#📚┃suno-school
1 messages · Page 10 of 1
I randomly remembered an old group that I liked about a decade or more ago. Used to listen to one of the member's podcast for a long time. I decided to junk about a bit and see if I could get something that came at least a little close to a song by their s I liked. I'm pretty sure I failed.
This though? This was intriguing to me. I'll definitely be revisiting the style and the memes as lyrics. I think there's potential there. What do you guys think?
This is called The Light has Gone Out (But I Still Work)
(Oh and in case anyone was wondering, the inspiration was Love Spirals, Stir About the Stars. And this was a second attempt when I decided to just have fun since I couldn't come up with a good poem like those lyrics are.)
is there a history of chirp songs I have created that I can look back into??
Who have idea to make ending of song ?
/Chirp
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.
Yes, here: https://app.suno.ai/
6281 in total, Genres, Styles, Subgenres, Fusion, etc. Would be interesting train a model like Suno with a database like that
Thanks for this! I really love it!
"Canadian House"
While I'm certain there's no such thing as "canadian house'... (unless it serves waffles)
I am 100% trying Norwegian Space Disco because I HOPE that one is real.... 😹
A quick search says....maybe?
Can we make only music without lyrics? If yes, how?
yes..., try:
- leave the Lyrics text blank
- use
[Instrumental]in the lyrics instead of lyrics - use Instrumental in the genre text
#crip
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.
Why is the output so high in treble?
Have you had success controlling it mocu
more
Since you removed that comment can I assume you're giving it a try after all? I finally got out of the work meeting and started listening to the Spotify playlist. A bit of a surprise that first song....
Did this one a few days ago and gave up on it...maybe I shouldn't have?
Granted the one I have above is rather energetic for the genre but still....
it is like a porn soundtrack from 1978, lololol
I have tried to prompt for 'burlesque' 'vintage stripper' et all for that saucy sax sound...
Never got it. 😓
Naughty filter at work?
lol, vintage words don't trigger the censor.
Just couldn't get that sleazy WAH-WAH saxophone
Also I made a song with 20+ euphemisms for oral sex....
I kinda think the kids complaining about the censor bot could try picking up a thesaurus.
You mean like this?
(I'm so sorry. Chirp just likes me...what can I say?)
I am so jealz!
'sultry' Hmmmm
If it helps, the b-track was just drums..
And I think the (hmmm) and the [downtempo] helps as well as the onomatopoeia but how it all plugs in together I couldn't tell you.
Actually, when/if one of the team members takes a look in here, could they take a poke and see if it is possible to isolate why this worked? I get that some of it is just sheer dumb luck on my part with the dice rolls, but can it be broken down a bit to see what bends in this direction? My prompt was:
`[intro]
(hmmmm)
[saxophone]
Wah-wah-wa
[downtempo]`
With the style as: sultry sexy songs
Result posted above was the first and only attempt and the first version produced.
(Or at least the left side of the two.)
I've got work meetings this afternoon, but I'll definitely check in later.
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.
Spamming the Genre box FTW 🏆
Melodic Saxophone, Melodic Saxophone, Melodic Saxophone, Melodic Saxophone, downtempo sexy groovy instrumental
the 'waa-wah' didn't work for me, but spamming did make a sax... that's a win!
Music
A lot more lighthearted than the one I was working on though.
yeah it's completely different...
(just glad to get a saxophone at all)
I see.
So I tried continuing it...it went ok I guess. But since I really didn't have any plans I don't know if it's just a bunch of noise or what?
I tried a bit of lyric but they all (in my head) sounded like something out of a 40s musical...which can be good I guess but not really my thing...
Though this unfinished work (for the same reason) may beg me to reconsider if maybe it is my thing and I'm just embarrassed about it...
😩 Agh! I made my song too long and it won't let me ❤️ the whole thing....
T-t-that's a thing?
Seriously. That's a thing?
I made an eight minute, twenty second song once.
How long is yours?
another 5 chirps after that...
So about seven or eight minutes.
Yeah I think there might be an issue. In addition to your own troubles, look at how this recent attempt just ended on its own. This is the beginning of the song...
[Intro]
is a word also used to mean a short bit of music and credits to introduce a video-segment....
I get a lot of these with [Intro] 10-25sec usually and a nice clean ending.
Huh...Yeah, I could see it.
That might be a great intro background music for a YouTube video title sequence or something...
yeah... that's what I'm thinking....
And they are often mini power-bangers....
Yes, yes they are.
Is chirp able to continue existing music? I am not asking aboet the discord bot, but rather the underlying software
ya, but we're not making that available for obvious reasons...
cool, just wondered
I asked for a blues sound and got rap???
lol, can happen sorry, llms.... maybe try again?
Possibly you have a lot of lyrics and it decided to speed up the pace?
Fingerstyle Acoustic Guitar
/chirp
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.
/chirp
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.
🙂 : Ok Chirp, let's make something dramatic and esoteric, like acid-rock guitar-feedback and mind expanding atmospheric, but pretentious and pointless too.... Just to see what prompts work.
🐣 : (pretentious atmospheric noodling)
☺️ : That is perfect Chirp, now END.
🐣 : (ends pretentious noodling) 🫡
🐥 : 🥁
🤭 : think you're being cute? I'll humor you this time... ⏩
😏 : But I'm not helping you with any prompts or genre. You're on your own!
🐥 : 🥁 🎸 🎵
🤨 : Not done yet... well... I guess. ⏩
🐤 : 🎶 🎸 🦕 🎵 ⚡
😒 : ... ⏩
🐤 : 🎶👄 🎸🧜♂️ 💫 🎵
🥺 : ⏩
🐥 : 🎵 🎶👄 🎸✨ 😗 🎺 🎹 💫 🎵 🎶
You test so creatively ❤️
what AI model does chirp use? Is it based on one of the facebook ones, or is it entirely custom?
how do you get out of a crappy sound when you're linking multiple files? Seem to have gone donwn a wrong alley and can't get the vocals to change with new genres
Some times that's the only way forward.
There's a context window you always have to be aware of.
/chirp
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.
/chirp
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.
/chirp
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.
Hi! Suno doesn't work for about 20 minutes
GPU outrage -- we are working on it
outrage or outage?
😅 😅
is /Chirp not working in the Beta-1 channel?
There is an issue being worked on...
It's outputting gibberish talk now!
@worthy grotto I think we broke it lol. It's outputting content but all gibberish you can't understand
😋 That's the best part.... when AI gets crayeative
not when I'm 5 clips into my song!
It can be frustrating! I know.
consider the last 10sec of the clip you're continuing from. That has a strong pattern for vocal it wants to finish. Try to come up with a short lyric that will fit.
or last resort, regen that chirp to get a different ending
so how would you break up this copy and feed it for best results? Going for a Dr. Seuss style story book rhyme: Thus, GiverGlee grew,
in joy and good cheer,
Thanks to the Smart Coin,
they all held so dear.
So if you want to give, but give smarter, you see,
Just get yourself a Smart Coin™, and you'll soon agree.
With this coin in your pocket,
you're charity's never alone,
For you're sharing the message,
each time that it's shown.
where is it cutting now?
that's all in one clip
pretty much the last clip before it started getting wonky
Here's the full lyrics:
server blocked my pasting the full lyrics
First clip: In the town of GiverGlee, far as your eyes can see, Lived the folks, who cared, oh so care-freely! They donated to charity, high-fived and cheered, But still, they wondered, "Can we kick it up, this year?"
Then came Keith, from Gray Water Ops, With a Smart Coin in his pocket, he took fundraising to the tops!
"What's the chip do?" young Timmy dared to inquire, "Does it dance, does it sing, or perhaps, spit out fire?"
"No," said Keith, with a smile so wide, "It turns every donor into a charity guide!"
You carry it daily, in pocket, purse, or wallet, Put the coin in their hands, and watch them be exalted!
This magical coin links their phone, no cap! A donation from here, is one simple tap.
The coin tells your phone, to open a browser. Takes you to the donation page, can't get easier!
The town buzzed and hummed, oh, they hummed with glee, "This Smart Coin™," they thought, "is the future, you see?"
They each took a coin, and what did they do? They became ambassadors, of their favorite charities crew!
They carried the weight, not heavy but light, Of their favorite causes, with all their might.
Sharing the mission, in palm or in hand, The Smart Coin™ was making kindness go hand in hand!
And if you tapped your phone, oh just give it a try, It's worth a try, don't worry, don't cry!
Thus, GiverGlee grew, in joy and good cheer, Thanks to the Smart Coin, they all held so dear.
So if you want to give, but give smarter, you see, Just get yourself a Smart Coin™, and you'll soon agree.
With this coin in your pocket, you're never alone, For you carry a piece of a charity's home.
Now, I know what you're thinking. I know what you'll say.
No one will do it! Not ever, no way!
But maybe they will, or maybe they might,
Or maybe they won't want to join the fight.
It only takes one to continue the cycle, Say yes to your plea, take the torch, and recycle.
You're helping our cause so what's there to lose?If it all falls apart we've got plenty of booze!
I didn't just say that, a Freudian slip,What I meant to say was keep a stiff upper lip!
Many will say no, get ready for rejection.But the one's who say yes, inspire love and affection.It's all part of the magic, let giving transpire.The smart coin in your plan takes your fundraising higher.
And don't forget, Keith's book makes it clear,"The Coin that Changed Charity" is a tale you must hear!So hop on to read, and get inspired,For a world of smart giving, that never gets tired!
Get a free copy of "The Coin That Changed Charity" on Amazon, with Kindle Unlimited, or grab the paperback or hard cover.Want more info or to discuss a custom smart coin for your next fundraiser? Visit Gray Water Ops dot com.
🤷♀️ I mean, there's no way for me to help. The AI has built up to some music motif in the last 10sec of your last lyric clip. the continuation wants to continue with that music/lyric pattern.
either you need to supply it some extra lyrics to move into (even 1 line might work)
or you need to go back, break up that last chirp into 2 chirps by making the lyrics shorter (in other words, change the lead-in so you don't have such a strong lyric-oriented motif)
ok
The hallucinations you've heard should give a clue to what it is expecting/wanted...
it seems that each additiolnal clip makes teh voice go higher and I can't seem to request a male vocal or change in the middle.
How about
male vocal, male vocal, male vocal, male vocal, male vocal, ...
right at the start of the genre description...
(also one of the devs has said the next(?) version of chirp is better at male/female vocal prompting, but idk the eta...)
Has anyone tried to put non-sung narration successfully?
Hello, I haven't been able to make it play a song in a specific key, for example: Fm. It also doesn't play at the BPM I indicate, for example: 128 BPM.
Does anyone know if it's possible? I would greatly appreciate the information.
Hi everyone! I have a question about Chirp, what is the output sample rate? What about the number of channels?
currently mono. Some stereo clips have been teased by the devs, but idk the eta on that release
maybe try Spoken Word...
the word for 'talk-singing' is Sprechgesang
Sprechgesang
I don't think using keys or bpm work with any reliability...
@neat nexus has had success slowing down and speeding up through prompts (maybe they will share some tips)
The magic incantation I use is [temposhift] and it doesn't always work. Ditto on [ritardando] which also doesn't always work. They do something often enough that I'm convinced that something is happening but it's not anything close to an exact science.
Like everything with AI related generation, you have to experiment.
You can also try prompting it like a script (film, musical, or opera) #🐣┃suno-showcase message
Hi, I am unable to renew my subscription
Is chirp going to be in local ?
I'm on a basic plan, which is free, and I only have 50 (or should I say i can only generate 5) chirps per day. Yesterday I wasted all my chirps, the next day (today) I can't make chirps now. Why?
timezones
Is there any dictionary on how to use chirp properly? I've noticed (I think) that () don't work and [] should be used instead... but is there any limit to what can go in the square brackets? Like [Verse 1] [Verse 2] etc
()'s are used for adlibs like for example:
"I don' t know what lyrics to type in! (type in!)"
[]'s are used for the part of a lyrics like "[Chorus]" or/and "[Verse]"
[chorus] and [verse] and common song structures work pretty consistently. Anything you put in brackets will be treated a bit differently by the AI overall. So you can try anything in there.
[fade out] and [song ends] are cool when they work, but more hit or miss. [Melodic Instrumental] is a recent favoriate
If you really spam the brackets, interleaved between the lyrics, you can sometimes confuse the AI in really fun ways. Like listen to the first line of this, it's sort of singing the words IN THE BAGPIPES #💬┃general-chat message
Actually, ()'s used as "(chorus)" sometimes works
You can even try {} and it sometimes does fun things. That's basically stuff the AI leaned on its own not on purpose, as I understand it. But [chorus] and [verse] were given an extra nudge from the devs. If you want to try things, think of something you might find in real song lyrics and how it is formatted.
I posted this as reply in suggestions but probably useful to have here. Some of this is from dev comments, other stuff is more of a personal guess.
Parts of the Chirp prompt are special or treated differently?
A lot of what the AI does with non sung text are things it learned by itself, and not a feature designed on purpose from the start. As far I know the tags like [chorus] [verse] in Chirp V1 was actually one of these random things the AI learned, and now in this version, V2, those tags were formally made into a feature, on purpose. Other stuff treated differently is:
- Commas in style field, separating concepts. Intent was to make the order of words not be too fiddly.
- [brackets] in lyrics field, words in brackets seen by the AI as a different than lyrics. That doesn't mean it never sings them but just generally.
- Some tags in brackets are deliberately boosted or added. [chorus] [verse] and some other tags (a list would be good...) had extra effort put in to try and get Chirp to pay more attention to them and work more reliably. Many other tags the AI learned by itself, though that doesn't mean they can't be surprisingly useful or even pretty reliable.
- Some common music styles had a little extra elbow grease too, not sure what. I think the big words, country, etc. Probably instruments? And reportedly more coming in next version, so more things that work sometimes now, should work most of the time.
How do I keep multiple artists.. if it's a male and female.
That started in the beginning
the last 10sec of a chirp are sampled to create the Continuation....
If your sound (whatever it is: singers, rhythms, tempo, bass) is not within the last 10sec, it will not carry over into the Continue ⏩ chirp.
There is always the option of regenerating 🔁 the lead-in chirp and hoping for the signature sound to be within the last 10sec..., but once the signature sound is 'lost' it won't come back. Chirp will create a new voice/groove/bass instead.
Aww I see thanks
/chirp
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.
(The following has been translated from Chinese to English using ChatGPT, so some wording may be distorted.)
Hi, I'm a Taiwanese user who just started using Suno Ai a few days ago. I have three questions I'd like to ask:
The first question is that I've noticed generating longer music requires multiple iterations and merging.
However, if I generate music more than twice, it leads to a highly inconsistent style for the entire song.
Is there a command that can make Suno Ai generate a consistent style? For example, making the chorus have an identical style.
The second question is about generating more challenging Chinese songs.
It seems that Suno Ai may not be able to generate songs smoothly if it can't understand the meaning of the lyrics.
Sometimes, there's only background music without vocals.
Is there a way to address this issue?
The third question is, how can I give commands to specify a male or female vocal for generating music?
AI = 🎲
- Chirp samples the last 10 seconds of audio, to Continue ⏩ the next chirp. Your sound (singer voice, base, melody, drum pattern) must be heard inside the last 10 seconds of audio, or it will not carry over.
(next versions (?) are said to improve the context length, but no announcement when it will be ready for public)
- True. Lyrics help to style the music.
I have heard interesting results with lyrics that are just repeated gibberish or 'sound' words (Ahh-AAAHHHH-ahhh, AAAhh aah, aaahhh...)
Chirp doesn't need to understand the words to find music and rhythm, but the words can influence the style of song. Lyrics about strong emotions usually make the song sound similar. (sad, angry, love, happy...)
Chirp uses an LLM (like a chatGPT but smaller)... some language examples here https://discord.com/channels/1069381916492562582/1154780350749351977
English is probably the largest vocabulary in the bot, but music terms are often international (same word for every language)....
- Suno devs say the next version will be better at this, too.
Until the new version is released, it might work to type female singer, female singer, female singer in the genre description box... repeating the desired prompt a few times to make it 'stronger'.
OK, thank you very much for your answer.
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.
riffusion is doing a similar thing... will this ever be open source?
I don't see it.
A short twelve second clip does not a song make...
could you please rephrase that?
Riffusion only seems to produce a 12 second part. No way to continue it.
It's definitely a riff but nothing to build on (yet).
(The above is my personal opinion based on information at time of posting.)
correct but it's still similar service
but will suno ever open source chirp like they did with bark?
Who says they ever have to?
that's why I asked
projects don't generally go from proprietary to open source. That would be unusual.
Far more likely Suno would sell themselves to a bigger company.
(read the TOS)
Riffusion used to be a text-to-song AI website, but the website updated, it became into a lyrics-to-song website. You can type lyrics and "hear them come to life", BUT, you can only type lyrics that has 5-20 words
Yeah you're right
yes i dont mean the lenght, i mean the text to singing+background music features are similar and was just wondering when they would become open source if they do
Idk man
Let me explore #🐣┃suno-showcase
Are there any plans on potentially allowing chirp to be installed locally similar to bark? I really prefer to use my own resources when experimenting with new software
I am trying to generate "instrumentals". Does anyone have advice to share that would generate a consistent maximum generation [40seconds/30seconds]? Here is a prompt I used that generated one 40 second sample and one 11 second sample: [jazz instrumental 16 bars][jazz instrumental 16 bars][jazz instrumental 16 bars][jazz instrumental 16 bars][jazz instrumental 16 bars][jazz instrumental 16 bars][jazz instrumental 16 bars][jazz instrumental 16 bars][jazz instrumental 16 bars]
Style of Music
smooth jazz
OK not at all scientific, but what is the "correct" approach to maximize generation time?
#audio-prompts
Sorry, how do I use that?
I posted a bunch of instrumental here: #📚┃suno-school message
I make mostly instrumentals, and I end up rotating through some methods:
-
leave the Lyrics blank
-
[Instrumental] in Lyrics
-
[continue] in Lyrics
-
'Instrumental' in Genre description
-
Genre and Lyrics blank on Continue
the MAIN TRICK is to be selective about the chirp you are continuing from... the last 10sec of that chirp is sampled to create the Continuation into the next chirp. The sounds, beats, instruments must be heard in that last 10 sec for t to carry over.
If that last 10 sec has a dip in volume or energy the next chirp often thinks its time to end the song. The last 10 sec might have a build-up of noise that will get worse in the next chirps.
- [Break] in the Lyrics of the new chirp can help clear the continuation of the last chirp's noise and singing
I can't hear your music, but I am guessing that if you count the bars in the 40sec track, it has a full 16bars plus about 1/3rd to 1/2 of the 2nd 16bars.
the Continue track is cut short to end just before the 3rd 16bars would begin. the 3rd chirp will start aligning with the 16bars.
(this is a common pattern I see)
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.
As the bot informed you, you will need to perform this command in one of the chirp-beta-X channels or you can DM the bot to perform the command in private. Also the bot is not as smart as you think, if you wish it to perform one of those songs you will need to find the lyrics and give it to the bot in that singer's style.
Good luck! 😄
/chirp
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.
Hi, I have a question. After running the algorithm many times, I've finally created music that I like. However, the music ends in a strange place. Is there any command that can make it generate an outro without lyrics?
Things to try (not all at the same time):
[Break] - interrupts singing or persistent ostinato sound
[Breakdown] - similar to above, sometimes more complex
[Outro] - usually starts repeating a sound for you to fade out in post
[Fade Out] - similar to Outro but will sometimes do the fade in chirp
OK thank you so much!
/chirp
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.
I think you guys were asking for major hallucinations before? How 'bout this:
Maybe?
Because I wanted to ask, is there always a context window? By which I mean does Chirp hum the previous ten seconds to itself beginning a segment every time?
Because I had a few that just jumped in and started singing.
So yeah, maybe Chirp really does love me?
Thank you, that helps
I was just "throwing things at the wall" with that prompt. What I meant to say was the initial generation gave me the two samples: one was 40 seconds long and the other was only 11 seconds (obviously not 16 bars)
ive had mild to good success with [reset to verse] as well
at the start of a continuation usually
/chirp
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.
/chirp
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.
I’m starting to think there are times where a stereo track vocal goes 90-100% phase out when converted to mono sometimes 🔊
I don’t have headphones rn to test the phase but o keep noticing really quiet vocals
a friend of mine made this with chirp and i'm kind of disturbed lmfao
I don't think there's an audio prompts channel? I'm curious whether anyone has a guide of all the metatags that aren't document, E.G. [break], etc. Does [bridge] have any effect?
some of the metatags were re-enforced by training
[Verse] and [Chorus] specifically... according to some old Suno docs.
other 'metatags' are trial and error. It's an LLM (like chatGPT) so it understands a lot of words you throw at it..., but throw too much and it gets confused.
LOTS and LOTS of examples inside the chirp videos... and you can compare how well they worked against the music that actually plays...
try searching for your favorite genre.... see what other people tried....
in my experience [Break] and [Breakdown] often work, but there's no real 'memory' from chirp-to-chirp other than the last 10sec of the previous clip.
Using [chorus] should get you chorus-energy, but you won't get the same chorus a 2nd time....
(I think [Bridge] works about as well as [Verse] and [Chorus] do... they help better than 50/50, but sometimes the AI does what it wants anyway)
So I had ChatGPT write a song and I tried my best to combine clips. Interesting slight tempo change when I used [break] since the samples were starting to get a bit distorted.
@thin flint liked a clip
Not sure how to paste embeddable clips in here.
What is the best way to get an instrumental break before the song properly starts, E.G. an instrumental intro followed by vocals so the song doesn't feel like it starts abruptly.
can download the clip, then upload with your comment?
We've been using [Intro] in the lyrics box with no lyrics....
but that sometimes makes a very short song (like a video segment 'intro', about 7-12 seconds then ends)
now I am trying to get the first chirp as an 'instrumental' with nothing in the lyrics box... but can then be a struggle to get singing in the Continue chirps....
(might have to waste a few chirps until it does what you want)
Idk why having a genre like “very heavy metal, upbeat, annoying” by itself works, but when I put lyrics that have the words “HATE” or “HATE YOU”, it trips up the filter
I don’t get how that works but uhhhhh
How in the world
/chirp
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.
does anyone know the tags which can be used to either change the songs vibe like if you repeat a word in (word) you hear it twice in like a echo but does anyone know every tag which can be used?
it's an LLM. The bot understands tens-of-thousands of words in 50 languages.
there is no list, it would be the dictionary.
Some things work better than others. Suggest you look through the chirp betas and showcase, especially the ones with similar genre to yours.
//chirp
That's what is so amazing about Chirp. The developers flat out admit they don't know how it suddenly started doing certain things but they have reinforced them when they happened, like you said. I think it's awesome that a large part of doing stuff is experimenting to see what Chirp will do.
This experimental stage is what I think gives our songs the life they have. It breathes something truly random into the act of creating a song. It's what seperates Chirp from so many of the other existing projects.
there is some info here....
https://suno-ai.notion.site/Tips-Tricks-Chirp-v1-969852c74c644c6b8262ec5c5be2325c
pure personal speculation just for fun, but often feels like Chirp (and bark) may be sneaking more information into the 10s than you think. maybe even quiet or unheard bits carry extra meaning even if you can't hear it. something analogous to these tokens, which seem at first to not be used in the output, but when you look closer you find https://twitter.com/TimDarcet/status/1707769575981424866
how do i change the ai singing into just silence and instrumental it seems like it keeps singing lol
Try using no lyrics or stopping the lyrics very early and only putting [melodic instrumental] or even [melodic instrumental][piano] on the line where any lyrics normally go. Sometimes you need just one line, but sometimes you need it on every line almost like lyrics. Either way may need to get a little lucky. You can also try instrumental in the style prompt, even repeated a few times.
thank you so much! is there also a Possability for rap mostly to make the ai rap faster and then slower?
That one I'm not sure about, I think I saw people discussing speed changes maybe a day or so ago in here somewhere. You could try using regular style prompt + adding slow, slow or fast, fast to the end or something. Also in general longer prompts (more lines of lyrics and also more words on a single line of lyrics) tend to be sung faster than shorter prompts. Not a strong effect but could help
oh if there is a typical music term for the genre of faster or slower style rap that would better than just fast or slow, probably
00000000
Try ‘acapella’ or ‘vocals only’
Cant seem to keep the same singers voice and tone from each continue, i leave the category and just change the words after clicking continue and its voice and tone is changed, if i fill in the same category it also changes. What am i doing wrong or is this a useless program past 40 seconds?
Same Voice Continued 6x ⏩ ⏩ ⏩ ⏩ ⏩ ⏩
I tried to type lyrics in Japanese, well the results were... bad. It's just an instrumental
can't see your full genre description... Did you include a description of the vocals or singer there?
2nd chirp has started singing. Do you want to try Continue ⏩ and see if the 2nd chirp is ready to sing the words?
The « style instrumental » is the best prompt to generate song ideas. But it feels like one style is tight to one band.
« Pop-punk » will be very close to modern pop-punk like « The Story So Far ».
« Nu-Metal » will be close to « Korn ».
Also tried with band names but it’s not great…
the genre was: "anime opening theme, upbeat and catchy"
Didn't say instrumental
Also i ran out of chirps
You didn't say a singer either. 🤔
That might make a difference
Sometimes when i continue a more meaningful lyrics on a a clip with a lyrics like "lalalala" and anything like that, it just repeats
I've started using [Break] at the top of a Continue, when I want to try to interrupt some noise (vocal, annoying instrument).
Has anyone managed to figure out a way to get Chirp to keep a beat vocally?
I was showing my Dad my version of a song which was in the public domain and he said that since it was missing that vocal intonation it really wasn't the song any more and felt flat.
I mean...I like my variation of it.
But he's right.
So any idea how to fix this?
It's this classic: https://www.youtube.com/watch?v=fun7CBPyeRk
Get beyond the opening "Tweedle-lee-deedle-lee-dee" section for an understanding of what I'm talking about.
Bobby Day performing Rockin' Robin, his classic hit song from 1958. Rockin' Robin tells the tale of a lively and cheerful bird who loves to rock and roll, with a catchy melody and upbeat lyrics. With Rockin' Robin, Bobby Day invites listeners to join in on the fun and dance along with the bird as it sings its heart out.
Bobby Day's rendition o...
This is my version, which I consider transformative and enjoyed making to a point.
The thing that really stuck in my craw though was when he compared my version to Pat Boone's version of Alice Cooper's I'm Eighteen.
Not a good look for me.
So any ideas (other than waiting for the next version and hoping it's there) on how to fix this?
🤔 add vocalization text in parentheses between each line?
Not sure how that gets me this vocal rhythm.
(clips from public domain song.)
I feel sometimes that there are certain vocal rythmns that I either don't know how to access or am doing it wrong.
Of course, I'm a sucker for a good voal rhythmn.
... rockabilly ...?
Editing screwed me up... ate my text. 😭
I called out as examples of vocal rhythms I am fond of and gave Red Hot Chili Pepper's Souls to Squeeze song, and the Wham! Everything She Wants song as examples.
.. rhythmic handclaps and fingersnaps...?
(writing that one down to try)
It's a vocal beat.
He rocks in the treetop all the day long
Hoppin' and a-boppin' and a-singin' his song
All the little birds on Jay Bird Street
Love to hear the robin go tweet tweet tweet
break up the wur-ur-urds
[Elvis]
Definitely an inspiration. I'll let you decide who inspired who. 🙄
As a genre? Hmmm...
I know they say that there are no performers as styles but sometimes it does seem to do something
it seems to work better on names that are >30yrs ago
yoink
steals this
🥹
Guys, should i describe tempo and music key to generate a song here?
I'm sure it can do something but so far I don't seem to be triggering this at all. 😦
You can try but AFAIK it doesn't do anything at current time so people tend to try to do other things to make those work.
Nothing at all like what I wanted but I have to revisit this one again:
https://cdn.discordapp.com/attachments/1157347834287362170/1169665555704324298/chirp.mp4?ex=65563aef&is=6543c5ef&hm=d8251846acdadc016bdb619b328ea2665977dbae79d24ef2654bb4a107540ba5&
in my experience, that doesn't work...
in general 'loose' control gives best results (letting the AI do most of the work, and regenerating with tweaked lyrics to fit the music you're hearing)
why you always getting the beats I want? 🥺
Chirp loves you better.
Now you're really going to hate me.
`[fade in]
[intro]
[rhythmic handclaps]
(clap clap clap-clap)
[rhythmic fingersnaps]
(snap-snap)
[funky bass]
[melodic instrumental]`
Style: Elvis, rock and roll
https://cdn.discordapp.com/attachments/1157347834287362170/1169666524588544041/chirp.mp4
https://cdn.discordapp.com/attachments/1157347834287362170/1169666525272223754/chirp.mp4
That second one has some really nice 'Bang a Gong (Get It On)' notes to it too...
I see. I wish guitar pro can produce realistic sound like this lol
i thought up another:
syncopated staccato vocal
Say that five times fast.
Forgive me if I take a bit to come back to this, I've hopped aboard the Bang a Gong sound and am riding it to see where it goes. Literally making it up as I go.
[cosmic surfer gif]
Think I'm going to stop this one here. Kept getting noise added to it and mucking about with the instruments wasn't really helping.
I suspect the handclaps and the fingersnaps may have doomed me from the start.
Either that or the "Elvis" did.
(Doesn't sound like an Elvis I've heard...)
/chirp
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.
I would like to continue my own song I made
I have the entire project file of the song I made
Unfortunately, the devs make it pretty clear they don't want to allow fakes and clones of radio songs, so it's probably not going to happen
(until a generative music model can be run locally)
/chirp
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.
/chirp
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.
Hi, if I buy Pro Plan 10$, do I still get 50 credits per day for free?
no, each plan is separate. it would be cool if pro could get bonus by logging in too, maybe post a suggestion
/chirp
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.
/chirp
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.
Hi, is there a way to get the ai to make an ending for a song rather than just stopping?
You can try several things.
One thing people do is create a part with an [outro] tag in the lyrics field to give it some space in the song to play off the ending. Then they do a final segment and spam the lyrics field with [fade out].
I tried the outro tag with no luck but not tried the fade out that’s a good idea thanks man
Will give it a go
Please fix the buttons. I cant continue
This is a known issue, please submit it to #🪲┃bugs-web so it is known to be more widely affecting people. Am I correct in assuming you are on an iOS device?
Ya
+++++
Hi, I'd like to ask a question. I've written a song with 4 lines of Verse and 4 lines of Chorus, and the generated music is slow-paced. It ends at the middle of the Verse. Is there a command I can use to make it continue generating based on the original content rather than starting over?
yes, you can use https://discord.com/channels/1069381916492562582/1149508588470022164 to extend the song... typically just add the left-over lyrics to the Continue pop-up.
Приор
I was wondering if there was a way to change the vocalist from male to female between versues for a duet?
right now that's still a bit tricky, you can try something like "female, female, female" in the description but will likely still be quite random. next version will have some better control around this
Dansong2023
/info
Why is chirp bot stuck at the pre-generation/lyrics phase? when attempting to generate more/regenerate I am told to wait for jobs to finish.
Has been over 10 minutes
Having issues myself. Also noted that "Death Metal" as a style was flagged as a prompt. I'd not had that problem the last time I tried that style for something halloweenish. style censorship? booo!
Is there a way to keep the voice from changing or remember the same voice from one continuation?
Yes.
Try to always be concious of the ten second window that Chirp has for everything. If your voice is outside that context it will be altered.
It's why I always try to maintain shorter lyric submissions.
Do too much or end with a music note that is too far away and you lose the voice.
As an example:
My Dad and I were working on this for a bit when I was showing him how the system worked. He wanted to end it and then had a last line. So I showed him how having too much music between a vocal would cause the vocals to distort and mutate.
Still a funny song and I definitely intend to see if I can finish it.
Speaking of mistakes in losing the context window...
I was happy with how this sounded at first but then I realized I would never get back to the music with this one because the window had passed.
Going to try to reuse those lyrics though, since I think they still have potential.
Logs look like there were a few issues early in the morning but seems ok now hopefully
Since Chirp won't let me continue my song, here is what I had with the busted part I was trying to replace added instead of my continuation where Chirp wasn't hallucinating...
слышь а че
Is there a way to pick the image or change the image that pops up with the music?
Nah, that's automatically generated.
How do you upgrade?
type /subscribe
it will take you to a Suno.ai page about the subscription options.
👋 is there a way to get the same type of sound when I buy a subscription and add suno to my own discord chanel?
here when I say continue with my old song, the AI voice is the same and I want to continue with it (I'm not talking about the beat)
everyone (paid/free subscription) gets the same bot, the differences are how much and how frequently you can chirp....
The trick to continue with the same voice is to be sure the voice is heard within the last 10sec of the chirp you are continuing from.
The last 10sec is where Chirp samples the audio (beats, voice, irritating noises) that will carry into the next chirp.
several of the jobs I submitted disappeared before they processed...is this a known issue?
Having some server issues today? Chirps are taking forever or disappearing altogether...paid subscrber
Yeah, looking into a few issues, hope to have a fix in 30 min or so (if something gets stuck wont count for credits.) Sorry for inconvenience
thanks MC
Generations should be more stable now. Sorry about that.
I love this track! Love to listen to it!
😃 ty
I'm trying to upload my song to Music Diffusion to publish it but I'm getting "Audio analysis issues"
check their docs? you probably need to submit in a specific audio format.
No, it's saying it recognizes parts of the music if that makes sense. It was citing specific areas that "matched"
did they happen to say the names of these alleged matches?
yes but I didn't capture it and can't find it now
M'kay... tbh this sounds like a Music Diffusion problem.
I've put in the lyrics but I only got music back. Does that happen often?
I don't know about often but it does happen sometimes. All you can do is reroll\and adjust a few things in hopes it helps.
Thank you. Do I need athe square brackets to indicate [Verse] and [Chorus]?
You don't have to do it but does seem to help guide things a bit.
One of my earliest Chirps, before I knew anything about what I was trying to do. As you can see I never even used a single [verse] or [chorus] in it.
thanks a lot!
Hi how to replicate voices like Like /trump /biden
Anyone have any tips for acapelllas that are good. Most of the time its almost atonal or just terrible
😆 try both 'acapella' or 'vocals only' in the description. you could also try adding things like 'ballad' or 'melodic' maybe? i feel like i usually get almost better melody when it's just vocals but not sure..
I'd also recommend using a cappella instead of acapella — didn't realize it was technically spelled that way and have personally seen better results once I switched my tags to that spelling
/chirp
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.
/chip
is there a way to control the speed or am I able to insert a pause, where I want it? Some lines seem to rush through them.
punctuation works: comma, ellipses
probably best results are when you deliberately break up your lyrics to span multiple chirps
less lyrics usually gives you better pacing and more 'singing'..., as opposed to talk/rapping quickly through a wall of text.
Absolutely this!
Chirp is sometimes like a child with a school assignment of a certain length that is overdue. They begin to write in larger text as the paper goes on to fill it up quickly. Chirp will do music and start singing and suddenly realize they're almost out of time. At this point it will speed up trying to make it all fit and ruin the song.
Piece by piece is the best way to go.
I love this and am a HUGE supporter~
Is there something like a seed value, which you could use to bring back a certain style that you liked on a new song? Something more specific than the genre?
not anything us users can access...
Might be one for the https://discord.com/channels/1069381916492562582/1154587937611190354
So does that mean four lines of lyrics per chirp? Or eight max?
Four, usually is as long as I'll willingly go.
This is complicated by my habit of breaking the lyrics into shorter sections too.
So while it may only be four lines, I could break them into shorter ones.
Scrolling up earlier there is an early Chirp I posted:
You can see what I mean by that.
/chirp
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.
not working in the channel. no pop up
Are you publishing a raw Suno track? Good for Youtube but I wouldn't publish it to the official streaming services, tracks are in mono for one thing.
Do you mean this example? Can't see your prompt...
You can throw the track in a DAW , and make it stereo
Interesting. I have Reaper but haven't really learned it.
I use fl studio, they have a free version of fl studio as well . They have plugins built in that are ok, when it comes to turning a mono track into a stereo track
/chirp
I meant this one.
guys, can someone tell me, where I can see the amount of my credits?
Hey Rinso! Sure thing 👍 just type /info in the channel or dm where you are creating!
thx!
Happy to help!
and can u tell me pls, how can i change a basic plan to something other? "Mange your account" button dont do any 😦
Certainly! The manage your account button should create a pop-up that will send you off-server to https://app.suno.ai/account/ where you can adjust your subscription as needed
Not directly Suno related but I just made a new Suno track and uploaded it to Youtube. I made a Youtube thumbnail for it inside ChatGPT but Youtube, as of the last few songs, hasn't accepted the thumbnails at all. They're PNGs, an accepted thumbnail type...
Yes, I get to this page. But where do you need to click to log in or register? Nothing happens when you click these links
The maximum upload size of a YouTube thumbnail image is 2MB. I would check to make sure that you are under that.
Give me a second to test run this, brb
just “sign in” changes to “sign up”, and so on in a circle...
That’s weird. Seems to be working well on my end. Try this…
-
Update Your Browser: Please ensure you’re using the most up-to-date version of the browser you are using.
-
Browser Extensions and Cache: Extensions can sometimes cause unexpected behavior. Try disabling them or switch to incognito mode which disables most extensions by default. Also, clearing your browser’s cache removes temporary files that may be causing issues.
-
Stable Internet Connection: A consistent and speedy internet connection is crucial for any upload or export tasks. If you’re on Wi-Fi, consider a wired connection, or at least ensure your signal is strong and stable.
-
Reset Discord
Thanks, I'll try a little later
How do I get a chirp longer than 30 seconds?
is the server down? /chirp isn't working in the beta-1 channel
Yes, we're working on it — will be back up shortly!
Welcome to OpenAI's home for real-time and historical data on system performance.
🙌 Chirp is back! 🙌
not for me yet :/
why does the pitch of the voice keep going up with continues?
my theory: it's the sampling rate.
There's just not enough info in the high-end (treble turns to mush)..., and leads to degraded sound quality over multiple chirps.
I know this was probably asked million times but What is the quality of the Chirp music? (compared to a 128kbps MP3)
Looks like it's 69kbps.
So better than radio but worse then CD.
vocal start so late all the time how can fix that
Is it possible to use famous synth sounds e.g. Synthesizer Roland JV-1080 -patch Bass Pizz ( e.g. Faithless -Insomnia was made by using this one)
would like to have same kind of vibe to my song
what happened when you tried?
Just low level blur sound
technically it's upper limit is roughly that, but in practice, since it's an end2end model it completely depends on what is "hard" for the model. so in many cases it can sound like lower quality. should get better as the model gets larger and better 🙂
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.
i've never gotten anything useable when prompting for a 'synthesizer'....
My theory is that 'synthesizer' is just not a well-defined concept that is easily explained to a language model (AI is already 'synthesizing' everything: voices, instruments, percussion....)
A dev told me this is probably more an issue with the (small) language model than the generative AI's ability to make sounds, and might improve as the LLM is updated.
I have better than 50/50 results describing an instrument and the quality of playing:
sparse plucked balalaika syncopated glass marimba electric bass arpeggio fingerstyle acoustic guitar
etc
Hi. Why i have "this prompt has been flagged?"
It could be an issue of the lyrics being too long or it could be that you've used a word or two that was considered naughty by the AI content moderator. Let me check and see if anything is listed so I can try to tell you which one it may be.
....
OK nothing is coming up in the feed under your name. Were you using any words commonly considered to be "swear" words or words known to trigger ChatGPT style AIs? I had one song get flagged for the word "stupid" in it so it can be that basic.
hi
I have a question
I subscribe 1 year for pro
I think it is not applied yet
because I cannot chirp anymore
hi, is there a chance you logged into web on a different email or discord? Just email us at billing@suno.ai with your discord handle and your email for payment and we’ll get it fixed for you!
(Replying to your message)
DM'd. Sorry for the problem!
It is OK I can wait no problem
because this is really cool.
now it works thank you for your cooperation.
Hi, I have a question. I'm writing a fast-paced song, and I want the generated music to have an intro in the first half. What command should I use?
/chirp
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.
Hi guys! Is there a way to easily generate a continuation that matches exactly the melody and rhythm of, let's say, a chorus that was cut off in the first generation? Especially if that first generation had the last 10 seconds clearly defined, or do you have to just keep generating until you get lucky? I cannot get the same "flow" of the last 2 lines of the chorus Example:
😕 🎲
I think it's either rolling the dice until you find something you can live with, or trying to edit the music in audio production software...
Depending on your subscription (how many chirps you have available), it's sometimes easier to force shorter chirps by feeding it fewer lyrics at a time... but that's no guarantee for consistency, just potentially fewer cut-off lines....
Thanks! That's what I was thinking. I hope they will soon increase the length of clips, at least by another 10-20 seconds.
Working on this! These are great tips
It would be great if there were a feature to "block" the instrumental intro when the lyrics consist of 4-8 lines. The instrumental intro could be limited to a maximum of 10 seconds to ensure ample space for, let's say, the verse and chorus. In the example I uploaded above, the instrumental lasts for 18 seconds—quite a lot! This excessive length encroached upon the space needed for the final two lines of the chorus. Those extra 8 seconds could have easily accommodated the remaining lines. Additionally, it would be excellent to have a "backward continuation" clip generator, allowing users to extend the instrumental intro introduced in the clip with lyrics. Furthermore, having the ability to create an exact copy of the "chorus" with the option to add more elements, creating a sense of growth, would be a cool feature. The structure remains the same, but with additional elements subtly layered in the background.
Love this one. Some of the later ones will take more R&D than others. We are thinking hard how to make instrumentals happen less
Just so long as it stays optional. While sometimes I don't like how long my intros are I do like my instrumentals...
/chirp
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.
/chirp
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.
Another awesome feature would be a browser for every publicly generated clip, allowing users to search by keyword or line lengths. For instance, you could search for something like "nostalgic" or a combination like "nostalgic, rock, soaring male vocal," and instantly get all the related results. It would also include a ranking system for the most frequently used keywords. This way, people could grasp the kind of results they can expect with specific keywords, discovering the most effective ones. Plus, it adds a fun element to exploration, leading to better results with fewer random generations. Something in the style of Midjourney.
Keep 'em coming!
The next cool feature could involve matching the style or vibe of a song by uploading, let's say, 30 seconds of audio. The AI would then give it a shot, doing its best within the database constraints to align the generated song with the original file. This would be especially helpful for those wanting to recreate a specific style without dealing with complex, vague descriptions. However, the potential issue with copyrighted music is uncertain; this suggestion is to explore whether it would be allowed. Another feature I'd love to see is a "slight/stronger variation" remix mode using a seed. You provide a seed from a specific clip generation, and the AI takes a swing at remixing it. It would work similarly to continuation, except that you can either remix the entire clip or choose a specific time to make a subtle, more significant change, or introduce something completely different, all while keeping the rest intact. This would be incredibly helpful because there are times when you absolutely love the first part, while the second is just meh, and you really don't want to continue from that.
/chirp
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.
Hi! Assume it's totally cool to upload the Chirp songs to Instagram/social etc yes?
Is there a list of tags for structure available. I saw [chorus] and [verse] mentioned. Is there something like intro, outro, instrumental ?
yes. Free subscription is non-commercial and attribution (the video has attribution). Ok on socials or background youtube
paid subscription gets commercial license. Ok on Spotify, streaming, advertising, for-profit media.
I paid for subscription, but bot still says "You don't have enough credits to use this command."
You can access your acct info through the chirp bot by typing /info ...
/subscribe will also take you to your acct on Suno website
the common advice I see is about making sure you joined with the same email as your Discord acct.
Might need to contact billing@suno.ai to get things sorted.
Yep please drop us a line there if you're still running into this problem after checking wetcircuit's tips, and we'll get you back up and running
hello there ! do you know if i can use a seed command ? as an example, i have a generation that i like a lot, nearly perfect, and i wanted to know if i can redirect the continuation of the sequence to keep the same tempo, the same voice, or even the same rythm
Unfortunately this is not available at this time though it is a requested feature often and is something I know the team is working on.
You could always try leaving a message in #💡┃suggestions-web to show there is continued interest in this feature?
how to one improve sound quality? SUNO has better sound quality, but when genereted on discord server sample rate is low
It's a pain point. The webapp is actually using the next generation engine, while we poor Disocrd users still have the more creatively powerful version but that one outputs at a lower sample rate. We're supposed to be getting that next generation engine Soon ™️ and so for now all we can do is wait...
😅 sorry yes discord will follow in a couple of days together with a more proper announcement on next model capabilities
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.
If you purchase a subscription on the suno website and get credits are you allowed to use them here on the discord or only the website?
should work both places, but they need to use the same email to connect.
Type /info to check your tokens here.
(popup widow only you can see)
If you run into any issues here just email billing@suno.ai
I have a question, is there any way I can somehow ask the AI to create a remix of a certain melody?
Is it possible to change language of the generated songs?
thanks for your answer ! looking forward to that !
I'm guessing the website version of chirp doesn't let you feed the seed either?
I'm under impression that chirp uses bark under the hood, but with special model, right?
🥲 not yet....
(or we need to wait for v2 to work here?)
getting seed would be so good, because it would open so many possibilities for small tweaks
I'm aware it won't gen the exact same thing, but at least on regular Bark (I've used it locally here), you can often change individual words and still get overall same'ish result for the whole thing
when getting some cool things on chirp, it often messes up individual words in some places, so could try to correct them with slight prompt change while using same seed
Is it possible to use our own lyrics with the new model right now or is that not available yet? I was gonna check the site later tonight when I am on my pc but if it's not there yet I will wait to test it
not yet... website is set up to generate with chatGPT lyrics....
Ok, I will have to wait for a while then, thanks for the reply!
oh
chirp v2 is actually stereo
(it's been mono so far)
just realized this by playing some clip on the showcase channel
I don't really hear huge difference on the audio quality itself but stereo effect does has spatial effect on the output
could be neat to have the option between the two
v2 still misses on the clarity of say, stable audio.. but I can't make stable audio produce any music that sounds even slightly interesting 😶
I wonder what the actual true sample rate currently is
I know these get stored at 48kHz but there's nowhere near that range used, or alternatively things get muffled by the procedure otherwise
🧐 should be possible to look at the audio wave and see how smooth the sampling really is...
resamping the output to 24kHz on audacity has minimal impact on audio quality
it does change the output slightly though
32kHz does too but I can't hear any the difference between 32 and 48
this depends a lot on the track too
so I guess they may use 48kHz range for real, but it just gets the distortion and bit muffled sound from other reasons
this is especially noticeable on any music genre that actually has distortion on the instruments
Heavy Metal Deathcore Dubstep for example
🤔 📝
is the chatgpt template for the lyrics gen using chirp known?
mainly wondering if it is something super simple like write {prompt} lyrics or if it's way more detailed
Theres probably some line thing that says 8 lines max or smthing
it's actually quite hard to make chatgpt obey that kind of instruction
but they could just crop it manually too
Write {prompt} lyrics only using one verse and one chorus using 2 quartets mayhaps something like this
the reason I ask about this is because if I know the whole template, I can hack around it and force it to feed my own lyrics... and get the try that v2 early on 😁
but I also don't know if they limit the chatgpt input field length on chirp
I mean it could be using its own version of GPT and not ChatGPT :>
I basically mean this here
this would work if the template is write {prompt} lyrics and you'd just fill in that middle part as instruction
it should work on any LLM really that is able to follow the instructions
but chirp could just cap the description short
How do you continue to expand a song in the website version?
Website version is in beta version right now, so you can't. Soon though, so I hear
Thanks for the info. Is it true the website is using a new version of the bot?
the website version has audio only version that is way higher fidelity than the one with video
...why
there's huge quality difference here
this is due to mp3 vs aac compression on the audio. it's not even about bitrate
avg bitrate on audio is the same on both, but AAC is NOTABLY worse, it especially trashes the high end hihats and like
even mp3 is 128 kb/s which suggests that the actual gen without compression or with higher bitrate compression would be way higher fidelity
to give comparison this is the chirp v2 from webside, video version
this is the same thing with manually muxed audio-only mp3 from the site
you can clearly hear the difference from first second
I'm going to do a frame challenge and say 'so what.' 🤷♀️
The video is for sharing on social media, where the majority of people are on a phone with speakers the size of a pinhead.
If you intend to do edits or want the better quality audio, why would anyone download an mp4 if there is an audio track available?
I disagree that it is 'night and day'
Open the waveforms in any DAW you can see they are not very different,
....but sure no one is contesting that a social media video with a different compression sounds slightly different...
I still say 'So what." just download the mp3 for yourself. It's a smaller file anyway.
I dunno, almost everyone I've played chirp gens make a note about it's bad audio quality, but they are impressed about it's creativity
and I do know the main audio quality issue is the distortion
you get that from that bad compression
I also wonder if there are some devices that can't play that mp3 audio track when muxed to the video
because... it takes exactly same space
just sounds way better
yes. AAC operates with less overhead. It's also rated better at lower bitrates (possibly lower than this), so yes.
the video should have BROADEST compatibility.
(work on a potato phone on the worst web browser)
the audio should be HIGHEST available quality.
(we don't get WAV... so, mp3 it is)
this is how it appears to be.
(obv I did not sit in the room when these decisions were made, but creating content that plays across all social media is an exercise in sacrificing quality for compatibility)
/CHIRP
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.
I get why no WAV, it's notable difference in file size, obviously I'd prefer that too, but I do see the issue on bandwidth, storage etc when used in service at this scope
FLAC is probably too niche to even consider
Absolute better to get as raw audio output as we can, but for the curious here, the distortion is most cases (the kind that gets worse over time) likely isn't from bad audio codec compression, you can see it even in Bark setting output to raw wav, it's more fundamental to AI in general and how transformers tend to accumulate errors, you can see it commercial voices like 11 labs, you can see it in ChatGPT, etc. You can see it diffusion audio models too, interestingly https://arxiv.org/pdf/2308.02560.pdf
so 320kbs mp3 is easiest compromize we could get atm
ah, I've used bark locally as well.. so I know it's not perfect.. but current chirp is notably higher fidelity model already.. but just listen to those two clips I linked here and you can tell the main issue is the compression between those two already
there's notable audio compression artifact from the first videos AAC compression
Might be just an ffmpeg flag that isn't set quite right or something, it should have plenty of bitrate
both have same avg bitrate, videos aac audio track and mp3
when I muxed the mp3 to same video, it's the same file size
mp3 is just superior in quality between the two
not saying you are incorrect, it could be bad compression setting
Right, but AAC should be better if anything. So maybe there is like a bug or something? If so would be easy fix. Whereas offering .wav as a third optino probably takes programmer time. Just total guessing
hmmm, apparently M4A can do lossless ALAC too
apple seems to use this format, so it's probably somewhat well supported
but will it play in the instagram browser...?
🥲
ah, I was more of thinking this for option to get lossless audio download
it wouldn't have to play on every device
but why pack it in an MP4?
(no offense to the people that programmed the lyrics video)
I imagine the uncompressed stereo 48kHz wav file size is nontrivial matter when people use it for this many songs
that being said, I don't know if Suno still stores the uncompressed audio once the gen is finished but I doubt it
what I mean is, we could just download an audio file in whatever compression scheme...
Or if it uses it's own file type, just download that...
we don't need it inside a video slideshow - that's extra steps to compress, extra steps to demux...
yes
I could see benefit of only having high quality audio file without any extras
I imagine the video is there to spread awareness of the service tho
No one needs Aunt Edna accidentally posting a 1400MB 1min goof song to her fb group... lol
I only want as high fidelity version as possible, regardless how it's shipped 😄
my surprise on the video vs audio quality here is because the audio track is literally the same size on both but video sounds worse
I didn't realize there was difference between these atm
so, thanks for pointing out that browser site
Compression doesn't work like that...
if you look at benchmarks, different compressors have sweetspots where they are better, but at other settings there is little difference (or even worse)
the reason for AAC is web compatibility, which might also include specific bitrate settings.
It's really rare when you can take advantage of 'best' compression schemes, just cherrypicking audio and video compressors and rates for best quality.
normally the goal is
a. compatibility
b. filesize
c. compression time
d. decompression overhead
...
z. quality
I did look around and general opinion does seem to be that AAC is supposedly better quality in general to MP3 when comparing same bitrate.. so there's something just off here right now
it's def possible.
one issue (not claiming I know better) might be the time it takes to compress the file. a single pass might be 'good enough' where as a double-pass might be better quality but take more than 2x to compress....
they are providing 2 'songs' in <2min... Suno seems proud of that stat. lol
🤔 idk if that's an 'live' stat that is an estimate of real eta...
What words can’t be used in a lyric? I am getting flagged for using a few run of the mill swear words in a metal song
Try using phonetics to help. (If you know what I phucking mean, friend.)
That can easily be gotten around at this time.
Vocal processing needs definitely tuning. Especially s-sounds are too much compressed. I'd prefer uncompressed song data, but I understand the technical issues. A good compromise could be less compressed and focus in vocals, more natural hearing.
this discussion is helpful, we may DM you at some point for more detailed feedback.
Uhh you prolly ran out on one end and the website didnt register it yet
Does anyone know how to generate pure music using chirp?
Well wdym by that Instrumental?
- leave the Lyrics text blank
- use [Instrumental] in the lyrics instead of lyrics
- use Instrumental in the genre text
(text reposted from an original post by wetcircuit)
how to make 3min track
#⏩┃continuation
Has anyone had success in generating a 'deep' male voice?
I have not had success with 'soprano' or 'baritone' – these don't seem to be words Chirp understands.
'Bass' gives me the instrument Bass.
Webapp or Discord version?
either, really...
begs the question: do we have any real promptability over the voices (or are there a set number of Bark-like voices inside Chirp?)
One last attempt and I give up on this and concede you must be right.
(Granted all my tests were in Discord.)
Ahahahaha! I did it! 😄
soulful jazz neofunk, deep rumbling velvet male vocals
( I may have been thinking of Ike Turner when I was generataing, hence the lyrics.)
Let me know if you can replicate it.
You have changed the compressing level? Vocals sounds much more natural than yesterday.
Actually, trying the webapp, I don't know if it pays any attention to directives on voices.
☹️ webapp is not reliable for genre or instrument prompts...
For testing purposes I gave ChatGPT the lyrics and used the music style as: soul, funk, deep rumbling velvet male vocals in the Discord version, of course.
After that I'll test it with the lyrics above in a different genre to see how it goes.
May have been luck.
Not getting it again.
I think it just blended well with a rumbling bass instrument...
suggesting it knows what you want...
(and maybe we are limited to bark voices?)
Pondering.
ungh... Discord prompt is so much bettererer 😭
I tried 16th-note arpeggio electric bass on Discord... perfect every time... (not a perfect song, but I got it 100%)
Trying deep velvet male vocals, since I was getting rumbling bass. I suspect you're right though and we have located a limitation. 😦
I mean..., if we could prompt for specific voices (by name)... that would be hecka useful....
Barry White voice plz!
https://www.youtube.com/watch?v=kGazEJKYe50
Soul singer Barry White, the sweet-talking, deep-voiced performer who rhapsodizes about love. His hits include "My First, My Last, My Everything," "Never Never Gonna Give Up," and "I'm Gonna Love You Just a Little More." Before he became a hit as a solo performer, WHITE put together the female vocal trio, Love Unlimited and founded the Love Unli...
(i need to pay a 10yr old to clone this)
lulz
Yeah Barry White would be better.
Or Turbo B, from Snap!
I extended it a bit. I like it but I don't know if I want to be the voice of Ike Turner...sounded kind of cool though.
fairly positive I have that same voice on many tracks...
📝 🕵️
It wouldn't surprise me if you did. I was more referring to the perspective I was doodling the lyrics from.
I don't want to be trying to see things from that man's POV about ..well just about anything, really.
Sorry for not getting this... how do I change language for generated songs? Just typing in the language doesn't cut it. I can't find where to switch preferences
Are you trying to do a dual language song?
write 10/12 lines about [Prompt] in v2 sometimes the song will end with outro instrumentals
Is there any way to keep a voice/style and make a new chirp with it?
No unless you get insanely lucky with the chosen genres
Any tips to get a voice change mid song it has gotten to scream and I want it a little more calm
Sample of the scream plz (also try using an instrumental or interlude)
Thanks for your reply. No a single language song. When registering to this channel I was asked for my preference language. I can not find the setting and only typing the Ai request in another language gets translated to English.
please see examples in https://discord.com/channels/1069381916492562582/1155499376869916782
or if I've guessed the wrong language there are many examples in language specific channels: https://discord.com/channels/1069381916492562582/1154612901030678578
Hi. I have an active subscruption. But bot says there is no left coins
Going to ask the obvious here, but did you use all your coins? (Sorry, it has to be asked.)
What does it say in /info for coins?
@neat nexus info says I have used all coins, but according to subscription I have 2530 coins
i need some guidance. I'm helping a musician. He wants to upload 100 of his original songs. Then have AI copy his songwriting style and voice to make new songs. Where do I start?
This is definitely a bug. Please email the team at: billing@suno.ai to let them know about this so it can be corrected.
Chirp does not copy songs or voices.
It is text-to-music generation.
Maybe Bark might be more closer to what you need? You'll need to set it up yourself.
Do you know what program does this?
Try looking into Bark: https://suno.ai/docs/faqs/bark
thanks
It took like 100 different generations but I eventually git something to work lol
@broken lotus it does Swedish but with half the song only instrumental. Also 1 of 4 translates back to english.
Don't know how deep you are in to this. Don't want to be a bother. Just trying to give proper feedback 🙂 over all it's awesome
When are they expanding the 40 second limit? I mean bark is like 6 months old and it still has the 15 second limit, so i'm not so sure if they really will expand the limit of Chirp any time soon... hopefully i'm wrong.
V2 first prompt is 1:20 at Max they can expand it
HWhen did the webapp get released? Last week it was only on discord...
v2 not officially released... but they're building the web app on it so we've been trying it this week.
Suno says they're adding v2 to Discord, and also the full Discord controls onto the website.... not sure which happens first, but it's in progress.
are you using Discord UI? that UI is more reliable rn.
The website UI is not complete...
Discord on Android phone
Ahh! so you should be able to just type in your native language... in the large text box, add your own lyrics.
(leave the ChatGPT box empty, that is the only thing I can guess would try to translate..., )
Hello. Please, help me with one question - can I create songs more longer than 40 sec? Is it for subscribers only?
They will expand this in the future, V2 at first it's now 1:20 and they will expand it
see #⏩┃continuation for now
what are the current credit top up offers for pro and premier?
you'd think those would be visible on the subscription page before you sub but they aren't
(so it's hard to make educated decision between the two subs, on which is better suited for you)
these are same for both tiers?
as far as I know.
It is same pricing from before October when premier did not exist.
tbh, current free tier is bit too good in comparison to pro tier (150 vs 250 monthly chirps), but premier does look interesting.. if the top ups had similar rate to monthly sub amounts, pro would be more appealing in general
Is there still no info when suno bot will be update to the web site version?
We're working on it, probably sometime this weekend or early next week
thanks
@light ether are u planing to create a paper about the chirp models?
Maybe one day...
did a lot of the architecture chage?
😭 i guess thats a no
Hmm, how do I generate a song with lyrics I chose on the website?
You cant as of the minute you should be able to soon as confirmed by the Team™️
nice!
looking forward
the song quality is SUPER high
this is a big, big, BIG step up
@light ether I have a few ideas to improve the UX on your site.
- Include the original prompt or add a button that allows us to recall that prompt somehow. That way I can take a prompt, repost it, and edit it a little as needed if I forget to copy and paste it.
- Option to delete the shittier songs? Some are just, y'know, bad.
- LOVE the AI art style you chose, but make a way to click and expand it so we can see it at a higher resolution.
- Unless I'm missing something, maybe if we download the audio, the AI art can appear for that song as album art inside the metadata?
Super!
- Since its good data for us, we are thinking about making heart like / dislike. Would that work? Then we can just have an option to hide dislikes somewhere.. Or do you want to nuke them from orbit?
That's quite interesting
That works for me anyway if it actualy personalizes per user :]
that's a good solution
it condenses the song to a tiny "hidden song" line
easy to skip over
but you can pop it back in if you want to check
maybe something like "Hidden Song [Short description here]"
Btw my clips does this already
so you can at a glance say "Oh yeah that's that shitty song I hid where Wallace and Gromit serenade each other"
also on the create page after 10 generations it starts getting rid of them from the page and makes em only seen in clips
btw, what are you using to generate the lyrics? gpt3.5? 4?
OH that actually kinda sucks
You can find all gens in clips still tho
Ah yeah we will fix that eventually to
lyrics failt but nice beat
a song with 80s wave style Talking about movie actors and their movies
Hey @vernal glacier, try that one on the web app at https://app.suno.ai
Would love to hear how it turns out
does/will chirp have an API in the future?
Somewhere down the road, possibly. As of right now though, no.
New to this app. Is there a way to prompt suno on the webapp (or in discord for that matter) in a way that guides genre primarily, rather than lyrics? Is there a way to prompt it in a way that will mostly avoid lyrics?
When the program is creating a song, are the instrumentals and voice separate generative models, or are they tied together?
You can just add instrumental to the description and keep the lyrics empty (discord only for now)
i already subcribe for Suno pro but in discort still free accout. what went wrong?
Probably different accounts. Please send an email to billing@suno.ai
any tips on how to stop the AI putting the "genre" in the lyrics? like, I want an aggressive rap or whatever and the AI will put that on the lyrics
If I like a song, any way to get it to complete the full song vs. being cut off at 1.2 mins?
I imagine once they inplement the own lyrics feature, they will also extend the duration to let's say 2 minutes since they're constantly working on extending the duration, as they say.
what i would love to see most is a way to take one of the generations and be able to tell suno to cut the song at a specific point, and then continue from that point, rather than only being able to continue from the very end.
some of my generations are amazing for the first half, but unravel near the end, and I'd love to tell it to continue from the best part so it keeps the same flow and style
Is there a way to continue a song made a few days ago, it was generated through the website before I started using it through Discord
Hello! I need help.
I buy PRO plan on website, but discord-chanel still give me just 5 generation per day.
How can I associate my account on website with discord?
contact billing@suno.ai
they will straighten it out
Maybe you can also have separate like/dislike about the voice. I get a lot of v2 songs with no voice at all or where it is nearly impossible to hear or understand. The music might be good though.
/chirp
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.
3 minute song!
#⏩┃continuation
Okay, I have a good clip going, but how do I get chirp to close out the song?
[outro]
^
can use it with or without lyrics, but it may also gen some nonsense lyrics there if you don't give it any
but yeah, you never know what you'll get 😁
Here is an instrumental I did a bit back...
You can see what I did to try things for yourself.
/chirp
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.
Thanks! Yeah, just giving [outro] got some nonsense lyrics. One of the gens had a lyric that had been repeated a lot in the song so that was ok. But sadly the best one got cut short. Will have to play around with it more when I have more credits 😅
That can happen. It's even worse when it happens in a perfect section of audio. Everything is great and then...WTF??? Where did that come from??? LOL
Final clip. Very rough sounding, but was the best gens I could get before running out. I think it starts out pretty good but seems to lose focus on the rhythm
I see it does understand tags to an extent, but it really doesn't seem to understand what a "chorus" is lol
chorus tag is pretty mixed bag right now, sometimes it works, sometimes it doesn't. I've also tried [pre-chorus] on previous verse but you can't really tell if it's just coincidence it put the chorus in right place at that point 😄
[bridge] also seems to work
(not for chorus but in general)
[Outro] also works :)
The only thing i Wish existed was [Breakdown] or smthing similar
is it possible to specifically ask for an instrumental song without vocals?
Web or Discord?
web
Try in the genre part of the prompt adding Instrumental with the rest of them
I tried it, but it still generarated lyrics and vocals, even though the genre shows up as "ambient soothing instrumental"
web always generates two songs. often one with a langth of 40 seconds and one with a length of 80 seconds. is the 40 seconds one V1 and the 80 seconds song V2?
That sounds like it could be a bug possibly. Would you mind mentioning it in #bugs so a developer can look into it?
what would be a bug about that?
If they're trying to keep both song options the same length. As far as I know it is supposed to be a situation where both options in the generation are v2 chirps.
I assume it's A/B testing to see how much more likes V2 gets compared to V1
@light ether is it A/B testing on the webapp or are they supposed to both be v2 chirps?
no - they are both v2, but it could be a quirk with instrumentals
it's definitely the most impressive music AI I ever tried out. really impressive how it can actually do lyrics. and how it can even somewhat do obscure genres like black metal, even if they come out significantly worse than more common genres
thanks! and next weeks update should make instrumentals much easier on web
nice! and are longer generations also planned? 1:20 is quite limiting especially with most songs being abruptly cut off at the end
will be able to "continue" similar to discord
so is the discord version superior to the web version at the moment?
for instrumentals, very possibly 🙂. but it's an older model with :40
so web has a better model, but discord has more "features"
ya. hope to get them around the same by next week 🙂
do "in the style of X" prompts work, or is it trained without any knowledge of existing music?
nope, genre focused (intentionally)
would be really nice if I could mention a bunch of black metal band names in the prompt and it would actually try to make something that follows their style
especially with metal there are like a million of subgenres 😄
it's funny how it puts something like (ooh-yeah) in the lyrics of a black metal song at the moment, which really doesn't fit that genre at all:
our lyrics engine goes hard on the ooh-yeah rn for sure
is it capable of doing harsh, growly vocals at the moment? I see it can do distorted vocals fine, but it always seems to come out as distorted and not really growly
:30?
Even tho for some reason "Heavy Metal Deathcore Dubstep" comes out nicely only issue i can find is that there is mo way to do Canibal Corpse level vocals
No sadly, ive tried
I'd love it to be able to do Amon Amarth style vocals
Suprisingly havent heard of them meanwhile me with the ~9600 song long metal playlist on spotify
Any thoughts on this???
I find it impressive how well it can do drums, those sometimes come out really well in metal songs
I'm sure you know Twilight of the Thunder God 😄
Maybe heard it via tiktok lemme look it up
wait 1 week. it's coming.
Sounds a lot like hammerfall meets brothers of metal, and yup seen em on tiktok
regarding the lyrics, instrumentals and especially the drums, this song is really very good. just the vocal style is too clean, but with some more controls over the vocal style and the ability to make it longer than just 1:20, this will be really impressive:
Yea (I can sense voice distortion and it reaching some sound threshold)
Hey guys how do i tie my discord account to my account on the suno website?
Send a message to billing@suno.ai about it they shold be able to help sort that out.
Thank you 👌
No problem
is there a option to imput a audio file and have it build a song bassed off that?
I just need this program to take a song I already wrote the music to and make a production out of it.
I'm afraid not. This is purely text to music.
is there any apps like that? it seems it would be a easier task for the ai.
I don't know of any but this doesn't mean one doesn't exist. Maybe someone else will post about it?
Hi - not sure if i'm in the right place or not. i subbed on app.suno.ai , and I am curious about the limitations on keywords for prompting and so on. is there any documentation? it seems there is some opt-out mechanic at work, or something similar, as referances to some musicians give errors, while others do not. would be very interested if there was some sort of guide for this type of thing. great AI, and would be awesome if you could have an option to have it produce stems at 6x the cost, or something.
hi, thanks for the sub! sorry, that there's no documentation for this aspect of model performance right now. there are a lot of legitimate reasons to make references to specific artists in a prompt, but also a lot of murkier reasons, and right now we're trying to err towards caution. Thanks for the warm feedback, and definitely hear you on the stemming feature
it's really quite noticeable that the overall quality of pop generations is way higher than the overall quality of metal generations. I wonder if that's purely due to metal being way more complex, or also due to there being way more pop in the training data than metal
yeah I don't know if we can really guarantee support for 13/8 polyrhythms by the next release 😅
a feature idea, that could be well into the future, i don't expect it next week 😄 but if it was possible to upload a clip, as you would an image in stable diffusion, to generate, say a brass section on top, i would be more than happy to let you use my clip for training, as an exchange
yeah that's a cool idea, though for technical reasons it would probably take some thought to figure out how to do it well. fyi there's #💡┃suggestions-web for proposed features
would probably be impossable to stop people uploading other peoples copytrighted material.. too many legal problems 😄
@light ether something I would LOVE to see on the new site: a section for browsing chirps people have made. If the user first hearts a song, it appears in the pool.
Random: Songs the initial user/creator/chirper hearted appear here randomly. If you heart a song you find, that helps it reach the top of the hour/day section, which can propel it upward faster.
Top section: Top rated songs go here and can be sorted by top of the hour, day, week, month, all-time. Maybe it randomizes the top a little bit too.
Bonus Request: Some sort of way to filter by song type/genre.
Thing to watch out for: Make sure the first songs ever upvoted don't just stay at the top forever, permanently gaining hearts and never being budged from that top spot. There needs to be a way for newcoming songs to have a shot at obtaining recognition.
In theory likely a lot more pop training lol
Hey, what are the differences between generating on discord and on the web app ?
Web has higher quality and longer clips however theres an audible threshold it currently hits. Discord has the ability to continue clips and to insert your own lyrics. Sooner or later both versions will be synced.
Where does the quality difference come from ? isn't it the same model ?
Suno Chirp Bot
BOT
— Today at 7:03 PM
This prompt has been flagged. Try again with a different prompt.
why alway get this message whenever i use chirp what went wrong?
Discord is V1 and Web is V2 currently they will be syncronized soon™️
One of your lyrics might have been a bit explicit or nsfw
Or prompt asked for something in that vein
On web, is it possible to create instrumentals? (no lyrics)
Yes however its a bit Iffy put 'Instrumental' into the genre section
Lol, I get a guy singing about an instrumental. I'll stick to Discord for now 🙂
There is no genre section on web. Only one prompt box.
Yup, thikn of the web version as a preview of the next version of Chirp. The versions will be synchronized soon.
How to change the generation language?
Genre Section within the prompt An Instrumental [Genre Insert Here] song about...
Hi everyone. Im using the web version. Is there a way to enhance the output of the audio quality ?
What do you guy usually do?
Is there a FAQ with theme, prompt, musical styles listed?
Not yet. I was considering compiling one quick but as we're right between a major version change it may be obsolete, so waiting for latest updates to fully roll out.
Fair enough, it's stupidly impressive as is so curiosity reigns supreme when thinking of new concepts 🙂
Makes the mind just race with the sheer possibilities, doesn't it?
Very much so. I've already made a Reggae song with african vocals and personalized lyrics to an artist friend who is now sweating bullets for real. They thought they were safe from AI for awhile!
are we not allowed to make it create songs in a similar style of other artists?
Got some prompts denied when quoting direct artists. Other times it’ll accept it but generate something not in the correct style. A bit finicky in my experience atm
Style? Sure. Exact copies using that artists' voice? No. It's not supported by design.
yeah but I say things like "a song rapped in the style of eminem" and it gets denied
I only want the style not the same voice
Are you using the webapp or the Discord version?
web
ChatGPT getting in the way there I guess.
Once you can use your own lyrics you should be able to generate them on your own and then use them in the app.
Since it can take some persuasion to get ChatGPT to do certain things.
I see thanks!
Also it may be worth seeing if any of the (extremely short form) jailbreaks will fit into the webapp for your request.
Don't know if it would work or not for you but it could be worth trying.
will try!
there is more to your prompt than that, if it's being blocked by chatGPT. 😒
no there is not 😒
I did not tell it to use the same voice I said to use the same style
you are correct. I apologize.
try adding 'about..." (carrots, or whatever)
apparently ChatGPT generates something it doesn't like and then censors itself 🤡 🙃
a song about carrots rapped in the style of eminem
Sounds more Vanila Ice than Eminem to me but not bad...
Oh! and I'd like to point out that as moderators we do not have access to anything more than you guys do. We're moderators, not developers. No special access.
(Yes that was a hint, LOL)
thanks will try tomorrow already used my credits today lol
damn not bad indeed
Lovers of Simlish, your wait is over. I got you.
Web Chirp Propmt A song using Lyrics from Yesterday by the Beatles, but no vowels are used in the lyrics
Not quite as cool as legit unconditional gen where there's always the dream the model actually makes genuine sense in the lyrics
@hushed tulip Another approach to get something like unconditional 'Simlish'
You don't need to use a real song like Yesterday for this.) Meant to reply to #📚┃suno-school message actually
Prompt: A song using Lyrics from Yesterday but written in unicode italic text (replace italic with bold or 'scary' for slightly different vibe maybe)
How to mention style like Hip hop energetic in app.suno.ai ?
try to make it into a sentence, since it will all be 'interpreted' by chatgpt...
here's a couple of mine that turned out (more or less) how I asked...
sinister groovy hot disco about being dipped in gold to look gorgeous forever
minimal hot synthpop, sultry male, lyrics about chauffeur attracted to his passenger the boss's wife
so i cant give my own lyrics ?
not yet on webapp... v2 is in preview, should be testing those controls over the weekend... Devs are estimating early next week (and also v2 will be here on Discord)
Duet Formatting in Web App try someting like lyrics are formatted like television script with character names like ROSS:
👏 👏 👏 👏 👏 👏
The lyrics are killing me
I hope this is the right place to ask:
I have read the whole TOS but still want answers set in stone.
1: How much right do you ACTUALLY get to songs generated? The TOS seems to say full but admins mentioning it here talk in a way as if its much more complicated?
2: Is there any way to get the rights of same level for a song you generated outside the paid period or are those forever unobtainable (as in rights)?
How did u get duet? What is prompt
That was it, more or less
You just wrote duet and it generates them?
In the web ChatGPT version where you can only give a basic command
A 'some music style' song formatted like television script with character names like ROSS:
Half the songs are basically ABOUT the Friends TV show so maybe don't use Ross though 😄
That’s epic!
I was expecting something more like BERLIN: I'M A MAN, lolol
🤖 🤖
Anyway to have V1 come out stereo
no, but v2 will come to Discord over the next days...
Any website or editing program you know to bring the quality up @broken lotus
anyone know?
The only thing I've done is using a source seperation tool or model to seperate lyrics from music, so I can boost lyrics volume. I'm curious what other people do. Do you have an example of the low quality part you want to improve?
can you email support@suno.ai? Someone will respond in detail there but may take a little time
Umm.. I use mp4 to wav website increase the volume 150 and select stereo for channel
I use Ultimate Vocal Remover to demux... but it is so-so with v1 content... v1 content just has problems due to the sampling rate...
There's things like Audio Super Resolution but I don't think it would actually improve this much, though maybe I should double check. Might be interesting to try downsampling to bad quality and re-upsampling as an effect.
it lets you separate the song into 'stems' which are separate audio tracks for the vocals, drums, bass, and other....
that lets you import those 4 tracks into audio editing software (like garageband) where they can be each adjusted to sound better, vocals can be cut and re-arranged... beat can be extended...
uh wow 🤩 I have BandLab
So ima test this out
this probably works then https://www.bandlab.com/splitter
yeah, same idea (but free)
you'll need to download some 'models' for it, which are a bit cryptic to know what's 'better'. I use the DEMUCS models, but there are a bunch you can install.
You have a music channel @broken lotus
I'm not a lawyer...
music publishing/distribution is crazy-regulated, entertainment media is California's biggest export, and California has the world's 5th largest economy if it was it's own country... they are politically powerful.
once they figure out Suno (et al) has just disrupted the music industry, they WILL be looking for ways to stop it (NAPSTER, etc)...
Discord is inherently social. Public betas allow us to take anybody's chirp as a starter and then continue it. I can make a song that is half my lyrics, and half someone I don't know..., and vice-versa.
Suno also can't prevent us from stealing song lyrics that are copyright protected and making a new piece of music with someone else's copyright lyrics.
Meanwhile, AI gen does not get a copyright.... this is potentially a legal tangle of rights and public domain.
the TOS attempts to clear this up, legally.
Everything we upload to Suno 'belongs' to Suno (unless it's something bad, then it's our fault lol). This fixes the community-sharing problem. If we don't want to share, we can DM the bot privately, but legally we still 'give' our lyrics to Suno.
as rights-holder, Suno can distribute (download, webview) our songs.
Suno grants us back a perpetual, royalty-free license (we will never owe Suno more money for our chirps).
free subscription gats an 'in-house' license (non-commercial) but they allow posting on socials with attribution to Suno (video contains attribution)
chirps made with paid tier are granted a commercial license.
(not the goal to punish free tier so their chirps can never be changed to commercial license..., the goal to prevent 'farming' free tier for good beginnings.... It should be obvious whether a (normal) free subscriber switches to paid and uses some of their free chirps (or continues them) verses some sort of automated acct that is farming dozens/hundreds of free accts to circumvent the free daily limit.
Suno might be sold someday. 😬 Our songs go with.
If music is your business, do not take my word for it. I don't represent Suno, am just a mod here.
Wow! thanks for taking your time to explain :)
Thats a solid read but question, when someone unsubscribes does the commercial use go with it?
interersting that both songs sung the sun emoji :sun_with_face::palm_tree: Feeling the heat, the sun shines bright as "summer" curious if just a hallucination, actually interpreting the emoji, or just reading the raw 'sun_with_face'. But I think I'll wait till we have better control because testing through ChatGPT wastes a lot of gens.
Waxky
I hope we're able to continue the existing web segments directly when that feature comes. I gotta build on BLANK SPACE #🐣┃suno-showcase message
the license is attached to the 'song'... it's whatever your status was when the chirps were made
(hence songs under free subscription are not granted commercial license, but I suggest that would not be a normal thing Suno (hypothetical lawyers) would go after under normal use)
God the mobile lag trying to figure out what yout ref to
Ķk (so remove the 4 stores i am unable to set to free)
How's that again?
Pretty solid info (if thats what your ref to) routenote doesnt allow setting free on amazon/apple music for example for some dumb reason
I've recently found that you can have specific words that exist on multiple languages and chirp just swaps the language to english for that one word, and then keeps going with the original lyrics language 😁
but if you put language descriptor to the genre, it's more likely to force the original language for all words
Huh funky
kinda wish you could "inpaint" those specific moments, like fix just the small bit in the middle
same with parts that it fails to gen right but rest of the clip is great
basically we'd need something that would take the start and end and could fill in between, instead of just taking the start and expanding on that
🤔 maybe a timeline... where we can cut the track(s) apart and rearrange, leaving spaces for in-painting...
would be amazing for reusing the chorus too
could make seamless transition to it later again
no idea if this is even technically possible with this current tech (not talking about discord bot's capabilities for editing but actual underlying model)
[Verse 1]
In Palm Desert's golden light, at ninety-three so bold,
A tale unfolds of Merrill, a heart that's finding its own road,
A lifetime's worth of love and dreams, now taking a new start,
At 93 years old, divorce won't tear him apart.
[Chorus]
He's living life, under the desert sun so bright,
In Palm Desert's warmth, he's finding his own light,
At ninety-three, a new chapter he'll ignite,
In the heart of the desert, he'll take his flight.
[Verse 2]
With each sunrise, he awakens to a world of grace,
Embracing change, finding solace in this tranquil place,
His heart may bear the weight of years, but it's still strong,
In Palm Desert, where he belongs, he'll sing his song.
[Chorus]
He's living life, under the desert sun so bright,
In Palm Desert's warmth, he's finding his own light,
At ninety-three, a new chapter he'll ignite,
In the heart of the desert, he'll take his flight.
[Bridge]
Though love may change, and paths may twist and turn,
His spirit's fire will forever brightly burn,
In the midst of cacti and the desert's endless view,
He'll rediscover life's wonders, anew.
[Verse 3]
With each passing day, he finds joy in simple things,
The desert breeze, the beauty that the night sky brings,
At ninety-three, he's proving it's never too late,
To rewrite the story, find love in a brand-new state.
[Chorus]
He's living life, under the desert sun so bright,
In Palm Desert's warmth, he's finding his own light,
At ninety-three, a new chapter he'll ignite,
In the heart of the desert, he'll take his flight.
[Outro]
In Palm Desert's gentle embrace, his spirit soars,
A 93-year-old named Merrill, with a heart that still explores,
Living life, finding love, in the desert's warm embrace,
In Palm Desert, he's found his happy place.
this is too long for 1 chirp to sing in 30sec....
break it up into multiple chirps, and use https://discord.com/channels/1069381916492562582/1149508588470022164
also type /chirp to call the bot, but not here. Use one of the beta channels.
Nvm cant edit the selected stores so thats fun [I did set prices to Lowest] for them..
Broadway duet formatted as angry art debate between multiple character names like TONY:
I finally got one where they sing together...
🥺 3rd verse
I think Chirp knows there should be 2 voices,
but ... can't ... quite ...
Do the team of suno have a post or page anywhere that explains how suno works and how it was trained?
#1151542418991239168 likely has a section on it somewhere
Chirp is commercial product, not open source so no papers or public research
🤫 is big secret
without knowing any details, I've been under impression it's more like specially trained bark model?
I mean, it was named "bark sing" first, wasn't it?
stock bark model can sing too, but it's clearly way worse in every aspect with this
You can get a sense of how lyrics still influence music style, a lot of songs written in assembly are robotgardens
Morse Code often just fails, sometimes is a sort of SimLish
Honestly there's a billion ways to get pseudo-Simlish
related #🗣️┃chill-chat message
Guys is there going to be an API for chirp?
If i had to guess not for a long time
How come my songs don’t go to 40 seconds anymore? They cut off at 29 but I thought it was extended
There's a bit of randomness involved.
The new model is capable of longer durations but currently is only available in demo form, using ChatGPT for lyrics on the webapp.
All sorts. Including original blends you create yourself.
All I can say is to try it and see what you get.
Is there a reliable way to have Suno not generate lyrics? I'm quite interested in specifically the musical side and less in the lyrics, especially since it will inject lyrics describing the style of music the prompt is supposed to be in which is odd at best and bad at worst
hji
I've had luck just leaving the lyrics sections blank, but ive only tried electronic music
How can i see how many credits i have?
'/info'
Put ‘instrumental’ in the style and keep lyrics blank
Why do some songs come out with no sung lyrics, even though the lyrics are written?
I can’t wait the limit to go above 80 seconds and to be able to write lyrics in version 2!
the voicebot was on a lunchbreak.
🤣
yeah... happens
you can add 'vocals' to the type which usually helps (once the field is available on UI)
Okay thanks.
Ah I take it then that the website itself is different than how you interact with it in discord? There is no fields besides the single prompt box on the website itself
I've not used this through discord and I'd rather not. Midjourney is already tied to it and that's an albatross even paying for private mode.
The website is in preview with a new version of Chirp (v2).
There's no extra charge for 'privacy mode' on Discord, you just direct message the Chirp bot and do your prompting in private.
(web site is gaining the full prompt controls and Discord gaining v2 this week.)
Does this mean the website produces better results than through discord?
v2 is stereo and has a longer song-length
the audio quality (sampling rate) is MUCH improved over v1 downloads, with an MP3 option that is (slightly) better than the video's audio quality.
⚠️ warning: inevitable frustration at getting better quality but less control
Thanks for the info, I'm cool to wait for the functionality to hit the website
and yeah all my exposure is in v2, I messed with this slightly a little bit months ago and it was neat, but this new V2 version has clipped over whatever it is that makes it sound really, REALLY good.
and since I've only used the website that would make sense
Thanks. I am still slightly confused. Do you mean Discord is alread using Chirp V2 and the website is just getting the new functionality?
I think the website is using v2, so the model is better meaning the audio you get is stereo and 'better' but the website itself lacks functionality in how you interact with the system vs. discord
I wasn't sure which is suppused to have better quality for now, website or discord
right... Discord is still on v1 ... (for now)
Ah ok, thanks!
but along with the website getting the better controls discord is supposed to get v2, so then there will be parity this week
based on what wetcircuit just said
neat!
Yeah it seems there are pros and cons of the website for discord, would be good to get feature parity across both e.g. custom lyrics input on website.
it's coming... 😭 "SOON"
No worries, appreciate all the work so far. I am already amazed how good it is.
how can i delete tracks on the web version?
when will suno v3 (the sites model) go on discord?
Fantastic application. Personally I don’t like Discord, I find it unintuitive. Therefore, I waited to try SUNO and have opted to use the web interface now it’s available. However, I have some questions. 1. How do I delete generated music. 2. Is the web interface chat based, as in, does it consider previous prompts while creating new ones? 3. Is there a prompt guide for the web based version? Any help with these questions will be much appreciated. Thank you.
- will make that option soon
- no - each prompt is separate
- we are working on one, but you can just ask about specific prompts in the Discord and folks generally helpful
v2 🙂 and later this week
yay!