📚┃suno-school

[obsidian lintel]

tried that didnt work either i just re wrote the whole verce but thanks anyway

[versed valley]

Welcome! For help, check out Suno Help

[fallow gull]

I tried to use the lyrics generator to create a song called "Superb Owl" and it kept changing it to "Super Bowl"

[versed valley]

Welcome! For help, check out Suno Help

[fallow gull]

I'm wondering how much text I should put in my first generation... If I put in a verse and chorus, I get <1m song start, but it often cuts off the last word. Sometimes it goes off on a tangent and fills out 2m with a solo or riffing on the chorus.
Should I put in more text and see how far it can get, then use the continue from to replace what was cut off and fill in whatever is left?
I'm just wondering what approach people are using to try and get complete songs.

[eager topaz]

I am putting 2 verses and a chorus. then seeing how far it gets, then cut off the lyrics up to the last recorded word, add any more lyrics and continue. It works most times

[fallow gull]

That makes sense. I'm still thinking like v2 where I had to make sure the lyrics fit in ~1m. Getting as far as possible in the first generation would probably help the issue with the song starting to fall apart in continuation 3.

[eager topaz]

Yes, I think so. V2 I was only putting in Verse and Chorus. V3 is a lot better but still throws some strange things sometimes.

[rapid bluff]

Suno or the moderator GPT API being used looks at things in context, would need to see all the lyrics to get an idea of where it flagged, "kick their ass" was probably it but could have multiple things flagging as well depending what you had in there

[obsidian lintel]

i can post it here if it helps you fine tune it but i just re wrote the verse and it worked... do you wana see?

[rapid bluff]

sure! can put here or send DM if you dont want it public, cool with either

[obsidian lintel]

(Verse 1)
In the depths of space where the Thargoids roam,
We've prepared our ships, we've polished our chrome.
With nanite torpedoes, we'll make our stand,
It's time to show them the strength of our band.

(Chorus)
We'll send the Thargoids packing, with our nanite torpedoes,
Clog the Titans' heat vents, and deal the final blows.
No more hiding in fear, it's time to attack the pile,
We'll take on the challenge and kick there ass in style

[rapid bluff]

this flagged or the one that went through?

[obsidian lintel]

flagged

it didnt seem bad to me...

[broken lotus]

maybe attack the pile...?

it doesn't seem bad to me either...

[rapid bluff]

yeah and/or kick there ass as well

but yeah seems super innocent

i get away with much worse lol

[obsidian lintel]

feel free to look at the flagged versions i tryed i few variations all got flagged

[broken lotus]

yeah very vague violence...
like sports talk is worse

[obsidian lintel]

well either way i thought it was odd so brought it up. i worked arround it now so i am good but hopes this helps you fine tune it a bit

[rapid bluff]

honestly it could be a false positive too, tends to do it sometimes based on text placement

[obsidian lintel]

oh ok... well you have the lyrics so i guess you can reproduce it and see why it was un happy lol

[broken lotus]

did you finally get it through?
did you end up sacrificing any lines...

[obsidian lintel]

i had to change the whole last verse

[rapid bluff]

like removing the punctuation at the end may resolve it

[obsidian lintel]

i tried kik and arce and changing their to there and changing attack to destroy but nope

so i just re wrote the last verse and it was fine

wasnt as good though (not that it was great to start with lol)

[rapid bluff]

im messing with it

[obsidian lintel]

is it doing the same for you?

[broken lotus]

verse seems to pass..., but not chorus...
but I can't locate a problem word in the chorus, so it might be the overall context...?

or more than one..

[obsidian lintel]

do let me know if you find out why just out of curiosity

[rapid bluff]

its a handful of things

final blows

[obsidian lintel]

oh now i never would have guessed that

[rapid bluff]

me either lol im still going i got everything but the last 2 lines

[obsidian lintel]

on an unrelated note but while i am givving feed back i seem to get a lot more giberish with the web site than on the dischord....

[graceful condor]

?

[obsidian lintel]

but overall i love this app so much!

[graceful condor]

I use clipchamp

[rapid bluff]

web uses V2 or V3 if your paid, discord uses V1

V2 does pretty well but can hallucinate sometimes

[obsidian lintel]

yeh v2 is a great improvment thanks so much

[rapid bluff]

V3 might as well be on acid, lol its still alpha but when it works it sounds amazing

[obsidian lintel]

well i hope you sort it out cant wait to hear it.. the second i get a job this is my first sub

[rapid bluff]

im not staff, but they will, been here awhile and these guys get it done

[obsidian lintel]

its really amazing how good it is most of the time i had somone ask me the other day how i found so many songs on spotify that use my name lol

[graceful condor]

wait, what's ur spotify name?

[obsidian lintel]

i dont have it lol

[graceful condor]

?

[rapid bluff]

lmao its crazy right! even if we were still in V1 id still be just as mindblown as i am now

[graceful condor]

barefoot, what's ur spotify?

[rapid bluff]

JHOMEZ mine

all made with Suno

[obsidian lintel]

they thought the v2 stuff was music from spotify

[graceful condor]

oh

[eager topaz]

Just used the "Continue from" feature for the first time and it is fantastic 😀

[rapid bluff]

first time?!

like in V3 or period?

[eager topaz]

First time ever!

[obsidian lintel]

lol magic isnt it

[eager topaz]

It sure is. I was unsure about how to use it before, but thanks to such nice people on here I found how

[rapid bluff]

oh wow! make yourself some full songs

[eager topaz]

I have really needed it much before

[rapid bluff]

keep the chorus in each generation and it will or should maintain the same flow

[eager topaz]

Yes, it does, tks

[rapid bluff]

https://app.suno.ai/song/2d8f7fe5-c3e2-4eae-84e9-1cc20efaf5a7

thats as far as i could get, something with attack the pile is not having it at all

[obsidian lintel]

lol thanks man

i hope it helps you or them fine tune the filter anyway

[rapid bluff]

yeah me too, idk how much of it is controlled by them, its a chatgpt moderation i believe, which seems to act out alot lately, not just with Suno but everything

[obsidian lintel]

interestingly i just fed the lyrics to chat gpt and it flagged them too but still used them as a prompt lol

[rapid bluff]

maybe ask it why, see if it says or even knows

[thin flint]

Someone should add some V3 info to the wiki if any info has been found, new info that is.

[raw sluice]

has anyone else noticed that after the update the length of each generated section is way less? some of mine are getting only 1 verse finished. Solution?

[eager topaz]

I was about to post the same. In Instrumental a lot of clips are only 15 seconds!

[noble summit]

Is there a paper I can about this text to music model somewhere ?

[frozen jetty]

I keep getting a prompt is likely copyrighted error and it is not! I'm just pasting the lyrics from the chorus from the beginning of the song to this second generation and it keeps saying that my prompt is freeking copyrighted which it isn't. I'm using the output that the lyrics gave me and it worked the first time when generating the lyrics for the first chorus but it won't generate the second chorus. Any fix for this or is this just some stupid copyright detection bullshit?

[limpid spade]

Check the lyrics if it contains even a title of a popular song or artist and tweak the spelling if you find it.

[versed valley]

Welcome! For help, check out Suno Help

[magic gate]

/chirp I remember the day it all went down
[Jazzy upright bass and finger snapping]
Woke up in a haze, nobody around
Sun barely up, sky still gray
Just tryna make it through another day

[Jazzy piano melody]

The beat drops smooth like pills on my tongue
Pop one, then another for fun
Lighthearted keys, major key
As I vibe out chasing musical dreams

[Record scratch]

Hook:

[Scatted vocals over jazz piano]
Life flashes by, gotta hold on tight

One day you're high, next day say goodbye
Life flashes by, enjoy the ride
Yeah, life flashes by

Verse 1:

They call me Slim J, improvisational MC
Blessed with the illest flow, straight up lyrical genius
Mastered the art, from the heart I create
Jazzy hip-hop tracks to make you feel great

Rhymes like walking on air, without a care

Mind full of colorful thoughts I love to share
Pills couldn't compare to music's redemptive touch
Now I'm writing happy songs 'cause it means so much

[Jazzy guitar solo]

Hook

Verse 2:

Sometimes I hear the crowd cheer loud in my daydreams
Remember the good times, making joyful scenes
But I got caught up in career schemes, had to pull away
Re-centered myself so I could find my own way

Anxiety comes and goes, still I compose
Positive poems for any thoughts morose
Feel myself again, music to amend
Now I have balance with family and friends

[Scatted vocals interlude]

Hook

Outro:

Appreciate y'all vibing with me
I give it all in these tracks, now I feel carefree
Hope my music helps lift you up
Don't forget...life flashes by

[Jazzy piano outro]

[limpid spade]

Chirp won't work in this channel #🐣┃how-to-chirp

[pale cedar]

The song is about unrequited love, sad

[digital talon]

Something else that's really nice about v3 is you can get really specific about the music styles. For example, I started this instrumental with the style prompt "Harsh synth 3 note broken chord loop"
https://app.suno.ai/song/b98be3fa-9d60-4a9b-98b1-e4d00f0b1888
https://app.suno.ai/song/7d5d5cd3-6956-4536-a2ae-82b01a3d32e0
https://app.suno.ai/song/ad437d6b-6163-4b04-856e-a2b6e994e3e1
https://app.suno.ai/song/13887b60-1204-46bd-af39-68388524d48c
Then you can add specific music styles to it like so:
Electro-industrial https://app.suno.ai/song/cf30f841-3a50-43f1-878c-63afcdb4ddda
Industrial https://app.suno.ai/song/b2213d42-a3ae-4653-a241-e8951e9221be
Disco https://app.suno.ai/song/128890f7-bfd3-4920-96ce-c58c892950fb
There's varying levels of success with the prompt, but it can help you get more specific sounds, and combined with the ability to cut songs at specific points you can get pretty granular with your song design.

For example, with this song I'm playing with the first 10 seconds were "Dark Single note digital choir", then the next 10 were "Dark fast digital sequencer up and down arpeggio loop", "Fast industrial guitar loops lead", then "sampling, looping, breakbeat, fast tempo, drum machines, Synthesizers, sequencing". I get the feeling that the AI can only handle so many different musical tracks at a time, so I'm trying to figure out if there's a way to clear certain musical tracks to make way for new ones https://app.suno.ai/song/0656d407-25e4-43bd-9659-1eaaeb0060e0

[exotic fiber]

https://tenor.com/view/hold-up-wait-write-that-down-take-notes-taking-notes-gif-6092441834999206288

[tight frost]

So say I generated a song with Suno and I want to replace the singing with a cloned voice, what's the best app to do this ? (since it's a custom song)

[runic dock]

I give the complete song at the first time. I see that the length of the generated section is important. And as you can continue from any time it looks fine.

[sleek jay]

Thanks, but with the keywords I tried, I didn't get any noise like on a record. When I use 'lo-fi,' the bass sound comes in.

[limpid spade]

Do you mean like vinyl crackling sound when the needle goes on the record?

[sleek jay]

Yes!

https://www.youtube.com/watch?v=m5Mz9Tqs9CE

[limpid spade]

I've tried a few times in Suno without success. You could literally layer that YT clip sound over a Suno song in Audacity as a 2nd track though 😄

[sleek jay]

Got it. I'll give it a try! Thanks for the advice.

[exotic fiber]

theres a FREE VST for DAWS for that.. let me look for it, cant remember the name right now

this one https://www.izotope.com/en/products/vinyl.html

very very simple to use, a lot of personalization and works on any daw or software that supports VSTs

[sharp merlin]

hey I found a bug, not sure how to reproduce this but on mobile I was generating with the instrumental box unticked and it was still generating songs with the instrumental tag until I refreshed the page. I had it switched on at one point but then I wanted to continue a song with lyrics and couldn't figure out why it wasn't working until I looked at the song closely

[broken lotus]

I will try to reproduce and alert the devs... thank you! 🙏

[cobalt hornet]

So I just opened my web based suno account and all of my previous songs are gone. How do I get them back?

[neat nexus]

Wow...I remember that from the olden days on Winamp...

[digital talon]

I've been working to extend this song bit by bit, adding something new everything 15 seconds or so, so the first 90 seconds of this song is about 6 or 7 different part, each trying to add a new instrument or style. With a bit of effort you can really take control of the flow of the song https://app.suno.ai/song/72502540-dde2-4f5b-970f-f2175a0ecce6

[mild timber]

Does anyone else have the problem that with the V3 Alpha, Suno creates a song, usually the verse and chorus, then ENDS the song on an instrumental and then picks it up again in a new start after a few seconds of silence?

Might be my prompting, but it seems strange

[untold briar]

Sometimes, but remember you can choose a custom continue time before the song "ending" starts, which should preempt that in the next clip.

[mild timber]

Any idea if does a hard cutoff at the time marker or blends the two tracks together?

[untold briar]

It blends, sometimes very cleanly, sometimes not quite as clean. but it's better than puttig the audio together yourself from the .wav versions almost always

[mild timber]

Oh yeah I tried that with V2, the song was good but not "great" ^^

[runic dock]

I had this too. Even without adding à single [End] In the lyrics for this part. Suno ends normaly the song and after a few seconds of silence restart singing again.

[mild timber]

Man I wish we could split the vocals and instrument into individual tracks.. mastering them in a DAW would be so much more easy

[digital talon]

I've had the same problem, I've had to edit the audio before uploading it to Soundcloud, but not much can be done about it in the website.

[digital talon]

Did you try something like Moises.ai to split tracks?

[mild timber]

Not yet, usually splitting a rendered audio track with the help of AI just muddies the sound even more, especially if you already have sound artifacts in there from the start (which even V3 still does sometimes.)

[digital talon]

It's not perfect, but it has helped me fix some issues in the past

[mild timber]

I'll give it a go, thanks :) Also fun how sometimes, there are no lyrics generated at all, it's just an instrumental :P

[tight frost]

Hey guys I want to create an "epic rap battle" type of video from a suno song. First I need to replace the duet singers with 2 cloned voices. I think revocalize an vocify both can do that ? Then I'll need to animate and lip synch an avatar. I've tried several tools before but not for singing or custom voice. Can someone recommend a process ?

[clever sky]

I'm having an issue where I am generating freat stuff byut just BURNING through credits becuase most of them seem to be incomplete and just hard cut or dont finish the lyrics. I know V3 is alpha nut it's such a shame to get what I actually wanted then not be able to use it.

[broken lotus]

Have you tried continuing a clip...?
click the [...] button next to the song you want to extend...

(up to 1min)...

then you can extend THAT clip,,, and extend the next one, until you've made a long song that ends correctly.

[devout crypt]

How to I make my music private where it's sent to my email?

[clever sky]

Yeah again, burning through credits to get it to match well from each track. The time it takes seems to be random, sometimes it will finish the lyric, geberate again and it hard cuts. its realaly sad but it made 1 great track out of 20 abd thats usually the hard cut track or missing lyrics

[neat nexus]

You make a post requesting this as a new feature in #💡┃suggestions-web and wait for it it be implemented. (Be patient this may take a while...)

Alternatively, you could make the music on the website and then email the completed songs to yourself manually.

[clever sky]

also whats tre different between Chirp and Bark, is the V2 and V3 on the main site? which is which?

[digital talon]

With v3 you can choose the timestamp where the new part starts, so you can cut off the wonky bits

[clever sky]

yeah but it still leaves the random issue, so you copy over the ;yric and it cuts off another lot etc etc and.. it's only top the second right?

[neat nexus]

V1 Chirp is available in the Discord for now. It will eventually be replaced with either v2 or v3. We don't know when this will be.

The website uses v2 or v3 alpha but the v3 alpha is restricted to those with paid subscriptions.

Bark is a voice only audio generator. It is open source unlike Chirp and can be found in places like huggingface. You may also find it and run it locally if your hardware is capable IIRC. (I have not used Bark all that much if ever so I know very little about it.)

You can find the v2 Chirp at the website here: https://app.suno.ai/create/

[clever sky]

Thank you, ive paid to get access to v3

[devout crypt]

how can I make my music genertions on discord private

[neat nexus]

Great! I think you'll find it greatly improved in many ways compared to the Discord v1 Chirp. Just remember it is still in alpha and as such you may encounter issues occasionally. Just the ability to resume from an earlier time is a killer feature upgrade though, so I think you'll like it a lot once you get used to it.

[neat nexus]

You can DM the bot and then do all your creating in DMs.

Just remember you are using an earlier version of Chirp with the Discord.

It will eventually be upgraded and V1 Chirp is not forward compatible with the newer models. Complete your songs soon so that you will still be able to stitch them with Chirp and not need to resort to a DAW to edit clips together.

[clever sky]

oh I LOOOVE it, i paid right away after trying once, It's amazing but its driving me crazy making gold and not being bale to finish it properly

[devout crypt]

Thanks @neat nexus let me try that.

[daring orbit]

Folks I read all the docs but I’m stumped on setting the musical key - is that possible?

[neat nexus]

Oh I know this feeling! 😄

[broken lotus]

I have been using
[refrain]
in the lyrics, by itself, on the final clip...

it improvises on the lyrics and (most times) ends the song.

[clever sky]

improvises?

[digital talon]

I found using things like d-major or c-minor actually does something but I couldn't say it's actually changing the key correctly

[clever sky]

Where do you find the commands list? thanks for sharing that...

[digital talon]

There's a Wiki here but keep in mind v3 might be a different beast https://suno-ai-wiki.notion.site/4e24aa838bf84fffaf383b1f9ed50748?v=755ec8954abd4e3f9f5c899950f0f315

For example, v3 definitely handles more complex style prompts much better than v2

[clever sky]

ok, with this is mind, AI arts has seeds, so if you get a track you love and want to use that exact style and voice again can you use a seed command to say make the [Outro] change you didnt in the first track and get a better ending?

[digital talon]

I've not found a way to identity seeds.

[clever sky]

The remix option dosnt seem to carry over a seed either

just the style genericaly

[digital talon]

I've found you can have a certain amount of success by continuing the original clip from early in the song, once the voice and main tune is established.

[clever sky]

do you add the missing lyrics to the start of part 2 from the end of part 1 that it missed or hard cut?

[daring orbit]

Eliot for the key declaration you said you heard a difference did you check the key?

I have software that can get the key from the file so I’ll check it today when I get a break from work

[digital talon]

I didn't, but I'll see if I can dig those examples out

[whole matrix]

Hello, Is there an API or similar way to generate the songs without using the website?

[final hazel]

No but it is officially on the roadmap

[whole matrix]

Perfect thanks

[brave gull]

is there any ETA? (or, where can I find said roadmap?) :p

[final hazel]

Not that I am aware of

[digital talon]

@daring orbit I had to make a couple of new ones
C-major piano
https://app.suno.ai/song/86d35c6d-d806-49ef-a76a-88db859ca069
https://app.suno.ai/song/ec76b949-e8ab-4983-8290-1f8fd1c686a9

D-Minor piano
https://app.suno.ai/song/749b7ecc-e5b0-48d3-a955-a9babd65acb9
https://app.suno.ai/song/0226c8c3-2a60-4c19-aa5d-12c0aab2585d

[clever sky]

how do I raise a ticket? Im getting a copyright notice now, halfway through making a track, this is really annoying !!!!

[neat nexus]

This is probably a glitch but please check the unofficial community wiki article on this subject: https://suno-ai-wiki.notion.site/Error-Couldn-t-generate-that-or-Try-another-prompt-d6e88752dd9144d6b2c14503ceca43aa

Assuming that your lyrics aren't subject to any of those blockers, it's a glitch. First I'm sorry for what you're going through. We've all been there and yes it's annoying. Especially half way through a chorus that it was perfectly fine with the first time.... 🙄

Assuming that it's a genuine glitch (which it sounds like it could be) try adding several lines of indeterminate length of spaces at the end of your lyrics. Add some adlibs (oh--, yeah!, etc) at various points as appropriate. Add metatags to the end for instrumentals or just make up tags, its about changing the profile of the song so that the AI will see it as something different than whatever it is that it is blocking.

[clever sky]

Thanks. weirdly, I clicked generate lyric, they stayed the same, but then passed through. Wonder if there are 2 different systems checking and one is more upto date?

[neat nexus]

Oh and if all else fails, try removing the name of your song and see if it goes through.

I can't prove it but based on personal observation I think it may be choking on song names sometimes. It might also have just been a deal where I needed to close the browser and reopen it, because that's how I discovered this one. So your mileage will vary.

[clever sky]

Thank you? are you on the "team" ? Have you had any luck pushing the peaks back down, a song that started like mellow and great is now screaming every lyric and producing a wall of sound and even just using verse dosnt dial it back down.

[neat nexus]

https://tenor.com/view/well-yes-but-actually-no-meme-aardman-the-pirates-pirate-gif-26563702

What I had happen was luck and persistance.

It can happen. I've had it happen.

But I don't know that I did anything myself.

[clever sky]

LOL, thanks 🙂

[maiden salmon]

Why is my song continuation getting "flagged" for moderation over and over?

Ive changed the lyrics 4 times, changed the continue point, nothing works

[rare pulsar]

If I upgrade my membership does the renew date change to the date I upgraded but one month forward, or like is there no point in upgrading at the end of monthly period?

For context I have 5 days left and currently have the 10$/month monthly plan

And don't realistically think I can use the 7500 credits in the next 5 days

[brave gull]

not really good with web technologies -- any idea how to interpret the content returned from audiopipe.suno.ai? eg. https://audiopipe.suno.ai/?item_id=7e5b40ea-27e3-42ef-82dc-350e09a3b8da

[strong hornet]

what's audiopipe? some sort of Suno api?

[brave gull]

afaik it's some sort of pipe that is used by the main website to read the song when it's not finished yet
first screenshot is player playing unfinished song -- second is finished

just trying to somewhat automate this, but I can manage with waiting till it's done and fetching the entire .mp3 directly

[iron thicket]

Hello everyone, on v3 Alpha I have tried several times to have an electric guitar solo in a metal song. More often than not, the solo is hidden in the mix (nice idea in it but barely audible). Did you guys witness the same thing ?

[strong hornet]

if it's streaming data you could fetch/read and when it's done 'streaming' save the MP3 file (in theory)

but not sure if it's that simple

[brave gull]

i'm more interested in reading it as it generates

but yeah, trying to fetch it just hangs (tried with curl and browsers)

[limpid spade]

Try [guitar solo, virtuoso]. I've had a lot of success with that for instrumentals.

[strong hornet]

does it not return anything? maybe you need a different header

[brave gull]

i'm..... an idiot. it actually returns it in real time.
it's a pipe.
i was checking like 20 seconds after it stopped piping the song 🤦‍♂️
so yeah I can just read that no problem

[strong hornet]

hahahaha well glad I could help

[brave gull]

now I just need to make some sort of queue and I can get myself a somewhat constant thing going

[strong hornet]

I'm just waiting patiently for an official API - one day !!

[languid lantern]

i just hoping they add frenchcore as a style option personally

[strong hornet]

omg JJD adventure (50+ million plays on youttube)

I listen to it every day when I went to school (big fan)

[eager topaz]

How to force an instrumental in a song? I've got a song with just a couple of verses that stretch across the first clip. Now I want to add a 2nd clip that is just instrumental. I've tried a few things but nothing is working. No lyrics and Suno is giving me about 20 seconds on clip 2 before it ends.

[carmine osprey]

I feel like I'm starting to understand the generation better; it appears the model is pulling a coherent song (or song fragment) and I've noticed "transitions" between what seem to be two base songs (usually its like > 3-4 mins in); the lyrics are tuned to the current song, rather than the other way around, generally it's really good, sometimes the high point of the song doesn't coincide with the word in the line that you'd expect emphasis on. In some cases, the song doesn't seem "compatible" with a singing moment, in which case lyrics have trouble generating with that segment. I'm not sure how to split [Verse]/[Chorus] sometimes; I think cutting it at the end of a line then starting with the next line kind of works, just not always mid-verse.

[rose ore]

mwhahaha, there's a radio edit subgenre

that's funny

and potentially useful

[eager topaz]

Jingle is an interesting one 🙂

[cold flax]

I Chirp available to run local? Is there info on that somewhere?

[teal raft]

Suno is not available to be locally run

[lavish badge]

correction: chirp is not, bark is
suno is the company that produces both

[teal raft]

Sure but ~suno~ is now what everyone is calling the musical production side of things, 'chirp' isn't anywhere on the website

Here on Discord sure

[lavish badge]

ig it's the same concept as people saying google owns youtube at this point.

but suno just creates a product, chirp and bark.

[teal raft]

I'm mildly curious to know what they call their stuff internally

[lavish badge]

I imagine the model names, or that pain in the ass and no the other one

[viral lotus]

I am just starting to generate on the website using version 3. I've generated about 15 times (x2 songs ) so far, and they all sound great. However, even though I give it two verses and one chorus, it most often cuts off either after the first chorus or towards the end of the second, for a total time of 30 seconds to one minute. Maybe four or five of the 30 generations so far have completed the two verses and chorus. Is there anything I can do to extend the generation length?

[devout gate]

Is there anyway to control the BPM or the key and is there anyway to actually stick in stone when it does the chorus and verse happen? sometimes on V3 I have problems with it oh bang when I say the choruses when the versus is there anything I can do to assure that happens. I tried to make sure all the syllables are or even to each other for the lines and I try to do four line to eight line segments but I still struggle with it sometimes just jumping into the course at random times

[viral lotus]

I just finished generating 30 versions of a song, and for about 10 or 12 of those, it would jump into the chorus two lines early, but always at the same spot.

[graceful condor]

hi everybody ! how can i export WAV version of my SUNO generated songs? I mean in good audio quality? and can i have the Playback version of that same song ?

[final hazel]

You can only export it as mp3, as it is only made by Suno as mp3 directly by the AI.

This technique is new, so there is still a lot of development and discovery for the devs to be done...

[graceful condor]

ok thank you for your answer

[digital talon]

Someone mentioned this before, but in case anyone missed it, "guitar virtuoso solo" gives excellent results if you're trying to get guitar solos.

[runic dock]

Now i give all my Verse the first time. Even if it cuts at the wrong time, when you do the continuation you can specify the time to continue from. Than you can specify for example the time at the end of the verse or the chorus taken entirely. It works pretty well

[runic dock]

Specify it in the music style

[digital talon]

Here's an example, especially the later 3 minutes https://app.suno.ai/song/1d7d3db3-a7f7-4639-801d-ef37cf71f80e

[lost surge]

Hey, is there an API available for suno?

[final hazel]

No.. it's on the roadmap thought.. ETA unknown atm

[lost surge]

Hey, thx so much - this year?

[final hazel]

Please remove your double post in the other channel @lost surge

[lost surge]

Hey, i am not sure, how to delete. I tried, but could not find a delete option.

[final hazel]

Hover over your text, go to the 3 dots on the right, it will open a new menu

[lost surge]

thx, it worked!

[final hazel]

@lost surge thank you

ETA is unknown as I said

[lost surge]

What means ETA ?

[final hazel]

estimated time of arrival

[lost surge]

Okay, understood, thx so much.

[lost surge]

I am not sure, how you handle mails, @final hazel . Had sent an email to the accounts you give in the app, but there was no answer so far.

[final hazel]

I am not an employee of Suno, I am a volunteer from Suno.

[lost surge]

ohh, i see. Is there anyone from Suno here? It seems like, everyone is kinda volunteer.

[final hazel]

They are working on the product.. Since everybody wants API's and stuff 😉

[lost surge]

Ahh, i understood. So, patience is the key.

Thx so much.

[final hazel]

Yeah.. they tend to be more around during USA timezone.. but mostly heads down working to make Suno better and more complete 😄

[lost surge]

Thx so much.. i understand!

[rose ore]

Hey has anyone else gotten the creepy laugh hallucination in v3 yet? So far, I've seen it twice and it's bone chilling

I mean, it's kinda cool

even if it is unsettling

[broken lotus]

is the creepy laugh in the room with you now...? 😬

[rose ore]

lol

no

silly

[runic dock]

But i dont see guitar virtuoso solo in the lyrics or in the music style.
Anyway i have tried it before. But it was in the music style and works fine

[digital talon]

I used the v3 instrumental option, so I just had to put in musical styles, nothing in the lyrics box.

[runic dock]

OK. Got it. I was seeing only "Psychedelic Rock, Punk Rock, noise rock, experimental rock, Avant-Garde, Surrealism"

[digital talon]

Yeah, that was the intital style I used, then I used guitar virtuoso solo

[runic dock]

We should be able to introduce guitar virtuoso solo in the lyrics . Right ? I am burning all my crédits trying to do that with no success. Any clue ? @rose ore @broken lotus do you know what the Build metatag is supposed to bring exactly?

[broken lotus]

probably need some kind of guitar description or guitar-genre in the style (to begin with)...

and then maybe something like
[virtuoso fingerstyle guitar]

[runic dock]

Yes i confirm that adding in the style at the begining does correctly the job. Now i have first part made with v2. The guitar was not in the style. I would like to add guitar in the continuation part. When i add in the style for the continuation suno is lost and does not sing my lyrics. I think it should be possible to add the guitar in the lyrics

[broken lotus]

the hallucinated lyrics is a bad sign, it means Suno has 'lost' it's place in the lyrics text, probably will skip any metatags too...

I use to try to stop that with a [break] at the top of my Continues..., but I guess now I try to fix the lyric transition..., usually by feeding it some more of the already sung part as a lead in...

[digital jungle]

How do I use REMIX?

[daring lava]

Maintenance seems to be finished, if anyone was waiting

[final hazel]

remix just means that you will reuse the prompt.. so lyric and music style

[digital jungle]

Is there a way to remix and keep the music, but change teh vocals?

[daring lava]

Are you using v2 or v3 (alpha)?

[digital jungle]

v3

[daring lava]

You can use the "continue from" function and continue it from before the lyrics

[digital jungle]

I can remove stuff from the lyrics and THEN continue

[daring lava]

You will need to modify this (it's below title) to the time you wish to continue from

[digital jungle]

what if it cuts off the last word or two from the end of the lyrics?

[daring lava]

Try to get it as close as possible to where you want to change the music. Try it out with the next verse. If you notice that's cutting off other words then include those as well in the lyrics section

[digital jungle]

I don't have an option Continue from a time under the title:

would also like thew ability to take a song I made with Suno a couple months ago and run it through again to improve on the music quality and clean un-muddy lyrics

[daring lava]

Did you select the "Continue from this song option" on the generation you are trying to continue from?

[digital jungle]

yes

I found it

it's only there after you click continue from. So I should continue from a specific time

[daring lava]

v3 is backwards compatible., so you can do that (kinda). It will change things but you can build off (continue) from a previous generation

[daring lava]

Yes continue from where you would like it to begin generating alternative outputs

[digital jungle]

v3 is the better version correct?

[daring lava]

without a doubt

[digital jungle]

I love Chirp. Here's one of my latest songs! https://youtu.be/XgLaGQFo_Fg

is there a way to take this finished song and improve on the clarity of the music and vocals?

and this one: https://youtu.be/E99bl--ebJM

[final hazel]

Please don't use this as a showcase channel.. we have #🐣┃suno-showcase

[digital jungle]

no I was showing you to ask if I can improve on the quality

[final hazel]

everybody has the same problem and no.. you can't really.. in time the quality will improve..

You can "master" it like professionals do, but it will never be perfect

[digital jungle]

yea I master all of my songs before I publish them / release them

[final hazel]

Suno will become better in time.. but already is amazing

[digital jungle]

I'm using Suno to create songs that tell the story of a patented fundraising tool I invented. Each song has it's own sound and message but all to promote a product I own the patent to

[final hazel]

you should share things like this n #💬┃general-chat, more people hang out there

I am sure some would find that interesting 😉

[runic dock]

OK. Can try that

[full kettle]

Does anyone find any specific prompts gets it to consider higher quality vocals/recordings? Sort of how you can use prompts like that in AI image generators to adjust which sample images it is considering?

[final hazel]

Keep things minimalistic. That helps. But studio quality is not yet there.

[full kettle]

Minimalist in terms of the genre?

Like not mixing a lot of suggestions?

[cold zenith]

Which software did u use for the video ?

[cold zenith]

She means don't put a lot of style in the prompt style. less u put better it will

[full kettle]

Thanks! I'll keep that in mind

[daring orbit]

Just spent the morning testing if you can define the key of the song and it looks like it’s completely random unless anyone else knows an input method

Also the keys it chooses are very complex and somewhat hilarious and unexpected- D Locrian, Blues E, Gflat Chrimatic

[upbeat jewel]

does anybody know how to tell suno that the song shall fade out (outro) and come to an end? I have the feeling, every continuation makes it longer and longer....

[final hazel]

Try removing your text in "style of music", and replace it with the word "end" (without quotation) and also remove everything from the lyric text box.

That works for me with V3

[upbeat jewel]

will try, thanks!

seems to work, super! 🙂

[final hazel]

https://tenor.com/view/sol-cheer-solana-cheer-gif-25196726

[odd pasture]

minha musica começou sem introduçao instrumental, tem como acrescentar depois de gerada, a letra é minha mas queria manter a musica

My song started without an instrumental introduction, I can add it after it was generated, the lyrics are mine but I wanted to keep the music

[limpid spade]

I've had success by adding "short, [end]" at the end of my style prompt for the last generation.

[final hazel]

yeah I know that one too.. but I have better results with V3 with my new method 😉
It even builds a different end than I am used to like that

[pliant cave]

Agreed on V2. The more tags you use, the worse it performs, particularly on vocals. V3 seems to be the opposite though; for the most part, it thrives on more specific style tags, otherwise, it tends to produce fairly generic but still quite good outputs.

[pliant cave]

Try the Bark model; it has plenty of creepy stuff like this.

[odd pasture]

help me

I created a song that I liked but it didn't have an intro, how do I add an intro that goes well

[neat nexus]

Editing in a DAW. Or creating the instrumental before continuing the song. There is at this time no way to add an intro after the fact.

[lethal abyss]

Is there a way to submit a song as an input? Also, is there some kind of guide for really advanced commands? Because it feels it's still very unprecise

[eager topaz]

No way of seeding atm, but let's hope!

[ionic trout]

Is the site too crowded rn or why cannot i generate?

[broken lotus]

the devs had to throttle back the service while they are tracking down an issue that was causing heavy load and slowing down gens (Not a technical explanaition, just how I understand what was told to me)

the site will come back for free accts too, but it may be a while (hours, maybe even sometime tomorrow) but it will be back for everyone.

[sturdy pawn]

Hi folks! I have a doubt. Can I generate soundtrack without lyrics?

[final hazel]

with V3 alpha.. for which you need a subscription

[neat nexus]

Some tips at the unofficial Community wiki: https://suno-ai-wiki.notion.site/How-do-I-make-an-Instrumental-352abcb0bf6a4f77b8fcf76a768f18af

[sturdy pawn]

Thank you guys!

[rose ore]

thank you

[limpid spade]

I've had some moderate success continuing from after the Chorus in Part 1 but repeating the full lyrics including [intro]. I haven't spliced them together in Cubase yet but I "think" it might work. It helps if there is a complete break between the sections even if only for half a second.

[sterile nacelle]

what is the 'trick' to get it to create a clip longer than 30 seconds at a time? or is it just a hope and prayer kind of thing?

[wet minnow]

for me it makes long clips when a lot of text is there

but sometimes not.

[sterile nacelle]

Not sure if anyone else is having this issue but the site is having display issues.

[wet minnow]

the server is under mainenance at the moment

[final hazel]

#💬┃general-chat message

[sterile nacelle]

? vague?

[thin flint]

What is your new method for ending? Is it similar to your "intro:" tag?

[half dragon]

She said put End in the style box and nothing in the lyric box... I can't wait to try that as I too had difficulties in ending

[sinful abyss]

subjectively - the number of hallucinations decreased significantly after maintenance (or I'm lucky, lol)

[sinful abyss]

although no - just lucky 😦

[deft girder]

have you guys tried switching genres completely between segments?

crazy stuff

[untold briar]

https://app.suno.ai/song/a2c3839a-1e38-4e72-abf8-7057ef5fd0f9/ love it. tip: keep your continue time close to the start of the clip, it makes the style prompt more likely to override the existing music.

[deft girder]

badass!

[final hazel]

See my text earlier #📚┃suno-school message

[green stag]

hi

anyone awake?

[ashen hemlock]

Me

[green stag]

hi

[ashen hemlock]

Hello

[green stag]

are you enjoying making songs?

[ashen hemlock]

Yes. Getting a good laugh making goofy songs

[green stag]

haha good 🙂

[ashen hemlock]

What about you?

[green stag]

i have a silly video of my song.. would you like to see it?

[ashen hemlock]

I'm still a newbie though so I don't like burning credits on trying things that don't work

Yes

Send it

[green stag]

its only a short clip.. but i just wanted to show someone haha

ty

🙂

[ashen hemlock]

For sure

I'll send you one too

[earnest citrus]

i want the ai to sing exactly the words i write, Is there a way to force this? I dont like it when some words are skiped or added...

[gaunt galleon]

hi, where do i exactly change the version?

[green stag]

tell me if u can see it

[ashen hemlock]

I see

[earnest citrus]

Turn on custom mode

[green stag]

haha cool ok 🙂

[ashen hemlock]

Nice

😂

@green stag https://cdn1.suno.ai/6d90b2a6-12db-43d5-b0f3-a4b58a70c63f.mp4

[green stag]

lol ty

[earnest citrus]

Below title. you dont have payd sub, so its not there

[green stag]

haha nice i have a candy song too

[ashen hemlock]

Haha yes

[earnest citrus]

Is there a way to turn of instruments, and only generate voice?

[final hazel]

Not fully.. you can try with adding "a cappella" in the music style

[green stag]

hi 3D

did you see my video

[final hazel]

Yeah.. I wonder why you posted it in the technical channel though.. it's a bit out of place

[green stag]

i told you its him 🙂

[final hazel]

thin ice

[final hazel]

with V3 it's skips more often than with V2.. But nobody knows how to force it 😉

[green stag]

the timer thing on v3 is perfect really

just use that and youll be fine

yall be good im sleepy

[limpid spade]

'Female vocal only, vocal stem' although most generations are very slow songs and as soon as i put in a music genre it adds instruments. https://app.suno.ai/song/83604964-ff11-4db7-a9e3-050c71989e7c/

[red mauve]

what is it with the overwhelming bass lately? it gets bassier with every take and it distorts everything

[broken lotus]

I was hearing the opposite issue (losing bass) yesterday...
Maybe the prompt...?

[red mauve]

too much bass example :

wayyyyy too bassy too:

[final hazel]

Maybe post the video url so we can actually see the prompt etc?

[red mauve]

it's not the prompt. these are all styles which are not bassy by nature :
darkwave, postpunk, shoegaze, noisepop, dreampop, tempo 160bpm, hush and velvety female vocals

[final hazel]

If you feel it's not that, then I think #🪲┃bugs-web might be a better place to let the Devs know..

[cold zenith]

Why when you separate the music into stems. Is there music in the voice part? Does anyone know? Does anyone have an idea for this problem?

[final hazel]

That's normal

Stems are suppose to be recorded separate.. but Suno delivers 1 file (mp3).. and there is no way if you then split it, that the stems are fully clean. There will always extra sounds from other stems in eachother.. it is called bleed

Especially if you use music generated by Suno, because it is not studio quality

Not yet..

[red mauve]

magic does not exist yet 🙂

[final hazel]

Also.. please do not make it a habit to make double postings. You also asked in #💬┃general-chat

[cold zenith]

sorry

[cold zenith]

Okay, thanks for your reply. I think the big challenge is here. Because being able to work on the tracks on your own would be really good. What would be incredible is if, when you download a finished track, there was a Stems option for pro and premiere. I don't know how the music is generated afterwards....But if the instruments could be generated as separately as possible from the vocals, for example in two separate files during generation. That would avoid this, I think. Besides, it would give you the possibility of keeping an instrument but changing the vocals independently.

[final hazel]

Asked many.. MANY times 😄
It's challenge though, so they are investing if they can create things like that

Thing is.. now the music are made in one file.. separating is not possible (yet)

[cold zenith]

i think they can do it, but it will give them a lot of them to do it. And think that they have other priorities too maybe.

[cold zenith]

Yeah because, if they do it, it will be a big rework.

[final hazel]

this is a big proirity.. but it is really not that simple 🤷‍♀️

[cold zenith]

But i really think that, this features is a must for other opportunities

without it, it would be very difficult for them to add several option in the generation

[final hazel]

Not sure what you want me to say next, rather than repeating myself 😉

[cold zenith]

No nothing i was just saying loud, what i was thinking at the moment haha

[limpid spade]

Haven't seen it mentioned as a metatag but 'sidechain' definitely works and probably 'sidechain compression'. It gives that throbbing effect. Not sure if you can be as specific to say 'sidechain synth to the kick drum' but Suno uses it a little bit. https://app.suno.ai/song/a7582d4d-1f53-4bb5-aa66-f8b767c795d4/

[ionic wind]

I'm sure this has been a topic before, but are there any best practices on how to end a song? My go-to trick so far has been to continue the song then replace "style of music" with nothing but "outro, fade out", which has worked with some success. However, at least half of my attempts don't end this way or end and then resume playing. Any tricks I should know about?

[limpid spade]

I use [end] in style for my final generation and get a good success rate

[ionic wind]

I'll give that a try, thanks

[limpid spade]

Note: i'm ok with resuming playing because i just trim the restart in a DAW or Audacity.

[final hazel]

If that doesn't work.. I have a different way. See: #📚┃suno-school message

[ionic wind]

This particular track is very stubborn when I try to end it. Probably comes from it being mellow and progressive, which often leads to lengthy passages with little variation

[broken lotus]

there are different demux options...
UltimateVocalRemover for instance has access to a bunch of models that separate in different ways...

As Sheep3D said there's no magic, but some styles work better than others...

[limpid spade]

I haven't tried UVR5 with a V3 track yet. It did struggle a bit for me with V2 but i'm sure the 128kbps didn't help with that. Might try tomorrow.

[broken lotus]

I saw this in v2 with long disco instrumental songs... the music can fall into a loop and it never moves on
(I ended up with a 13min song that did nothing new after 6min... I did not win the battle, I gave up LOL)

I finally decided this was an 'opportunity' to create a long fade-out in post edit....

😅

[ionic wind]

Yeah, I'm doing my best to keep these tracks self-contained within the app for sharing, but sometimes there's only so much one can do, it seems. Apparently, this AI model is as prone to repetition loops as text models

[cold zenith]

I'm gonna try it now then 🙂

[broken lotus]

probably a 'feature' to help keep verse/chorus patterns...

[limpid spade]

If your songs are skipping to the chorus before it should or mixing up where a verse ends and the next section begins, the best tip that's worked for me is to check the cadence or meter with your lyrics (number of syllables).

I've seen songs where the lines of the first verse were 3,5,7,8,7,7,7,7 and yeah Suno basically took the first 2 lines as being one and got out of whack. Doesn't have to be perfect but probably within a syllable or two does help. #noguarantees.

[ionic wind]

A lot of the time when I'm continuing a song, the system makes something completely random (disregarding my prompt entirely) that has no relation to anything I've made. This on v3 alpha. It's frustrating to waste credits like this, so what am I doing wrong (if anything)?

[twin dragon]

Hello! Can someone tell me what the Remix option is for?

[final hazel]

reuse the prompt

just click it and you will see what it does.. doesn't cost credits unless you click "create"

[twin dragon]

OK I see, thanks! 😉

[graceful condor]

currently there is alot of neon lights and flickering lights in random lyrics right now : D

i got chirp drunk! https://app.suno.ai/song/5923ceb1-0cbc-4109-948e-0bb73a0ab71d

quality is really flaky on v2 today. sometimes v2 totaly nails it and sometimes it feels like its on a 16 bit realaudio stream.

[graceful condor]

is chirp about to hallucinate.. get some really interesting results today

hey guys, I am using v3 alpha and trying to continue song and or i get lees then 5sec and refund or very weird results that are not continuing song - anyone else having same problems? Is this the right channel for this kind of questions?

[final hazel]

Are you trying to continue at the end of a part?

[graceful condor]

yes

its acting the same on v2. this is actually nice as you dont loose credits.

there might be some in the lyrics that stops it

[final hazel]

I have that behavior too if the part is getting to the end.. try to start a few seconds earlier

[graceful condor]

ok

thx

manged to get two tracks in one. last part is reversed vocals at half speed : D

https://app.suno.ai/song/f9791d09-54c7-4985-9588-f3cf9ad00f85

[final hazel]

lol

Thats a positive way to look at it

[graceful condor]

kind of wierd.. it dosent show any lyrics on the url .. also thumbnail glitches ;D

at 00:55

[final hazel]

that's great bug to post in #🪲┃bugs-web
Please help the Devs iron out things like that

[sharp shard]

Help?

[final hazel]

https://tenor.com/view/ba-dum-tuss-jokes-gotcha-burn-ouch-gif-5645546

Funny ;P

[green stag]

hi

[final hazel]

Hey nick.. chat is in #💬┃general-chat 😉

[green stag]

omg im so sorry

[eager topaz]

Doing Instrumental and clips are going crazy. Some are empty, some are 5 or 6 secs, some 2nd clips are 2 minutes! Sometimes the 2nd clip music is completely different. 🤔

[daring lava]

I'm assuming you are using v3 because you said you are getting 2min long segments. Would you mind sharing the "Style of Music" prompt you are including or maybe a link to an erred generation?

[eager topaz]

Yes, V3. I am getting errors with all different styles, but next error I get I will share the style.

I must admit that I was mixing genres to get interesting results! Could that be the source of my problems?

[daring lava]

Hard to tell without seeing what's being used to prompt, but we encourage everyone to play around and be creative. If you happen to find anything that has a repeatable bad result please share in #🪲┃bugs-web so it doesn't get lost in traffic here

[eager topaz]

Ok, no problem. I totally love some of the results I'm getting. The errors don't really concern me too much. Just thought I'd raise it here.

[daring lava]

Glad you did 🙂 Definitely always on the lookout for more intel on things.

[abstract grotto]

Please help me

Hello anyone there

[eager topaz]

How?

[limpid spade]

I think Swag got answered in the General channel

[gentle musk]

Yesterday's update busted it!

errors, can't continue songs... uggh

[limpid spade]

Not sure it's busted for everyone. I've created a lot today with zero errors.

[gentle musk]

I've refereshed multiple times, tried to create new... same error

appears the song I was working on is dead, can't continue on it. I do old ones

errrrh

[limpid spade]

It could be words in your prompt?

[gentle musk]

won't even do random

[limpid spade]

Title? I guess it's hard to diagnose without seeing all that is included.

[gentle musk]

it's working now

[lunar quail]

Hi,
is there a reason why songs are often clipped at 50sek,1min...even there is more text... very randomly
for example: https://app.suno.ai/song/d7f7a23d-d850-4411-8ee6-e8ad7c349179

[calm crater]

Hello yall 🙂 Out of curiosity ... if music and voice are generated separately, but still work well together, does that mean that music is made first, and voice is generated in adaptation ? Or that a harmonic grid is given to both before they start to work separately ?

[wet minnow]

if i have understood it right, everything is generated together at the moment. the qestion to get seperate tracks (or stems as they are called today) comes up quite often. i am sure the devs are working on that or at least have seen that the people want that 😄

[willow wren]

Paying user now and i am really impressed. Working as a Deep Learning Researcher since many years, i can say the quality of the V3 is a boost. Made 2 German Songs and they Sound really great 😃

[ionic wind]

What the? I just had a generation that was almost perfect for what I was looking for. Listened to it while it generated (until the 1:45 mark or so), then listened to something else until it finished. Went back to listen to the final version only to find that it was now only 25 seconds long, started in the middle, and jumped to a completely different melody halfway through that. The music I listened to and wanted to keep is nowhere to be found, even after reloading. I've never seen this happen before.

[fallow gull]

An interesting observation that adding "trumpet" to the end of the style swings the genre back to Military March. Without "trumpet" it tended to be more towards the Heavy Metal genre. YMMV
https://app.suno.ai/song/27e60abb-2fe7-43cc-b451-f9584bf46b04

[fallow gull]

I thought it was just my memory, but I'm pretty sure I experienced the same thing last night. Listened to the song while it was loading, then went back and it was truncated and messed up.

[lethal abyss]

What are your tips to get very particular and experimental sounds through Suno? Thanks

[neat nexus]

Try.

Try again.

Tweak.

Tweak again.

Get mad.

Try some more.

Rearrange the wording. Move words around. Add new words, remove words.

Try again. Keep trying.

Take a break.

Access a GPT and probe for details about the style and the sounds. How they're made. What they sound like. Who makes sounds like these. Copy and rearrange some of this information and add it to your prompt.

Try.

Try again.

Keep trying.

[final hazel]

1. Experiment with Descriptions: Be creative and specific with your text prompts. Describe the mood, instruments, tempo, and genre you envision. The more detailed your description, the more likely Suno will generate the sound you're aiming for.

2. Use Meta Tags: Implementing meta tags like [chorus] and [verse] can help guide the structure of your music. This can be particularly useful if you're aiming for a specific arrangement or want to experiment with unconventional song structures.

3. Style and Genre Variations: Suno allows for style inputs, so you can try combining different genres or styles in your prompts. For instance, mixing elements of classical with electronic music for unique combinations.

4. Continuation Feature: Utilize the 'Continue from' feature to develop a piece further or take it in a new direction. This can lead to unexpected and experimental results as the AI develops the music based on the initial segment.

5. Instrumental Mode: If you're looking for experimental instrumental sounds, using the instrumental mode can yield unique compositions without vocal elements.

6. Version Selection: Different versions of Suno (like V3 Alpha) have varied capabilities. V3, for instance, offers better audio quality and longer clip lengths, which might be beneficial for experimental music.

7. Feedback Loop: Use the feedback system effectively. Liking and disliking songs helps improve the model and might influence the kinds of sounds it generates in the future.

8. Creative Prompting: Sometimes, unconventional or abstract prompts can lead to interesting and unexpected musical pieces.

Remember, experimentation and iteration are key in achieving unique sounds with AI music generation tools like Suno.

[lethal abyss]

Thanks, 2. What are meta Tags except chorus and verse?

[neat nexus]

https://suno-ai.notion.site/Using-Metatags-34944efe09ec41d693e314f13af44695

Be sure to have Chirp selected and not Bark.

[lethal abyss]

thanks

[neat nexus]

Also not all tags will work. Some tags work some of the time and get sung other times... 😅

[final hazel]

So true...

[graceful condor]

Is there a way to get Sino to not sing my prompts or do ad libs where it adds unintelligible vocals at the beginning of the track?

[final hazel]

To address the issue of Suno adding unintelligible vocals or ad libs at the beginning of a track, consider these strategies:

1. Use of Meta Tags: Inserting specific meta tags into your prompts can guide the generation process, potentially influencing the model to adhere more closely to your intended structure, though it's not guaranteed to completely prevent the model from adding unwanted elements .
2. Instrumental Mode: A suggested workaround is using the instrumental mode or a similar option to get an intro. After obtaining an instrumental intro, use the "continue from" feature to add lyrics. This method can help ensure that the beginning of your track starts with music rather than vocals or ad libs.
3. Addressing V3 Issues: In the context of V3, challenges with intros and the model sometimes starting tracks with words or acapella are noted. This indicates that some issues with the beginning of tracks are known and being addressed .

[neat lake]

I'm sad, even when using V3, most of my songs came out very muddy and low quality for most of the time, but before my credits completely expired, I managed to generate a few incredible results which are more like V3. Seems it is completely random, but I'm assuming that things will improve with the official release. It is crazy how it can generate both something beautiful and high quality, and next day generations can be muddy and bugged at the same time 😄

[final hazel]

Your experiences with V3 alpha of Suno, where outputs vary in quality from high to muddy, reflect its developmental phase. V3 Alpha is being refined and may struggle with specific prompts, leading to unexpected results. A useful approach to address some of these challenges is the "get full song" workflow hack. By using this feature before continuing, you can help the model retain a longer memory, potentially improving the continuity and quality of your music creations​​​​.

[ionic wind]

I find that the sound quality can vary a lot between genres

[broken lotus]

👆

[graceful condor]

What's the get full song workflow hack?

[willow wren]

Today i made the opposite experience: the Memory Hack made it more hallucinating the former Part and Not the actual verse

[final hazel]

The "Get Full Song" workflow hack in Suno refers to a method where you first use the "Get Full Song" feature and then continue from that point. This approach helps the model to retain a longer memory, "potentially" improving the continuity and overall quality of your music creations​​.

[graceful condor]

@final hazel thanks I'll give that a try

[ionic wind]

Uh, I just generated a superhero theme, and the cover art made by the system very clearly features two copies of Sonic the Hedgehog. I suspect that's not what the devs want their AI to generate due to copyright 😛

[final hazel]

please put this in bugs and report in the app as inappropriate 😉

[viral lotus]

johnm68 Do you mean you only feed it verses during the first generation? And then you'd need to edit the finished generations?

I've generated dozens more (of the same song) and nearly half the time, it starts the chorus exactly two lines (half a verse) early.

[runic dock]

No. I feed everything the first time. Chorus and Verse. Suno stops where he wants but i see a longer génération. Than doing continuation i edit the part left. Doing this i get more consistency at least for the first part. It is not easy to make suno sing the same way during continuation. This process burns lot of crédits to have consistency

[errant grove]

Why on earth do I get that for my own text? Any idea?

[limpid spade]

I've had it for just 1 line. "how will I know" and "bring me to life". Try some kreative spelling on werds ewe think it mite not like.

[limpid spade]

Checking number of syllables per line can help suno match Verse Chorus timing better. #📚┃suno-school message

[tight zephyr]

@final hazel Hi, can you explain more how to transition from one music style to another?

[final hazel]

The best way to transition between styles is by using the "instrumental" option, as it allows you to focus solely on the style. Here's a workflow:

1. Start with the "instrumental" option and specify the desired music style.
2. Choose the preferred generation and select "continue for this song," specifying the exact second for the next part to start.
3. Switch to custom mode, input the lyrics, and generate until satisfied with the result.
4. To switch to another genre, return to "instrumental," remove the previous music style, and add the new one. Adding "solo" or "intro" can sometimes help.
5. Repeat steps 2 to 4 until the song is complete.
6. To conclude, clear the "style of music" box and only input "end" without quotes.

Side note: Sometimes, you may need to spend a few credits for the transition to occur, and occasionally, styles may be challenging to combine...

[tight zephyr]

Thanks for the extensive answer, Ill start experimenting with this!

[supple frost]

I just noticed that the clips made using the bot, on the website are labelled as V1. I thought that V2 came to the bot already?

[final hazel]

kindly remove your double post either here or in #💬┃general-chat

[wide jasper]

Is there an issue with V3 not being available?

[final hazel]

Yes, there has been an issue reported with V3 Alpha not being available for some users. This is often related to billing issues. If you don't see the option to toggle to V3 Alpha, it's recommended to check your account page and ensure all invoices are paid. If everything looks fine on your account and the issue persists, it's advised to send an email to Suno support for further assistance. You can check your payments via "Subscriptions", then "Edit Billing Details" on your account page​​.

[fallow gull]

The style of "perpetual canon" which is a "round" like "Row, row, row your boat" is interpreted as some kind of death metal. I'm guessing it is the word "perpetual" since "infinite canon" was not metal at all.

[eager topaz]

V3 in Instrumental is not generating 2nd clips and credits are deducted although Credits refunded message shows, refunds are not generated. Anyone else having the problem?

[broken lotus]

(copy any prompts first)
then refresh the page.

I think the counter on the website doesn't update to the actual credits (It's not actually checking every time), but a refresh will set the real number.

[eager topaz]

Many thanks. I reported it as a bug, but perhaps it wasn't.

[broken lotus]

It could be called a bug... lol

[eager topaz]

True 🙂

[eager topaz]

I refreshed and it added the refunded credits, tks!

[trim yarrow]

It is possible to shorten clips? You can often hear the tearing

[untold briar]

In V3 you choose a custom continue time instead of the end of a clip. So that can create a shorter ending

[errant grove]

Why do I get this message for my own text? Its not copied or anything:

[final hazel]

The message "couldn't generate that. Prompt likely copyrighted" is a generic error response from Suno's system. It may occur for various reasons, not necessarily because your text is actually copyrighted. This issue has been acknowledged by Suno, and they have expressed a need for more specific error messages. Unfortunately, the development of more detailed error messaging is still ongoing.
In your case, since you are certain that your content is original, this message could be a false positive. You might consider altering your prompt slightly

[exotic fiber]

this in kind of dumb, SUNO cant relly on Chat GPT ways of copyright flagging. its not possible to copyrigt a line of a song, so, no one can use bring me to life because is in one song already? Please SUNO, look for a different way to protect the app from copyright content. 🙂

[trim thorn]

I was wondering is there away to get at the midi for individual instrumentation before the mix is output

[daring lava]

At this time that isn't a service provided by Suno, but it's a great suggestion. Please feel free to add it to #💡┃suggestions-web at your convenience 🙂

[ionic wind]

Obvious but useful tip for those of you using the "continue song" function: remember that you can change the prompt each time you do this to introduce or remove elements from the song. I often do this to add an instrument that I feel would heighten the way the song has developed. This can confuse the AI sometimes if you make drastic changes, but as long as you give it enough time to get used to the song first, it should be fine.

[nocturne hornet]

Really enjoying Suno so far, but the only thing holding us back in the low fidelity of the audio files, resulting in "muddy" recordings. Does anyone know how to get more professional quality audio and clarity? Thanks!

[ionic wind]

Use a different genre. The tech limitations hit differently depending on what you're making. Try something like acid jazz, for example. The quality is almost crystal clear. And then try opera and hear the shrill frequencies caused by the low fidelity. You just have to figure out what works best for you.

[final hazel]

I'm glad you're enjoying Suno, kaphene! The "muddy" audio quality you mentioned is a known issue, especially in the current V3 Alpha version. Suno operates at a 48000 Hz sample rate and 192 kbit/sec, which, while decent, isn't up to studio quality yet. This is part of Suno's development and innovation process. To achieve a clearer and more professional quality audio, the Suno team suggests experimenting with music that has minimalistic effects, like unplugged styles, and simpler genres such as hip-hop. These genres tend to produce less distortion and clearer audio within the current technological constraints. Keep exploring different styles and settings to find what works best for your creative vision, and stay tuned for future updates as Suno continues to improve and refine its technology.

[nocturne hornet]

Thank you for the update. I understand your suggestions, however, those are not the styles of music we are working on. Additionally, it is the rich depth of the instrumentation and layered vocals that really make Suno head and shoulders above the rest. We are doing some of our own experimenting and will respond further to the community if we discover and hacks.

[nocturne hornet]

1. Are both this Discord and the Suno.ai website generating using the same version of V3 (alpha, beta, release)? Or is there a difference?

2. Are my generations in this Discord and/or the Suno.ai website automatically (1) public or (2) private? Is there a difference in the answer based on using Discord or the website?

[broken lotus]

Discord is using v1..., last I heard it is still planned to be updated eventually.

the (unlabeled) model on the webapp is v2.

paid subscibers have a dropdown option to select v3alpha.

each model is a step up in quality, clip length.

generations on Discord beta channels are public... DM the bot to chirp in private

[low laurel]

i can't have the continue from here: normal

the 3 dots have no effect

[neat nexus]

It is known. Please see #🪲┃bugs-web

[tight zephyr]

Do you have any examples for styles that work? I'm trying to change from trance to opera or vice versa and tried two dozen times but no luck, with steps 1-3 above. Or maybe I'm just asking too much.

[ionic wind]

the AI isn't that flexible with genres, so some things like that are basically impossible right now. At least as far as I can tell

[fallow gull]

I was doing some testing on on Irish tune styles since v3 can generate some impressive songs using Irish Folk with fiddle or tin whistle intros and fills. However it is not able to understand specific styles such as if you wanted a Jig, Reel, Hornpipe, etc. You might get a Jig or Reel, or some other rhythm but it seems to be random.
Suno understands Polka, but you get a stereotypical German/Czech Polka instead of a more subtle Irish Polka.
Suno understands Waltz and you may be able to get a decent Irish Waltz with some generational effort.

[round hawk]

How can I link my discord account with Suno Bot to another Suno subscription (I'm not using Discord login for Suno).

[final hazel]

try this (not sure if this still works but let's try!)

To link your Discord account with another Suno subscription, follow these steps:

1. Type /subscribe in Suno's Discord server or in your DMs with the Suno Chirp Bot, and then press Enter. Alternatively, you can directly visit app.suno.ai/account in your web browser. If you choose the direct website approach, proceed to step 5.

2. After typing /subscribe, you will receive a message from the Suno Chirp Bot. Click on Manage Your Account.

3. A pop-up titled Leaving Discord will appear. Check the box to trust links from app.suno.ai and click on Visit Site.

4. You'll then see a Sign in pop-up. Here, click Continue with Discord.

5. Finally, a pop-up will appear asking you to Authorize. Click on it to complete the process.

This will allow you to manage your account and link it with your Discord.

[round hawk]

Ok, so this isn't what I'm asking. I already have a Suno subscription, it's using as Gmail login. If I login with Discord, that's another account...

[final hazel]

Yeah I had a suspition. I personally don't know. I think Suno needs to help you with that. My advice is to mail them [e-mail]

[round hawk]

Thanks! Yeah, I figured I'd have to end up doing that. It's strange it doesn't allow you to re-link accounts via the chirp bot..

[versed valley]

Welcome! For help, check out Suno Help

[final hazel]

yeah.. probably because each is a different account in that regard.

[thin flint]

Yay. I made a pretty sweet medley. #🐣┃suno-showcase message

[broken lotus]

on the website... click on your profile pic and choose Manage Account
you can connect a different Discord account

[eager topaz]

Is there a way in Instrumental, V3 to prevent vocals? I don't put vocals in Style but often get them.

[untold briar]

You can add the word "instrumental" to the style prompt as well, with a comma, "blah blah, instrumental"

[ionic wind]

What are we supposed to do about the bug where attempting to continue a song instead produces something totally random that is clearly not even an attempt at continuing what you started? It's starting to cost me a lot of credits.

[untold briar]

There were some recent tweaks to continues that may have improved this? Do you example from tonight I could see? (I'm not sure how much or what was fixed, exactly.)

[ionic wind]

would I have to make them public to do so?

I had several instances of this today for one song alone

[untold briar]

No, you could keep them "link only" and DM me privately the links if you like.

[ionic wind]

will do

[untold briar]

How recent is recent? I believe there was an update less than 24 hours ago, but more than 12 hours ago.

[ionic wind]

I generated these about 3 hours ago

now your question is making me wonder if all the trouble I've had getting my last song to stay consistent even with the melody is related to that update. Will pay close attention to what I generate today

[untold briar]

I'm not certain, but I think this bug was NOT fixed recently. But the bug is the type where it works or it doesn't, not 'sort of' working. When this goes wrong the song sounds totally different.

[thin flint]

I don't think these are bugs, this is just model issues where the only way to fix them is more training data.

[round hawk]

Bingo! That was it! Thanks!

[fallow gull]

You can't

[fallow gull]

Ok, slightly longer answer that I have done with v2. I made a spelling mistake in the first generation but it was a really good generation. I used continue and re-used the same lyrics with corrections. Then I just kept the second generation. You can't complete a full song like this though, it will only work with a song less than 2m. V3 may be different though. Sometimes if you try to repeat the same text, it skips/disregards the lyrics. YMMV.

[runic dock]

The feature has been suggested several times in #💡┃suggestions-web

[mint salmon]

In custom mode, is the lyric box solely for lyrics and metatags? I noticed that people advised to be descriptive with what you want, but I am not sure where such a description would go? Would that be the style box?

[broken lotus]

Have a look through the new wiki site...
Some answers and suggestions here:
https://www.suno.wiki/faq/metatags/

[eager topaz]

O.k., tks, I will try it. All the vocals are foreign voices and I don't know if they are nonsense or insulting!

[tawdry flint]

Hi everyone! Does anyone know how to generate extended song clips using Chirp V3? In my experience the longest clip I had is 56 seconds and the song is just stopped in the middle of the verse. Moreover, many songs I generate are just stopped at the beginning of the verse and the whole clip length after that is just 12 to 18 seconds. Is it possible to tell Chirp to somehow finish singing the lyrics I provided (or at least sing more than 1 minute)?

[final hazel]

V3 is indeed known for sometimes struggling to utilize the full two-minute length, particularly with music that includes lyrics. Instrumentals, on the other hand, often do reach the two-minute mark. It's been noted that the amount of text can influence the length of the song. If V3 runs out of text, it might extend the song by repeating lyrics, creating new content (hallucinating), or cutting off before reaching two minutes​​.

A useful tip is to include more details in the "style of music" field when generating songs. This seems to have a correlation with the length of the music produced. So, try to input more information into the "style of music" section and experiment with different styles if you're encountering short clips. Keep in mind that V3 Alpha is still under development and it might take creative liberties with song structure. The Suno team is continuously working on improving V3 Alpha, especially in areas like following certain prompts and song mixing​​.

[warm iris]

I noticed that in V3 he doesn't finish the words completely. As if in a hurry.
And 80% have an accent in another language (mine). V2 had 15-20% accent with almost no accent.

[final hazel]

Please post this in #🪲┃bugs-web so that the Devs are aware of your feedback 🙂

[tawdry flint]

Thanks a lot for explaining! Will try to experiment more with different styles.

[fallow gull]

I found that adding "finale" to the style for the last generation can get some really interesting endings. Suno will still cut off the last words or otherwise mangle the generation leaving you with something unusable, but interesting nonetheless.

[runic dock]

Sounds like lot of things makes suno ends properly ... except the documented way : the [End] metatag. 😀

[warm iris]

What time zone does it work in? (billing team)?

I paid yesterday but credits 0

already

[untold briar]

EST. You can check your payments on "Subscriptions" page then "Edit Billing Details" on that account page​​, maybe there is a failed transaction there.

[warm iris]

It’s weird, but I received the bill on March 1 and not March 8, I paid yesterday. But nothing has changed

[untold briar]

If you can't find a failed payment you'll need to email `[e-mail]' to sort it out. But check carefully the error is easy to overlook.

[warm iris]

I sent them a check and a message yesterday. But still no answer. I would like to know in what time zone the billing suno team works.

[final hazel]

I personally suspect that your renewal of your credits is on the 8 of march, but they just sent you the invoice in advance. Further more, I know it was Sunday yesterday. So that's probably why nobody has responded yet. Keep in mind it's still early in the USA atm.

[light ether]

Can you DM me your account info and email? We'll take a look

[feral pumice]

any news on an API for chirp been released ?

[broken lotus]

no news... not even rumors.

[ionic wind]

Am I the only one struggling with stuck AI generations right now? I've got three generations just spinning and going nowhere

anyone? It's up to four generations now, and nothing happens. Just completely dead.

[cold zenith]

for people that generate music less than 5 minutes, there's voice.AI, that can split your song into : " Bass, Guitar, Drum, Other, Piano and Vocal ".

for free

[broken lotus]

I have also been getting clips swallowed with a Clip Got Stuck...

[ionic wind]

Yeah. A few got through after a few retries, but the ones that got stuck got stuck permanently. Very weird.

[tight zephyr]

Anybody have any suggestions for giving more clear directions for the song parts? I'm trying things like '[Verse1: Soft and inviting]: In moments low, when shadows fall,' but in this case it will still sing soft and Inviting in most cases

[broken lotus]

I try to keep it to 3 words or less...
https://www.suno.wiki/faq/metatags/

maybe drop the 'verse', or use commas?

(a comma?)

[tight zephyr]

Thanks making it shorter and using commas helps!

[exotic fiber]

https://tenor.com/view/bananas-dance-dancing-grooving-gif-10743961

[nocturne hornet]

Is there any way to tell Suno to make the vocals: (a) dry (no effects), or (b) without harmonies?

[daring lava]

Hey kaphene 🙂

a: haven't found any consistent results with this, though I wouldn't be surprised if that level of control is coming down the road.

b: Yes to a certain extent. For example you can do things like duets/choral performances and such, and when the voices mix they can sound super dope. But... You can't direct the traffic so to speak. The cars just run smoothly together. But again, down the road, there is no telling the level of control there will be. I'm expecting a lot

[fallow gull]

I just did 6 generations (12 songs) and they all had the same female voice.

[buoyant osprey]

would anyone know how to get this beat?https://www.youtube.com/watch?v=JQ4qdmnoQ6A&list=RDJQ4qdmnoQ6A&start_radio=1

I was thinking synth hip-hop rap pop but haven't gotten anything close.

[alpine venture]

can I talk to someone in PMs about credits in general? I use a LOT and it occurred to me that (I haven't done this) making two accounts would be much much cheaper than buying credits. It does say you can contact about custom plans. Basically, I'd like to do that.

I get that this is by design, and it's clever, but I'm not rich haha

[thin flint]

I enjoyed creating this one:
Genre Medley #2: Blues, rap, Electro Swing, progressive heavy metal https://app.suno.ai/song/cc798c36-69d4-4c5b-9b9f-1589c0b1234e

[civic parcel]

I pro user, I am using V3 for make song. I last area V3 give me be bad lyric or soud. So I Spacry the ara for next generation, But suno not mach with my next generation lyrics, of cause it is my scecond part generation for that song. SO How can I fix it, and match my licrics?

[broken lotus]

these tips might help:
https://www.suno.wiki/faq/making-music/my-words-are-wrong-after-continue/

[versed valley]

Welcome! For help, check out Suno Help

[final hazel]

I personally have 2 accounts (premium) too. Paid from the same credit card. There is no problem with that.

[alpine venture]

oh cool, I didn't realize that

[ashen night]

Any suggestions on how to sucessfully make a prompt from a duet woman/man à 'Endless Love' or similar, ie woman answering man and vice versa ?

[final hazel]

Creating a duet in Suno, like a "man/woman" interaction similar to "Endless Love," can be a bit unpredictable. However, you can certainly try a few techniques:

1. Use the Style Prompt: You can specify "male and female duet" in your style prompt. This sets the intention for Suno to generate a duet, though it's not guaranteed.

2. Use Metatags in Lyrics: Within your lyrics, use metatags to assign parts to the male and female voices. For example:

   • [Mary] This is my line…
   • [John] But now I am singing…
   • [Mary and John] or [duet] There is a chance both of us are singing together…

   This can encourage Suno to alternate between male and female voices or attempt to have them sing together.

It's important to note that while these methods can trigger duets, they are not always reliable. The voices may swap roles or sound too similar. Also, keep in mind that attempting to create duets may require multiple tries and could consume more credits.

[limpid spade]

@final hazel You had that in your clipboard already, didn't you??

[broken lotus]

She is a cyborg...

[final hazel]

I have a bot trained with answers.. it helps me a lot 😉
I just tweak it per case

[ashen night]

That was awesome, thank you so much!

[vivid cedar]

it seems there's a max length as to which "Continue from here" will have in its context, right?

For example:
[0:00 - 1:00 Catchy verse and chorus]
[1:00 - 3:00 Something else where the verse or chorus is not present]

Anything after 3:00 it is impossible to get a reprise of themes or melodies from 1:00. It feels like there is currently a limit, I am guessing around 2 minutes. Is this documented?

[broken lotus]

in v1, the 1st clip is 40sec, continuations are 30sec, and Suno confirmed it listens to 10sec
in v2, 1st clip is 80sec, continuations are 60sec... exactly double. I don't think Suno ever explicitly confirmed the context window...

in v3alpha, 1st clip is 2min, continuations are 1min.... Suno has not confirmed memory window

[vivid cedar]

thanks for that extra context. I'm using V3 alpha and based on 1000 credits i spent trying to get it to reprise a theme... I have a hunch the memory window is probably around 2 minutes 😄

[neat nexus]

Two things to try:

• Try making a complete song and then continuing from that. Seems to help it "remember" things.

• Try seeing if [CODA] works to make it do that.

I've had limited success with CODA to the point that I'd like to use it more to see if it does help as much as I think it does but I keep forgetting to use it.

[vivid cedar]

interesting, thank you

[nocturne hornet]

Thank you, David. I know what you mean. The music that I am being able to generate in Suno is far above the other platforms. My only challenge is that all of the vocals, regardless of style, sound like the have stereo chorus and/or Flanger on them, making them "muddy" sounding, not clear or sharp. Additionally, it also really makes it hard to separate the stems for further mastering. I am also adding to the suggestions the idea that when the music is being generated, that it produce individual vocal and instrument channels, really making Suno a professional-grade platform.

[ionic wind]

This could be confirmation bias at work, but it feels like whenever the servers start to struggle (like right this instant, for example), the quality of any tasks that actually go through drops dramatically. This happened both today and yesterday, and practically all the generations that come through while the servers struggle are useless to me because they start veering off into territory the prompt and the song history never asked for.

They start introducing completely new elements to the music in ways that only rarely happens otherwise

I'm looking at 12 useless generations right now, some of which I had to wait for quite a while before they even processed.

[broken lotus]

This is not how the internet works. the creativity does not 'leak' out when we wait for gens to complete... it's just a job, that waits on a queue...

[ionic wind]

That's not what I mean

[broken lotus]

However, the waiting may play a large role in how we feel about the results.

[ionic wind]

I'm a senior software developer

that last statement I can agree with, at least

cloud services sometimes have to shuffle tasks and processing power around in ways that can cause unintended (and unexpected) bugs

and that's how something like this could theoretically happen if it isn't confirmation bias

[turbid topaz]

It seems like it goes through periods where it spends less horsepower on processing yes

it's probably a queueing and priority thing... maybe a metric for how long it is allowed to take

and if it can't scale up to handle them at its best

it scales back each request

that's pretty normal realtime systems type stuff

[outer whale]

Something is not working well with some Lyrics in spanish, the AI still singing "subtítulos por la comunidad de Amara" and that is not written anywere.

[turbid topaz]

"if the queue gets big enough, it spends less cycles in order to keep the wait times down" is a standard scaling strategy

especially if it has hard or firm realtime constraints for resource allocation

it will try to hit its metrics in order of priority

every approach to a timesharing system has its tradeoffs and compromises

the usual goal of an on-demand implementation is "scale up when you need it, but scale back down to keep costs and usage low when you don't" so if it can't scale up fast enough (even the smallest microservices take nonzero time), or if there's no room because it hit a hard limit on resource availability, or. soft one on cost, things will get queued to use what it has in the moment

and if its directive is "spend no longer than X time based on the number of queued requests" you will get less refinement/attention paid to the processing

and the way to do that is

1. Drop requests
2. spend less time/resources processing

1 is usually worse for a customer-facing product, so 2 is the usual choice

then again, it gets to where you can't avoid 1 pretty easily as well

[turbid topaz]

https://github.com/openai/whisper/discussions/928

chatgpt bug heh

that line is present in a lot of non-English text

because so much of it are subtitles for videos that were translated by the amara.org subtitling community, so GPT thinks it is supposed to be there in non-english text lol

like when DE or MJ put "artist signatures" on stuff ha

[fallow gull]

I wondered the same about busy periods and quality. In the technical paper for Sora it was noted that quality is better the more compute resources are used. It was a potential explanation for sometimes getting poor results from Suno.

[turbid topaz]

pretty much anything that is open and cloud based resource sharing is gonna do something like that

requests/sec processed, max processing time, that's the stuff backend engineers are constantly tweaking and arguing about the tradeoffs

scheduling is a big problem with a very wide solution space, and the more we all rely on cloud services it just keeps getting bigger with more nuance

[neat nexus]

@vital quarry I'm sorry you're getting caught by the automod. Do you mind trying to post that in sections? Or I can try posting it for you? Or even better yet, could you maybe adapt it for the community wiki?

https://www.suno.wiki/

[vital quarry]

I wanted to post the best prompts that I used for the latest music on my website Filmmusic.io

[turbid topaz]

just as a curiosity, what constitutes a "play" for counts? is it a minimum listening length or just how many times somebody clicked "play"

[vital quarry]

I can send you the markdown file that I have for my documentation. I can send you.

What do you mean with play? On my website? On my website a play is counted once per user.

[turbid topaz]

that was for the Suno people

[fallow gull]

It's basically the play button and the song starting from 0, but there are a few protections. It is incremented on repeat as well.

I don't think links in Discord trigger a play increment. I think you are playing the MP4 video. I think you need to open the link in the web interface and play from there in order to get a play count

[turbid topaz]

yeah I figured it was on the web ui

[broken lotus]

I checked the spam filter...
it was blocking because the post is long...

if you could split it in two it should post...?
🙏

[turbid topaz]

just didn't know what exactly counted

[turbid topaz]

I appreciate your dedication to Continuous Delivery, Team! Warms my CTO heart!

[wanton aurora]

just tried v3 and I'm blown away. is there a technical report or roughly what the model looks like?

[final hazel]

It's great to hear you're enjoying V3! While we don't have a detailed technical report available for public access, I can provide a brief overview. V3 is an advanced AI model that has improved audio quality, faster song generation, and better expressiveness in music generation compared to earlier versions. It supports instrumental tracks and has expanded language coverage. Additionally, it allows for the continuation of songs from any point. It's important to note that V3 is still in development, so it may take creative liberties with song structure and could struggle with certain prompts. This model represents a significant step forward in AI-driven music generation, continually evolving and adapting to provide better outputs.

[eager topaz]

I'm having a lot of problems in Instrumental with short clips being generated. Burning a heap of credits with this problem. Short 1st clips well less than 2.00 mins and 2nd clips being just 7 or 8 seconds long. No pattern to it, very hit and miss.

[final hazel]

I understand the frustration with generating short instrumental clips. Here are some possible reasons and solutions:

1. Switched Title and Music Style: One common issue is accidentally switching the title of the song with the style of music in the input fields. Ensure that these are correctly entered in their respective places.

2. Original Clip Influence: If the original clip used for continuation is fading out or has silence towards the end, it can result in very short follow-up clips. Try using the 'continue from' option and select a different starting point for the continuation.

3. Input Content: Ensure you're providing enough content for the AI to work with. If the input is too short, the AI might not have enough material to extend the song.

4. Style Details: Adding more details in the "style of music" field often leads to longer generations. Try inputting more information into this section and experiment with different styles.

5. Trial and Error: Sometimes it takes a few attempts to get the desired result. Don't hesitate to experiment with different inputs and styles.

Remember, AI music generation can be unpredictable, and finding the right balance sometimes requires adjustments and experimentation.

[sage turtle]

The "style of music" prompt area seems way too small... it's hard to describe harmony, vocal style, instrumentation, effects, etc, in those few characters. Does it need to be so short?

[neat nexus]

Did you know that you can skip spaces with commas?
An example:

melodic instrumental,hiphop,afrobeats percussion,timeless trap house

It won't help a lot but it may be enough to squeeze another word in for you.

[sage turtle]

yep, gets a few more characters, but not nearly enough

[neat nexus]

Part of the issue IIRC is that the more you add to the style the faster the vocal decay. It doesn't stop me but it's a reason to limit that input.

[sage turtle]

very interesting, thanks for that. I'm trying to put some stylistic cues in brackets in the lyrics with mixed to poor results

[neat nexus]

Yeah, that can work sometimes, but it can also get misread and the AI will sing the directions.

[sage turtle]

yeah a song really loses its dramatic impact when you hear "she cried to the stars violin hits oh why"

[rose ore]

technical question: Assuming that a thumbs up means a take is good, and a thumbs down means the take is bad... how does the computer know which parts of it are bad? Like, for example, I've had hallicinations that sound incredible, and there have been takes that I rated as bad, but had some issue with the performance. I guess what I'm getting at is, how smart is it there?

[fallow gull]

Hopefully they aren't using the rating for anything other than organization or ranking in the trending section. I would hope the the thumbs up and thumbs dows are meaningless, because it is the only organization mechanism we have and not reflective of the quality. I have thumbs downed a lot of really excellent generations, they just cut off to early or had some other issue.

Having AI train on itself can be a bad idea. Studies have shown it causes the AI to converge to a certain style and essentially get worse.

[graceful cliff]

there are playlists now

[graceful cliff]

not very smart but again the bar is high, any bad is bad

[rose ore]

Sucks too, because some of my hallucinated tracks where it misreads and garbles my lyrics have been some of the best things I've ever had musically

[graceful cliff]

😦

we live in a imperfect world. hopefully only with perfect music.

[rose ore]

I know tony, life is just.... you know 😭

[peak topaz]

I spent 2 hours trying to get a duet earlier. closest I got was an autotune sounding echo

oh wow. yeah I got no where close to that

I tried splitting it up and specifying male/female in the style, and even then I'd usually get one line as the specified one, and then it'd morph back into the original lol

this is as close as I got

but after that it wouldn't switch back

[thin flint]

Where'd you find that?

[thin flint]

I'm interested in them; feel free to DM. Always interested in learning new Suno power tricks.

[ionic wind]

I'm only getting shortened clips right now, cutting off before the full length of 2/1 minutes. Anyone else seeing this?

[limpid spade]

Nope. If I get shorter clips it is usually from not having enough lyrics, with the chosen genre, to take up 2min.

[limpid spade]

My best duet was a V-day accident.

[valid bone]

Hi ! I have subscribed to suno... But I was wondering if there is any API available or anyway to login through CLI / python ?

[broken lotus]

No API... not yet anyway.

[placid cape]

is 2 minutes pretty much the cut-off for a song?

[valid bone]

In the bark's git it says that if we need more info on chirp's api to check a URL... but the url is 404... Anyway...

[final hazel]

Suno's song generation lengths vary between versions. In V2, the first generation is up to 1.20 minutes (80 seconds). V3 can go up to 2 minutes for the first generation, but this can be shorter depending on the amount of text provided. If V3 runs out of text, it might extend the song by repeating lyrics, generating new content, or ending early. For both V2 and V3, any continuation generations beyond the first part have a maximum length of 1 minute.

[broken lotus]

this might help?
https://www.suno.wiki/faq/making-music/how-do-i-make-the-song-longer/

[placid cape]

it does, but it seems like the entire song is also limited to about 2 minutes

[valid bone]

I really need to login through CLI but this google login is a mess 😦

[final hazel]

the wiki url Wetcircuit provided explains you how to continue

[placid cape]

I'll try it on Discord; the website just cut it off suddenly

[final hazel]

Yes, music get's cut off. That is normal. Then you will create a second part, Follow the instructions on the screen. In the menu you have the option to create the next part.

And in the end you can create a whole song of all the parts. It get stitched together.

[placid cape]

gotcha - I'll work on it. Thanks!

[frigid sedge]

where is a link to understand the syntax for lyrics? I see things like [chorus] , etc

[broken lotus]

this is a good introduction:
https://www.suno.wiki/faq/metatags/

[brave rain]

If I didn't use any of my credits last month, is there any way to get them back or get a credit?

[final hazel]

Unused monthly plan credits do not carry over to the next month. This policy is in place because Suno.ai allocates computing power based on a fixed monthly credit usage, and accumulating credits could lead to unpredictable demand and operational challenges​​.

[brave rain]

Any possibility in getting a refund for that month then, seeing as I didn't use any of the allocated credits, and am continuing my membership?

[final hazel]

Sadly that is not how it works. I understand your frustration though.

[frigid sedge]

Is there a way to just get narration or story telling? Like if I want to do a childrens book?

[peak topaz]

I managed to get a narration to kind of work after ~10 tries

[peak topaz]

https://app.suno.ai/song/1afd2e2a-c6b9-4487-bb43-e5e2e5e0ec18

this was the best I could do after like 30 minutes of tinkering lol

[neon marlin]

i used this style "deep male narrator speaking clearly and emotionally over soft slow piano" and only music played, any suggestions?

[peak topaz]

took me something like 10 tries per section to get it to work

[neon marlin]

so there is not a standard style

[peak topaz]

no idea, I only discovered this AI existed yesterday

lol

suffice it to say it took some effort to get that narration

back to duets... a back and forth duet, or a harmonized duet, within the same chunk of music seems impossible... I did nearly 100 attempts last night messing with trying to do that with ultra limited success.

I eventually decided to do a different style (each verse a different singer), and even then Suno wanted to continue the original singer's voice despite emphatically specifying a gender in the style.

Trick I ended up using was to put a couple of seconds of instrumental in between the voice changes. still not 100% success, but went from like 1 in 10 being what I asked for to maybe 1 in 2-3... managed to make a mostly complete 3+minute custom song this way.

the longer the instrumental part, the better chance I had of switching to a new voice. but, realistically wanted minimal delay there, so cant just extend it forever

splitting things up a bit also let me specify things like [Chorus, emotional swell] and it seemed to react to that somewhat consistently

[final hazel]

The quest to create a harmonized or back-and-forth duet in Suno is filled with both opportunities and challenges. Here's a deeper insight into how you can approach this:

1. Nature of Duets in Suno: Duets can occur both spontaneously and through specific prompts. However, the reliability of these duets in terms of distinct and consistent voices for each part is not high.

2. Techniques to Try:

   • Genre Style Prompts: Consider using genre-specific styles in your prompts, such as 'Broadway Musical' or 'duet'.
   • Metatags in Lyrics: Assign male and female names to different song sections. For example, tagging sections with [Mary] and [John] might help the AI differentiate between the two voices.

3. Unpredictability and Challenges:

   • Inconsistencies with Gender Assignments: The AI doesn't consistently adhere to gender assignments in vocal tones, so your duet may not turn out as expected.
   • Voices May Swap Roles: Be aware that the voices may interchange roles or sound too similar.

4. Balancing Creativity and Credits: While creating duets can be enjoyable, it can also consume a significant number of credits due to the need for experimentation and repeated attempts.

5. Trial and Error Approach: You might need to try multiple generations to achieve the desired duet effect. This process requires patience and a willingness to experiment with different combinations of prompts and metatags.

Remember, Suno is continuously evolving, and the AI's capabilities in complex tasks like duets are still being developed and refined.

[peak topaz]

yeah, saw this when I searched discord for "duet", but had extremely limited success... literally did 100+ generations trying to make it work hehe

even did a bunch of the basic Mary/John one with like half a success

[final hazel]

I am by nature a persons who loves to puzzle, so do it frequently like that. It learns you how to talk it's language 🙂

[peak topaz]

yeah, best single-clip switch I got was this one I posted last night, after many many attempts: #📚┃suno-school message

was basically perfect, but it refused to switch back to male no matter what I did

[final hazel]

Nice!

[peak topaz]

full version of what I ended up with last night after ~2000 credits 😅

only two singer changes at ~1:10 and ~2:40

[final hazel]

Great job!

[glacial heart]

Hi everybody¡¡

[weary bolt]

Is it posible to edit a generated song ??

[final hazel]

Currently, Suno does not offer a feature to directly edit a specific part of a generated song. However, they are actively researching a "paint in" feature that would allow users to select and regenerate specific sections like a bridge, chorus, or verse, without changing the entire track​​.

You can however continue the song from everywhere you want if you are on the paid subscription.

[graceful acorn]

Anyone getting issues with V3 (Alpha) only doing around 10 seconds on continued songs?

[icy light]

how do you guys get the V3 option?

[graceful acorn]

Paid sub.

Might be only a few people in the subs, not sure.

It's definitely worth it. First generation is often around 2 min and it follows most of the full song lyric. But continuing after that I get a lot of stuck generations or very short 10 to 15 seconds. Just above the refund threshold. 😆 (only tonight, yesterday it was fine)

[deep socket]

will suno ever create midi and seperate tracks, so that we can import the songs into a daw and work on them?

[upbeat folio]

I belieeeve they've officially said it's a goal. In any case it's definitely a very highly requested feature.

[deep socket]

also, is there any documentation on what prompts and commands work best with suno? I love that its quite random but it would be cool to know how to take as much control as possible

[upbeat folio]

https://www.suno.wiki/faq/

[deep socket]

I´m glad its highly requested. It would be super useful. Sometimes Suno makes killer tunes but it might run a verse too many times or hit a wrong note or not resolve into a tonika when I need it too

[deep socket]

thanks!

[deep socket]

This documentation is very lyric oriented. I only prompt instrumentals beause I dont think ai song is good enough yet. I hope some guideline prompts direct instrumentals as well though

[upbeat folio]

Yeah the instrumental switch in the UI is new with v3 so the documentation is also a work in progress I'm sure. In the meantime, you can probably find some good advice by searching "instrumental" in the discord server. Is there a specific question you have?

[deep socket]

I basically just wonder what suno understands and not. Can I feed it bpm, key, scale and musical structure info and how do I do that effectively

Does it understand what an AABA stucture is for example?

If I would know what data is has been fed, I can communicate better with it. I study musicology so I got quite a lot of different approaches I can choose from

for example. It seem to understand much better what "late 80s videogame music" is rather than the genre name "chiptune". That makes me wonder how it works. it seems to understand other genre names just fine though, so it seems like lesser known genres with fewer written descriptions are harder for it to create. In that case it should be trained on Ludomusicological data. The data is there, but it might be harder to get to. IDK

[limpid spade]

https://www.suno.wiki/ even the devs acknowledge that there is no comprehensive list of styles or genres Suno understands. Try what you think and see or check out the Explore page, Showcase and General for what others post.

https://www.suno.wiki/faq/style-and-lyrics/styles-and-genres/ A list but not exhaustive.

[graceful condor]

is there a way to mass delete generations you've made or do you have to do it all one by one

like, i literally go through 1000 for each song that i make

[full kettle]

lol

[broken lotus]

I think the latest update allowed for multi-select... Try shift clicking on a bunch in a list, then try the trash... I think they should all delete?

[bleak lotus]

/chirp text to music

[mint salmon]

Would I be correct in saying that Suno only uses the song you are continuing from as reference for the next generation. So part 3 will only be using part 2 as reference?
if you wanted better continuity, you should 'bake' your parts into a whole song every so often so that it can refer to earlier parts more successfully.

[lethal wren]

Question, can crop a generation , and continue from the cropped version somehow?

[final hazel]

In Suno, the "Continue From" feature primarily uses the immediate previous part as a reference for generating the next part of the song. If you're aiming for better continuity that references earlier parts of the song, it's recommended to use the "Get Full Song" feature before continuing from a specific part, as it allows the system to consider more of the song's history​​.

[final hazel]

Yes you can if you have the paid version.

[lethal wren]

i have the paied version, amazing, thanks!

[final hazel]

Enjoy!

[lethal wren]

❤️

hmm, i don't have this option when i'm looking in the ap, do you ned one of the more expensive versions?

[mint salmon]

do you have custom mode toggled?

[lethal wren]

yes

[final hazel]

If you're unable to see the V3 option in Suno, it's most likely due to an issue with your last bill not being paid. Ensure your payment details are up to date to access V3 features​​.

[lethal wren]

i have it set on V3

[final hazel]

Go to subscriptions and click on "edit billing details"

[lethal wren]

ahh

fixed it

appenertly the son needs a title

[final hazel]

👍

[lethal wren]

thanks again! ❤️

[ionic wind]

What happened to the embeds today? Some of you manage to embed your tracks in the #🐣┃suno-showcase channel, but mine just stay regular links now. It worked until today.

[final hazel]

which one?

[ionic wind]

I've posted two links today that didn't turn into embeds (at least on my side). I also see other people's links also don't resolve, but there are some working embeds without links. I can't figure it out. Something has changed.

[final hazel]

Can you give me a link to we can try to figure it out together?

[ionic wind]

Well, here is one of the links I posted earlier today in that channel: https://app.suno.ai/song/ef4a25a0-d28f-4f96-a8ff-f4180d232ed0

see, no embed when I post it like this. Before today, it would create an embed automatically

[final hazel]

https://app.suno.ai/song/d1368fba-d186-4f4f-b6b1-0c1a9afd2b57/

Okay.. I see it happens now with others too... I will let Devs know

[ionic wind]

Thanks!

[stable pollen]

Are there any good resources discussing how the technical aspects of how Chirp works available anywhere online? I found the Github repo for Bark, but nothing for Chirp at first glance

[final hazel]

Likely because it is not open source 😉

[stable pollen]

aw, that's too bad

Have you released a paper or technical report on the model or is it completely closed?

[daring lava]

The team hasn't released anything of this nature (yet)

[stable pollen]

Oh, okay, thanks anyway

As I understand it, Chirp is a fork of Bark, so I can probably have a look at that to get a general idea on how Chirp works

[sturdy quest]

yeah it would be so cool if even something the same level of v2 was opensource... do you guys have any idea about the approximate amount of VRAM necessary to run something like v2 or v3?

[daring lava]

I have no idea, definitely out of my wheelhouse.

[sturdy quest]

well with bark only taking about 5 Gb of VRAM to run these days, I am positive it won't be long till a very good open source model will be available for everyone to use

btw, I thought the website had no credit limit during this stage, so I finished all of them without even noticing... so I was wondering, the generations we make here can't be private, correct? Just making sure I got this right

[balmy vector]

I've done one part of a song with V2 and another with V3. The volume in these parts are very different. What happens when I choose "Get whole song"? Will the difference persist or will it be solved?

[daring lava]

The "Get whole song" function doesn't master it by any means, the exampled results will be what you get.

That being said you can use the "continue from" function to trace back through your v2 back to a point where the volume doesn't seem to fluctuate/decrease and build from there

[daring lava]

If you mean via Discord, you should be able to invite the bot to your server and do it privately there

[ionic wind]

It feels like "Get whole song" applies some kind of normalization to the clip as part of the process. It definitely seems like the volume tends to increase after doing this

[daring lava]

I’ll test this when I get home

[ionic wind]

It's something I only discovered yesterday, so it might be a new thing

[broken lotus]

I have heard it too, and sometimes the volume seems normalized the other way (single clip is louder), and sometimes I don't hear a volume difference...

[thin axle]

Well that's super helpful! I feel like if I continue from a timestamp it forgets everything it's done, but usually if its an entire clip then it keeps things mostly cohesive, melodies often retain, etc.

So basically just codify it with "get full song" before adding? is there a way to get full song only up to a cropped timestamp in the final clip?

Using paid version webapp

[final hazel]

When you make the full song, you can just continue from anywhere with this tool:

You're right in your understanding. When generating new parts of a song, Suno's "Continue From" function uses the most recent part as its reference. This can sometimes result in the loss of continuity from earlier parts of the song. To ensure better cohesiveness and reference earlier sections, you can use the "Get Full Song" feature to compile the whole song up to the desired point before generating additional parts​​.

[thin axle]

Definitely appreciate that, and will use it. The bigger issue tho, is that if I use a timestamp instead of the full clip, its the same thing. Like for instance, I have my first clip, and trying to continue from 1:20, and its picking random section, adding new melodies, just twists it up pretty bad. And I've probably gone through 10 iterations and not one actually picks up where it left off. Maybe it's just a limitation right now

[runic dock]

From time to time, Suno resembles a racehorse, powerful but blind. It can go very fast but in the wrong direction.

[humble laurel]

is there a way to copy a similar sound to another song that you have made in suno? Using the same keywords will generate vastly different results. I generated a song with a really dope style and I want to use it for other stuff but I can't figure out how. I know the continue from feature will do that but if you hit 'get full song' it will bake all the previous parts as well. Basically I want to start over but with an identical style

[final hazel]

In Suno, generating a new song that closely mimics the sound of an existing one is quite complex, as the AI's generation process involves a degree of randomness and interpretation. You can attempt to use an entire song as the style prompt and then continue, which sometimes provides better continuity, but it doesn't allow for starting anew while maintaining the exact same style. Currently, there is no feature to "go back" or revise tracks in a way that you can retain the style but start over; you can only continue tracks​​. Your best bet might be to note down the exact settings, keywords, and style choices you used in the song you like and apply those to the new track while understanding that some variations are inevitable due to the nature of AI music generation.

So, the one thing you can do, is use the "continue from" option an try to start the song from the start (so best if it has an intro). This "sometimes" helps to captor the voice and start a new song with it. Here a short how to use the continue from anywhere.

[humble laurel]

Thank you so much for the detailed reply! I will give this a shot

[north beacon]

Im using my own lyrics. I try to use the continue song from set point of time on previus clip, but very often for some reason when I have at least like 40 seconds of lyrics in it does like 9 or 18 second clip cutting the lyrics short? Why?

I tried doing only a verse at a time and it still for some reason cuts the lyrics out every now and then

[runic dock]

How can you use an entire song as the style prompt?
Replicating a style seems impossible.
several suggestions have been made here to control the degrés of randomness of the suno generation. LLM have the temperature parameter. Suno should provide something similar that can allow users restricting its freedom.

[obsidian tendon]

Huh I wouldn't have thought of that. I thought Get Full Song was more of a finalization but what I'm reading is that, Suno does better at analyzing more of the song vs continue from and thus the results are better.

[final hazel]

Currently, Suno's capabilities do not extend to using an entire song as a style prompt or accurately replicating the style of a specific band. The system's design focuses on generating music based on user-inputted parameters, but it's not configured to analyze and mimic the complex nuances of an entire song or a band's unique style.

[obsidian tendon]

This is an example of still needing to understand the principals and foundations of what the AI Tool is designed around. If you know music theory, structure, and read' into the tagging in suno.wiki. You learn quick the tool is best at making parts and piecing it together outside the app. Like in Audacity.

I seriouslly think that a Music Producer with Technical Skill set can one day create music indistinguishable from the old ways

[runic dock]

I have misunderstood what you previously said :
" You can attempt to use an entire song as the style prompt and then continue, which sometimes provides better continuity,"

[obsidian tendon]

I think they were saying you will find better results

From their experience or understanding

[azure urchin]

Do I need to include lyrics or could I just leave it blank and get an instrumental

[final hazel]

Ah.. sorry.. I am Dutch so sometimes I might also be not fully clear 🙂

[obsidian tendon]

You also need to toggle Instrumental I believe

Leaving it without lyrics and no toggle, I guess it would hallucinate lyrics. I've seen it mumle jibberish if I got the syntax wrong on a tag

Like [Verse} where I placed } incorrectly, reads it as lyric and tries to add it

[final hazel]

Hallucinations are just like ChatGPT making up things and lying .. that will happen from time to time depending on various factors.

[azure urchin]

Seemed to work fine justt leaving it blank. Didnt see a toggle.

[obsidian tendon]

I like if one day I could feed it a recording of an original guitar piece and help me build from there. I can come up with pieces but not a whole song, instrument wise.

Another Tool would be AI transcribing the music into tabs or sheets to learn how to play in real life

That would be dope

[final hazel]

Yeah that would be great, until people violate copyright.. and that is the big pain point if they would develop this.

[obsidian tendon]

This is true. I would want to work with a localized model and not one publicly or even privately held by Suno. That would be the only sure way to avoid 1. Having my piece copright violated because the data exists 2. lesser change of violating existing copyright by using a custom data set than the default of Suno

Which brings up a question I've been curious about, what data set are they using; has that been shared anywhere?

I'm curious if it's doing so well from enriched data of patterns and audio clips or is it learning from actual music, songs, etc

[final hazel]

Oh but they are exploring the possibility of allowing users to prompt with their own beat or recording, yet there are many pitfalls to consider. Moreover, as long as you are creating within their tool, they have responsibilities to prevent abuse that could violate copyright laws. The only other option remaining would be to make it open source. However, for a company like Suno, how would they cover their expenses? Would you give away the blueprints for a great product so easily?

[obsidian tendon]

Yeah the path is not clear. AI tools right now is like the Internet before ecommerce and social media. We have an idea of the implications and in my opinion, it's always around what the person(s) does with it. Not what the tool is capable of doing. But we like to put blame on it versus ourselves. Humanity, go figure ha

For now, I like the idea of using Suno to input how I feel when I wake up and get a theme song for my day. Mental Health Application 😉

[runic dock]

Suno has the potential to be a disruptor in the music industry. Therefore, making it open source might not be the best idea. They possess an advantage that should be maintained over competitors. A more sensible approach appears to be offering an API similar to OpenAI. This would allow for greater control by users while keeping Suno's intellectual property proprietary.

[tranquil abyss]

Suno seems to be trained with material from Epidemic Sounds or am I tripping? I've heard melodies from several tracks from there especially the 2010 pop ones

[rustic kayak]

Are there any plans for giving chirp / Suno an api so music could be generated / fetched programmatically?

[final hazel]

It's on the road map... ETA unknown

[obsidian tendon]

Would the API make it faster and easier to create tracks, upload to Soundcloud, Beatport, etc then $$$?

[final hazel]

Evertything is possible in the future 😉

[rustic kayak]

That's great to know. Thanks!

[obsidian tendon]

Hopefully there is some regulation or label system in place by then. So two types can co-exist, music created old ways and AI music. Both talented muscians and talented techies should benefit from this

[undone cypress]

trying to figure out how to better micromanage prompts merging different genres so that they remain harmonically interesting... often if i take one style and merge it with something more poppy, it'll think i want those standard diatonic chord progressions which is precisely what i wanted to avoid with the former

adding something like "chromatic" doesn't seem to be much help

[final hazel]

The first day V3 alpha was released I made this: https://app.suno.ai/song/8db92e67-6b31-4726-a0ea-08fe35b571f6
Is that what you mean?

If not.. getting styles to work out together the whole song, is trial and error. And you could go through others peoples songs to see how they do it here: #🐣┃suno-showcase

[undone cypress]

i think "jazz" might be one of the innate exceptions lol. it varies a lot from style to style. in my experience adding "math rock" will make just about anything sound proggy as hell

[final hazel]

mine starts with jazz.. it progresses to hip hop.. to pop.. to blues.. to country

[hybrid heart]

This may have already been discussed but it's hard to find in here. If I have a paid subscription, can I use discord chirp bot to generate longer songs in v3?

[magic gate]

Is anyone getting results like this lmk what u think

[bright quiver]

Quite often Suno sings some lines too slowly (or too fast) and goes out of rhythm. Are there some hints to show Suno how fast I want it to sing this part like maybe adding "|" to imitate musical bars, or maybe dashes "-"?

[rose ore]

Okay, so riddle me this ai wizards: Why is it that Suno can remember entire bridges from five extends ago when I don't want it to, but it ofen can't complete single words on a seam?

[lofty gyro]

i cant finish my song with lyrics

[final hazel]

No, Discord has V1. V2 and V3 are only on the webapp.

[north beacon]

Im using Suno from your website with google chrome. V3 with custom mode and lyrics, does the AI understand commands like [Crescendo] [Key Change] [Tempo Increase] ?

I watched the metatag list and it seemed very short and I was thinking is that truly all that it can understand?

[final hazel]

While Suno's AI is capable of interpreting dynamic control elements like [Crescendo], [Key Change], and [Tempo Increase], it operates as a large language model with a degree of unpredictability. There is no certainty that the AI will consistently use these tags as intended. This might mean you'll need to spend some credits experimenting with different inputs to see what works best. The AI can be random at times, adding an interesting element to the song creation process, but it also means you might encounter inconsistencies​​​​​​.

[north beacon]

Is there a full list of the commands that it can understand somewhere? The list on tips and guides is quite short and doesn't involve any of the commands I mentioned for example

[final hazel]

No, because that list could be anything. It's like asking if you know from yourself how many words you know. A language model can talk to you like ChatGPT and knows even words in all languages. People use all kinds of words like [scream], [whisper], [sing], [rap], [boom], [explosion], [sung happy], [hook].. It's up to your imagination.

But there is a wiki the community made here: https://www.suno.wiki/faq/metatags/

[north beacon]

Okay thankyou!

[broken lotus]

Try not to think of these as 'commands'..., that implies too much.

They are 'prompts'.

Someone once compared metatags to plastic surgery 😁

"a little, in the right places, can be fantastic..., but too much all over is not going to give good results."

[north beacon]

Yeah okay!

Another question does capital letters inside the prompt mean anything for example [Chorus] or [chorus] will it understand it the same?

[final hazel]

same 🙂

[knotty kite]

No ..maybe? Also , there are ai DAWs we can use to import the mp3s and get the individual tracks extracted and treated like individual midi notes

[north beacon]

Okay great, one last question. Is it okay to use tags in middle of lyrics?

[Verse]
In the server's glow, a question forms
"How do meta tags in your app perform?"
[Curious] Seeking knowledge, clear and vast,
In Discord's realm, the inquiry's cast
[HelpSeeking] Guidance sought from those who know,
To make your app's metadata glow
A simple ask, in the digital night,
For understanding, [Enlighten] for insight bright

Or should I do it like this

[Verse]
In the server's glow, a question forms,
"How do meta tags in your app perform?"
[Curious]
Seeking knowledge, clear and vast,
In Discord's realm, the inquiry's cast.
[HelpSeeking]
Guidance sought from those who know,
To make your app's metadata glow.
A simple ask, in the digital night,
For understanding,
[Enlighten]
for insight bright.

[broken lotus]

🤷‍♀️ 😬

let us know what you discover.

[vestal star]

Would also be interested. I would love to make a songs genre and feeling change between verses and chorus.

[final hazel]

You can learn a lot from others in #🐣┃suno-showcase

[cold zenith]

hello everyone, hope you're okay

[livid notch]

when i want to generate a song idea in prompt maker then the genre &/ the song subject don´t gets displayed in chat gpt 4 prompt maker then rest is fine but the essential thing style and Thematic don´t show up @final hazel

[final hazel]

Sorry.. I never use the prompt maker. If you think it is a bug feel free report it here #🪲┃bugs-web

[eager topaz]

I use RipX DAW Pro , extract to stems then edit the vocal track. Vocals can be edited quite easily, unlike a standard audio file. The biggest problem is that the quality of Suno is not great, so there is a lot of excess noise to be deleted, although that has improved a lot with V3. RipX DAW Pro can be downloaded for a trial to see if it suits your purpose.

[vestal star]

Usually the prompt istn't shown or am I missing something?

[final hazel]

if you open the video, you see above the music style, and the lyrics with the prompts.. so it does

[vestal star]

That's really all there is to it? OK, I'll check.

v3 is only available to paying customers?

[final hazel]

yes

[unborn turtle]

Is there any way I can get a single bassline riff with no other instrumentals? Then build of it to add snare or drums, then build that intro an electronic music song ?

[dry oar]

Yep, same thing. Hopefully in the future more options appear for prompt engineering so we can fully shape the song the way we want it

[obsidian tendon]

Idea to suggest a UI change, less clicks to access by bringing it to the first thing a user clicks after the song generates

[pearl falcon]

is there a list of all the commands you can promt with the lyrics? i want to create a spot in the song for talking

[versed valley]

Welcome! For help, check out Suno Help

[twin saddle]

I’m curious who to talk to if I’d like to make a personal data access request if it was possible?

[obtuse flare]

just adding memes to explore section songs at 1am

[clear reef]

How do you make your songs longer? My songs always stop at 1:20 and never finish its lyrics

[ionic wind]

You need v3 for longer generations. They go up to 2 minutes and can be lengthened

[runic dock]

"Continue from this song"

[warm bolt]

https://youtu.be/s5--nRB4ykI

[tired briar]

/chriso

[narrow gorge]

Hi, I made a song through Bing with plugin and I really liked it. But im stuck as I want to continue it. However the chat session has ended, so i cannot continue it from there. Or make a new chat session and continue. not sure if subscribing to Suno would work either. I can only link to the song piece and download it. Any ideas on how to finish it or make next verse?

[thin axle]

Is there tech support? There's a UI bug that makes this thing nearly unusuable whenever I continue a song.

It makes the section with the generated clips TINY THIN

And there's no way to resize sections.

[light ether]

We will take a look

[willow cosmos]

Hi guys, im having trouble getting the create tab in the web app to load anything at all. ive tried different browsers and devices. any idea what im doing wrong?

[broken lotus]

I am guessing...

There is a recent 1-time popup to agree to not use copyright lyrics....
it is attached to Custom Mode on the Create page.
maybe your browser has blocked the popup, and the Create window won't load without clicking agree...?

[granite helm]

Hello, is it possible to enter a chord progression so that the AI uses it for a song?

I saw someone who could do it

[runic dock]

#🐣┃suno-showcase message

I ve seen this. Dont know how it works

[thin axle]

yea, that example definitely did not follow her chord progression lol BUT can imagine (and hope) that in the future we'll get some harder parameter controls like chords, keys, stems, tempo, etc.

[runic dock]

Yes. This is one of the Highly desired features. Hope we'll have it soon.

[fallow gull]

It doesn't work. People have put chord progressions in songs, or try to specify the key, but it is more of an accident if the song follows those instructions.
Suno does seem to understand "minor" as in a minor key though.

[empty jungle]

question, can you use the lyrics box in custom mode to drive composition and notes played in an instrumetal track? I did this and got a pretty cohesive track that sounds like it follows the directions

[broken lotus]

sometimes my lyrics don't sing but the instrumental seems better for having a bunch of lyric tokens to make the music.

I think the real question is not whether you CAN do it, but how well it followed those specific instructions

(or if it did something random but still more interesting than nothing)

[empty jungle]

yeah, it may have been a happy accident. The 2nd option track generated sung most of the instructions to comedic effect

[blissful raven]

Thoughts on Claude -3 lyrics potential @broken lotus

It actually can critical analyze what your lyrics mean and give a really good description verse by verse aswell

I haven't asked it yet to write me a song yet but it's really better than gpt4

[noble harbor]

Is there a way to make an intro to a part? Basically “Continue From This Song”, but have it start before the clip instead of after

[versed valley]

Welcome! For help, check out Suno Help

[neat nexus]

Not currently, though there is a rumor that such a feature may one day be available.

[noble harbor]

Good to know, thanks

[warm oak]

Is Suno down? I've been waiting for 10 minutes for my tracks.

[dusk kettle]

I think it may have went down temporarily cause mine have been loading for the last 20 minutes or so and looks like some are just now loading in.

[warm oak]

Good to know. Thanks!

[dusk kettle]

And the ones that loaded in and didn't disappear from my screen are not close to the track I'm continuing from. >.>;

[warm oak]

Now mine have disappeared.

[dusk kettle]

Yeah some of mine disappeared but when I refreshed my browser some came back. lol

[warm oak]

I'm just starting a track so maybe they'll appear later. My previous attempt went off the rails.

[dusk kettle]

Yeah they should start to reappear if you refresh your browser cause mine started to.

[warm oak]

They did reappear. Not sounding too good. A lot of harshness in the track.

[dusk kettle]

Yep. I had to do another couple generations again to get something good.

[sleek shuttle]

where are the new meta tags for V3?

[north arrow]

Hey guys. Is there any way to prompt for a consistent voice when generating music? I have not tried this on Discord, but I have tried on the Suno website. I have only found that I can get the same voice in a song, and in that way manipulate myself to getting a consistent "artist" so to say in that way, but not the greatest results. Would be awesome if we could create our own voices with Bark and use them with Chirp

[low cloak]

Ok so I made a 2 minute track with Suno last fall. Now with v3, the sound quality is noticably better. Is there a way to run chirps generated in v2 through v3, like generate a clip using the same seed and prompts, to make a better quality version of it etc.?

[cold zenith]

continue the song V2 but put V3 and get the whole song

[low cloak]

That won't help in this case, I need a rearrangement or mix of what already is there, not a new song; I've already made a music video that matches the beats and tonality of the one generated from v2 chirps

[cold zenith]

What you need to do is take your old V2 music. Click on continue and validate the V3 option, putting in just one instrumental for example, and if you get an error it's no big deal. This won't change your music, but it will put it in V3, improving quality, etc. It's the same thing I told you to do, but I guess I didn't express myself well (I'm not a native English speaker).

[balmy moth]

Greetings, is this the right place to ask questions about the generation of suno? I want to generate an instrumental song, but when I press "continue from thi song" it only generates about 5 seconds of silent audio. Is there any way to work around this?

[runic dock]

I found Suno truly impressive when executing some guitar solos. I have just tried the wha-wha effect by just adding wha-wha in the style. The result is awesome. Thanks to suno devs. Unfortunetaly Suno went on singlng something else after few minutes of silence at the end. This is still an issue. Check it here
https://app.suno.ai/song/94218f2a-64e8-48ae-9d75-ede197613985/

[low cloak]

OK that clears it up! Thank you!

[daring lava]

Hey there 👋 thanks for reaching out! This is a great channel for such questions. If you’d like to post what you are working on, it’d be easier to figure out what’s going on

[eager topaz]

It happens and it's very hit and miss, sadly. I haven't found a workaround, just re-generate and use more credits 😦

[warm coral]

I had a similar problem, it's replicable if the first part had lyrics and you remove them from part 2 or click instrumental. It just fades out and stops.

Sadly it's wasted credits but only thing you can do is regenerate and put in more credits

[meager stump]

its worth it in the end tho cause u make such a banger

[dusk vine]

Anyone have any tips on how best to break up lyrics for a continuation? I am using Verse/Chrous metatags. V3.

[dusk vine]

Another question. I usually use custom lyrics, but when trying with the simple prompt it included the prompt text in the lyrics of the song!? Is that normal? In general Suno is being a bit strange today. Getting back 20-30 second clips at max.

[broken lotus]

Simple is a different kind of prompt, I think... you are not describing the music and style, but the story and mood.

Let Simple pick the style.

[sleek shuttle]

where are the updated meta tags for v3?

like [chorus] etc

its not updated in the docs

[broken lotus]

have you found new ones?

[sleek shuttle]

i saw a youtuber use some new ones

[broken lotus]

If you remeber what they said, let me know.

[eager topaz]

This is all I have. Some people use their own:

Rhythm: [Syncopated] - Controls the beat's pattern and pace, bringing rhythm to life.

Melody:
[Ascending] - Shapes the sequence of notes, forming the song's main theme.

Harmony: [Gm] - Underpins the melody with chordal support, adding depth.

Dynamics: [Crescendo] - Adjusts the music's volume, from soft whispers to powerful crescendos.

Texture: [Polyphonic] - Layers multiple sounds, creating rich musical landscapes.

Timbre: [Warm] - Colors the sound, defining the unique quality of each voice or instrument.

Voice: [John] [Chorus] - Specifies vocal characteristics or identities for song sections.

Outro: [Fade Out 0.5] [Song End] - Concludes the song, guiding its closure with a specified effect.

[intro, short instrumental, build, verse]

[verse, release]

[chorus, build]

[verse, build]

[short instrumental]

[virtuoso finger picking guitar]

[virtuoso guitar]

[thin axle]

Whatever fixes were added, now the newly generated are 100% hidden whenever another file is in the "continue from" section

[sleek shuttle]

thaank you

[torpid mason]

Can I run chirp locally on my pc?

[final hazel]

Unfortunately, Chirp is not available for local installation on personal computers. It operates on Suno's servers and is integrated with their web application and Discord bot. This design choice is likely due to the complex computational requirements and the need for constant updates from the Suno team.

[torpid mason]

Thanks for the information. Is there any API available for Chirp?

[final hazel]

Not yet.. in the future maybe.. it's on the roadmap, but that is all I know

[torpid mason]

All right, thanks

[warped orchid]

Hey guys, having some challenges getting a short intro build up at the beginning of a song, tried v2 and v3 and including [Intro] on line 1. Pretty new to this, what am I missing, I see lots of songs on the Sumo site with instrumental build ups before vocals

[final hazel]

Have you already seen the community wiki page? https://www.suno.wiki/ @warped orchid

[warped orchid]

No, thank you, let me take a look

[light ether]

Sorry, this should be fixed soon but hasn't yet

[noble apex]

hey anyone else having issues with v3 when it comes to getting the full song after u extend it ?

[thin axle]

Thanks! currently having to clear out previous track now (forgetting timestamp) to get back to the list of tracks it just made 🙂

Is this the best channel for comms?

[noble apex]

this work in the lyrics section or the style of music section?

[light ether]

you can try #🪲┃bugs-web as well for UI / other bugs

[light ether]

We just shipped a tiny bugfix - if you refresh, does that continue form look better?

[thin axle]

Yes! So much better 🙂 This was after refresh, so no settings or active-use, but at least all 3 columns are visible now while using "Continue From" 🙂 thank you! Great response time too! Please thank your dev team from me 🙂

[broken lotus]

It's much better!

[fallow gull]

Love this feature showing the chain of your final song.

[lofty thunder]

Looks like image generation might be down from the web interface. Last several songs haven't generated a thunbnail.

[slate epoch]

For awhile now.

[torn geyser]

I have the same problem. Can't generate anything at all

[mint owl]

i made a 9 part song about 24 hrs ago but wasn't able to get the whole song, just keeps loading and then says "One of the clips got stuck" (cant get the error to appear now so can't remember exactly what it said)
thought i'd wait and see if it works later on but still not working 😭 i have a feeling its because i trashed the very first part and started working from Part 2, somehow its delinked? cant see a part 1 highlighted anywhere even though i never perma delete anything in my trash
i've tried restoring everything from the trash, combining the song from the trash itself, i tried remaking the song but it just doesn't sound the same lol
any ideas?

[untold briar]

Could be a bug with the trash... hmn

[mint owl]

im pretty sure i've found what i had used as the part 1 but its not even highlighting that its linked

so sad because i've had the way the song sounds stuck in my head all day so i'm so anxious to put it together LOL

i'll wait for the error message to appear then submit it as a bug i guess