📚┃suno-school

[daring lava]

These are soooo good!!

[neat nexus]

Yeah but complete beginner's luck.

[daring lava]

But the results

[neat nexus]

This is a bit of a HOWTO I threw together for what I did:

#🐣┃suno-showcase message

[untold briar]

btw, I'd be very interested in songs that transition from music to live comedy, or vice versa. I've had trouble getting that result

[neat nexus]

I had a few that did this on me but none that did it deliberately because I was trying to avoid that sort of thing.

[untold briar]

A few went from a jazz club vibe to comedy and back (and the jazz had a live club feeling, like the comedy), but neither the jazz nor the comedy was high quality. I was ideally shooting for like, comedy and then BOOM somebody bursts into song, or breaks out a saxophone, and then a line of comedy. Oh I did have one clip that was a bit like a saxophone performing comedy, with very weird "laugh" sounds that were not crowd sounds, but not music. So maybe it is possib.e

[royal cave]

My song switch language (or gibberish) when i continue a song. Any tips?
The first sound, he reads it clearly.

[daring lava]

I'm working on doing a star trek captain's log

[neat nexus]

https://tenor.com/view/steely-dan-gif-14267876810067722943

Unfortunately that's all that you can do. Try again.

[daring lava]

Not quite but will come back to this

[versed valley]

Welcome! For help, check out Suno Help

[royal cave]

Well, I'm going to burn all my credits, but after 20 attempts, it's still the same gibberish 😦

[neat nexus]

How are you doing the song?

Are you adjusting the lyrics as the song progresses?

[untold briar]

Try a blank prompt, or a few words (spoken, recording, field) here and there, and lyrics like a script.
https://app.suno.ai/song/dd07fe2f-2e80-4518-a19d-d3f3ad48bbee/
https://app.suno.ai/song/46c9208f-4589-4b13-9b6b-15623d767029/
https://app.suno.ai/song/480507f3-c79c-4054-9d57-db593297cc46/

Keep it simple, chirp delivers. Or make it complicated. Sometimes that works too.

[low remnant]

I went all the way to Kyle for the male and still got female at least half the time. Tried Lita for the female and Ozzy for the male with no luck also LOL

[royal cave]

it cut off in the middle of a verse, so I continued the text without the tag. Then I tried putting a verse tag back in. I tried with "intro", tried to add a empty line

[neat nexus]

This is the routine I follow:

Try depending on where you are in your lyrics (if you have a lot more to go) just continuing and running the same lyrics you have. This can sometimes make it pick up where it left off. A large percentage of the time it will start over in the middle of the song. This is just the easiest thing to try to see if it works.

Next try removing everything already sung but the last good line, then try continuing. This sometimes works. It's how it always seemed to work in v1 Chirp for me.

After this try creating a line break (empty space) and continuing from the line that was interrupted. This has a good chance of working.

Then starting with the line that was broken. It sometimes will "remember" where it was and pick up from there.

Finally when all else fails, batten down the hatches, remove everything up to the word that was interrupted and hope for the best and throw the dice. You will likely encounter some weirdness and you will need to keep trying. But if everything else has failed it's this or...

...or you go back up a chain in the song and try again giving up on the good part. It hurts, I know, I know. But sometimes it's the only way forward.

...As you can probably tell this is a bit of a pain point for me and I've done these often...

Don't guarantee these will work but it is the process I follow.

[royal cave]

Great, I'll give it a try, thanks

[untold briar]

@daring lava You might appreciate this 25 minute monster (be sure to watch video, unspoken prompts are content too), the initial clip is simple mode injection, with I worked into the theme (for the first 8 minutes, after that it's just chirp science) https://app.suno.ai/song/1b38239f-0149-4936-b93a-6d8511fb0c4c/

[neat nexus]

Do you mind if I borrow your text for an episode on How is it Manifested?

!chron We choose to go to the Moon. We choose to go to the Moon... We choose to go to.. In a world where everything is plain. We choose to go to the Moon. We choose to go to the Moon... We choose to go to.. There's a magical toaster that's quite insane. We choose to go to the Moon. We choose to go to the Moon... We choose to go to.. It's made of apples, shiny red and green.

[untold briar]

Sure. And if you need any pictures of toasters made of apples, I will never run out.

[neat nexus]

!chron is the command for the pesudo History channel prompter and I'm really curious to see what it would do.

[untold briar]

Honest to god, sometimes I used that test prompt (and in the style) to recover a bad quality degrading chirp song and it somehow resets it (while changing the music style). I feel like it has some weird magic, but it's probably just because it is using a billion brackets or something mundane.

[neat nexus]

I dunno "We choose to go to the moon" seems like a phrase that resonates.

[untold briar]

Oh yeah, that's true, in fact that's where the phrase came from. It was a style prompt, because it's one of the most sampled EDM audio clips, from some stat I saw.

In Chirp V1, with longer style rompts, using a longer speech excerpt, to try and get a style like either a speech, or Apollo astronaut chatter

But in V2, only a tiny bit fit, so it became just that

[neat nexus]

Well I've got to go now for a bit. See you all tonight.

[daring lava]

Wow! Thanks for sharing! Definitely very appreciated! Will check this out tonight 🙂

[thin flint]

As a temparary workaround, use Narrator to sign in. The sign in modal is accessible with Narrator. Once you sign in you can switch back to Jaws etc.

[final grail]

hi! I've just subscribed to pro plan, but it still says I'm out of credit and go manage my account and under /info it says that I'm on basic plan

[fresh hemlock]

i'm trying to reduce the volume of the voices with prompt commande Voices: 0 db, voices: low level, voices: muted with no success, someone have successful with voice level adjustment, and to remove the chorus on the voices?

[wanton wigeon]

Wadadinkadatchucklenuts?

this gibberish prevents the song from generating due to it being copyrighted

[tight zephyr]

Can you explain this one for other folks wanting to try that?

[rotund beacon]

3

[brave gull]

trying to get something that sounds like this https://www.youtube.com/watch?v=XgD-3JBGvt4

any tips regarding style?

[polar arrow]

/ؤاهقح

[versed valley]

Welcome! For help, check out Suno Help

[polar arrow]

/Add me a melody to this song

[thorn parrot]

Suno website gives error. In the suno website when i click to create a song i am taking the message "An error occured while generating the clip" . I tried many times and always get the same error today

[broken lotus]

I am experiencing the same, since about 20min

(wasn't checking before that)

[thorn parrot]

Exactly the same

[low remnant]

I'm getting mine through. Not that there's no problem, but some generations must still be working.

[broken lotus]

slow indusbubblegum, lively synthpads, melodic electricbass, dreamy female vocal

I misspelled 'industrial bubblegum' (my copy/paste didn't fit) 😓

Gave me a hook that is all hallucination 🥺 the syllables of my hook fit, but that's not what they sing.... and the hook is super-catchy.

https://app.suno.ai/song/d34c4a15-0e59-41e9-a552-369c12eb9261

🎶 'Lahsie-Las... Al-pistah!'

[neat nexus]

Pretend that it's exactly what you intended it to be and deny anything else if anyone asks.

Hmmm... [joyful hook] 📝

[flint light]

does anyone know how to fix th credits refunded issue? i have 4 segment working fine, the 5th just keeps failing

[broken lotus]

the problem might be at the end of your 4th segment. It sometimes ends the song, and adds 1 or 2sec of dead air at the end of the clip.

Almost impossible to continue from that clip, because it continues dead air, then cuts off (a known issue, that's why we usually get automatic refunds for very short clips)

unfortunately no way to workaround it (yet). Have to go back to clip 3, and generate a new clip 4.... or wait for a comming soon feature that should allow us to continue from a specific time inside a clip.... (may be several weeks before that becomes available, Devs have discussed it for a while, only recently I saw an interface screengrab of it, so it is in development)

[flint light]

thank you i might have to continue from clip 3

wiill that ruin the current combination from 1-3 thoug?

[broken lotus]

no... it is like choosing a different branch in a 'WhichWay' story.... the path still leads back to the beginning clip..., but from here on your song will be different

(no way to combine branches, or skip around within different branches..., well, we can edit in a DAW)

in other words, 1-3 are the same

[flint light]

ok great thank you, thats good enough for me for now. really hope suno continues to update

[terse cape]

Hi guys, I have a question: if I mistype some lyrics in my chirp message, can I edit them or recreate the clip with the edit worked in? Or do I have to basically create another clip? Can I tell, the bot "what I meant to say" ?

(sorry if this has been asked)

[slate epoch]

New one.

[terse cape]

Alright, thank you:)

[lone abyss]

yo

whassup guys

Can anyone tell me a way to make the AI use a woman singer?

[neat nexus]

Have you tried female vocals in the style prompt?

[lone abyss]

i've tried woman singing before

i will try that

btw

There is a way to make the IA "say" something before the music starts?

like a intro

[neat nexus]

Like an [Announcer] maybe?

[lone abyss]

like

sometimes

when the music will start

(mostly rap)

the singer say something

like when eminem will start the music, you know?

"one shot, one opportunity"

[neat nexus]

Try:

One shot, one opportunity...
[Intro]

[funky beats]
[horn sample]

[Verse]
XXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXX

Or something like that?

[lone abyss]

ok

give me a sec

will try it

[low remnant]

I think I had some luck with [Spoken] as well. Can't think of any examples to search for in my creations though.

[lone abyss]

i will try that

the cat example worked

kinda good

[broken lotus]

What happens if we Continue...
and then throw part 1 in the trash...
and then try to get whole song?

[neat nexus]

https://tenor.com/view/que-gif-27530657

[untold briar]

Continue 11 credit refunded 3 clips clips to make a 33 second song. Delete every even numbered clip. Now you have interleaved gaps every 3 seconds in the history. Now continue clip 11.

[low remnant]

Bard was just making up some tags on this one, but I put them in anyway. (Trumpet Fanfare) was interesting when it wasn't spoken. Also [Playful Vocal Improvisation]. Not sure how repeatable any of it is, but if anyone is interested and wants to test them out further. [Funky Bass Groove] only seemed to do anything when it randomly screamed "Funky Bass Groove!" mid song and made me laugh. https://app.suno.ai/song/0f8c4ed0-9bd8-47cf-a8f4-7b75dc04a624/

[neat lake]

Hi guys, please understand it's not my intention to spread negativity, but what has happened to the quality of music recently? I used to make clips and would have to decide which one to continue, now if I manage to generate something of good sound quality then it's a success and it takes ages. Anyone can explain what this could be due to? I am mainly using the web app btw

I have two clips to compare, both generated from the same place, but one was from a month ago and one from today

When I say quality, I strictly mean the volume and how clear it sounds, I'm sure most people can hear the difference

[untold briar]

As far I know everything is 100% the same as one month ago. My guess is that what's happening is random variation, but maybe not in a way that is obvious, because the type of music and the quality depends a lot of the song (the prompts, and the history) you are working on.

If you are continuing from the same specific clip, there is a smaller range of what kind of music you will out of the AI compared to a fresh song. This is especially true when doing a continued clip, which is highly dependend on the clip it's continuing from. If for pure random chance the previous clip is trending things into the not-so-great-range, it feels like 'every' song you make is not-so-great.

The same dynamic applies (though less so) at the prompt level even on a fresh song. If you have a style and lyrics prompt in mind, and keep retrying it, it could feel like all the songs just aren't hitting now. It's not like the prompt is bad, it's just random chance that that particular prompt tends to yield less great results.

[neat lake]

That makes sense, but then how do I preserve my credits on creativity, when a lot of them go into the trash? Sadly that's how it seems to me recently

[untold briar]

But if for example you generated 10 fresh songs with 10 completely different prompts (style and lyrics) you would get a range of quality over those 10 songs that is the same now as it was one month ago. But when we use Suno Music, we rarely use it like that. Typically people are working on a song or at last a specific concept for a song, maybe making smaller tweaks, but not radically changing everyting every time they generate. So at any particular project, you are in a smaller set of possible outputs, and sometimes that subset is more less satisfactory than others.

[neat lake]

To be fair I generate the same genre, same style and everything time after time until it clicks, but well it doesn't 😄

[untold briar]

The first scenario is the much more significant one. If continuing from one particular clip repeatedly feels like it always is terrible, don't keep throwing good credits after bad. It's better to abandon that branch, and back up a clip, even though it feels bad. In my 25 minute song I had to go back 8 clips once. Ouch.

[untold briar]

Oh yeah, the first you can try is using different descriptive words to describe the same genres. Google musical subgenres of the genre you are looking for, and add those. Add some emotions "angry, sentimental" whatever. Add an instrumnet.

[neat lake]

Thanks, will try to mess around with some characteristics to se if it packs the punch I need

[untold briar]

I look at @broken lotus 's style prompts. cinematic seduction, melodic bass, sultry percussion, slow synthpads, heat 😓 You can feel the heat coming off the textrea wow https://app.suno.ai/song/553f6453-bdb8-4b45-9347-4f5d27659743/

Getting stuck on one clip and throwing a billion credits at it when every continue is terrible is by far the quickest way to burn all your credits. Even I have done it and I constantly tell people to never do it. You just have to remember you can make more good clips to continue from, the one you keep trying to work with isn't the one good one you will get. Also in the future you may be able to continue much more effectively with more controls.

[broken lotus]

I just spent 10min trying to figure out how you could see my prompt 😅

https://app.suno.ai/song/698c5c22-a9d6-461f-8bff-12ffda895329

[untold briar]

I should have checked where it was posted, it was the first example in Discord search and I was like GREAT ONE I'll use that without thinking

I use sultry a LOT, that is suprisingly good in almost any style really

[broken lotus]

🤣 all good.... but I was literally like 'woah, he is sooo accurate...'
hey waitta minute.... 🤪

[broken lotus]

and 'sparse' ...

'Sultry and Sparse' is my favorite female cop show from the '80s

[bronze fiber]

Quick question, I have Chinese lyrics, but Suno generate Cantonese singing not Chinese singing, how should I resolve this issue?

Is there any kind tech expert can help?

[willow junco]

Is there like a wiki page to help guide sudo in isntructions as "[something]" in the custom lyrics? Most things I add in there simply get ignored. What words or formats work??

[broken lotus]

Here's a community guide for using metatags.

https://www.notion.so/Metatags-22a69f63b1814aa9afbec9d21f7b55c4

[fathom sedge]

What is the [end] label for? More than 100 attempts and not a single trigger with the completion of the composition. At the same time, he makes endings in arbitrary places at his discretion (maybe even in the first piece, and after that it's useless to try to generate continuations, it turns out to be a mess).

Why does the neural network choose a duration of exactly 1 minute, if it is obvious that it does not have time to place the entire vocal, and the maximum duration of a piece can be 80 seconds?

[rare forum]

i generated a song in suno website with my lyrics, the music and the vocals came out great. i want to replace a word in the lyrics and generate the song again without changing the music or the voice. Can anyone please tell me how to do it
Thank you

[eager topaz]

You are not able to do that in Suno.
The way I do word or letter replacement is to rip the song to separate tracks then using RipX I can edit the vocal track. I can't add an extra word, but can only use a word that is already recorded somewhere else in the song. I then copy that word and replace the word that I don't want.

[rare forum]

Thank you

[low remnant]

If it was a continuation track, continuing again will usually be similar. But the first segment of each song is golden and impossible to get again. You could try continuing the track you like, put a [Break] in and then the correct words, then edit out the original first segment manually with an editor app when you finish the song. I've never tried that myself but some of the heroes around here have reported success with it. Personally, I just accept a "wrong" word in the first segment unless it's so bad that starting from scratch is the only way I can live with it.

[inland berry]

Hi. Is it possible to preset the language for lyrics? Automatic language recognition works incorrectly - first, there is a part with a very scary accent, and then gradually the neural network catches the desired language and adjusts. I tried to use tags like [ukrainian] but it does not work

And the same for emphasis - I tried to use the standard symbol ˈ then tried capital letter - it did not work, and the AI still emphasized in the wrong places

[untold briar]

In Russian some users reported accents worked better in Chirp than capitals for stress, I don't know if this applies to other languages, but may be worth a shot. For example:
сто́ит(costs) and стои́т(stand)

[inland berry]

sometimes it works (accent) but sometimes not. Thanks anyway for the answer

[minor forum]

well, i always use language in style

for example if i want trap

in russian

i will write russian trap

if your lyrics still do not work, you can use english as well, for example

if i want to use letter Š

and the ai says S instead of Š

you can write it as SH

[inland berry]

will try tomorrow 🙂 thanks

[minor forum]

np, hope it works

[fallen radish]

Has anyone found a more-or-less reliable way to get a spoken word intro countin... as in something like:

[Intro, Spoken Word]
One
Two
One Two Three Four

[Verse]

I've tried all sorts of things that used to work like putting [Spoken] and [Spoken Word] in front of lyrics and none of these seem to work anymore

[flint light]

are weoing to be receiving an update soon on the next build?

[broken lotus]

we are getting some quality-of-life improvements and features soon... (we got a Trash for unwanted clips this week, as example)

on the 'soon soon' roadmap (weeks?) are also the public and explore section of the website, renaming clips, possibly an early version of 'playlists' or 'radio' (not sure the difference)

and Continue From Clip Time...

v3 is being trained, but is still some time (months?) away...

[tender helm]

That 'Continue from clip time' would alone be a godsend feature, so many times I had the perfect clip ending up being ruined in the very last second

[neat nexus]

I never tried this and now I'm sorry that I seem to have missed out on it. Dang.

Let us know if you find something that works, ok?

That's a classic opener that you'd think ought to work.

[low remnant]

https://app.suno.ai/song/f468047f-7f25-42a8-acbb-51e6bd2ad208/ It's really hit or miss. But that was on the 3rd generation.

[amber notch]

How to make phonk music

[broken lotus]

step 1 – type 'phonk' into the search
step 2 - listen to what other people have tried
step 3 - steal the good prompts and go on from there

[neat nexus]

step 4 - show off your demi-god abilities by posting what you did here so those who come after you will know how to do it too and bless your username

[shadow timber]

Have you tried putting it in parentheses, like [spoken count in] (one… two.. one two three four)? I wonder if that would make it less likely to sing the count in

[fallen radish]

I've tried a bunch of variations like that... it seems to ignore all of it

[flint light]

think some very basic functions like being able to continue at a specific time stamp, create voice templates so that you can utilize the same voice consistatly across multiple songs, being able to consistently use prompts like into, instrumental, and layering ( layer1 verse, layer 2 background singer, layer 3, instrumental)

[broken lotus]

I am not a dev. I am just repeating what I heard from a dev.

When I say 'we are getting' I mean us users/subscribers.

[final hazel]

your the "Dev"-il in disguise... 😝 🕺

[tired schooner]

please anyone tell me why song ends in 1:20 sends eventhough i put [End] tag on end of the song

[low remnant]

Because Suno does what it wants. Some day it might be more compliant, but some of us will long for the old days when it was luck.

[soft wyvern]

/chirp

[flint haven]

The /chirp command is not available in this channel

Go to the #🐣┃chirp-beta-1 channel to use the /chirp command, and check out the #🐣┃how-to-chirp channel for step-by-step instructions for making music using Chirp on Discord.

[flint light]

yeah ive noticed suno does what it wants

very annoying to have an AI that doesnt follow instructions

especially when your paying for a premium service

[low remnant]

Not trying to go full a-hole, honestly. But if the service isn't meeting your expectations, you have recourse.

Suno is new, it's pretty unique. And honestly, I think the creativity that it shows helps me to find gems of songs I never would have if it just listened to me. It can certainly be frustrating as heck sometimes, not going to lie. But the service isn't breaking my wallet and it's perfomring something I can't find anywhere else.

[willow junco]

It would be really cool if custom could also read requested chord progressions and bpm.

[neat nexus]

Some people claim that they've been able to tell it chords and it does it often enough that they think there's something there. You'll have to try for yourself of course.

[flint light]

There is no issue with being able to have random creativity and functional control.

[tender helm]

I'd love to have a 'kick the AI' button when it decides to do its own thing again, but I figured thats just the state of the current technology. But if developments goes as fast as with the AI image generation last 2 years then we don't have to worry long about things like that 🙂

[flint light]

definitly im sure the devs are aware of 99% of our suggestions already

and will do whats feasible according to cost

[rapid bluff]

The amount of progress Suno has made since I've joined this server in July, well it was only bark that you could get to make very small 12-15 second songs using the poem command and it was horrendous quality but yet still amazing that it was possible. What they have achieved so far since then I would have thought back then was years away from being possible

[trim iron]

https://chat.openai.com/g/g-ldbohmO2c-compose-me

https://app.suno.ai/song/f1e66804-4994-4209-8828-c7104c0ad065

This song was made using GPT linked. It included tags for chords and the music follows it.

[inner steeple]

Since I couldn't make chirp gen anything from my own prompts... I did what every desperate person failing few dozen times would do.. and snatch the voice from successful clip 😁 so.. I took that squirrel clip and continued from it

[neat nexus]

Can I share this elsewhere? I know a guy who'd love this. He's a big Unreal 5 fan.

[inner steeple]

sure go ahead

it probably doesn't really land here that much, unless there are game devs around

[neat nexus]

Or people who tried and found themselves waaaaay in over their heads.

[misty dust]

Any way to get danceable electro swing? Here's my attempt at a full electro swing song, and yeah (even tho the first pro-chorus is killer to my ears):
https://app.suno.ai/song/51ba86bb-5189-4d8d-9816-6474bd2e5ebc

[slim cobalt]

Hi all I have created 2 parts of a song. I cant or don't have the contination button on them. Why? How do I continue them?

[broken lotus]

if part 1 was made on Discord, part 2 cannot be made on the website...
they are different AI models (until Discord gets the upgrade)

Maybe there is more to your issue, if you will post some more info?

[tender helm]

If it's on the website: Press on part 2 and select 'Get whole song', It will give you a new song with part 1 and 2 combined. That new one you can continue from.

[trim iron]

https://app.suno.ai/song/7f101110-4310-45a7-aa2b-cd92f19484dc

[queen token]

I'm creating a song without lyrics and am creating an outro
at 39sec ...oh man...creepy 😂
https://app.suno.ai/song/17c4b267-db3c-4827-ae4e-b83cd60ad684

btw: is there a resource anywhere with working meta-tags ?

[neat nexus]

These are mostly from the v1 Chirp days, but they should mostly work with the current version on the webapp too. Be sure you select Chirp and not Bark. (By default it is on Bark.)

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

[runic dock]

Is it possible to combine both bark metatags with chirp ones in a song
Let say i want an applause or a laugth inside the lyrics.

[broken lotus]

O...M...G... 😆

[neat nexus]

You can try...just don't be upset if it doesn't work. And accept if you're chasing something not supported that you may spend a lot of credits.

[broken lotus]

do a serch for those words..., they get tried and discussed a lot.

my working theory is they can be triggered in the right context. A few days back everyone was doing 'standup comedy club' and could trigger [laughter] on cue...

But just [applause] at the end of a song almost never works...

but in the right context?
live jazz concert... or something similar

[sour mesa]

Any tips for vocal consistency? Im trying to have same vocal on different songs

[neat nexus]

If you learn anything, definitely share it here. I'd dearly love to be able to do this on purpose.

[fallen radish]

Sigh... all my continuations are ignoring lyrics and spitting out "I'm gonna be your friend" repeated over and over.. then they lyrics seem to start eventually... https://app.suno.ai/song/94bbac32-9b4c-42f2-ac60-cf0944e45d62

[willow junco]

I think it wants to be your friend, but I could be wrong.

[fallen radish]

It's just a sneak prelude to taking over the world... first music, then missiles 🙂

[fallen radish]

Big sigh again.. all forms of [Intro] seem to be causing 100% instrumental clips again

[fallen radish]

That would be cool... preserve the vocalist but not the rest of the style

[amber notch]

can Iuse paypal to buy premium?

F I want to buy premium with paypal...omggg

[eager topaz]

CArd only, I think. I don't see a PayPal option. I would prefer PayPal.

[trim iron]

Add something like [Female] [Male] [Duet] to the song style to force a type of vocals

[trim iron]

https://app.suno.ai/song/94d61d99-252e-4f62-8a5c-4f7ee8f58dbf

[mystic granite]

Anybody else ever have issues with being unable to continue from clips, and having Suno refund credits? I've got a bunch of clips that I'd like to continue from but simply can't, and can't figure out why

And is there any way to force continue? I've got one in particular that I need to be able to continue one more clip from to complete a song, and I don't think it's gonna happen

[willow junco]

I haven't done too many yet but have run into that too. Their instructional suggests to try [Break] to get it out of the weird state maybe try that?

[slow whale]

The flagging is way to much now!

Can't seem to create songs now everything is being flagged, even posts here. Are we living in "woke" world?

[neat nexus]

The automod didn't like your use of the word teats.

[junior lark]

There is a lot of noise in suno's sound quality. It is difficult to release it as an official music source. Is there any solution?

[eager topaz]

It's very difficult to produce a Suno track that is "release ready". I rip the downloaded song to separate tracks with RipX DAW Pro then edit the vocal track, usually the worst track and the most important. I then manually remove every excess piece of noise before finally exporting to my DAW and producing a master. At best, I can mostly produce a demo of the song, with very few % of songs that I would be happy to publish online.

[round briar]

how i can edit the lylics of own song but same music

[willow junco]

Does chirp use facebook's audiocraft musicgen, or just trained on a bigger library, or is it a separate code entirely?

[trim iron]

Recently updated (2024/01/28)
https://chat.openai.com/g/g-ldbohmO2c-compose-me

[tired schooner]

plese tell anyone if i give english lyrics it compltes the full son but wen i give my native language lyrics it only composing 1:20 seconds song eventhough i use [end] tag? any solution?

[manic cradle]

Thats cool. It seems like it ignored a lot of the meta data.

[manic cradle]

What DAW are you using?

[peak yacht]

whats meta data?

[broken lotus]

https://www.notion.so/Metatags-22a69f63b1814aa9afbec9d21f7b55c4

[low remnant]

Ok, anyone ever had it decide after a dozen generations that something wasn't original? And even when you undo the recent edits to something it did before, it keeps saying that? And 100% it is original, not trying hard to duplicate any song style just spearfishing in music land.

[potent snow]

my song has two issues, 1) the chorus is repeated twice, consecutively, I’d like to split them so that one chorus is at the end because its so powerful. how do i do that? i tried remixing, but that just gives me new genres of style. 2) there’s a lot of popping in the background, how can i clean up my song for distribution?

[low remnant]

Nevermind, I don't know what word or phrase it tripped on, but I hit remix on an older one, and whatever it was was gone and I can generate again.

[rigid quail]

Chirp knows how to produce some Techno / Hardstyle sounds if we play enough with the music styles.
But...
When I use metatags such as [Break], [Interlude] in order to "calm down" the fever, it seems it doesn't really works. The "drop" cycle still continues.
If I use only [Break] in the lyrics, it will stops the sound after few seconds then nothing else.
Even while trying to generate more continued tracks in order to have a luck that it respects what I want, it's such complicated for Chirp to understand.

[fallen radish]

Anyone found a style prompt that give's hard-core 70's disco, as in Saturday Night Fever type. I've tried everything that seems obvious and I can't seem to hit the mark

[broken lotus]

I make a lot of disco.... here's a recent one...
https://app.suno.ai/song/c7d428cd-885c-49f2-bd69-7d3e0187139f

sinister disco, muted electric bass, muted disco guitars, synthpads, Diva vocal

[fallen radish]

Thanks, I'll try some of those tags... ok so got some awesome sounding disco... exactly what I was looking for... sadly, it's 1m20s of instrumentals

[eager topaz]

Reaper 7 is my production DAW. RipX DAW I only use for ripping to stems and cleaning up the Suno audio.

[eager topaz]

You will need to edit in a DAW to cut one chorus and paste it where you want it. A bit tricky, but can be done.
The only way to clean up your song is to rip it to stems (separate tracks) then clean up each track in RipX DAW Pro. You can get a free trial.

[eager topaz]

Happens to me often. I just keep trying and it often works fine again. Sometimes I go away and come back the next day and it works o.k. Strange thing!

[trim iron]

Some disco for you
https://app.suno.ai/song/1205c020-3b72-4a3c-88f8-e510f530345c

[runic dock]

Awesome. Thanks @broken lotus . But is there a possibility to have some explanations on what some of the undocumented tags like pre-chorus or disco hook are supposed to do?

[broken lotus]

here is a community guide for metatags....

https://www.notion.so/Metatags-22a69f63b1814aa9afbec9d21f7b55c4

[honest isle]

Love the song, was this all Suno or did you do further edits in a DAW?

[broken lotus]

the Haunted Ballroom disco song...?

That's 100% Suno (my dumb lyrics)

[vague light]

Running into an issue where it keeps wanting to end the song on a fade out when I want it to continue the song. Every attempt to continue the song always keeps generating a 5-10 second long fade out. Any way around this?

[honest isle]

Are there plans to make Suno work in "BARS" instead of time? is this a crazy idea that's just impossible to do? also, folders to keep songs organized would be very helpful.

[broken lotus]

your last clip might have a volume drop at the end, and the Continue clips are continuing that fade....

It weird to get identical fades since [fadeout] is one of the un-reliable things to trigger. A problem ending on the Continue From clip seems like the culprit.

No workaround atm. your options are to go back a generation and re-gen the last clip, hope for a better ending on it...

Coming 'soon' is a feature to Continue from Clip Time... where we can choose a time before the problem fade.... No eta on the feature, but it seems to be in alpha testing rn (expect several weeks at least)

[vague light]

Ah darn alright, so far it's generated 10 fade outs, and no desirable continuations

Sucks because it's one of the coolest sounding things I've generated with Suno

[broken lotus]

when that feature comes out, you'll remember this track and go back to it...

😅

Hopefully by then we'll have a search...

[vague light]

I hope so haha, thanks for the advice though, much appreciated.

It'd be really nice if there was a way to tell it that it did a bad job like a 👎 to train it that it did something wrong and to do something differently next time.

Oh and I think you might be right, there is a very slight dip in volume at the end of the clip

[wary zinc]

Anyone knows, why Suno is broken now? I tried generating some black metal (preferably sombre, melanholic and depressing 😉 and the best it does is some merry hard rock with female vocals. It used to do terrific black metal songs before

Also sound quality is very low know, it sounds like 96 kbps MP3 from 2000.

[graceful condor]

I want to give you a good idea here, in my opinion it would be so good if you could put your voice on suno AI, I think this idea is very good, suno AI would be more successful than it already is around the world 👀

I'll give you another example as an employee like "RVC" where you could train your voice and put it in Suno AI to generate music with your voice or that of another artist in this case👀

[lunar quail]

I had also the same idea.. and tried using RVC over generated Suno songs.. this works in most cases .. even if you hear sometimes digital artifacts.. you can listen to some songs I created in the public gallery https://www.aimastercraft.com/imagegenerator/gallery

[eager topaz]

Getting lots of messages today that clips got stuck while generating, credits refunded. Also lots of clips being generated are music only, no lyrics. I'll try again tomorrow.

[untold briar]

I posted this the other day, and it's probably worth a repost.

As far as I know everything in Suno Music is the same from day to day, there are not small adjustments being made to make things better or worse, even though it often feels like that.

My guess is that what's happening is random variation, but maybe not in a way that is obvious, because the type and quality of music you get depends a lot on the song (the prompts, and the history) you are working on.

If you are continuing from the same specific clip, there is a much smaller range of outputs for the music the AI makes compared to a fresh song. A continued clip is highly dependent on the clip it's continuing from. If by pure random chance the previous clip is trending things into the not-so-great-range, it feels like 'every' song you make is not-so-great that day.

The same dynamic applies (though less so) at the prompt level even on fresh songs. If you have a style and lyrics prompt in mind, and keep retrying that prompt, it may feel like all the songs you make just aren't hitting now. It's not like that prompt is bad - it's ultimately just random chance that that particular prompt/style tends to yield less great results.

Instead, if you had generated 10 completely fresh songs with 10 completely different prompts (style and lyrics) you would get a range of quality over those 10 songs that is consistently the same.. But when we use Suno Music, we rarely use it like that, writing 10 completely fresh prompts every time. Typically people are working on a song, or at least a concept for a song, making smaller tweaks, but they don't radically change the prompt every time they generate. So at any particular project, at any particular moment, you are in a smaller set of possible outputs, and sometimes that subset is more less satisfactory than others. And at those times it feels like Chirp is having a bad day.

The first scenario (continuing from the same specific segment) is the much more significant one. If continuing from one particular clip repeatedly feels like it always is terrible, don't keep throwing good credits after bad. It's better to abandon that branch, and back up a clip, even though it feels bad. In my 25 minute song I had to go back 8 clips at one point, ouch.

[rose ore]

It's frustrating how the strongest musical compositions suno puts out are those 120 second blanks. They're just fire.

[eager topaz]

I've completed around 40 songs with Suno in the past month and almost every song is fire. So many are good commercial songs. I'm processing songs to get the best quality I can, but it's a slow process. I will pitch performers and/or publishers once I've got good quality demos.
I love my playlist of my Suno songs and so do my friends.

[rose ore]

Nice. I'm glad you're enjoying it.

It's a hell of a service

Probably the most creatively satisfying of any tool on the market right now, I think.

[eager topaz]

I haven't found anything better, or anything that comes close to it.

[low remnant]

It seriously feels like a next level compared to where I thought the tech was.

[rose ore]

Yeah, not even close.

[eager topaz]

So impatient to see what Suno can do in 12 months' time!

[vernal meteor]

Sometimes I continue on refunded 5 second clips a couple of times... they might suddenly start back up again

[lucid pulsar]

hello is it possible to continue old songs we've made with chirp? I want to continue an old one (like 2 months ago I would say). It doesn't work on the website. Thanks in advance

[broken lotus]

clips made with v1 (on Discord) are not compatible with v2 (website). You should still be able to continue on Discord for a little while longer (until the Discord version is updated)

[tight zephyr]

Make sure to display the era: https://app.suno.ai/song/524d1857-684a-433f-b748-a422841fe4d3

[broken lotus]

I think the audio quality is sometimes too 'vintage' when I add a retro-decade.
I find it works a bit better by describing the instruments and intended feelings.

(although sometimes sounding 'vintage' is the goal)

[lucid pulsar]

Thanks is it possible to render the full song in the end in discord? What is the command?

[broken lotus]

use the [❤️ 1] or [❤️ 2] button.

[radiant aspen]

How do you control time triggers? For example, Like a conductor conducting.

[wraith vigil]

Same here! It feels like something has been changed with the AI.
The past couple days my generations have been pretty disappointing, I've even went back and re-used Lyrics and Genres that were giving me consistenly decent generations and now they're outputting pretty mediocre clips.

[mental garden]

How does remix work?

[broken lotus]

remix is just a copy/paste of the text prompts...
you'll need to switch on Custom Mode or there is nowhere to paste the prompts to.

[low remnant]

I'd give it a dollar if remix also remembered which segment was the continue from segment instead of making that a two step process.

[broken lotus]

as long as it's the clip that's in the website player, it's relatively painless to do the 2nd click from there....

But if I forget to load it in the player, or the player loads the next clip... 😭

[low remnant]

When I was doing my recent batch of tests on ending a song, I decided to try to "fix" one I had never been happy with. I could easily find the part 4 ending of the song I wanted to replace, but then to find the part 3 it came from was on like page 10 of my library. Lots of page flipping, would have been nice if I could have just hit remix from that last disappointing segment and it remembered. Not a huge deal, but certainly a QoL.

I really need to clean out my junk generations now that it's possible

[broken lotus]

yeah, this...

Just moving a bunch of 'other gens' to the trash would help. 😓

[low remnant]

I have over 61 pages in my library. Not looking forward to purging one song at a time. LOL I appreciate the added feature but cleaning old stuff is never fun. I'm doing a bit better nuking as I generate new stuff.

[light ether]

Definitely feel ya, working on multiselect/multiaction in a future update

[cobalt hornet]

Let's say we want to upload a song to spotify, is there a way we need to save it where our information is embedded into the song? Just signed up today and haven't played around too much with it. Thanks.

[gentle rose]

I have a few questions if someone dont mind helping me.
In a subscription do the left over credits rollover to the flowing month?
And regarding the generated songs, are they a one time generation only or are the same songs being generated again for other users also?

[neat nexus]

Credits do not roll over month to month. They expire at the end of their period. One exception: Extra Credits which are purchased separately and only come into play when the monthly allotment has been used up. These Extra Credits only then become usable and will carry over month to month until used up.

Songs are original music. They are generated for the user based on instructions in the style prompt and in the lyrics used.

[eager topaz]

I use a free app, Mp3tag, where you can enter artist name, album, composer name. This info stays with the MP3 file and acts as the copyright. You can also add an image.

[gentle rose]

Thank you. So even the lyrics generated by Suno are original and the same lyrics are not being also generated for other users?

[neat nexus]

It's AI so anything can happen. But it's not the intended way for it to work.

[broken lotus]

we have a running joke about the lyrics chatGPT will write... (if it's not told to write something more specific).

So no, your lyrics will NOT be original, unless you make an effort to add your own creative instructions.

[bleak tundra]

How to get back to my private room

[rose ore]

Fancy seeing you here. Hi! And they're terrible and tropey, more often than not. It's not so bad if you brainstorm with the ai for a skeleton and work from there.

One thing I found, when running the plagerism checker, is that chatgpt seems to grab lyrics from quora and reddit posts, like little bits of sentances here and there. It's so quirky.

[pastel knot]

Any chance that something analogue to inpainting will be possible in the future? I would love to edit single mispronounced words. The best workaround I have right now is to use a third party vocal remover tool to separate the music from the voice, then recreate the lyrics as a second part of a song and then mix it up manually.

[deft girder]

192kbps now????

i just realized

ive been converting to 128

on accident

[ivory badger]

It's almost impossible to get instrumentals only anymore, anyone having that issue?

[deft girder]

@ivory badger i can still get it pretty easily

try [Instrumental Music] for lyrics or something similar

obviously easier for certain genres

[untold briar]

If it's a continue, that's likely the explanation. If not, can you post a link to a song clip that failed?

[gentle rose]

Hi i have a question regarding styles. If ask chirp to give me lyrics in the style of a particular singer does it do so? Or is it restricted to do that? Also if I may suggest, It would be great if in the future it could be made so Song Key and Bpm could be displayed for the song so we have this information when further editing in DAW, it's a pain to have to figure it out. Or are we able to set that in the prompt?

[gentle rose]

Oh so Suno is using ChatGPT to generate Lyrics? Oh Wow! - Is there a tool to check if they are being plagiarized?

[cobalt hornet]

@slender basin thanks. Is anything else needed? I mean lyrics and such? Or do we just upload the music to these sites?

[violet bramble]

Is there a way to continue generating a song when I liked it, but it wasn't fully generated?

[neat nexus]

Do you mean #⏩┃continuation?

[rose ore]

I would LOVE to see something like that. Also, like there was this song I did last week that was awesome, There was this fantastic piano section in the middle that only came up in the second verse. Would have been nice to juggle that around for the first and third

[rose ore]

Here's the one I use. Only caveat, is use the tool before publishing your lyrics somewhere like bandcamp
https://www.check-plagiarism.com/

[gentle rose]

Thank you. This is very helpful. ☺

[rose ore]

Because I'm a dolt, this one happens to me all the time, lol

[gentle rose]

That would be amazing if they could add that but I would be thrilled with much simpler things like: folders, key and
bpm, bars instead of time, a way to tag a particular vocalist, a dropdown of a list of styles to choose from a drop-down of tags to choose from.

[strange crow]

some example of [Tags] used for putting more intrumental without filling the music with instrumentals?

[urban pollen]

The website will no longer generate vocals I have written in the music it creates. It worked great yesterday, but today only instrumentals are being produced. Any suggestions? Has anyone else experienced this? I wasted by credits for the day.

I reported a bug for each track that did not produce lyrics. Hopefully that helps.

[rose ore]

Question: Does anyone know of a good tutorial on how to assign instrument sounds to midi tracks in Reaper DAW? I'm stumped.

[rose ore]

Yes, sometimes when you over prompt the genre, it goes blank. Try to keep it to three descriptors or less.

[untold briar]

Check chirp-showcase. People use [lots of musical words in brackets] describing solos and instruments, also evocative adjectives, moods, emotions. Or even just a bunch of random un speakable characters like ''''''' and .... to fill out the lyrics. Alternatively if you really over do the "instrumental" style tags, sometimes the clip will remain instrumental no matter what you put in the lyrics field. This is a problem at times, but if you want an instrumental you can write anything there and it can often have an effect. (In the best case, the lyrics can sort of guide the instrumental melody, it plays an instrument almost like it's singing, but I mostly get that with luck.)

[urban pollen]

I literally used three words for the genre and eight short lines of text for the lyrics and no lyrics were produced. Yesterday I entered far more than that and it handled it fine.

[rose ore]

That sounds frustrating. I haven't logged in today yet.

[eager topaz]

It will restrict the input if you try that.
I've had no success entering a Key or BPM either in the lyrics as a Metatag or in the styles.

[eager topaz]

Sorry. I can't tell you as I haven't uploaded to Spotify.

[untold briar]

I tested musical keys and they worked better than random chance, but so unreliably you have to run a sample size and run the numbers to detect it. But there could be prompts that are somewhat better. There is a notation for notes, that is interesting, let me find an example/

[gentle rose]

Oh wow, that's a great tool. Don't be too hard on yourself, that provably happens a lot now that ChatGPT is in the picture. I'm defo checking mine out today. It will really suck if it finds any cause my songs are already finished. I had no idea that ChatGPT was being used.

[untold briar]

Musical Notes. DOES NOT WORK OFTEN. So don't burn credits trying things, unless you specifically are in the mood to do Suno Music experiments.

Hap[C4_1]py[C4_1] birth[D4_2]day[C4_2] to[F4_2] you[E4_4]
Hap[C4_1]py[C4_1] birth[D4_2]day[C4_2] to[G4_2] you[F4_4]
Hap[C4_1]py[C4_1] birth[C5_2]day[A4_2] dear[F4_2] Su[E4_2]no[D4_2]
Hap[Bb4_1]py[Bb4_1] birth[A4_2]day[F4_2] to[G4_2] you[F4_6]

[eager topaz]

Are you talking about adding an instrument VST?
Reaper is my DAW, the best of many I have tried over the years.
These are the best tutorials for Reaper:
https://www.youtube.com/@REAPERMania

[broken lotus]

when I tested the birthday prompt, I got 1 gen where the notes were approximately the correct intervals apart, like they were shifted down the scale, and the tuning mode was like Lydian or something (something wrong)... but the interesting part was the notes sounded roughly the right distance from each other....

(the other gen was not so clearly this).

makes me want to build verses that are more melodic..., whatever the notes just get them moving around...

[untold briar]

I'm interested if using a similar format notation can just make the singing performance a lot different, can you shove in multiple notes in a single word? Been meaning to test just never got to it.

Oh that reminds me. You know the worst possible style prompt? I think I got it.

[broken lotus]

Are ya gonna SHARE or are you protecting us from the forbidden knowledge?

🥹

[untold briar]

What if you want to generate the most useless chirps ever? Long enough to not be refunded, but the music content is always just one single note? Well I got you, "note" in the style field with blank lyrics and POOF go your credits:
https://app.suno.ai/song/2b2c7f98-42ac-4310-ba5d-da769a05cbf0/
https://app.suno.ai/song/ae3b244a-1212-4d0d-986a-c39cb1929874/

[untold briar]

I can't search my library so I had to generate TWO NEW USELESS 'NOTE' CHIRPS!

[gentle rose]

Thanks for the info. That's unfortunate about the Key and Bpm. It would be great if we could set and be able to change the Key at any point of the song. Or be able to set chord progression through out the song.

[Intro D, Dm]
[Verse Bm7, C#7, Em, D]
[Pre Chorus Am, Fm7, D, Cm ]
[Chorus, Bm7, C#7,Em, Dm7]

[untold briar]

Actually maybe as a song ending, you might actually prompt that, theoretically

I didn't even test what it does in a continue actually

Ugh god

[broken lotus]

🤣 I have to pick myself up off the floor. idk why I think that's so funny...

[Dark Suno Humor]

[untold briar]

"note" as a continue, with blank lyrics, doesn't seem any different than just blank lyrics blank style (generally just a fade out)

[gentle rose]

Can't possibly be the best if trying to find how to connect a vst to a track ends up being a question. Lol

[eager topaz]

Lol! I suppose it depends on how many DAWs you've used.
Reaper has a bit of a learning curve, sure, but it doesn't take long to get your head around it. Follow Kenny!

[eager topaz]

VSTs come under the FX button on the track control panel

[gentle rose]

I'm just messing with you. Reaper is a great DAW for the cost, with some great features and I would be all over it if it was as modular as Bitwig is or Reason.

[untold briar]

maybe some occasional intereseting effects. using multiple notes, some seem to apply to instrument, some to vocals, at same time. or sliding around notes a bit. It's not following the notes, but it's changing things I think.
https://app.suno.ai/song/e1d753e4-9357-477e-a745-deaa19e82151/
https://app.suno.ai/song/efab40bb-ae2d-45ae-ae48-f0c0c72d3363/

[hot star]

/chirp

[flint haven]

The /chirp command is not available in this channel

Go to the #🐣┃chirp-beta-1 channel to use the /chirp command, and check out the #🐣┃how-to-chirp channel for step-by-step instructions for making music using Chirp on Discord.

[broken lotus]

that 2nd one.... 🥹

I love when we trying to figure out some technical prompt and it outputs a beauty....

[rose ore]

It's pretty cool. There are a couple others, but they're all pretty comparable. None of them are great, but they're all pretty good. And sometimes, when you're dealing with lyrics that are entirely ai generated, they do find things. So it's good to have something like this around, just in case.

[rose ore]

Yeah, my guy at Sweetwater recommended Reaper, and I'm loving the ui of it. It feels like it should be really easy to take my ripped midis and convert them over, I'm just not quite sure how to do it yet. By default, they don't play at all. Sound registers, like it's playing, but I'm guessing it needs a sample or something to actually render the sound itself, which, oddly, kinda makes sense.

[trim iron]

This song shows some ways of using [meta tags]. The GPT [compose].me was modeled to output specifically for Suno. Any feedback is appreciated.

https://app.suno.ai/song/7f101110-4310-45a7-aa2b-cd92f19484dc

https://chat.openai.com/g/g-ldbohmO2c-compose-me

[eager topaz]

I'm using RipX DAW Pro for ripping to audio. Ripping to MIDI is not very successful and I'm guessing it's because of the quality of the Suno tracks. They are pretty poor when you listen to the ripped audio of each track. I mainly just edit the Vocal track

[rose ore]

I'm also using RipXDaw. So far, it seems to be doing okay okay on guitar and rock and country tracks. It's got weird quirks, like putting harmonica in the vocal track, for example. But it gets really weird on dance tracks. I'm thinking this is less of an issue with suno, and more a problem with the onboard ai. The next one I'm going to experimen with is FL Studio, which seems like a more polished piece of software overall, does the same kind of ai ripping, and seems to be less quirk prone. I'll have to come by and let everybody know how it goes.

[thin flint]

The resulting metatags this GPT produced made some interesting hallucinations. I did line space between the verse and chorus sections, not sure if that matters.
https://app.suno.ai/song/c576d3e7-e030-43f0-8030-bd1fa6abcc0c
https://app.suno.ai/song/17e22f4c-018b-4b37-957c-812441c7d1c3

[trim iron]

i noticed this on the current revision. I am working on revising the structure a bit. the [ ] with the : is hit and miss right now. Try reducing them to a single word and limit the amount of tags to 2

[trim iron]

i should have the update done by the end of the week

[thin flint]

So instead of:
[Verse] [G7] [Piano: Upbeat] [Voice:Female] [Texture:Vibrant]
Change it to
[G7] [Piano: Upbeat] [Voice:Female]?

The brackets I'm sure have to be in there somewhere.

[eager topaz]

I look forward to the results from your experiment with FL studio. I tried in RIPX with one of my self-produced ambient audio tracks, ripping to stems and then exporting to MIDI and it did an o.k. job, better than with Suno, but still far from what I'd call perfect.
But the only thing I use RipX for is to edit and clean up the vocal tracks from Suno and it does that well.

[rose ore]

It's really nice for editing out hallucinated vocals, especially at the beginning of tracks.

[thin flint]

This is why I prefer to have Suno generate an instrumental lead-in to vocals if I can.

[eager topaz]

[G7] [Piano: Upbeat] [Voice:Female]
That gave me 2 clips in the key of A. One was a very Latin beat with female singing in foreign language. The second clip was a piano ballad, female vocal singing in English

[eager topaz]

That's true. It's also very good at copying one word or even 1 letter and replacing a poor one. Great for getting rid of background hiss and noisy breaths.

[eager topaz]

Yes, I do it often, even if I put in some words that I can delete after ripping. Oooohs and aaahs, mmmm, I use them all and take them out if I don't want them, but I have the instrumental intro.

[rose ore]

See, I actually really like the fact that suno is breathy. When you make music with vocaloids, they don't breathe, and it feels uncanny, but in a different way.

[eager topaz]

It's more the background hiss and distortion that is the major problem. I don't like Vocaloid robots!

[rose ore]

I do. The new generation with the more realistic voices is nice. What I don't like about them is all the rules, and the whole concept of "canon characters" which is bullshit.

[radiant aspen]

Is there a comprehensive meta tags list and syntax? (aside from the sparse one on the suno site)

[tired schooner]

please anyone tell me why song ends in 1:20 sends eventhough i put [End] tag on end of the song

[jovial narwhal]

how can i make a song that is longer then 1min?

[tired schooner]

how i get more sharp clear realistic voice is there any particular word or tag?

[graceful condor]

is there any possibility to get a WAV Quality of the music that i produced?

[broken lotus]

https://www.notion.so/Metatags-22a69f63b1814aa9afbec9d21f7b55c4

[thin flint]

I think that whole cannon thing is the fandom coming up with stories and lore because anime culture.

[rose ore]

Yeah, but it's built into the licensing

[rose ore]

Oh that's handy. There are also a few unreliable ones that are handy like, [pause] and [keyof] that work sometimes

[broken lotus]

Ooh! haven't tried Pause...

Did you see the other day when everyone was doing 'Standup Comedy Club' and triggering [laughter] on cue...?

[rose ore]

no! I have to try that

[broken lotus]

(so good)

you could do a LOT with that kind of track... heh heh heh

[rose ore]

The other day I broke the meter calculator and triggered a dinosaur growl. Was quite proud of myself.

[broken lotus]

Whaaaaaaah?

LOL

(starts writing dinosaur love song...)

[rose ore]

Yes! When you break the meter bad enough, the system seems to search for context cues. And that's interesting
https://soundcloud.com/lynncoleandtherobots/the-manual

🙂

[neat nexus]

Chuck Tingle intensifies

[rose ore]

lol

[rose ore]

There should always be at least some chuck tingle in all of us

[neat nexus]

🤣 ....wait...not like that. Not like that!

[rose ore]

lol

[broken lotus]

😂

[rose ore]

I likee the experimental stuff

Need to do more of it

[neat nexus]

Experimental stuff is fun.

[rose ore]

Confused ai's make the best stuff

[dusk seal]

is chirp model available

[fluid oriole]

Hello, I've tried contacting support about a technical issue and haven't gotten a response. I purchased credits on the website but it's telling me on discord I do not have any credits. I've checked to make sure my discord account is linked and it is.

[jovial sundial]

Hi 👋 guys is it possible@to get API access.

I have a fun Valentine usecase I want to explore with Suno/Chirp

[neat nexus]

There is not at this time any news about API access but it is something that the Team has in discussion for the future.

[eager topaz]

The answer is "No". What Suno produces is a low quality MP3 and even converting to a WAV won't improve the quality.

[willow wren]

Quality varies depending on the style. Pop generated me higher quality than Metal. I use opensource Transformer Models and Split into Voice, Bass, Drum and Rest. Predicting MIDI from the non Voice Tracks works well for Bass and Drum, Rest varies. So i can mix and adjust all in Garage Band. Sounds much better then.

[eager topaz]

Yes, you can tweak it. I rip to separate stems (tracks), get rid of all hiss and stuff in the vocals then master it in my DAW. You can improve it, but not get it better than MP3 quality.

[sinful abyss]

still no 100% method to get proper intro?

[light ether]

Hey there, sorry about that - can you DM me your account info (I can't find an account under this particular discord ID)

[light ether]

Working on this one still - not yet

[rose ore]

Interesting suno function that works fairly reliably across both genre and tagging "goes hard"

[raw grotto]

Why I might be getting this error again and again on suno app?
How to avaoid?

[broken lotus]

i see that error when my lyrics are too long.

[final hazel]

Yeah.. you probably used "remix" from a "whole song" you created before. Then it will post the full lyrics which are too long @raw grotto

[placid gorge]

@broken lotus @neat nexus Spam in every channel saying RAID

[neat nexus]

https://tenor.com/view/ban-hammer-futurama-scruffy-gif-20750885

[short drum]

Its impossible to generate songs with the word "europe" in it. great

[fallow gull]

Europe in the song body or the genre? I had some good results with eurovision if that helps.

[graceful condor]

Unfortunately, I have the following problem.
The first part of a song was really great. Unfortunately the 01:20 are in the middle of a chorus. No matter what I enter, either the sentence is not continued, sung again from the beginning or the style changes completely. even if I work with continue.

One more thing: I am currently using the Suno app. Can anyone help me which command I should enter or how I could solve the problem?

[wary matrix]

This happened to me and I just put the same Chorus at the start of the next and it ran with it correctly. Then again, it also did some weird things with it a few other times. 😂

Just make sure it's tagged [Chorus] in each so it knows to pick up where it left off. (I think)

[rustic gale]

I had a lot of moments where 1:20 ended mid sentence OR even mid word.
I just try to catch where it stopped and try to continue from syllable or word which should have been next.

But u still need to generate few/ Cos it not always works out.

Example
This poison, I know I need to beat, <-- ended here
To find a path, to feel complete. <-- what was next, so i started "continue" from here

And if mid words then example:
This poi<-- ended here| started "continue" from here -->
son, I know I need to beat,
To find a path, to feel complete.

even had somehting like this, and it worked.
supposed to be: Partners in a land of bytes but ended midword: Partners i

[graceful condor]

I've tried so many things but none of them were really good. I now have ins which sounds a bit strange with three transitions but is acceptable. somehow the style of the song is a bit different than at the beginning and the chorus also sounds very different.

sounds like a lot of work from both of you

[rustic gale]

Well it is if you want to make "complete" thing XD

[wary matrix]

#📚┃suno-school message

Helpful link for the tags

[rustic gale]

Some songs took me ~1800 tokens to finish XD

[wary matrix]

I'm just trying to figure out how to get similar sounding voices for another song. I got one done up that I really like the vocals for and I want to try and do a follow up but goodness... It's all over the place.

[rustic gale]

Yep. As i said. up to 1800 token used for 1 song sometimes XD

To find what i want and "finish" it

But result are nice, I love it.

example:
https://app.suno.ai/song/e56e37dd-a99e-4ff2-ab9b-abe8a84107cc

I jsut wish it didnt sound like bad compressed mp3 XD. But oh well. AI will improve 🙂

[wary matrix]

That's a lot of tags for the music type. 👀

[rustic gale]

Yep. I play a lot with them, the more you add the more you can pin down genre/style/instruments u need

[wary matrix]

Type O Negative vibes, haha. Cool track!

Yeh. I had a REALLY awesome continuation but the quality got horrible midway through and it never recovered.

[rustic gale]

yeah. Sometime you also get 0 words but AMAZING beat 😦

Not that XD sec.

[wary matrix]

My absolutely mental experiment, lol. If you listen, it's around the solo. It's horrible.
https://app.suno.ai/song/156c9425-4282-40a7-96f6-35603b00f29d

[rustic gale]

This:
https://app.suno.ai/song/ec3f151e-7818-400f-961b-56397f07df66

0 words but beat is SICK XD

[wary matrix]

Definitely a sick beat. I wonder if it's because you didn't use the tags with []

[rustic gale]

possible. But in the end i still was able to produce this:
https://app.suno.ai/song/6562a503-ec84-49f1-9c53-1b4b1e41fddc
Its close to what i was oping so i was fine with that XD

[wary matrix]

Interesting... The tags are so confusing. Like I managed to get Intro Guitar Solo to work one of ten attempts.

Unfortunately the end of it was not great so I had to keep retrying... Alas... Nope

[rustic gale]

will play more with tag as i dont use them almost at all XD I just generate 10 variations each time and listen to all XD Then repeat XD

[eager topaz]

That is a great track. Suno is amazing when it behaves! Even if you use 2000 credits, it's a super cheap song that you can be proud of. Imagine the cost of hiring musicians, singers, studio, etc.

[elfin yoke]

any guidance if i want to create a song with two distinct voices?

[low remnant]

Good luck. It can be done but very low reliability.

put duet in the style. and put names in parenthesis (john) (jane) after the [Verse]/[Chorus] tag. It assigns genders randomly. Sometimes it pronounces the names iwth or without changing the voices.

[shy badge]

I finally found the pages. I want to continue from this generation as it is. But can't

https://app.suno.ai/song/7ace664d-ae66-4207-ada5-2c06734a46af

[exotic fiber]

question for the devs... what do you think about this? any chance to rebuild/migrate chirp from transformers to mamba? apparently its way better than transofmrers including audio generation.

https://www.unite.ai/mamba-redefining-sequence-modeling-and-outforming-transformers-architecture/

https://arxiv.org/abs/2312.00752

[placid gorge]

If im reading right same issue with Heavy Metal

[tired schooner]

When I give English lyrics it finishing entire song. But when I give my own language, it is complete in 1 minute and 20 seconds. Why is this?"

[neat nexus]

Any luck with “You-rope” instead of Europe?

[runic dock]

Just discovering today that one of the requested features has been added to suno :Deleting a song. Great job than to devellopers. But is there any release note ? What are the new features released?

[fallow gull]

"Move to Trash" already exists in the web interface.

[broken lotus]

there was a #🚀┃updates message

[fallow gull]

Is using [Verse #] the same as just [Verse]?

I had some issues with the verse melody starting to diverge and was wondering if it was due to using [Verse 2] instead of just [Verse]

[eager topaz]

I've had the same thoughts, about using verse number or not and to me, there doesn't appear to be any difference in the clips produced, whatever you use.

[cinder blaze]

How many "chirps" can i do on discord? I went to the suno site and it gave me 50 credits and i can only do 5 tries (10 credits each)

[broken lotus]

the free credits reset every 24 hrs.

[cinder blaze]

is there an time exactly? because i used my credits yesterday around 9-12 pm and it hasn't reset yet

but if I use discord, does it consume my credits the same? i few months i remember using chirp (here on discord) an entire day and it didn't tell me that i ran out of credits

[runic dock]

Got it. I am not very familiar with discord. Now i am able to check mini-updates

[runic dock]

According to mini-updates, the feature was added on january 28th. So it is really new.

[vale anvil]

I have a technical question: is there any tool that could make the voices generated in Suno Ai not sound so generic??

[broken lotus]

please don't ask the same question in multiple channels.
You were answered in #💬┃general-chat message .

[sinful abyss]

Is it impossible to get back vocals in "continue" after instrumental track?

[broken lotus]

Very difficult…. the Continue feature listens to the end of the last clip… if there is no voice there it is likely to continue without a voice.

I have gotten it to start singing by spamming
Female singer, female singer, female singer….
But this is not reliable….

(In the style prompt)

[coarse cairn]

lets look at an example

[Verse]
xyz........

"""
<here I want to put some flute voice>
"""

[Chorus]
abc

===================
is there any way to achive it

[broken lotus]

Is ‘flute voice’ the sound of a flute?
Try
[flute]
Or
[melodic flute]
Or
[flute solo]

If ‘flute voice’ is like human voice and flute sound combined, i am not sure it is possible, sometimes I se prompts where an instrument is described then some ‘lyrics’ that is the sound of the instrument…
[flute]
Hooooo hoo hirroooo whoooo

[spark ginkgo]

hi guyz is there a way to make a duet or specify the sex of the singer in the duet ?

[neat nexus]

Please do not spam the Discord with multiple posts in multiple sections asking about the same question. The answer is in the other place you asked. Please consider this your verbal warning.

[spark ginkgo]

sorry

i didnt read that rule

[neat nexus]

It's usually a great idea to read the rules of any Discord you join for situations just like this one.

[spark ginkgo]

yeah it's my bad

also what's the difference between bark and chirp. i didn't understand

[neat nexus]

Bark is a voice and sound effects model and Chirp makes music and vocals. They are two different models. The model for bark appears to be not working here at the moment though you can find implementations on huggingface, since it is opensourced. Chirp is a closed source model only.

I have never used bark IIRC, so my knowledge is limited on it. I only use Chirp.

[coarse cairn]

yep, you can name the character but not very reliable, like

[Bob]
hello, I love to sing like a man

[Alice]
Don't care until you dare

[Bob]
oh, I see you do care

[Alice]
but, if you are there I could spare

[vague light]

Anyone else running into this? If I try to just make an instrumental track, sometimes out of nowhere a voice will shout "Thanks for watching!"

[neat nexus]

Yeah, sadly it's a known "hallucination"

All you can really do is try again and hope it doesn't come back.

[vague light]

It's ruined a bunch of my tracks that otherwise sounded great lol

[neat nexus]

Yeah. It can be super annoying. Ruined my finish in this song.

[vague light]

Oh God lol

Yeah it all sounded great up until that

Super weird bug

The weird part is that its always the same phrase

Also is there a way to prevent it from just creating a 4 second continuation? It keeps doing that for me.

[neat nexus]

::winces::

Ah...did the last part before the continuation have a low volume or break? It can cause Suno to think the song is done and there is nothing that can (currently) be done about it. The Team is working on a tool that will allow a custom resume point which will hopefully fix this. (I have so many parts that I intend to see if I can fix them once this is out...)

[vague light]

It seems kinda random tbh, it'll generate a 4 second continuation alongside a 20 second one sometimes, other times it'll do both of them as 4 seconds

A custom resume point sounds great

I get a ton of them that have a random little volume dip at the end resulting in all continuations being a fade out

[neat nexus]

This is my personal one that I carry bitterness for. It was perfect and then it faded at the end.

[final hazel]

Yeah.. happens to the best of us... I had it too

[vague light]

Oof yeah, sounded great till the fadeout

[gusty raft]

How do I make a song

[neat nexus]

#🐣┃how-to-chirp

[eager topaz]

It's extra work, but ripping the song to separate tracks allows you to edit out the offending voice. I use RipX Daw Pro. It's not cheap, but there is a trial you can use. There are online places where you can do rips, but you don;t have much control for editing.

[fallow gull]

The text to voice is not able to pronounce Ithaca. Comparing to Eleven Labs, it was pronounced perfectly.

[rose ore]

Behold my sly narrative trickery

[thin flint]

Made a song with some new prompt stuff to try. Intro and Outro worked without any regens. https://discordapp.com/channels/1069381916492562582/1083129341870878780/1204255085799149638

How does one link messages from other channels in here without copying the message text? Can I right click on the message and pick an option?

[untold briar]

The 3 dots menu has a "copy message link" option. Oh looks like you got it right.

[thin flint]

Prompt trickery or stitching tracks with DAW trickery?

[thin flint]

Good song 🙂

[rose ore]

I was thinking lyrical trickery. Quite proud of myself.

🙂

[eager topaz]

You've done well! I find that Suno, because it produces such good results is really inspiring. In fact, it's hard to stop! I find myself thinking lyrics all day long, trying to find great hooks.

[rose ore]

Yeah, same. It's crazy. I haven't exercised these muscles in years, and I'm feeling really excited about it

[thin flint]

A bit meh at the beginning, but I like the vibe. Tried to go for a live performance sound. Sounds a bit like what it would sound like if the bar was empty. https://app.suno.ai/song/8231218f-4bf2-40b3-99e5-5dc46197f4d2

[eager topaz]

I'm using words I wrote 30 years ago and didn't do anything with them. Suno makes me sound like a pro 😆

[neat nexus]

I'm sympathetic to this. I keep hunting for my old notebooks with half finished stuff to see if I can finally make them come to life.

[eager topaz]

Right! So inspiring. Thank you Suno team 😀

[neat nexus]

Agreed.
👏 Thank you Suno Team.

[rose ore]

It's totally changed my song writing process. Also, it's a really confident performer

[mortal frigate]

.

[spark ginkgo]

it doesn't work 😦

[peak veldt]

https://tenor.com/view/pitch-perfect-comedy-rebel-wilson-crushed-it-well-done-gif-3459487

Not sure what that sly narrative was all about, but those lyrics resonate. And finding a good Sonu rendering makes it all worth it... Nicely done!

[shadow timber]

These lyrics are actually really good, I could see this as a hit

As someone who never wrote a song in my life two months ago, I have to say Suno is teaching me a lot. About the third time it sang the same lyric awkwardly, I started thinking, hey, how can I make this not awkward? And it went from there

[low remnant]

When the lyrics come together, they REALLY come together. It's like a concentrated flash fiction. You've got like 350 words to tell a compelling story. And keep it in a rhythm with some sort of rhyming scheme. When it hits, BOOM

[rich thunder]

does suno have API , where can i read about it?

[low remnant]

No API at this time. At least no public one. I bet Bing has API access 😛

[neat nexus]

Yeah, well Bing is Bing.

This is all that I know:

We don't have a publicly-available API at this time, but it's on our roadmap.

Please note that it is a direct quote. 😉

[low remnant]

My road rage song yesterday, with just [Spoken Verse] ended with a great Screamed lyric worthy of Suicidal Tendencies.

[dusk pulsar]

this is amazing. are you going to try to blend it with the video?

[broken lotus]

Cinematic Synthout, ...

gives an atmospheric feedback ambient sound (that is not necessarily synths)

they all seem to have this delay/loop quality...

[vague light]

Honestly this is a great idea. Thanks for the suggestion, I do have a Lala.ai account so no problem there.

[untold briar]

Random Thoughts, if differences in Chirp Music checkpoints are like the differences between Bark Coarse small and large, switchign between them is a bit like a style transfer with some sounds

Mainly large to small thoug

[cold zenith]

what's are your best tips guy for having a clean voice without auto tunes or noise ?

[broken lotus]

Dark Surf

'surf' has that great sproing guitar amp string sound...
(but finding it hard to get it to do anything else...)

https://app.suno.ai/song/fa0280cc-aba9-405b-adb3-bdf374fbf2de

[willow blade]

Why is the Voice not clear

it's like bad autotune

😦

[broken lotus]

because it is literally a synthetic voice... Did you think they have a room of actual people singing on demand...?

Also, check your prompts, not all genre sound alike.

Try listening to OTHER PEOPLE'S results in showcase. You might learn some tricks from the ones you like.

[willow blade]

Okay but also if I only get Instrumental

[low remnant]

I think if you upgrade to Pro, the room of people totally sings on demand for you.

[willow blade]

I upgraded already but yeah

[broken lotus]

do you want https://discord.com/channels/1069381916492562582/1166166196467548280 ?

[willow blade]

yeah I tried both but it's not stable enough. I think you need another software to clear it

but its better than others

[broken lotus]

Hard to understand you.

Typically people post to ask for help, We get a little impatient at people who show up just to announce their opinions.

We have opinions too Bro.

[willow blade]

Okay but where can I share my opinion then. It would be nice if there is a feedback channel

but nvm

[shell jungle]

was the site updated? it sounds like v3?

[broken lotus]

https://discord.com/channels/1069381916492562582/1154587937611190354

[broken lotus]

no, but the valentine's day song seems to be an unannounced v3 preview....

[shell jungle]

hm, cuz i am making a song and i got blown away because it sounds better then v2

i can send you the song when its done

[shell jungle]

here.

[broken lotus]

very clean up-front voice... 🥹

'amazing song'...????

😞 I am working too hard on my prompts.... LOL

[shell jungle]

You think its v3?

i just tried https://v-day.suno.ai/ and its v3

[broken lotus]

this is how v2 started... limited prompt..., no way to continue... but a clear jump in sound quality...

leading to v1 ennui 😭 torn between better quality with no control, or (suddenly noticeable) worse quality but where we can get it to do what we want (sometimes)

as @viscid wharf says: 'PUT IT IN A SONG!' 😁

🎵 "torn between two versions..."

[azure saffron]

Wow, I tried this V song and it seems like the quality has significantly improved. At least now I get clear sound without any noise or crackling...from the first attempt.

[thin flint]

I don't know how to submit my information to Suno on this preview page.I hit getting started, next, and that's about it.

[broken lotus]

🤔 Can you try a different web browser?
you don't need to be logged in (but it asks for your email)

maybe somehow your main browser is blocking it loading the next screens?

[thin flint]

I'll try. On Chrome.

[azure saffron]

After experimenting with the potential V3 update, I can share the following observations:

• This version is significantly more stable, with considerably less noise and fewer artifacts. It delivers improved results with fewer attempts.
• There's a noticeable enhancement in composition quality, resulting in more captivating songs.
  Having used Suno extensively since its V2 iteration, my primary concern was the persistent noise, particularly when experimenting with unique styles. This posed a significant challenge.

V3 has refined genres like pop, hip hop, and EDM, which were already quite good. However, there appears to be an issue with overcompression in songs, evident even in "intense styles." This overcompression, especially with a pre-loaded dynamic range, leads to noticeable distortion. Moreover, in genres such as hard rock or pop rock, both male and female high vocals tend to sound distorted, especially in choruses.

But Damn, we are step closer to release ready quality!

P.S. Style and mood Prompt: "I met Victoria at (Death Metal Festival, Tomorrowland EDM, Hip Hop Concert, Jazz Evening etc) and she (pushes me forward, makes me headbang, inspires me, makes me sad etc)

[broken lotus]

this was exactly what I was about to ask... 😆
thank you!

[thin flint]

How do you switch genres etc with this? You only have that form to fill right?

Also how'd you figure out the prompts?

[azure saffron]

I just used my imagination. Also, from experience, I know that the outcome from Suno depends not only on the prompt but also on the lyrics themselves. So, it was logical that all these fields like 'where did you meet' and 'what makes special' would influence the lyrics and the overall genre. I created metal, hip hop, EDM, and a few other genres just by typing different 'meeting locations,' and also got different moods by playing with 'what makes my Valentine's special.

[thin flint]

I got it to work. Interesting the number of line spaces within the lyrics this time...

[azure saffron]

And the duration is longer 🙂

[thin flint]

It is. Wonder if we can do consistent intros once V3 UI is actually out.

And outros for that matter./

[azure saffron]

oh yeah....I suppose instrumentals will also have a better quality. Let's see 🙂

[tired schooner]

can anyone tell me that is V3 gonna coming soon? can we get good voice quality?

[neat nexus]

No idea on timeline. But if you want a preview of what v3 is shaping up to, check out the valentine's day site: https://v-day.suno.ai/

[tired schooner]

ok thanks bro

[viscid wharf]

i do really love how it sounds

[cold zenith]

What are your best tips guyz ?

Please share

I just try v-day this is insane

[knotty kite]

How do we combine all parts of a song on the Suno site?

[near patio]

3 buttons then "create whole song "

[bold glade]

Is there a way to combine part 2 and 3 of a song and discard part 1? Do I have to delete part 1?

[normal patrol]

When can we get the Wav?

[knotty kite]

Hm, no. I don't see that anywhere. Please advise. 😦 That is why I'm asking

[near patio]

Are you doing the three dots on the "part 13"?

[knotty kite]

yes, and many others

[knotty kite]

yeah

[near patio]

That's awfully strange. If you go all the way to part 2, does it show up?

[knotty kite]

No it doesn't show up anywhere. The menu is the same everywhere as the screenshot I provided

[near patio]

That's weird. I don't know. Ask in #💬┃general-chat because they will usually tell you anyway

[knotty kite]

Yeah so it doesn't show up for me but it shows up for you?

[knotty kite]

see I dont have the option to continue from the clips or get the whole songs'

[neat nexus]

Are these made with the webapp or were they originally created with the Discord version of Chirp?

[eager topaz]

Can't get .WAV, only MP3

[knotty kite]

discord

[neat nexus]

V1 Chirp (Discord) is not forward compatible with v2 Chirp (webapp). You can only continue in Discord.

Be warned that eventually the Discord version will be updated and it will be come impossible to stitch songs together from the earlier model without a DAW.

Finish your songs from v1 Chirp and start using the webapp before the change over happens!

[cold zenith]

Do you have tips guyz ?

[eager topaz]

Jump in and start learning! It's great fun, it's cheap and it's excellent.
There is plenty of info on here to get you going and start making some great music. enjoy!

[runic forge]

Use simple mode to get a hang on the styles and emotions to get some ideas. And then go wild in custom mode. And remember that sometimes good things take multiple tries.

[cold zenith]

I'm using it for 3 months by now :), but i want to know if people has some tips to have better results

concerning the auto tunes, noise, and the voices etc

[hollow steeple]

Hi everyone, how to get past "Couldn't generate that." message?

[cold zenith]

maybe change the word that is problematic

to find which one...

[hollow steeple]

Ye that's the problem

I really dont understand whats wrong with my lyric

[broken lotus]

what are you trying now? how do your typical style prompts look?

[cold zenith]

thanks for your help. My prompt are usually like that : " MODERN AGRESSIVE DRILL TRAP DUBSTEP EDM ". I put lyrics and separate them like that : " INTRO VERSE CHORUS VERSE PRE-CHORUS BREAK BRIDGE OUTRO BREAK "

[broken lotus]

MODERN AGRESSIVE DRILL TRAP DUBSTEP EDM

🤔

1. do you ever try commas to separate...?
2. Do you ever try describing the voice...?

I don't know this will 'fix' it, but I do notice (anecdotal) when I don't use commas in a longer style prompt I seem to get 'harsher' sounding results (like the EQ is all over the place, sometimes too hot, sometimes very quiet) – I feel it's a tradeoff for being experimental and 'genre-bashing'....

When I do use commas in a longer style prompt (usually to separate instruments and details) the results seem to be a bit more 'balanced' in the EQ.... with the trade off of sounding a bit more mainstream...

This is also anecdotal, and these aren't my genres so I have no direct experience mixing those....

[graceful condor]

Hi everyone!!! I wanted to ask you if:

1: Is it possible to download WAV instead of Mp3?
2: I hear the audio a little "saturated"... does this happen to you too?

[cold zenith]

No i never put commas or describing the voice except saying it's a female or male voice. Do you have an exemple of what your prompt are ?

[cold zenith]

i don't know

[untold briar]

I agree with your observations generally. And more extreme than not using commas is using entirely made-up words. even more experimental and atypical, and even more of a tradeoff on occasion really harsh results

[broken lotus]

Here's one of mine grabbed at Random...

torch-lounge, melodicbass, dark electro-pop, sophisti-guitars, sultry male vocal

[cold zenith]

i've found out that Crunk music is not in SUNO by the way, i tried a crunk music it never works

[untold briar]

The music probably is there, even if Crunk the word (the prompt) isn't.

[broken lotus]

(I am also making up words, but perhaps not AS experimental as Jonathan's...
😉

[untold briar]

Though honeslty I don't know what Crunk is, I'm assuming it's not SOOO niche, obviously there is a limit I'm sure.

[cold zenith]

hum i understand. U even said how the singer s sing

[neon mantle]

THe system doesnt seem to know a lot of urban genres, so you have to play around with it

[cold zenith]

This is what CIARA used to do at her two first album. Lil Jon, used to do it too

[cold zenith]

yeah.

[cold zenith]

When you put something like " sultry male vocal ", does it makes a lot of autotunes and a lot of noises ?

[untold briar]

Any descriptive word may be worth trying, as (something) (vocals/singer). For example, a lot of these are from the random style generator.
ethereal
operatic
powerful
gritty
silky
soothing
fiery
haunting
rugged
velvety
nostalgic
soulful
whimsical

Not sure which most strongly vocals a lot, versus the music, like sultry seems to.

[broken lotus]

nothing is ever 100% 🤕
but here, you can hear how that prompt turned out...

(lyrics are also a big factor in the pacing and stuff)

https://app.suno.ai/song/6ad95539-154c-4f7d-a882-50189dab3e1f

[cold zenith]

👍

Thank you both of you for helping

[cold zenith]

It sound great

[untold briar]

It uses up credits, but something that often helps is to try generating your lyrics with a blank style prompt. This will tell you what the lyrics themselves are doing to the music (if anything obvious) which if you know about, you may want to prompt for or against it (or even use fake lyrics as the first clip)

If your lyrics are very evocative, sometimes a blank style prompt actually just works great, as well.

[cold zenith]

You mean let SUNO find out the style etc

?

[untold briar]

If there is no style prompt, the song will still be somewhat matched up to the lyrics - what style would the lyrics. Usually this is a small effect and not noticable. But not always. For example if you write Christmas song music, and use no style tags, it will probably often still sound like a christmas carol

If the lyrics look like rap music, the song will tend to be rapped. This used to be a big problem in earlier versions, where any song that had a lot of words looked a little bit like rap, because rap is fast, and has a lot of words.

[cold zenith]

interesting

I didn't noticed it because, i always fill the style part

[untold briar]

Us old timers are used to this, because believe it or not, the style prompt didn't always exist. We used to have to try and prompt music styles with only lyrics.

[cold zenith]

i wasn't there at the beginning, so you saw the evolution

[eager topaz]

Yes, those things are problematic, more so depending on the style. I rip the Suno final song into separate tracks, using RipX DAW Pro. Then I edit, mostly the vocal track, which has the most problems. I edit out all of the noise as much as I can. Sometime the drum track is bad and I cut and paste in my DAW. Sometime I add extra piano when the piano track is not good. It's pretty hard the way that Suno is at present to get a release-ready track, but I try! Hopefully quality will improve.

[cold zenith]

true

[knotty kite]

“Aggressive, high-energy”

[eager topaz]

No, it's not possible to download WAV, only MP3 . Yes, saturation is common, more so depending on the style. Some days is worse than others. Hard to edit it out.

[cold zenith]

I just saw on google what is RipX DAW PRO. i didnt know it

[cold zenith]

really ?

[eager topaz]

It's a fire app, not cheap, but there is a trial available. It's so good that I sometimes take out just a single letter or copy and paste one. Very hard to do that in a normal audio editor. Give it a try!

[cold zenith]

okay i'll try, i saw there is an AI to split the music. You know you can do it with bandlab too ?

there is a free studio on Banlab

[eager topaz]

I've heard that you can rip in Bandlab but I don't know about the editing.

[cold zenith]

you can do everything on it

but as you had buy Rip u should stay on it

if you masterize it more also

[wide jasper]

look at my article, you can split the track and convert to .WAV: https://medium.com/@flencrypto/how-to-split-music-into-different-parts-45555797ccbb

and it open source

[fallow gull]

Does anyone know of any GPTs that produce output similar to the SunoAi lyrics generator?
What I miss most from Chat GPT and some other song writing assistants is the inclusion of vocal ad-libs and embellishments. Like (oooh-yeah) or repeating a key word in the song.

[eager topaz]

It seems that vocalremover.org will only give 2 tracks, vocal on 1 track and musical backing on the other. I see only that option, and no option for .wav file. Am I on the wrong site or did it change?

[cold zenith]

i never understand what the button remix does

someone could explain to me ?

[neat nexus]

I think it was another site that split things further. IIRC they charged for too many splits.

Have you tried this one yet? https://ultimatevocalremover.com/

[wide jasper]

me neither

[eager topaz]

Me either!

[neat nexus]

::sigh::

[neat nexus]

Remix is poorly named, a lot of people feel--for precisely the misunderstanding you're experiencing.

All remix (currently?) does is reprompt the exact same text used in a previous attempt. It does not get back the same style of music or vocals. It is no different than just making a new song wit that text.

[wide jasper]

regenerate

not remix

then

[neat nexus]

I've suggested it be changed. ¯_(ツ)_/¯

Unfortunately that is not something that was considered.

[final hazel]

vocalremover.org will give your more stems if you are a donator on his patreon.

[wide jasper]

oh i see!

https://www.lalal.ai/

Try these guys

[wide jasper]

sorry didnt see this open source one...ignore me

[neat nexus]

More options are good.

[wide jasper]

my new ADHD rap written by me, and took about 1000 credits will be ready soon! its awesome

[wide jasper]

heres a tiny bit of it

i havent edited it yet

[normal patrol]

How did you make a Wav out of the Suno song ？

[tight zephyr]

Allmost of of these vocal removers are based on Meta AI open source Demucs project, here. https://github.com/facebookresearch/demucs

[eager topaz]

You can do it in any DAW. Render as a .WAV

[cold zenith]

yeah totally i don't see the goal of it

do you guyz how to fix the fact that when u click on continue from this clip, it doesn't change the rhytme of the Chorus on the other part ?

[runic dock]

New music with same lyrics.
You dont need to regenerate the lyrics
But agree . 'remix' is somehow a poor naming.

[runic dock]

Personnaly i use to generate several versions till i get one good

[cold zenith]

thx for your reply ^^

[cold zenith]

If you want a good end guyz, dont hesitate to put : " [fade ending] " at the end

( By the way is there a channel to share our tips ? )

[broken lotus]

I think this channel... 😅

[neat nexus]

https://tenor.com/view/blues-brothers-mall-blues-brothers-this-place-has-got-everything-gif-14885536

[cold zenith]

okay because i thought it was only for technical problem ^^

[broken lotus]

😅 mmmmaaaaaaaayyyybeeee

[cold zenith]

hahaha

[cold zenith]

Did you succeed to add a trumpet in the continue ?

[sly cipher]

Suno Writer should, I made it last night by vipcreatorsclub.com

Has all Suno documents within it too

And long instructions on creating a hit song, it's a work in progress any feedback welcome 🤷

https://chat.openai.com/g/g-02IWjW9i4-suno-maker

Incorporates music psychology and all kinds of treats lol

[lucid badger]

Can we do the same for the beginning?

[cold zenith]

but does it give us the style etc

[sly cipher]

[outro] always works for me, I'll try fade next time

[sly cipher]

Outro does and completely ends the song usually

[cold zenith]

Yep outro does but sometimes, i don't know why it keeps giving the chorus and cut in the middle

[sly cipher]

Cause that's what an outro normally does lol, it's the final hook line most the time, not always

[cold zenith]

aaaaaa

didn't know that

for me Outro it's just like " the last verse". The ending verse

[sly cipher]

It's usually, not always short instrumental, then the chorus hook, then end

Fade works too 🤷

Sometimes I'll add 1 more continue and put [Instrumental Outro] for a 15-25sec ending 👍

[cold zenith]

yeah because when u did too much continue sometime the quality of the voice change so rudely

[sly cipher]

Then I remember my YouTube stats and only like 4% make it that far so fuk'em lol

BLOP END MIC DROP

[versed valley]

Welcome! For help, check out Suno Help

[sly cipher]

Can you use bark in Suno? Hmmm wonder if /bark works?!

[cold zenith]

https://suno-ai.notion.site/Tips-Tricks-Chirp-v1-969852c74c644c6b8262ec5c5be2325c

interesting, they did not mention that's it possible to add a new instrument after

[sly cipher]

The default writer uses echo, concrete jungle, broken chains, into the light, and getting coffee too much... If you let it just do it's thing, they start sounding the same. Sorry, I'm rambling... 🙄

[cold zenith]

he listen to tauren wells " echo "

listens

[sly cipher]

https://cdn1.suno.ai/cfd2ae7a-e54c-452c-865c-726d68c3e951.mp3

Satanic... https://cdn1.suno.ai/461ac1d7-66eb-4e80-86bd-b8bd93adebbc.mp3

[cold zenith]

what do you think guy

guyz ^^

i'm put it in the general

[sly cipher]

Something has changed, I feel like I'm using V1, I can't get anything to sound good like I was... Maybe I'm trying to hard 🙄

Not getting the smooth vocals and transitions I was getting, it really is acting like V1 actually

No intros, grainy, sounds like pitch correct is on and I'm in a vaccum... Someone give Suno some CPR, it's Friday come on...

Probably took our quality for the V Day thing 😁

[cold zenith]

Try to disconnect and reconnect

[willow wren]

You can run Demuc totallly free using a T4 on Colab. You get Voice Drums Vocals and other.

[tacit socket]

When I type /chirp I'm not getting the enter lyrics box? Anybody can help? thanks.

[neat nexus]

Are you doing it in the #chirp-beta-X channels?

[wild lark]

Does Capital letters affect anything like genre "death metal "Death Metal" ?

[neat nexus]

Honestly, I'm not sure. Test it!

[wild lark]

Alrighty.

[broken lotus]

when you type /chirp you should see a pop-up of all the bot commands...
enter on /Chirp and it will load into the discord chat/prompt/text thing (where you normally type)

hit ENTER again and you should see the Chirp prompts popup...

[jolly minnow]

chirt

[neat nexus]

#🐣┃how-to-chirp

[jolly minnow]

/chirt

[neat nexus]

It helps to spell it correctly. And to follow the directions in #🐣┃how-to-chirp

[cold zenith]

The website should add a feature to choose the page of our history because scrolling back to the beginning is long

[wide jasper]

@normal patrol

[sly cipher]

Search Bar!

V3 Beta drop, V3 Beta Drop!

[cold zenith]

?

[cold zenith]

Is it out ? Or are you asking for it's release ?

[cold zenith]

guyz what's your technique for when you click on continue and it doesn't continue on the la sentences, it invented a word before that doesn't exist

[dim cloak]

Hello all, I have a lot of songs that I want my main characters in a story to perform, how can I get consistent voices so my singers will always sound the same?

and also how do I make it so I have a particular race sing the song, it doesn't seem that suno wants to recognize color and race exist

[broken lotus]

I start with trying the NEXT word of the lyric... seems best when Suno is singing slowly....

if Suno leaves a long pause or inserts the wrong words, I start EARLIER in the lyrics, and overlap a few lines from the previous clip... Suno will 'hear' the lines that were already sung, and may smoothly sing the transition....

When Suno is singing quickly, I sometimes must add the whole lyric block (without any metatag)...

it seems to be related to how fast Suno is singing the words, but that is only my anecdotal observation and I often need to try one technique, then the other

[cold zenith]

sometimes it's weird, it give you some lyrics in other languages loool

[cold zenith]

I really don't know sorry

[broken lotus]

yeah, there's a chance it will hallucinate something for a while..., and then go back to the lyrics we wrote as a new verse or something...

[broken lotus]

It doesn't see the 'race' of the singer, but will sometimes add an accent for certain genres...

and sometimes it will decide the singer needs an accent based on the lyrics... (songs about farm animals often get a 'country' accent, even if the genre is not Country)

We can add prompts to describe the voice:
sultry male singer
gets me a similar voice in the same genre (but not always)...

You might try RVC and a cloned voice, generate something close with Suno, then replace the vocals with RVC....

[dim cloak]

cool i was thinking about the rvc thing but splitting audio and voice isnt the best yet. I really hope AI and its developers evolve and understand that race isnt something to be eliminated but celebrated. Seems we always go to one extreme to the other.

I don't see a topic focused on future plans and suggestions. or the Roadmap.

where is there a section to make suggestions and get votes on said suggestions? I would like to make the following:
ability to understand more direct bracketed instructions ie
[acapella]
[female chorus]
slow intro/outro
specific instruments
change genre
musical iterlude

also the ability to download vocals and instrumental and background singers separate
consistent singers
save voice or see the seed so you can build an entire music cd. (my short film is about singers)

[broken lotus]

there's a https://discord.com/channels/1069381916492562582/1154587937611190354 channel...

I'm pretty sure Suno is on the same page with the bot getting better, and the audio quality improving with each generation.

[dim cloak]

thanks I see it now (#suggestions), Yea, suno has a lot of potential. I really like it and am blown away with some of the results i have gotten which is why I would like to build characters around that particular voice.

[neat nexus]

It's definitely a feature some of us would like very much to see. Trust me you are not alone in that.

[cold zenith]

one of the main difficult thing is to switch man to woman then switch again

i've tried the betty mark, thing you told me once. The systeme doesn't care this time

[untold briar]

Well, with custom continues, there is a manual way of getting it right. At a base level you can keep picking the spot where the voice should have switched, and retrying. But also you can fully 180 your style prompt to a different gender for each verse in the song, so that should work on more stuff overall

[cold zenith]

do you have an exemple of the good practice ?

[untold briar]

I should have clarified. "with custom continues (a feature which does not yet exist, but it supposedly coming soon)" I can't even know if that method works yet, just a guess!

[cold zenith]

juste because i put the word monster, they didn't generate my lyrics. It should be woke free

[rustic elm]

I'm using the app at website, but it always generate 2 songs. Is it normal? I want it to generate only one version.

[neat nexus]

It is normal.

You get two options to choose from. Choose the one you like best. If you don't like either, try again until you like one. Now continue from there. Then when you're happy with the second part, continue again. Once you have reached the end of your song, get whole song to stitch it together.

It's not automatic. It's a process.

[sly cipher]

What's the end goal for Suno? Just curious what the vision is beyond what we see now... 🤷

[exotic quartz]

Is there a public roadmap somewhere?

[final hazel]

I think the end goal is fluid. Nothing is set in stone.
Atm there is not a public roadmap

[sly cipher]

Like is it going to become a sync tool for music managers, a custom song generator as a service? What's the intent besides entertainment. Just curious... 😀

I wish I could white label it once it's at sync music quality level and become a branch with a different name pitching a different market 😇😇

[final hazel]

Sounds like a great #💡┃suggestions-web

[spare frost]

Hey guys, I love Suno, is there a way to generate A Capella?

[neat nexus]

Yes, use a capella as your style prompt.

[spare frost]

I tried that but it always adds a guitar or some other music. I'm using the website version, should I be using the discord version?

[neat nexus]

No, it is possible in the webapp version.

All I can do is say to try some prompting in the style prompts. You may need to roll more than one try to get it.

[spare frost]

Could you send me one you've done? I've tried 10 times and can't get just vocals by themselves

[neat nexus]

Give me a bit to try generating one. I don't tend to do these myself. I know others have done it so it is possible.

https://app.suno.ai/song/e28b146b-2c4f-44be-a496-05e9e6a47695

My Little Dream Girl

(C) 1915 L. Wolfe Gilbert, Public Domain

Too me four attempts to fine tune.

Style Prompt was: a cappella, warm soulful male vocal, vwithout accompaniment

@spare frost?

[craggy magnet]

On my Artist account, cheers Hepcat! I'll try working around these!

[neat nexus]

Probably try a different type of vocal though... That one was a bit harsh as an a cappella.

[craggy magnet]

I'll going for a powerful female vocal for a drum and bass artist

[neat nexus]

Probably go with soprano then.

[peak veldt]

I liked your article, but I couldn't find a way to forward a YouTUBE URL into the Upload section.

[neat nexus]

Are you deleting and reposting this every five minutes or something?

[peak veldt]

NO, I grabbed the wrong reference, sorry to be a pest. 😉

Replied to the wrong line repeatedly, will try to avoid that in the future.

[rose ore]

If anyone's interested, I just figured out how to mostly reliably trigger a harmonica solo on cue. [harmonica] seems to work 50ish% of the time

[unborn pewter]

Is it possible to add some special feature in suno create

1. Male & Female voice button
2. Song beat
3. Voice age
4. Voice clarity
5. Voice Language

????

[low remnant]

LOL, damn you. Now I'm going to have to write a whole song around finding a place for putting a harmonica in. I already went there because I decided I wanted a song with a fiddle in it.

BTW, has anyone noticed that Suno pronounces some words different when all in caps? rebel is rebel. But REBEL is ree-ble. Ether is also different in caps though I don't remember how it pronounced it, just that it was weird. Anyhow, might have some corner use cases...

[rose ore]

Oh, the other thing you want to do is include the instruments you want to solo with in your genre prompt, if they're more exotic than guitars or drums

[broken lotus]

I see this with voices too... If the voice was described in the Style, but doesn't sing the 1st clip, I can make it sing later by spamming and prompting...

But if there was no voice description in the style, I've never gotten a voice to appear later....

[rose ore]

oh interesting. like are you talking about prompting followup renders like continuations? I haven't had much luck with that, but I don't make a lot of instrumentals, either

[broken lotus]

yeah, on continuations...

[deft hazel]

We should all make a document on the "tricks" we use and I will compile them into an advanced guide

A collaborative effort from expert suno lovers 🙂

[low remnant]

It'll need a big disclaimer, all in caps, probably spammed a few times that they work SOMETIMES. Otherwise, some of our channel-mates will disparage the effort, LOL. But I'd love a repository of tricks. Part of why I'd love folders in my Suno library. So I could have "testing", "failed experiments" and "might kinda sorta work". I thought about putting some of the tests into the wiki but they're all so tenuous that I wouldn't feel good recommending them.

[deft hazel]

Hey, a book of works most the time is better than no book

[low remnant]

I don't know if I'm more excited about learning all new tricks for v3 or sad that some that work in v2 certainly won't work any more in 3.

[neat nexus]

That sounds like a great idea!

https://suno-ai-wiki.notion.site/4e24aa838bf84fffaf383b1f9ed50748?v=755ec8954abd4e3f9f5c899950f0f315

[ornate lark]

punjabi

[deft hazel]

Anyone have a notion business account and can export the suno wiki in PDF for me??

I'm training a chatbot on the data to make suno hit songs..... 🙂 trying atleast

I am combining wiki with hit song data to see what I can come up with

you goto 3 dots, export then check "ALL SUBPAGES".... anyone? ❤️

[wide jasper]

Oh yes, when you browse files paste the YouTube link into the file name section and click open of that doesn’t work you can use the method I DMd you

[cold zenith]

you can with your browser, print and select PDF instaed of your printer

[cold zenith]

Sometimes the system doesn't want to do want you want. I ask for a Gospel choir, it doesn't want to hahah

[vernal meteor]

we are all slaves to SUNOs whims

[cold zenith]

this is crzay i'm asking for a female voice for the song, it keeps give a male voice hahaha

I GAVE UP, i'm going with the voice male

In 50 try, it gave me Once the female voice hahaha

[fallow gull]

When you say "asking" are you using "female voice" in the style or metatags? Putting it in the style seems to work better.
I wonder if genre norms play a part in bias? For me Rap seems to be more likely to be male, Pop more likely to be female. Jazz Piano more likely to be female, but Big Band is male. I don't have any stats, I'm just going by my perception which can be biased as well.

[vivid cedar]

Any idea on what the devs are working on? Is fidelity being worked on?

[cold zenith]

i put Female vocale, it doesnt work, i put female voice netither.

[broken lotus]

if your genre is typically male-heavy, like metal... you are fighting inside the prompt. It will not work everytime.

You can add 'female vocal' extra times to weight it more in the ai.

'Female vocal, female vocal, female vocal...'

[neat nexus]

You can also try "cheating" and define the vocal qualities like soprano vocal and see if that helpo?

[broken lotus]

yes, v3 is another jump in quality, according to most people trying the V-Day preview...

[inner steeple]

v3?

ah, I see it's coming

finally being able to cut and continue from the middle

that'll be handy

[vivid cedar]

I am so excited for this

[vivid cedar]

where do we get development news and stuff? i'd like to read everything regarding developments

[unreal hatch]

i wanna high quality mp3 or wav...

[fierce orchid]

Hi! Is there a way to achieve the exact same style, mood, voice, and instruments from a song I created before?

[neat nexus]

https://tenor.com/view/lottery-lotto-gif-12706150

It's possible to be extremely lucky.

Another possible thing is to try taking an earlier song that you completed, go back to a segment that has the music and vocal you like and try doing a [break] and then resuming with new lyrics. If you manage to get it to stop and then restart and still keep the music and vocals you can continue as usual and then just clip the previous beginning from the original song once you obtain the complete song. But even this is very hard to get. So luck is definitely involved.

[fierce orchid]

I will try, thank you!

[cold zenith]

so i'm making a music, but there only music for 1min20...it'a rap music hahaha

[neat nexus]

Right. You need to continue it.

[cold zenith]

But this is crazy today, i even put only a girl , it keeps giving me a male for a RAp

[royal citrus]

What ever happened to Chirp V2 being released for Discord servers?

Their announcement said it was going to be on Discord the week after last Thanksgiving

[broken lotus]

try combining them in your style prompt?
female rapper
sultry female rap
sometimes it will put a female singer in a male rap... (backup or guest artist for a verse, lol)

[radiant aspen]

v3 you say? What does it entail?

[daring lava]

Stay tuned to #📢┃announcements 🙂 When the team gets it prepped and ready, all will be explained!

[radiant aspen]

What's in the rumor mill so far?

[daring lava]

It's NEXT LEVEL

[radiant aspen]

lol

[wild lark]

OH BOY i hope so! i love some of the metal im creating would love to get something better!!!! ill keep spaming chips because its helping my depression hardcore

[heady hearth]

Is this the server? Seems I've lost older stuff with the library finishing at page 8. Is this a hard limit?

@wild lark 👹

[wild lark]

?

[heady hearth]

@wild lark I'm looking forward also to v3, to generate even more brutality

[wild lark]

Agreed!!! I make heavy deathmetal hope it improves

[heady hearth]

@wild lark sorry, tacked on the emoji@you to my other question about library capacity

[wild lark]

i have no clue tbh

[heady hearth]

@wild lark dang, all my dubious creations, evaporated

@daring lava hey maybe you can help: is there a max capax to the library, 8 pages maybe?

[daring lava]

I have yet to meet a wall... if there is one, I haven't seen it

[heady hearth]

@daring lava cheers, I wonder what I've done😭

@daring lava just like that, pages 9 plus appear🤗

[daring lava]

Keep up the great work! Very awesome to see all of you creating!

[pliant cave]

Mikey was right. V3 has exceeded expectations. Excellent work, team.

[sour mesa]

It is launched? I see nothing on #📢┃announcements lol

[pliant cave]

Look again, it's there, in the V-Day announcement.

[lofty merlin]

Is it just me, or does the V-Day v3 model sound way more robotic in the vocals than v2 usually does? 🤔

[pliant cave]

I see a vast improvement in the clarity of the vocals, and I haven't had a single "no vocal" track that was typical of a "bad" V2 run.

[lofty merlin]

Don't get me wrong, it's definitely less prone to full failure cases or glitched vocals... it just sounds like every generation has autotune turned up 😂

[jade mist]

When will V3 be in the web app?

[cold zenith]

i did it didnt work i also pute in the lyrics : " [female][female vocale] " didn't work

[lofty merlin]

I don't know if it's different for rap, but if I put "female vocals" at the end of the style/genre prompt (not the lyrics), I get female vocals 99% of the time, for a variety of genres I've used

[jade mist]

Nothing to hide it’s V3. I get much longer song

[scarlet mica]

Is there a way to get it to immediately hop into the lyrics rather than devote 1 to 2 minutes to pure instrumental?

[cold zenith]

okay,

On the other hand, I've just noticed that suno will generally follow what was said at the first prompt. In the continuations, it will very rarely miss something you've changed in style.

[scarlet mica]

How do you get the full song from discord's chirps?

[heady hearth]

Hey is there any way to rename a piece? I've tried continuing and changing the title, remixing the entire song and doing the same - no dice. I ask because I foolishly named 2 distinct projects the same.

[cold zenith]

Anyone, does it already happened to you that a complete verse is skipped for the next one ?

[final hazel]

yeah.. things like that happen sometimes.. 🙂

[final hazel]

not yet

[heady hearth]

@final hazel ok, thanks🙂 My half-a-butt work around was renaming the files after downloading, but doesn't fix wrong/unintended title when sharing a link😶

[dry umbra]

0

[cold zenith]

I spend to much credit on that female voice, i gave up hahaha

[frosty rampart]

Is there an overview here of the different styles of music that are recognized?

[cold zenith]

here : " Here is the wiki, I recommend reading every page as a start..

https://suno-ai-wiki.notion.site/Style-Genre-List-f7d1bb28193c4eac910f03db3f602508

[pliant cave]

@light ether @open magnet

Here is what stands out immediately so far with limited V3 use:

1. Instrumental-only tracks issue: The unwanted behavior of instrumental-only tracks, where the model refuses to generate vocals from the lyrics, seems to have disappeared entirely. It appears they really trained/fine-tuned that out. I haven't encountered one of these issues yet.

2. Increased bitrate: The bitrate has increased to 192kbps from 128kbps at 48k stereo. It sounds amazing.

3. Clearer instrumentals: The instrumentals are clearer and more cohesive. The arrangements feel more deliberate and less like guesswork from the model, which is a significant step up from previous versions.

4. Extended context window: The context window or (max tokens allowed) now extends up to 2 minutes, which is longer than the previous 1 minute and 20 seconds. This is a welcome change, especially for casual users.

5. Improved 'ad libs': The model handles 'ad libs' in parentheses much better, producing results that are closer to expectations. This was an area where V2 often struggled.

6. Clearer vocals: In general, the vocals are clearer and better integrated into the mix. V2 and V1 sometimes produced tracks with vocals that were too quiet, especially when too many genre tags were added. Now, the vocal volume seems consistently balanced, regardless of genre tags.

Overall, I'm seeing improvements across the board, but the most significant is the reduction in "bad" generations with poor vocals. Kudos to the team for this progress!

Of course, there will always be someone who preferred something about V2, given the wider user base now versus last year. One suggestion could be to have a toggle button in "custom" mode that allows you to select the latest V3 model or the previous V2 (as per a comparable service like Midjourney), and to set a deprecation date for when the toggle will be removed. Meanwhile, 'simple mode' would always point to the latest model, similar to how OpenAI handles planned model deprecations, etc.

[wild lark]

Bro I hope the vocal sound is better my instruments are way to loud compared to the bocals

Vocals

[pliant cave]

I suppose that is somewhat subjective, but the only way to truly know is to test it with a decent sample size. I ran about 100 generations, and I believe that's probably the bare minimum before I felt confident in providing any opinion on noticeable changes, considering that the randomness or luck factor can significantly skew results with small sample sizes with this kind of tech.

[wild lark]

I run heavy death metal… but still sometimes the vocals are so quiet

[frosty rampart]

My credits are almost used up. I have subscribed to the annual subscription. I haven't been able to create an optimal song yet. it's so difficult and I get so many different bad variants. It seems like it's random. I think it's a shame that I don't get more tokens for an annual subscription. It's so difficult to create something nice. What can I do to keep trying?

[daring lava]

If you’d like to provide some examples of your process/prompts used/what you are trying to achieve, I’d be happy to try to help figure this out 🙂 just tag me in the post so I see it

[frosty rampart]

I do it via the app version. is it better via discord?

[daring lava]

Browser/app is my preferred method

[pliant cave]

It absolutely is random, @frosty rampart, that is essentially the basis of how the technology works. So it becomes something of a lottery in terms of sometimes hitting the jackpot and other times coming up with something destined for the trash bin. This is the reason why Suno traditionally generates two samples for the same request, as it really increases the odds that at least one will be decent. V3 seems to tilt those odds more in your favor.

[tired schooner]

when it will be out?

[deft hazel]

Does anyone konw if the thumbs up or thumbs down helps train the SUNO model for better outputs?

[daring lava]

By habit, I pretend it does. Though I can't be sure. If it doesn't now though , then maybe later 🙂

[broken lotus]

atm, the thumb-down is useful while creating, it hides the bad clips.

(maybe not super-useful, but)

[crisp otter]

Hi guys, how can I locally install chirp??

[final hazel]

you can't

[fallow gull]

If you want an non-Discord experience, use the website: https://app.suno.ai/create/

[runic dock]

When hopefully the first trial has an awesome music but you realize that your lyrics are very poor , you would like to " keep the same music and edit/change your lyrics".
The possibility of keeping the same music but changing the lyrics : this is the feature i would like.

[pliant cave]

I have no idea; I don't have any information beyond just my experiences from limited testing. My guess is that they are still iterating V3 since I did notice a slight recurring bug of some outputs looping back on themselves from the beginning at the end of the generation. I noticed it more than once, which usually means it's not a fluke but a "bug".

@untold briar Have you noticed that? No one does more testing than you, probably even more than Suno employees. 😅

[untold briar]

If you know the clip is the 'last' clip, which it always is in the valentine's day song, then my guess is you'd prefer to generate 'too much song, more than the lyrics' instead of 'too little song, not finishing the lyrics' because there is no possible option to 'continue from this clip' and finish the partial clip on the vday site. So probably just a preference/tweak.

Also remember the upcoming continue feature, also prefer the same thing. too much song is better problem than awkward cut off clip, likely.

I really like it. like listen to this song past the end of the text, it gets better and better https://vdaysong.com/ycknwenj

[pliant cave]

Absolutely agree, I'd rather get more from the model than less, especially if I can just clip it out at a convenient, correct endpoint. I suppose from a model perspective and an end product perspective, as a user, you would ideally not want that behavior and would stop it at the right spot every time. So I assume it's just an iterative fluke of V3. V2 had a few looping issues, although slightly different, they were corrected at one point.

[green spade]

Is there a better guide somewhere on what people are seeing success with as far as metatags? The documentation doesn't seem great and doesn't provide a lot of examples.

[fallow gull]

vdaysong used j-pop for one of the random styles. Problem is the lyrics were entirely in Japanese as well.

[shadow timber]

You have the craziest hobby 😅 that’s lovely. What program is telling you it won’t do standup? I have a bunch of standup songs from Suno

[thin flint]

Lol I like that too. What did you prompt to make it do that? I tried to make a song with it that'd do that, but I got the error about filtered content and it just wouldn't make the song.

[shadow timber]

If you input your prompt in Japanese it will give you Jpop in Japanese too, I was doing that yesterday

[fallow gull]

My vday prompts were all in English. It gave me 2 English songs and one in Japanese.
After I translated the words back to English, it is the correct theme.

[shadow timber]

I think it may be name based, it really wants to make Ellery as a name into Spanish songs. This one is my favorite, keep listening after the end https://v-day.suno.ai/?s=4bed92b1-4152-4fbf-97f8-c185d88a66fb&r=RWxsZXJ5JlJoZWRk

[fallow gull]

OMG, I hope that "En" language tag comes out in v3. I tried a song that alternates lines between english and french, but it sang the french with bad english pronunciation.

It can do great french if the entire text is french, but if it is mixed with english, it's not great.

[fading hearth]

is it possible to generate the same song in other languages?

[eager topaz]

It's my experience that many of the Metatags are hit and miss and what seems to work for one user doesn't work for another. All a lottery, I guess!

[crystal sapphire]

Is anyone else having issues with continuing songs where every result brings back something totally unrelated to the original music? Started today. Yesterday, I got a song that came back in an unknown language and English lyrics. BTW, are we all just trying to figure things out on our own with no help from Suno, even making up our own wik to answer our own questions while paying them for a service? They haven't answered a single email. And why is a chat server meant for gamers (Discord) being used by all these AI platforms instead of starting with a developed website? What is the meaning of life?

[graceful cliff]

meaning of life I think is 42

sorry for yr experience -- that is a known bug, and I promise that will get addressed with future model updates.

but there is no update on our side that will cause your today's experience suddenly change from yesterday. There are randomness in all generations, and sometimes it is just unlucky.

It is our job to help you to get music generations as lucky as we can. And we are busily trying. So if we don't reply to most messages, it is not we are ignorant...

[shadow timber]

To answer your other question, most of the AI platforms have fewer people on the team than you might imagine, and they prefer to spend their time and money developing AI, not websites. Also, discord helps with onboarding by providing a central location where users can socialize and learn from each other. (I’m not affiliated with Suno, just another user.)

[crystal sapphire]

OK, but this is happening to every song today. None of the music generations are matching to the original clips and I'm out of action for now. I don't want to burn credits when it's not working properly.

[untold briar]

Are you continuing from the same clip? Could be this going on, it feels like bad luck over and over, but it's more that the clips you are continuing from are already doomed #🪲┃bugs-web message

[crystal sapphire]

It's happened with three separate clips in a row. When trying to continue each one, the results are coming back totally different - melody and background music. I've been using the app for over a month and completed multiple songs. I know I'm not doing anything wrong on this end. You're probably right about the clips being doomed as they were all generated as initial ideas on the same day.

[untold briar]

Are they short clips, by any chance? That may also play a factor. (to clarify, I mean is the clip being continued from less than a minute here)

But yeah, in general, the biggest problem (source of burning all your credits, at least) is not even knowing you have a doomed clip (can't continue from in a useful way) until after you spend the credits. Probably the worst case situation IMO

[crystal sapphire]

Yes, and I rarely get even 1 min.

[untold briar]

Some cases of it should get better in future.

If the clip is less than a minute it may also be less stable contuining, more likely to do something different. Though if it's close that shouldn't be factor, unless it also has a 'song ending' feeling at the end of the audio. In that case the AI sometimes treats the continue more like 2 different songs. So the output feels very different.

If it's 3 clips though that could simply be random bad luck on the 3 clips, that part. And then it's not clear that every 'roll' on the clips is doomed because something causing a problem is in the past of audio even if often you don't hear anything odd

[crystal sapphire]

I'll just wait a day or so to see what pans out. If V3 is on the way, perhaps some of this will be fixed. I've burned the monthly credits. I'm on extra paid credits that won't expire and my new month starts in 3 days. I need to work on learning Midjourney anyway.

[shadow timber]

MJ is fun right now with the new sref 😃

[thin flint]

@placid gorge When is your surprise song or whatever going up?

[peak veldt]

I will agree with your assessment of the tool in its present state. I am burning my credits at a much faster clip, than I ever did with the 50 FREE CREDITS a day... but truth be told, this rolling the dice on decent continuations is a pure crap shoot.

Listen to this continuation below -- it is supposed to be an OUTRO ending. At 34 seconds into the clip, you will hear a creepy conversation of some dudes talking in an empty room, and chuckling... gives me the willies it even got generated. I would pay another 10 credits to have the sample (it came from) removed from the library.

https://app.suno.ai/song/e4c3a9ff-fea5-4927-8b74-4ec9db71d997

[placid gorge]

Been up sense midnight

[shadow timber]

That’s a perfectly fine outro, it comes to a nice clean stop, chop the part you don’t like off, no problem. A bad continue is one that keeps you from making a good song. If you expect Suno to be a stand alone maker of perfect publishable songs with no edits I think you are going to be waiting a while. Since the training data will always have people talking on their recordings of music.

[modest flicker]

Finding my chirps is nearly impossible. How is it done?

[leaden plaza]

Is the webapp on suno.ai on V3 already?

[cold zenith]

no

[shadow timber]

Rumor has it it’s v3 but they won’t confirm it

It’s just the valentine that’s v3

[final hazel]

it is using V3, it's been confirmed to me (the valentine thingy)

[leaden plaza]

I just did a song using the webapp. Omg. It's amazing

Too good for usual v2 standards

[final hazel]

The normal app is still V2... cannot wait to really play with V3 instead of making a valentine song

[peak veldt]

https://tenor.com/view/oh-come-on-steve-harvey-steve-on-watch-uh-huh-rolls-eyes-gif-16988383

That's a perfectly fine assessment to have in defense of the tools in their present state - but it doesn't stop my willingness to offer another wasted 10 credits to make it go away permanently as a future perfectly acceptable result in a continuation so it doesn't happen to anybody else.

[shadow timber]

If you got a perfectly fine completed Outro out of it, it’s not really wasted credits

[broken lotus]

You wrote text after your [Outro] and you're upset there was talking after your Outro...?

🤷‍♀️ I mean..

It can't possibly be that you gave the bot conflicting instructions and it attempted to do what you asked... 🙄

But as Allie says, it's SOOOO EASY to fix you lose our sympathy when you pretend you can't.

[peak veldt]

https://tenor.com/view/rigole-gif-25807483

[peak veldt]

https://tenor.com/view/what-huh-confused-shocked-disguested-gif-16058540

I am trying anything that works... and evidently, I haven't found that one something that consistently gets a proper OUTRO, fading or not... The truth is, these tools are very hit and miss via prompts and meta tags, and to be defending them as if they consistently get the desired results is anything but Ludacris.

And I don't own any secondary tools to modify these results... I much rather, promote and present the delivered results in one package that leads users I send them to see for themselves.

[shadow timber]

Audacity and bandlab are free, and SoundLab and GarageBand came preinstalled on my phone…

[broken lotus]

Fair..., but trimming the end off a media file is something any phone can do with the built-in free software...
(it does not require a DAW)

I've been using
[refrain]
without any lyrics after... it tends to sing stuff from the last clip
(the hook, sometimes the whole verse with different music)
*different meaning it drops some instruments or sings it like an outro...

and no, it is not 100% perfect, but I don't trust self-driving cars either...

[peak veldt]

Oh, that's good to know... but I rather have it self-contained in the one deliverable it can produce.

[peak veldt]

Hmm, REFRAIN (without any lyrics following) - that's good to know. I'll give it try, even though, it still is hit and miss.

[tired schooner]

hold you urine until V3 come

[neat nexus]

https://tenor.com/view/dog-dancing-dance-excited-dog-pee-pee-dance-gif-16473864

[honest dagger]

I think paying subscribers should get early access to V3 (yes I'm super impatient) been blown away by Valentines songs

[charred bolt]

is there a guide how to add chords ?

[open magnet]

maaaaay be happening 🙂 but still a few things to prep for that

[honest dagger]

Shall keep my eyes and ears peeled! Thank you!

[daring lava]

Not at this time

[broken lotus]

I've seen people put chord notation as metatags...
It seems to do... something.... but I'm not sure it is very accurate...

seems more like relative changes, like it often goes up and down semi-correctly, but not necessarily the actual chords.

[daring lava]

Looks like everything is running smooth now at https://app.suno.ai/

[runic dock]

What is the différence suno is making between [chorus] and [refrain] ?
Sorry my english is poor.

[broken lotus]

In many HUMAN languages this is the same word.

in Americanese... (at least according to AI) a [refrain] combines the idea of 'outro' with 'improvise based around the hook/chorus/any lyrics from the last clip still in memory"

I sometimes get a complete repeat of the last verse, but set to a different arrangement (not just a hallucinated repeat)

[remote creek]

Hi team, I since this morning I have some glitch generating song, I am a paid customer and now eI only have the music with no lyrics but I do add my own lyrics

[broken lotus]

https://www.notion.so/Why-can-t-I-hear-my-lyrics-6dea581cf3d5414d968fccd3f940994a

[charred yacht]

Hi guys there is any api to generate this songs?

[broken lotus]

no API..., haven't heard anything about an API from devs...

[charred yacht]

ok thanks!

[neat nexus]

The message I've seen every time is this:

We don't have a publicly-available API at this time, but it's on our roadmap.

[charred yacht]

nice to know thanks!

[broken lotus]

Today's Secret Metatag:

[pregnant pause]

[wraith vigil]

Hey guys, I paid $30 dollars for 10k credits last week. I’ve just logged on and now I’ve only got 50. Can anyone help with this?

[graceful cliff]

emailing [e-mail]