamps for syncing the lyrics of a generated song, in a program like Adobe After Effects

Right now it needs to be done manually, that could save alot of time for people that want to make animated lyric videos, those do well on social media

[formal bear]

would you pay extra for such videos? 😄

[worn pelican]

For the timestamps? Yes I'd pay extra

Because the video engagement would make me more ad money

Right now they are slow and tedious to make since I don't have the timestamps that match the lyrics

[formal bear]

ok but what if the timestamps aren't perfect, like everything

how much error tolerance do you have

[worn pelican]

Not perfect is still 10000x better than doing it manually

Its alot of pausing, setting markers

[formal bear]

honestly I agree

ok let me work on it tonight...

sigh

I thought ppl would just be happy with v3

https://tenor.com/bhZsq.gif

[worn pelican]

Oh its great, I am trying to make it better by making videos

For the viewer, they want that visual aspect

Thanks for listening 🙂

[formal bear]

you are welcome

thanks for the feedback

I need that motivation to work

[untold bramble]

when v4 alpha 😡

(jk)

[formal bear]

sigh

https://tenor.com/view/oh-no-homer-simpsons-hide-disappear-gif-16799752

[sage mirage]

lol

[delicate cedar]

https://cdn.discordapp.com/attachments/1021157626089848842/1076230395672150047/memes_homer_disappear.gif?ex=66071ca9&is=65f4a7a9&hm=19dbf89239c29c0608b53652082c185e5b73d06dcfaad6b080e4225eb0d9d6b2&

[sage mirage]

I'm not sure what you guys are doing for the social side, tony has mentioned emojis and things, BUT:
Since people can 'continue from' other people's generations, it would be neat if that generation kept tabs where it came from

That way you can see who started it, added to it, etc

Suno obviously has the ability to keep tabs on clips used to make whole songs

So naturally keeping that across users would be cool

Plus it might incentivise specialization. Some people might like to make hooks

or lyrics

Would also be a buffer against "THEY STOLE MY STUFF"

Since you couldn't, it says right there

[blissful kindle]

Are you guys going to add in a stem splitter for vocal swaps? If not, could you lol. Also. acapella vocal generation based on audio input and lyric text input would be amazing. I'm not sure what your setups like, but the way I theorize your setup, the acapella generation should be feasible. Maybe linking it to the beat maps you can extract with librosa and etc? I don't know.

[tawdry gazelle]

I'd love some attention to the breakcore genre for the model. Especially the more upbeat watery DnB type.
Doesnt quite get the vibes when prompting it out, often coming out pretty directionless and muddy.

[sharp sundial]

^^ Goes the same for Extratone. It doesn't exist. I would assume as there isn't a lot of that around.
Or, we can get some BPM control. It would be a great quality of life improvement. 👍

[blissful kindle]

For examples of how BPM control could be done, MusicGen Plus is a fork of MusicGen and it has BPM control added in. Lott of us were racing to bypass MusicGens limitations, I was the first to bypass the 30 second gen limit and took it past 30 seconds, but the maker of Plus beat me to coherent music several minutes in length. Required some looping and other tricks, but it works lol.

But BPM control seems functional in that repo

[pine tree]

Finish the song with the next music generation. Determine the number of seconds for the entire song, ending the song within 30 seconds. Decide the number of loops for the ending melody.

Use the space key to space the melodies. Use punctuation to space the melodies. mark to intensify the melody. To twist the nuance of the melody, use the ? mark to twist the melody. May the various symbols give you control over the melody.

-3° lower hamori, -4° lower hamori, -5° lower hamori, +3° upper hamori, octave hamori, etc. A setting that changes the flow of the song when the name of the music genre changes in the middle of the song. Should I write "minus3chorus"?

[tawdry gazelle]

I'd love some more specific instruments being trained on. Ive seen mention of things like Theremin and digeridoo yet they aren't able to be replicated.

It would be EXTRA cool if you can specify multiple instrument combinations in a reliable way; like Imagine the duet battles, or specific percussions, ect.
Often I find myself discovering a lot of awesome instruments and wondering quickly what they would sound like together, rather than opening up FL studio and trying to make something work.

[midnight shoal]

Cam you please remove the deep inhale at the beginning?

It's kinda repetitive

[lofty sonnet]

Need to work on reverb tails

sometimes they glitch when you don't expect them to, especially in dance music where there is stylistic sidechain compression

especially white noise fx and sweep sounds

[lost sapphire]

I really miss v3alpha , why must we not have access to it anymore?
I purchased a Pro account and developed a technique that would craft exactly what I wanted within 500credits of generations which made me confidently purchase the Premier subscription.. then v3 came out on the same day & now it's like I'm shooting in the dark again. 😐

So far, Definitive prompt instructions are highly ignored with v3. The technique i developed no longer works, im burning credits and getting no where close to the control or direction i discovered with v3a,

I strongly advise that we should have access to all models (v2, v3a/v2.5 and v3) as a paying Premier customer.

This moving goal post strategy is affecting the creativity suno wants to foster. Especially when there is no manual for this product, this is totally disrespectful of a paying users time and effort.

[lofty sonnet]

I would love an affiliate program!

[spiral acorn]

I thought I noticed something like this too, but it might be in head... I've barely done any generations with v3 lol. So I might have not done enough generations to even notice if this was happening for sure. Also even if there are subtles changes we might just have to adapt.

---

I would love if there was a 40,000 credit option, maybe $60 🤩??
Or even a custom slider that scales up.
I know we can already contact for a custom amount but I'm not sure if I can commit to the exact custom amount every month so that's why I was mentioning this.

[rancid bronze]

I used about 5000 credits during the last 2 weeks with alpha 3... I ended up with max 50 songs that still have several glitches but are ok. Suggesting clients that 5000 credits would be enough to create 1000 full songs is impossible, especially if you use own text where you need a lot of different retries, because suno doesn´t pick up correct instructions and is quite random. Also often you are stuck because Suno flags the song for moderation. So from my personal experience, for 8$ a month you would get ~25 songs and not 500, which is still ok. If other users have different/similar experience, please dm me, as I´m currently preparing an article about current AI in the media industry. Are there any devs from Suno here in this chat or ist discord just for the community?

The price is ok, but the content moderation annoyes me the most, are there any ways to bypass it via api or something?

[hushed canyon]

Since all of the generated songs contain "white noise" I wonder if noise suppression could be applied post processed, if this white noise is a base layer artifact which is applied per progression. Than having isolated base noise for 500ms or so and use this for denoising the end result 🙂

Also could the delete button be moved there? So it's harder to hit it by accident on mobile :3

[unborn token]

V3 desperately need Intro support. very very hard to get it to add an intro in certain genres, verging impossible. in rock and metal it adds an intro often (with prompt), but over 1000 credits on pop and not one single intro. not even one.

[sweet hawk]

Please add [angry] and [excited] emotion in Bark AI

[nova wolf]

What's the possibility of you making something that works like EmvoiceApp or Vocaloid? Being able to plot notes, write lyrics, and then generate with your seemingly endless supply of voices would be epic.

[fiery linden]

I agree the number of songs they claim you could get should definitely be reduced. For me, if things go smoothly, it takes 200 credits a song. But often it takes me 500 credits to get a good song because I'm very picky with the voice.

[dawn idol]

I know this is asked a lot here, but please add more payment methods. I live in Germany and people tend to use PayPal instead of credit card. I'd really like to invest a 10ner a month with you, but since you unfortunately only offer a single payment method, which I and many other people don't have, it's not possible for us.

[winter atlas]

I'd really like to see possibility to download the song as different tracks as well, to be used in post-production or using just the vocals etc. Like Output co-producer is doing with the loops.

[ebon veldt]

do you have some OpenAPI specification, so that I can use it in e.g. my GPT asistent?

[warm atlas]

can you add the option to import midi file , that suno turns into a instrumental track

cc @formal bear @obtuse tree @team

[urban granite]

i hope in future we can download just vocals from the songs we make? separately instrumental and vocal? 🙂

[steady scarab]

I understand your point, but let's keep in mind that it's still in beta, so aspects like cost, profit, market share, etc., are still being refined. In my opinion, if a high-quality song costs between $8 to $10 in Suno, it could be a game-changer for the industry! This would be after implementing all professional requests, of course. However, for consumer generation, the Pro package seems to be the most efficient way to create content affordably.

[rancid bronze]

im ok with the price, It´s just the false advertisment parts that is to much off for my taste. Sure generating only 50 songs fpr 10$ doesn´t sound so promising, but it would still be a ok deal.

[grim geyser]

for us music literate people would be cool to be able to enter in our own chord progressions using chord notation.

[primal olive]

Suggestion: an option to officially "end" a song

We don't want song to get cut off, we want smooth ending when we want it

[grim geyser]

I was thinking that earlier

[worldly falcon]

A way to flag single clip songs as "full song".
As of now the AI improved so much that at times the song will finish itself beautifully without need to string more clips together, and I'd like of a way to flag the clip as "complete" so that I don't mistake it for a random generate clip

[grim geyser]

also could be nice to have other markers that mark the chorus and verses

[ember mountain]

Also a good Idea might be If you have a nice Song If you can make this Song into a live Version of it with Fans chanting and singing

[grim geyser]

or to reuse the seed you used for your track generation, so you can try to keep the same beats. if we could re-use the seeds, we could also set the time stamp to start generating from so that we can potentially finish a song that cut off in mid sentence.

[grim geyser]

Riffusion has a feature you can use a microphone to put time stamps on every word.

[pure night]

Opa minha sugestão é o seguinte que tivesse um opção de encerrar a musica, "tipo um botão", e também se tivesse como encrementar uns ritmos com funk brasileiro, Arrocha Brasileiro, Pagodão, axé e muitos outros, e também se tivesse um previa de cadas ritmos que iriamos escolher, assim ja teriamos uma base como sairia o ritmo escolhido, sei que ainda esta em bet, mas pra mim isso é oq esta faltando no programa.

[wanton leaf]

Hey, will be great if you can add Describe option (to drop any audio so AI will analyze it and give a prompt to make similar song)

[eternal hazel]

Definitely hearing more of a sizzle / XM RADIO style sound to all the mixes in V3 vs V3 Alpha. The Alpha had it too, but V3 feels almost like it's stuck on as an effect that never goes away where as alpha had it in more towards end of song or in alternate generation (one generation would have it, the other one wouldn't)

[unborn token]

A long shot, but give an option to gift credits to people

[rancid marsh]

Hi I am new here. It’s impressing. But as Sounddesigner I wish for a paid account that as result I do not get just mp3. I preference wave files. Additional for paid is it possible to get separate midi channels, as well as separate audio channels per instrument and voice.

[harsh storm]

My siggestion is to have varying length of the continutation of a song, so it can determined when the lyrics will fit. A lot of credits are lost because it jst cuts off mid sentence. And it doesn't continue the sentence in the next part. Or add an option to continue it in the next part.

[eternal hazel]

just making a guess here (cuz i'm also an audio engineer / sound designer) but I'm betting the developers would not even be able to provide midi based on how the model works. it's literally constructing the song at the wave-sample level, so the AI doesn't really know much about the actual music it's creating. It's just morphing audio based on what it was trained on.

[delicate cedar]

I've been advocating for something like this for months now and while I've been told it's a good idea I too am still waiting for it.

I think it's a great idea. A wonderful way to give thanks to someone whose work you like or to encourage people to take chances, since it's a free generation. It'd be a way to get more people to do experimental stuff.

[tired frigate]

Will they add a feature to train the model on specific songs?

[stiff wagon]

Features Suggestion:

1. Allowing adjust melody and note and re-generate , or upload midi or voice melody for generation as base
2. Search for certain Keyword and creator's song, Top view and like leaderboard Daily, weekly, monthly
3. Comment Section for songs
4. Profile Check for song's creator

[unborn token]

I agree. I have credits left a day before my sub renew. and I see people using their free daily and still active in chat, why not gift them some

[urban granite]

and bpm of song thanks

[nocturne locust]

I'm struggling with the content moderation. I can't for the life of me figure out what is triggering it.

[grim geyser]

@nocturne locust you have to make sure none of your prompts contain band names.

[hybrid crane]

paste it in chatgpt 3.5, say this text is flagged for moderation. What could hvae triggered it.

[nocturne locust]

thats a good one

let me check

[nocturne locust]

it doesnt lol which is so frustrating, its some lyrics a friend wrote as a joke two years ago

[hybrid crane]

paste it here.. maybe we can help you

[nocturne locust]

oh god, I see

it seems so mild

[tepid trench]

As @eternal hazel suggested, the audio is generated as one whole thing, so it's not currently possible to separate the different channels/parts and I doubt MIDI is involved. Also it's previously been said that wave files would not offer any advantages over mp3 as you would not get better sound quality etc. but just a larger file size.

[hybrid crane]

yeah.. but chatgpt is here the moderator on your lyrics.. and they can be overzealous

[nocturne locust]

yea I figured what it is lol. it won't allow even the mildest forms of offensive language, refers to violence ect. it's gotta be very clean

thanks for the chatgpt suggestion

[hybrid crane]

@nocturne locust tip: something you can trick it by spelling it different..

[grim geyser]

weird, I'm not having any issues, does it have any lyrical cliches that are used a lot?

[grim geyser]

yeah like phonetic spelling lol that fixes lyrical issues I get where it doesn't pronounce it the way I want.

[hybrid crane]

https://tenor.com/view/tkt-smart-gif-20642718

[half sail]

Dear Suno team - could we get the "Continue From" field to feature reaction to mousewheel and feature an up and down arrow to adjust the time by a second? thank you 🙂

[sharp zodiac]

please add purchase via crypto we do not have access to international payment systems.

Hello everyone!

I've been using Suno AI for generating songs from my lyrics, and while I'm amazed by its capabilities, I've encountered a particular challenge I hope to find a solution for, or perhaps suggest an enhancement if it's not already available.

My main issue lies in achieving a consistent rhythm that matches my vision for the music. While Suno AI allows for specifying the style of the music, the rhythm often comes out quite random. I understand that the results are inherently stochastic, but I'm looking for a way to guide the rhythm more precisely. Ideally, I'd love to be able to provide a sample song to the AI to "learn" the rhythm from it and then generate new music based on this rhythm, albeit potentially in a different style or voice.

I've been experimenting with different styles and attempting numerous generations to get closer to what I'm envisioning, but it's quite a hit-and-miss process. Is there a feature, workaround, or perhaps a different tool that allows for more direct control or influence over the rhythm based on a sample song? Or, if not, is this something that might be considered for future updates? The ability to train the model, even if in a limited capacity, on a specific song's rhythm would be a groundbreaking feature that I believe many users would find incredibly valuable.

Thank you for any guidance, tips, or insights you can provide on this matter. Your work has already inspired so much creativity, and I'm excited about the potential to push these boundaries even further!

[prime widget]

Ok I don’t know if any suno peeps actually see this or where to better submit wishes for upgrades, but I am going to just plop them here when I think of them. I would love more control in the custom area, specifically thinking about profiles for performers. When I create a great song, I would love to be able to isolate the performers and save them, tweak them and use them in other songs, etc. it could be amazing actually.

[rancid marsh]

@eternal hazel @tepid trench Thanks for reply. I am just a hobby user. And it’s not a major problem to separate divergent voices the AI put together into wave files for channels. From a single separate wave file to produce a midi than inside yours DAW is not a deep knowledge adventure. Why not putting out STems? This is my old outdated DAW a Fairlight EVO and only used for Hobby purpose. As I said I did tested but for good creative results I hope if not midi for separate stem output.

[sage mirage]

Suno doesn't 'do' stems. The file that Suno outputs is all it outputs, a complete master. Welcome to AI.
If Suno were to output stems it would be a separate process done outside of Suno itself, mirroring exactly what you can do with the finished Suno master right now.

A lot of music people have shown up lately and how they do music isn't how Suno works. Suno only does the final master, one file. You have to work backwards from that.

[native stag]

can we have an option for the generations.. i hate creating two tunes with each track.. can we have the option of 1 2 3 or 4 as i hate burning up credits.. with tunes that did not complete.. and can we also have a full list on the webpage for all the [ commands ] i love the v3 alpha but hate making 2 tunes per track when i do not need them.. but mostly would love a command list on the webpage.. i am still getting used to the software.. but the [ commands ] on main web page would help users a lot..

[rancid marsh]

That Suno does it not do now is not saying something about future… here is a suggestion discussion board.

[tepid trench]

I don't disagree with the suggestion. I would love to have separate stems and MIDI to work with. This was what one of the staff commented recently about separate stems.

[hybrid crane]

It's by popular demand. They are aware. Look for stems in search.. you get HUNDERDS of requests. But it's just not that easy. They want to deliver, just have to figure out HOW. For that the quality needs to be better first as @tepid trench showed you in an original post from the main Dev of the models.

[sage mirage]

It's just feature that's fundamentally against Suno in a way that makes it where the solutions now are as good as Suno is going to be able to give you later. Unless they fundamentally change how Suno works at the base level.

They most certainly can do it, but like tony said, that isn't good enough for Suno to do atm.

Because it's probably a very large change to Suno's operation

[sage mirage]

For this one, the devs have specifically said that the cost, on their end, is for the GPU power to render Suno's output. It's having the GPU at all, and two songs is how the efficiency is laid out. If they were to allow 1 generation rather than the 2, it still costs Suno the same amount of $ to render the one generation, they don't save money this way. It would be more expensive, for Suno itself, to give you a single song rather than the two because they would be 'wasting' that GPU on a non-generation.

Two at a time is the most efficient cost-wise, and there is no savings to be passed on to you if they allowed one. You'd just be getting one less output.

If you want 4 though, just hit "Generate" twice

It's weird though, if you aren't used to how most generative AI systems work

[native stag]

still a nice list of the commands on the main webpage would be nice. for the [] commands.. or is there one already there that i just cant find

[lofty sonnet]

try LLM like chatgpt

[sage mirage]

https://www.suno.wiki/

[native stag]

ta

[sage mirage]

I'm going to step behind "wow LLMs are weird" excuse again, but commands like this aren't 100%

Suno is an opinionated slot machine

It's why you get two, the single generation might not be what you are after

The 'path' to a completed song you like is a tree you generate from, picking generations and moving forward. Every complete song I've generated has a long tail of unused clips behind it.

I'm more on the particular perfectionist side though, and I spend the credits for it. That's ok though!

[prime widget]

I have a million ideas using V3 and I know this channel is clogged but listen up: this thing could be like photoshop for music writing and I hope you guys are working toward the goal of not just gen ai creation, but the ability to use and create scores and charts and edit them, alongside profiles of performers, instruments etc. I basically want an ai symphony I can conduct with precision. It is a tall ask but not so tall in 2024.

[lofty sonnet]

combine with ableton in a workflow, it will still take you far with midi extraction and sampling capabilites, but yeah in-painting will be huge

[native stag]

thanks for the info.. i am loving the site.. mostly i use the webpage.. and thank you for a fast responce to my questions..

[eternal hazel]

Yes, but suggestions need to be within the scope of the context of what the tool is doing. The way it works is simply by mimicking data. It has no clue that there's a concept of "a band" or "various instruments", or that any of those things are "playing notes". All it knows is how to do a lot of math to create a single wav file based on some inputs.

For suggestions like "let's have it make stems" to be a worthwhile discussion, the technology would need to literally operate in a different way. Aka, they'd have to start all over from the very beginning with that as a consideration from the start.

Lastly, not here to deflate the excitement or enthusiasm here - just trying to help you understand that as smart as it seems, it's very dumb. I mean that in a really literal sense, not in a critical sense.

[sage mirage]

You can get stems now though, with Suno outputs. You just have to do it outside of Suno

Like a lot of things, really. Like snipping dead air

[eternal hazel]

yes, but those aren't real, those are also just approximations of separation. If Suno offered the same thing, users would believe they're getting real stems, not approximiations. It would be a faux feature.

[sage mirage]

Right

And tony has said time and time again they aren't in the business of 'faking' it

Same with the .wav request

The data isn't there, if they offered .wav output you'd just get a fat file with no more data in it

i.e. a 'fake' wav

[eternal hazel]

Yes, congruent with their current offering, since the technology is inherently limited to the compression limit they started with, they give it to you at that limit. Wav would just be falsely inflated.

[sage mirage]

Which shifts allll the way back to 'wow LLMs sure are weird'

[eternal hazel]

what I'm curious about is, was the training data also limited in scope to this compression limit.

[sage mirage]

That I haven't seen them mention or talk about I don't think

If the compression was a in put thing, or something that they do when they shove it in the model itself

[hybrid crane]

It can almost play any genre, and mood, any language.. that model HUGE.. but no information about how, what, where is public

[tepid trench]

To get proper stems, they'd somehow need to train the model from isolated instrument recordings, i'd expect. And possibly also train it on stems that have been split from other music (which wouldn't be "clean" sounding enough to be useful, probably). Then they'd need to generate the individual stems per-song, somehow ensuring it all works together, then mix, master etc. Probably a LOT more GPU intensive for the same end result.

[eternal hazel]

This is probably a stupid theory, but it almost sounds like they trained it by listening to pandora. It has a "Streaming Radio" sound.

[sage mirage]

If you needed a ton of audio data

That's where my head would go

youtube, pandora, spotify

I suspect we'll never actually know because that's not the sort of thing you can talk about without getting mobbed

[eternal hazel]

I'd say my favorite thing about the current product is that it's highly inventive while still maintaining a human-level element to the structure. Other AI models just do really whacky stuff that doesn't make any sense. Suno makes creative but constructive music that would actually pass to a music fan.

so for that reason, it is literally one of the craziest song writing tools ever invented. pop out your song idea and go record it for real

[tepid trench]

YouTube would explain the hallucinated "thanks for watching" that some people have had

[sage mirage]

Yeah

[native stag]

i love it when it hullucinates or gets drunk now and then it can come up with some rather unpredicatable but enjoyable stuff

[sage mirage]

Also my favorite part. Suno has an opinion

That's the seasoning it has

[native stag]

also one thing i have notes' when you use [fade and end etc or anything to say stop.. the alst picture it generates is always a bird. wish we could change that.. or an option to redo the artwork 🙂

[sage mirage]

That's a legacy thing, since Suno's outputs used to be 'chirps'

So, birds

[native stag]

hehe sound 🙂

[sage mirage]

If it has nothing else to pull data from, it will usually default to birds

[native stag]

shame no command like [image nice logo with the end] to have it gen the last image

[sage mirage]

george mentioning stems, for those of you asking about it.
#💬┃general-chat message

in general

[native stag]

but looking forward to the day when suno gets to the stage when it can also animate the images.. and make videos.. i can see it happening in time.. not this year but it will happen one day.. 🙂

[sick marsh]

Comments section for songs? Or at least maybe profiles...

[sage mirage]

A moderation nightmare

tony has said maybe a specific list of emojis

To avoid eggplant spam

Specifcially on comments

[sick marsh]

like we aren't already? what's a few more places to be. haha

but yeah, maybe like github has for emoji

or i guess facebook? like someone knows what facebook even is.

[sage mirage]

facebook was mentioned, yeah

[sick marsh]

another thing unrelated to this entirely: this seems like the most serious #channel, which is not very serious, but..... should I be reporting songs that...... very explicitly are imitating artists & the lyrics claim to be them?

[tardy sparrow]

Idk if there's a proper template for feedbacks but i feel like there's still way too much reverb and echoing especially in instruments like cymbals, claps and similar, often vocals as well, even tho i tried with prompts to get rid of all the reverb and echo I couldnt

[nova mica]

I'm pretty sure this is the most suggested feature already, but please we need more payment options! Why are you guys locking yourselves out of so many potential costumers? Credit cards just aren't as common around the world. I am from Germany, I'm an adult, and so is everyone I know, and I know of no one from here that has one.

It's sad not being able to support and partake in something that I like. :(

[hybrid crane]

Hey shisa-ni, suno understands your point about credit cards not being common in Germany. Suno is actually looking into more payment options to reach users worldwide.

[sick marsh]

I got a friend in Germany who uses revolut? is the issue credit as opposed to debit??

you could load a rev card in like 45 minutes

lol

[shy relic]

there should be a promotion channel for people that are actually getting music distributed/getting popular on yt

[hybrid crane]

#🐣┃suno-showcase is used for songs created here and also youtub videos @shy relic

[shy relic]

yeah but there's a ton of noise in that channel, im talking about a showcase for things that are getting popular or more mainstream

ex. post a 50k view suno video in there, a lot less traffic but higher quality content that people might actually look at

[hybrid crane]

Too many channel too.. so yeah... noise belongs to a discord servers. People love to talk 🙂

[shy relic]

not saying to replace the higher volume channel, but if i want to find genuinely popular suno content itd be cool if there was a cool way for me to do it on discord

i agree that people like talking a lot though!

[worldly falcon]

A way to lock you songs so that only you can continue from them or remix them would be cool imho

[fringe orbit]

Hi guys! First of all. A great Big thanks for all the fun you give to me. That's incredible! One thing cool could be to filter library with a key string as we work on multiple songs. Another one coud be to specify the type of instruments for the instrumental option. Just drum, just voice, drum and basse,etc...

[sterile yoke]

Heya guys, awesome work, V3 is fire!

Just an idea: it would be nice to be able to invite friends, and to earn some credits for it! It could be something simple, like 50 or 100 credits for both, maybe limited to one day so people don't exploit it. It would encourage bringing in more people :)

Other suggestion is to implement a search function to the website, so it's easier to find genres like rock, pop, eletronic and etc. There are plenty of good songs out there

[north dune]

Can we do negative prompts?

[hybrid crane]

no.. not atm

[north dune]

Ok thanks

[sacred shard]

Probably a common request, but it would be great if it was possible to set the random seed so that it becomes possible to tweak the lyrics whilst it keeps the sound. Even better would be separate seeds for the melody and voice parts (but maybe that is not even possible the way the model works)

[north dune]

I just mentioned this to a friend the other day. I think that's a great idea.

[sacred shard]

It can be quite frustrating when it gets the melody just right but then it gets the pacing of the lyrics wrong because of a syllable too many in a word.

[north dune]

I was able to change a lyric using the continue function. After a couple more generations it changed the lyric and kept the sound, even repeated the chorus.

[sacred shard]

Ah so you set continue from 00:00? I have to try that, thanks!

[hybrid crane]

In Suno, each song generation is unique due to the inherent randomness in the AI model, as Suno doesn't set a specific seed for generation. This means that tweaking lyrics while keeping the same sound, or having separate seeds for melody and voice, isn't possible in the current framework. The AI generates music with a degree of unpredictability and creativity, which makes each creation distinct​​.

However, to tweak things.. Suno team is working on "paint-in" to change certain parts.. opefully we will have it V4

[north dune]

Yeah, right at the spot you want to change, and then I put the word that I wanted in the lyrics box

[sacred shard]

That's interesting to hear - I assumed that underneath works a diffusion model which to my knowledge can always be seeded and be made repeatable.

[hybrid crane]

yeah a big misconception. 🙂

[north dune]

Is there a resource I can use to learn more about how this song generation works? I've generated some stuff that sounds almost too real and curious how that works, it's like magic.

[hybrid crane]

Nope.. I assume to prevent people who understand MML too to recreate...

[north dune]

Alright thanks

[frigid timber]

I am glad there are still newcomers like you that show up and realize how amazing this technology is, regardless of the certain buggy issues it has, it does sometimes feel like "magic" lol

[north dune]

Yeah, I love music, always have. This makes the whole process much easier, I was in awe for the whole week haha.

[frigid timber]

I still find myself in awe sometimes lol, especially when I find new sounds people discover and how well they can be mixed into other sounds you would never think to put together

[feral aspen]

Not from literally 00:00 but from any time before whatever lyrics you want to tweak started can work. The shorter the clip that more likely the new gen won't resemble the new one. (Which in some cases is desirable, you may want a new melody but in the same style or voice, for example.)

[north dune]

yeah, I like using mixed words like jazz rap, or classical dubstep. It's very interesting.

[gaunt wind]

barbershop grime

[north dune]

whoa

[hybrid crane]

remove your link.. no self promoting

[high smelt]

womp womp

[hybrid crane]

@fringe knoll can you please assist me?

[lunar bolt]

it would be great to have access to the mp3 and wave as well as the track by track with the voices without effect and the unmixed prods to be able to mix it ourselves.

[hybrid crane]

You are a day in suspend for self

[vocal portal]

FIX YOUR DAMN CENSORSHIP!
Sorry for the shouting but its getting beyond a joke.
I had my local LLM make a quick song about Cannibals to test if suno would be ok with that....it was so I created a more detailed one, based on the Idea of Kelvin being the evil one that suggests we eat the cannibals (This is for the game Sons of the Forest, sequel to The Forest, i should point that out... kelvin is a death mute that follows your written orders, so the idea its his idea is an amusing one...)

I bounced back and fourth with the ai came up with a good one.

Suno says "oh no you cant generate that, think of the children!"

I added references all over "video game" "THIS IS A GAME" but nope its not having it.

I am getting so very annoyed with this, there is nothing truely offensive in the lyrics!

(Verse 1: THIS IS A GAME!)
Upon a windswept sky, my copter did climb,
With Kelvin by side, no words passed betwixt us fine;
Crash landed we did on an island so vile,
Infested with cannibals, sportin' toothy smiles.

(Chorus)
Kelvin whispers, "Feast!" yet voice stays confined,
Torn am I, caught twixt his thoughts and mine own mind;
Shall we dine on these fiends, or seek peaceable ways?
This silent debate, nightly beneath moonlit haze.

(Verse 2: Video Game Song))
They charged us, those ghouls, under star's gentle light,
Clubbed high, blades sharp, prepared for a fight;
But hunger ignited within us a flame,
Our survival instinct, they couldn't tame.

(Chorus)
Kelvin gestures wildly, eyes gleaming with glee,
I waver, unsure if his plan's right for me;
Eating these savages, revenge served piping hot,
Or shall kindness prevail, ending this bloody plot?

(Bridge: Video Game Song))
Fog horn blew low, o'er the island vast,
An eerie lull before tempest unleashed at last;
Cannibals vanquished, feast layed bare,
Only then did I notice poor Kelvin's vacant stare.

(Chorus)
Silent Kelvin, hands moving swift,
For Kelvin is mute too, tho' deafness takes most drift;
Ok I admit it, It was my idea,
Not that you will tell anyone, that is clear!

@feral aspen sorry to ping you, but can you add the above lyrics to the report you sent to the higher ups over the censorship issues. Its looking like my tokens will expire before its fixed but maube it will help

[delicate cedar]

(╯°□°)╯︵ ┻━┻

You're absolutely right...

This is horrible.

[vocal portal]

it is how i feel i must admit.... not anyones fault... (well ok who ever decided to have the AI do the checks for decency should feel ashamed of themselves)
Sorry for the shouting.

[delicate cedar]

Unfortunately the realities are what they are. We moderate all prompts through ChatGPT because it would be extrememly cost prohibitive to listen to every submitted song ever made. And ChatGPT is a prude who has been engineered to be extremely "safe" for the kids.

Also...consider how long it would be between prompts if someone listed to every part...

But you're absolutely right.

[vocal portal]

To be fair, the censor has refused songs made WITH chatGPT so i think theres more going wrong

[delicate cedar]

It's just....this is as good as it can be right now.

But please believe me when I say the Team is definitely trying to tweak things towards as much freedom as possible. And to use the least amount of moderation possible. So please, please keep submitting these to the reports that Jonathan Fly makes. There are efforts to make this better.

[vocal portal]

I would say that only reported songs should be checked, currently Sunk is punishing all users for the sins of a few. (or even only checked if "public" is clicked) thise that abuse the system should be IP banned

[delicate cedar]

But no, you're not crazy. You're not over reacting. This is atrocious.

I had ChatGPT kill a song once when it refused to allow me to continue after my lyrics called someone stupid. (The protagonist was calling the love interest who failed to see them as a person stupid and ChatGPT lost its mind.)

[vocal portal]

I have cancelled my sub, my tokens expire on the 24th (or 23rd if my renewal was to be the 24th) at this rate I wont be able to use them as i mustly make funny songs for my mates about games.

[lunar bolt]

it would be great to have access to the mp3 and wave as well as the track by track with the voices without effect and the unmixed prods to be able to mix it ourselves.

[vocal portal]

Yeah I spent 7 or 8 generations to get the ideal first part of a song, only for Suno to go...Nope you cant have the second part.

[delicate cedar]

Mind you my example was quite a while back...it wouldn't necessarily do that today. We're talking v1 Chirp here.

[vocal portal]

for me it was a couple of days ago.

[delicate cedar]

All I can do is suggest that you keep checking back in and keep trying to be creative as you seek to get "crap past the radar" with your lyrics.

[vocal portal]

I loved suno before the censorship

[vocal portal]

Hey! my songs arnt crap :p (i know what you meant : )

[delicate cedar]

I knew you probably would catch the reference. But for those who don't know...

WARNING: TV Tropes link!
||https://tvtropes.org/pmwiki/pmwiki.php/Main/GettingCrapPastTheRadar||

[vocal portal]

Gonna let my sub expire go back to being a free user, once the censorship is fixed (if it is) I will re-sub. Voting with my wallet, but also the fact I cant use the service tbh

[delicate cedar]

I know we'll be sad to see you go. Having people trying new things is in part what makes things better because when stuff like this happens and people report it then it can be worked with.

[vocal portal]

And i really dont think anything in those lyrics are that bad or even close to bad. Touchy subject I guess, but we are talking about a game.

[delicate cedar]

That's the problem with trying to "make AI safe" it results in the AI being unable to understand the difference between reality and games because the people who insist on the baby-proofing only see the world through a Hollywood filter where Skynet and Hal 9000 are just around the corner.

[vocal portal]

Reported it a fair few times, even emailed suno with chatgpt logs from where it had lyrics refused. No reply there. Jonathan Fly did take a lot of effort to help and we sort of found a way around the issue... that then stopped working almost immediately, not his fault ofcourse.
At the end of the day I shouldnt be spending all my time editing the lyrics down and changing it to get it past the censor (I would understand if I was using the vulgarities we get on the radio these days)

[vocal portal]

The media needs to stop comparing LLMs to skynet and fear mongering, in no way is LLM the same level of AI as that... not even close... I would argue that its not even AI. A fun tool yes but Artificial Intelegence.... I had to add the l to a few of the Kelvins on the song it did generate as it got confused lol

[delicate cedar]

https://tenor.com/view/preaching-to-the-choir-preach-it-sister-happy-gif-14328684

[vocal portal]

Right I am off for a coffee. Please if you do find the team member that decided current AI was up to the job of moderating our songs, hit him (gently) with a rolled up newspaper and tell him "NO BAD DEVELOPER". It needs to be done or he/she will never learn

[delicate cedar]

🤣 ...I'll bear that in mind...

[tepid trench]

It would be nice. Currently the audio is generated in its complete form from the start so can't really be separated cleanly. Also the quality the songs are generated at would not be any better in wav format. This is what I've read from previous discussions about this here.

[vocal portal]

@delicate cedar Ill just leave this here (hey I might as well use the credits before they expire)
There is an Irony though, as it also has a bug... the lyrics did not fit 2 minutes, so it repeated the end chorus. So here is a #💡┃suggestions-web We could do with a key word we can add to the lyrics that when the Ai sees it , it knows thats the end for the lyrics (a stop sequence)
https://app.suno.ai/song/cb52e633-d5ab-4ed5-aab6-c6927ff30396/

^that song is meant a little tongue in cheek, I still hope it makes people smile. even if the issue is a major one causing me a lot of stress. Ok now 100% off to hunt coffee

[lunar bolt]

I understand, but yes, it would be great if it could generate multiple formats - track by track - and let us choose whether we want the final version or just the individual tracks so we can mix them ourselves if the mix doesn't fit

If I generate an instrumental via LIA with the subscription, can I sell them on BeatStars, etc., or not at all?

[grim geyser]

another thing I thought of is a special symbol that the AI interperts as a vocal rest.

[late shard]

im sure its been mentioned but, if we could get access to the split stems that would be amazing

[sage mirage]

Suno doesn't use Stems, your best bet is to split them yourself with an outside software

[bitter tusk]

#💡┃suggestions-web Hello, I have a few tips for improvement:

1. If you generate a song with lyrics in 2 minutes and just want to improve the lyrics and match the voice and sound, it would be cool if there was a function like Part2 Text, everything else stays the same, the sound and voice would be great. If that was added it would be great for everyone too because with new part it's gone.

It would be cool to also download wave files etc. in different files. For a video format without text, just a cool 2 minute animated text or a cool AI film video music cover etc. would be cool if this function came in, please include this in your noitz. Thank you

[keen sable]

I am sure that you know what you are talking about but still that sounds strange. Maybe it is because I am so used to my own workflow and assume that suno is working the same way. Still it would be of great use to get the stems, dry stems without any effects 🙂

[blissful kindle]

You guys need a music app for phones so that users and non users can play and stream the music like spotify and etc. Create stations and have things sorted by the ai by their genre combinations, let the rating system on the site and the apps dictate top plays. Charge a nominal fee initially to support it's development and growth and increase a little later for profitability.

This creates your own eco system.

You could allow free users and support it with ads.

Paying Creators should probably have streaming included.

Creators could pay a premium for Spotlights which would advertise them to free users and on the site maybe?

[sage mirage]

Suno is an LLM and outputs at the master. It doesn't built it by instruments, it outputs at the full master based on its training

It IS strange

And very much unlike how it's done normally

There are no stems, it doesn't use them

The way to get stems is, as been mentioned before, using outside software to break a Suno output

If Suno gives the ability for stems in the future, this is how it will be done. Broken apart from a completed master

Which you can do right now outside of Suno

Same with people asking for .wav

The data isn't there. If you could download a .WAV from Suno you'd get a big file filled with nothing

It's the training that is the limiting factor, not file format

V3 can't output any more than it can output, it would sound the same as a .wav

They could do it but it would be more like a placebo

And Suno has said they aren't in the business of selling fake stuff like that.

No API yet, no stems yet, no .WAV yet. The model can't give you those things that means anything.

Well, the API is more of an infrastructure thing, but still, not yet.

That said you should still ask for it, they read this suggestion channel daily

Just understand why these things aren't available yet

[formal bear]

I can't answer these questions better than you 🫡

[rancid marsh]

Fully agree. He is not someone that is doing compositions. I work regular with 700 channel templates as hobby user and the compositions and arrangements are a reasonable start for a faster workflow but soundquality i preference libaries of 400 euros just for one Violine. For a 2 minutes song I sit sometimes 6 months just for an arrangement. So there is no problem with paid service but there should be a different level. I do not talk about cheap Spotify trash sound quality. In fact… I preference my own libaries. And the tests are very positive but equally somehow sounds very often muddy .

[sage mirage]

Stare at the LLM void and see how people interact with it and you get used to how normal people approach strange, unknowable things like LLMs

It's neat seeing how music people approach it though

Thinking it works classically

[blissful kindle]

Every team has had different approaches and it's been pretty cool to work with the different systems over the years.

[sage mirage]

Suno's strength is in it being weird

[spiral vessel]

Can suno generate lyric videos?

[sage mirage]

The missing lyrics from the videos is a temp thing afaik

In order to get v3 out faster

[formal bear]

I answered this last night

someone guilt tricked me into working on it

so yeah

[sage mirage]

lol

[formal bear]

it will be ready sometime soon

soon I mean not this hour but also not in 10 days

[median kernel]

An idea: What about adding timestamp to instructions so you could fine-tune when something happens?
For an example: {1:20} [Music swells]

[sage mirage]

You can already sort of do that as long as you're cool with losing everything afterwards

For an in painting like thing, that's for later

[blissful kindle]

Riffusion (They have vocals as well now and are based off of stable diffusion) used Spectrograms, Harmon AI focused primarily on frequencies, MusicGen/Audiocraft allowed for input of original audio for continuation, Jukebox used a version of the wheel of fifths theory, It's been pretty wild.

[formal bear]

I like how you think about it 💯

[formal bear]

yep

audio is hard

[sage mirage]

So rewarding though

Out of all generative AI, it's audio that's it for me

[formal bear]

yeah who doesn't listen to music

[sage mirage]

It should be images considering my background but

[sage mirage]

I know some people. Well, "people"

[blissful kindle]

Definitely, this ended up being harder than video and large language lol

[median kernel]

That is kind of slower process.

[formal bear]

lol I am not totally sure about video, video is certainly interesting

[sage mirage]

The temporal aspect has to be the biggest issue, yeah? Video or audio

Different-but-same

[keen sable]

And the best composing buddy I have found on top of that. 🙂

[rancid marsh]

Rolling my eyes because of such expertise including my footnails sorry I post here inside a suggestion channel not to listen what is now… I post here because I hope that a developer read it. And to let them know what is necessary in my mind for a paid service.

[formal bear]

but his answer is right...

like I said, I will give the same answer

[sage mirage]

and the only reason I know it is because tony (and others) have said as much

I've just been around

[blissful kindle]

Video was solved a bit ago locally lol, they're rolling it out commercially now 😛

[sage mirage]

Just in case you weren't aware @rancid marsh tony here is Suno

I am not, I'm just some guy

More and more music people are coming in though and they tend to ask the same questions and request the same stuff is all

and that's ok!

[keen sable]

And you are some guy that is helpful. it is apreciated

[sage mirage]

Appreciate it, thanks.

[formal bear]

yeah but global...temporal, that's only sora so far.

need all the GPUs 🙂

[blissful kindle]

fair

[feral aspen]

It's totally cool to make feature requests, no worries, that's what this channel is for. Just providing context for why stems might look like a simple feature but actually aren't any easier than doing it for any audio in the wild. (But maybe in the future with research, who knows...)

[sage mirage]

I can never have enough vram

Nobody can

[sage mirage]

Doing it the classic LLM training way seems like it would be wild. You'd have to somehow train parts of a composition? Just on individual instruments? Is there even enough data for that?

i.e. stems in the classic way music is normally built

To get an LLM that would output stems natively

Maybe a midi thing that then would be given to a suno-like thing

idk I'm way beyond my knowledge

[formal bear]

but analogy to NLP, scale eventually solves everything

[rancid marsh]

He discuss a suggestion and I discuss not suggestions at least. I do just sounddesign as hobby. And doing folleys for film. This means just for footsteps I use to work with it with a basic set of 120000 wavefiles…. And if they do not match I create my own more. Or fieldrecordings of birds. My Libary are 10000 bird sounds just in a quality that I need just for the birds 8 TB of space. My husband did his medicine exams passed with 5 years programming AI. One server and I use 3 of them have each 1500 gb Ram and 96 gb of cpu. As I say I just make simple sounds.

[feral aspen]

That is interesting. I wonder if there's value in a 'footsteps only' audio model...

[formal bear]

we are def not outputing super resolution yet, as you can probably tell.

[rancid marsh]

This I am asking for. If I make from time to time music I just listen carefully. But I thought of payment than I like to see different things where I am myself can put own creativity inside. Here for example just a simple restoring. With Stems there is a base you can work out easy that the sound is matching with a video. If there is a song just it’s close to useless. https://youtu.be/C6D2eZ1XGqA?si=CNLrsvRKliSGUmM4

[formal bear]

do you just want instrumentals? you can click the instrumental button and generete without vocals, still not studio level sound quality though

[steady scarab]

Honestly, this experience reignited my enjoyment for the music.

[rancid marsh]

No the Charme of Suno is at least the multilinguale possibility. In my own Romanian language. This means I can sing a part myself and use a matching male part from Stems so it needs this separation.

[tepid trench]

That's the thing, if you have to train from individual instruments there's probably far less of that compared to finished music, especially if you consider how different parts work together. I'm pretty much a newbie when it comes to knowing how AI in general works but I'd imagine it might be possible to train something to identify and pick out certain instrument sounds, then use that as part of the training process to do stem splitting of the source material, before then combining these when generating output. It'd be a lot more heavy on computation overall, as there'd be more steps involved. That's how I envisage it anyway.

[rigid eagle]

Lately Suno has not really respected the lyrics I write, I understand it sometimes depends on the tags (some tags are more prone to respecting the lyrics than others), and I also get that it might be necessary for it to be more creative, buuut, it would be nice to have some slider to configure 'how much' we want it to respect the lyrics, if that's at all possible.

[amber gazelle]

A way to search people's songs via name or title

Ideally, using an app on your phone outside of Discord or the site would be nice

More than 5 songs daily (maybe 10-15?) for non-subscribers too?

A way to choose the song you just created to add vocals to it or remove it... after you listen to it, sometimes you like the music more than words, or you like it so much you want to add words now..so it'd be nice to choose after the songs were chosen/made

Ability to revitalize an older version of a song (renew or "enhance" a v2 song in a future version of the software

A way to comment on people's songs on the site

[hardy falcon]

what time does the credit refresh happen ?

[hybrid crane]

Free daily credits refresh at approximately 10pm EST, 3am UTC.

[hardy falcon]

oh so thats in 2 hours. thanks

[mint thorn]

opensource 😈

[sudden valve]

I would love an option to generate MUCH shorter snippets for less credits. As somebody who used Musenet to generate 2-3 seconds of a song at a time and LOVED it, I miss how amazing that kind of granular control over the composition can be.

[vocal portal]

Would be nive, would solve a lot of isues with the Censorship model. Alas its never going to happen, they are doing it as a business not a labour of love.

[mint thorn]

No i want opensource only for make inspiration of real song. and i can pay 100€ for a solo version only for me. i dont want copyroght or something like that... i just want make new wave and chirp can't.

One again... go hear :
Rounhaa 2b2t
Jima grotatou
Les Grands Enfants Wall-e

And you see fast... this type of song is impossible to create with chirp without a little inspiration... link or vocal note. but i know copyright gngngngn i speak with other guy and i know they never add this. and never go on opensource.... x)

i try like 2800 crd in one night to make a song similar at juice wrld IMPOSSIBLE and i lose all my crd for nothing 😄

[vocal portal]

I just want to be able to play around with it to make mates laugh, with out it constantly telling me that my lyrics are not allowed (wehn they are not even remotely offensive)

[mint thorn]

ah for the cencuse that on the other hand they still go through chatgpt lol

[vocal portal]

opensource would allow that, and one day an indie dev will likely be able to do something similar, until then we just need to wait.

[vocal portal]

Yeah, and it even refuses lyrics by chat gpt at times

[mint thorn]

ah yes absolutely xD

[vocal portal]

i really feel this current approach to censorship has killed it. (atleast for video game based songs) such a shame.

[mint thorn]

My dream for having the ULTIMATE Chirp is just the possibility of adding a voice note or a link for inspiration. And don't copy everything otherwise it's rubbish. this is my dream. because I can't pay a beatmaker, sound engineer etc.

I think the censorship can be fixed quickly

[amber crystal]

Hello, I don’t need this subscription anymore, please refund it! thanks, thanks! Sorry to trouble you!
Hello, I don’t need this subscript

[feral aspen]

For anything about Suno accounts or billing, you need to email [e-mail].

[bold basin]

Maybe in the future, have a chance of we record a music, and the AI use as a input to create inspired on our record .

[sudden valve]

would love to make continuations for our own uploaded music

[small raft]

bring back lyrics in videos? why so many steps backwards in the past couple weeks?

[sage mirage]

[small raft]

first basic mode is neutered and doesn't generate two unique outputs anymore (but still costs the same), now videos don't have lyrics anymore.

[sage mirage]

On the other hand, v3 is fantastic, lyrics are clearer than ever, the lyrics box respects instructions a lot better

Is your suggestion 'make it better' or what

The lyrics not being in the videos is a temp thing

[small raft]

my suggestion is to not remove functionality that already worked, which has now happened twice since I signed up.

[sage mirage]

You know why the 2 song thing has changed

We've already told you, because I remember it. You were the first one to ask about it

pat told you why

[shy relic]

(why, i wasnt there)

[sage mirage]

I was mistaken, it's wrapped around gauging user experience

[small raft]

There was no satisfactory answer to "why."

[shy relic]

makes sense

[sage mirage]

It only effects simple mode

[small raft]

I speculated it was cost reasons (one GPT call vs two), but I don't recall that being confirmed.

[broken zinc]

I'm having a hard time getting it to generate harmony vocals that aren't an entire choir. It seems like it's figured out how to use large vocal ensembles now, so it wants to put them in everything. Even a country song, where you might just want one or two additional singers. Is there a way to ease up on the large vocal ensembles for background harmonies?

[sage mirage]

Because now simple mode works like custom mode

In that you get two different generations with the same lyrics and style of music

[sage mirage]

The discussion kept going, sorry you missed it.

[shy relic]

i dont get why anyone would want to use simple mode anyway

[sage mirage]

Some people do, and that's fair

But it doesn't line up with how the vast majority of users use Suno

[small raft]

My problem with that change is it increases cost to the user to get more unique outputs, and wasn't even made clear. Was done without warning.

A material change in how things worked at signup.

[sage mirage]

and pat apologized about not making an announcement

[small raft]

Apology doesn't change that it still costs double the credits that it used to for the same result.

[sage mirage]

Sorry you feel that way, but a vast majority of the users of Suno have used Suno like this the whole time

Because they don't use simple mode, they use custom mode

[mystic bloom]

why i dont creat part 2 song ?

[sage mirage]

and now simple works exactly like custom

Perhaps they may revist simple mode, I don't know. But the change was deliberate, and while you feel it's a step backward, that is your point of view and not an objective fact.

[small raft]

I mean, nothing to do with how I feel. It used to cost X credits to get two completely unique songs. Now it costs 2X. That's basically a rug pull after they got the $. heh

[sage mirage]

As for the lyrics in the videos, that's a temporary thing that tony said he is working on right now.

[shy relic]

if part 1 like, ended already then part 2 will be silent, the clip will be short, and your credits will be refunded

[sage mirage]

You keep saying that, but I disagree.

[small raft]

How exactly can you disagree with a provable fact?

[shy relic]

i think adding the option to only make 1 should be added, if that's in the works than this becomes a moot point anyway

[sage mirage]

Because you claim it's a step backward

[shy relic]

its a step sideways imo

[sage mirage]

Right because it's now in-line with how custom mode works

[small raft]

Simple Mode != Custom Mode. They don't need to be "in line"

[mystic bloom]

i creat part 2 new lyrics.

[sage mirage]

They are if you are running a giant GPU-cycle machine

[small raft]

More relevant: They were not "in line" when I signed up and forked over $ for the service. Then all of a sudden simple mode takes twice as much $ per unique output.

[sage mirage]

Regardless, the lyrics in the videos will be coming back. It's temporary in order to get v3 out to everyone.

As for being happy with the service, that is up for you to decide. I, too, am a paying customer and I disagree with your 'step backward' comment. You don't speak for all users, you certainly don't speak for me.

[small raft]

ok, ignore the step backwards thing. that's irrelevant.

[sage mirage]

It sets the tone

You know that.

[small raft]

No.

On day 1, $X got 2 unique songs. Later, after having paid, the same functionality takes $2*X

[sage mirage]

Then don't renew? I don't know what you want me to say.

[small raft]

I mean, at a minimum I think this should be easy to acknowledge as reality of the situation.

[sage mirage]

The reality of the situation is how you use Suno is atypical vs. most of the users

[small raft]

Claiming nothing was changed for the worse is just gaslighting. lol. Doubling the cost of use is certainly a negative change.

[sage mirage]

There have been a few other people that also asked why it was changed, so you aren't alone

Uh huh.
The lyrics will make their return in the videos.

So don't you worry about it. 🙂

[small raft]

Already acknowledged that? Ok, cool. So what about the remaining issue?

[sage mirage]

What issue? The change about simple mode was decided by the Suno team.

And the lyrics will return

[small raft]

A couple of weeks ago, 10 credits would get me two completely different songs with different lyrics and even different styles. today, that same 10 credits gets me two songs with identical lyrics and style, and tends to produce nearly identical songs. So it now costs 20 credits to get the original functionality. How is that not effectively a bait and switch?

[sage mirage]

It wasn't a promised feature, unless you can pull up something somewhere that says otherwise.

bait and switch, gaslighting, whatever. Suno decided to change it, they did, that's that.

[orchid forge]

cry baby cry

[tepid trench]

I always found simple mode to give poor lyrics in all cases, so much neon and leaking of the genre/style prompt into the words. Personally I prefer to get 2 songs, similar style, custom lyrics, and then I can pick which one I think sounds better. If I want a different style, I generate a different style. Seems better to have 2 similar songs to choose from IMO.

[small raft]

So it's cool for a company where you prepay for the service to just all of a sudden without warning cut the value of your prepayment in half? lol