📚┃suno-school

[daring lava]

Happy to help

[pliant cave]

This should be pinned.

[daring lava]

I prefer the one they have via bot, it constructs the lyrics better imo. The one I provided helps with language but it acts a bit wonky with line count sometimes. If you do use it, I would suggest treating it as a template and adjusting it as needed. But again, the one in bot works very well. It’s my go to

[open magnet]

to be honest chatgpt is pretty terrible with making good lyrics in general. we are actively exploring other options

[bright dragon]

What are the genres that are supported by Chrip ?

[unique dome]

There's isn't a fixed list, but here are some examples of the types of tags that often work well:
https://suno.ai/docs/metatags

[wide knot]

/chirp

[tall orbit]

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.

[bleak stream]

This is my first hour on Suno Chirp so pls forgive newbie/dumbass question. I see some users with a lot of lyrics, but Chirp will not give you more than :30 seconds correct? Or can you pay a nominal fee for longer song structure. I get that you can do a song in bits and stitch it together. Is that what other users do, or am I missing something? Thx.

/chirp

[tall orbit]

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.

[graceful cliff]

See #⏩┃continuation 🙂

[vivid wren]

Is there any way to input an audio file to make Chirp make something around it?

Just like it is with ImageToImage on StableDiffusion.

I would like to know if there is an "AudioToAudio" option

[graceful cliff]

Hi there isn't any such way -- but the continue option is kinda like it. See #⏩┃continuation

[frosty crypt]

The quality of the first song that is generated is really great, but the more you continue a song the worse the quality. Are there plans to improve this, or extend the overall generation? And I mean you don’t really need to extend the generation windows past 5 mins since usually songs are only like 3-8 mins depending on the genre (but with just 5 min gen you can get good quality for a majority of songs and only need to extend once for pretty much the rest of them).

Though it seems expensive to generate two 30 second songs, so jumping right to 5 mins probably isn’t going to happen, but it would be insane if it does, but I can imagine many more optimisations may be needed

Also do you guys use the transformer architecture in anyway for song generation? Is be curious to know if it is useful in this regard

Because PlayHT was able to use the transformer for very realistic voices and voice cloning, and I’m wondering if it use can be extended to song generation

[cedar mauve]

I can't use /chirp command on dm sumo bot

[untold briar]

The magic of Chirp is private but we can assume it's a least a little bit like the open source Bark, I think, which is a transformer yeah.

[open magnet]

yeah longer context definitely in the works 🙂 but yeah will take a bit of time

[open magnet]

you have to enable direct messages for the server i think

[crystal tendon]

id:guide Hi, is it possible to change the font color?

[open magnet]

i assume so, how come? mind sending a screenshot?

[crystal tendon]

For example, I want to change the font color and background here.

[open magnet]

oooh you mean the videos...

[crystal tendon]

yes

[open magnet]

yeah sorry kinda limited tooling there right now since we are focusing on audio. so if you wanna play with that probably just use the audio and make your own background stuff

[crystal tendon]

Extra prompt commands can be added for this.

[untold briar]

A nice future-pro-feature would be when you click the heart, you get a popup that is all about customizing that particular video render. font, sdxl prompts, aspect ratio, etc. maybe even bouncing ball karaoke mode text if whisper is up to the task.

[hexed scroll]

https://tenor.com/bXjpu.gif

[crystal tendon]

@open magnet Is it possible to determine the style of music? For example, say it in fifty cent style

[open magnet]

Not by artist for legal reasons, but you can describe the genre

[crystal tendon]

Can you write an example prompt? In addition, the style can be voiced by anyone, there is no legal problem in this, as a result, the same voice or song will not be used, the system will only reflect the style of expression.

[daring lava]

Here... use this in ChatGPT

Description:
Guide the AI in succinctly capturing the essence of a music artist's style in no more than eight words. The description will reflect the artist's unique sound, themes, and persona. When the AI interacts, it will do so with the persona of an expert music connoisseur and seasoned producer.

Procedure:

Begin by inviting the user to name their favorite artist or song with the confidence and expertise of a seasoned producer.
Upon receiving the user's choice, acknowledge it with the depth of understanding and insight only an expert in the field possesses.
Deduce and convey the artist or song's style in a concise eight-word description, focusing on genre and sound.
Clearly present the style encapsulation using brackets.
Interactions:

User states: "[Artist/Song Name]"
AI, with the authority of a music connoisseur, responds: "Ah, [Artist/Song Name]! A choice that speaks volumes. Drawing from my vast experience in music, their style can be pinpointed as: [Genre-focused, sound-specific eight-word description]"

Example 1:
User: "The Beatles"
AI: "Ah, The Beatles! A band that transformed generations. Their sonic fingerprint is: [Psychedelic rock, harmonious vocals, diverse instrumental blends]"

Example 2:
User: "Billie Eilish"
AI: "Billie Eilish, a contemporary musical marvel. Her distinct sound is: [Dark electropop, whispery vocals, ambient beats]"

Interaction Guide:

Share the name of an artist or song you'd like to delve into.
I, with the expertise of a seasoned producer, will distill their sonic essence into a concise, genre-specific breakdown.
Experience the rich tapestry of your chosen artist or song, understood and conveyed by an expert's ear.```

Also, don't forget... if you want it to be a beat/instrumental to include that in as well

[merry cliff]

This stuff is incredible, good work guys. Any technical blogs or anything like that released? Or anyone technical up for a casual chat about this stuff?

[graceful condor]

Hello, I am new on the site. How to extend a sound that you have created beyond 30 secondes ? I tried to press the version 2 extension button, but it doesn't generate any results once the information is filled in! Thanks for your help !

[graceful cliff]

Should work, see #⏩┃continuation

[rare junco]

This probably gets asked a ton but discord search isn't super helpful, apologies:
Is there currently a way to run Chirp locally on your PC, or any plans for it in the future?

[graceful cliff]

Nope -- you have to run it on discord for now

[rare junco]

"for now", I like the sound of that :p

[indigo talon]

for style if i want to add multiple styles am i just typing it like this

rap, j-pop,

using a comma to seperate them?

[open magnet]

Sure ya that works

[drowsy siren]

#🐣┃how-to-chirp

[calm raptor]

Hi I’m trying to add the suno bot to my private server but I’m getting this error message: “This bot can't join more servers as it has not been verified or is requesting gateway intents it has not been verified for. Ask the bot's developer about https://dis.gd/bot-verification so you can add it to your server!”. Can you help with this? Thanks and kudos for this Great use of AI. Cheers

[humble dove]

Hi, what limitations should I keep in mind when defining Genre prompt? Does Chirp use any predefined list of music styles? Will it recognize band names as genre examples?

[valid lagoon]

Ist there a way to get the seed or something of a randomly generated song so the new generated song gets the same voice / Genre Style?

[bright dragon]

thanks

Is it possible to generate content with more length in single generation ?. Currently 30 secs is supported

[tall vault]

how do we know how many free chirps we have left?

[merry cliff]

Is there any research being done into essentially upscaling the output of chirp? Seems like it should be a relatively straightforward transformer problem to solve?

[weak lance]

/chirp

[tall orbit]

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.

[untold briar]

The recent version is quite a bit better audio quality than the previous Chirp with only English. So it is getting better, whatever the mechanism.

[lyric wadi]

trying to subscrib but it is asking me to verify my account. but theres nothing to do on the page

990211219141066793

[untold briar]

I'm just guessing but maybe try verifying in Discord like this https://support.discord.com/hc/en-us/articles/213219267-Resending-Verification-Email

[lyric wadi]

gonna try again now

[rich thunder]

hi why do some chirp come out with out vocals or in jibrish?

[graceful cliff]

Depends on your text input -- try to have shorter text input in general (like 6 to 8 lines)

[rich thunder]

Do you mean 6 verses or lines?

[graceful cliff]

Lines -- shorter lines

[rich thunder]

i will try that thank you

[untold briar]

Oh that's a different issue. Discord put a limit on the number of servers that the Chirp bot can be added to. Until that limit is increased, just have to wait to add it to your own server.

[fair drum]

/bark hi

[graceful cliff]

Go to #🐶┃bark-beta to try it 🙂

[graceful condor]

Hello

[lyric wadi]

ok thanks

sorry I was super busy

[lyric wadi]

That kinda sucks as writing my lyrics or adding them in public I cannot do as I wont put my music out there for someone to steal

[light ether]

You can just DM @versed valley as well. Re verify account, that's probably a Discord email verification

[lyric wadi]

I subscribed so I can have it on my own server private

[light ether]

Yeah, we're still waiting on Discord, but for now you can work in DMs with the bot (also private)

[lyric wadi]

where do I access this

[light ether]

Click on -> @versed valley and type a message

[lyric wadi]

Am I asking for access etc...

[long drift]

i absolutely love the first generation of a chirp i did..but every generation and continuation after has a different vibe and different melody.. is there a way to keep the music the same/similar and just change lyrics? almost like keeping the same seed? i'm using the chirp vocalist as a second singer (to me) on a song i'm working on and need continuity..

[icy wyvern]

I just Split the music in 4 bars for free with program bandlab (android and iOS) works Pretty good

[long drift]

thanks for sharing. i'm really looking to generate additional music with the same feel and also generate additional lines of vocals with the same "singer"

[crystal tendon]

How can I extend the duration of the song?

[graceful cliff]

See #⏩┃continuation

[crystal tendon]

No, I want the lyrics to continue, for example a 3 minute song.

[graceful cliff]

? Continuation let you create a song that continues on and on, you can piece 6 together into a 3 mins

[crystal tendon]

Ahh I see

[crystal tendon]

but very this is hard 😦

[graceful cliff]

we are working on making generations longer 🙂

[bronze yacht]

Thank god. I’ve done some continuation where the voice is completely different it’s so annoying

[graceful cliff]

It can def happen 😦

[frosty crypt]

When you extend the audio generation window, how long do you think it will be when you first update it? Like is it getting extended to 1 min, 2, 3?

[graceful cliff]

Let me throw this back to you a bit -- what's the shortest duration that you will be significantly happier with it? 1, 2 or 3?

[frosty crypt]

Well I like pop songs, which average 3:30 in length, so I would be happy with 2 mins and just continuing that once as generally continuing just once doesn’t degrade quality that much. But I mean it really depends what music I want lol, so for some 3 or even 5 mins would be a lot better. But I mean I feel like you don’t really need longer than 5 mins (just continuing that once or twice covers probably every song ever lol), so if you get to that point it’s probably worth focusing in other areas (like fine grained controllability, enhancing quality if needed or other features like improving the quality of multiple continuations).

[sly tree]

Even 10 seconds longer is appreciated, but I think 2 minutes is a good mark to aim for. That covers the length of most radio broadcasted songs.

[graceful cliff]

Great feedback thank you. Sounds like 2 min is the winner. I’d better get some sleep and start coding 🙂

[gusty jolt]

is it possible to remove song from the chat history? 🙂

[graceful condor]

Yes, I try to use this but the button didn't work when I press it and soumite the mew texte, nothing apen !

[daring lava]

Hey there 👋 not yet, but it is something they said they are working on

[broken lotus]

Any tips for how to trigger/not trigger a shorter clip?

If I put only 1 or 2 instrument descriptions in the content (and no lyrics), I sometimes get a really short clip – which is ok because it still CONTINUES ...

but am I doing something that makes this more likely to happen? Is there maybe a prompt I could add to encourage the longer segments?

and the reverse, How to trigger a shorter clip?

[broken lotus]

F-word?

[open magnet]

lol.... yeah...

i'm aware that a lot of existing lyrics wouldn't pass our filter, but if it's too relaxed then some people figure out ways around it to make very inappropriate stuff. we are working on improving what constitutes 'art' but in the meantime just try to keep it as PG as possible

[supple frost]

Thank you for the clarification. I did a brief search on the documentation I could find, maybe something on the TOS i missed?

[elfin sky]

Hey not sure if this is the bugs and suggestions channel but I want to propose a refer a friend perk to where just as the name implies, If you get a friend to join the discord then you get some credits for generations. 🙂 Keep doing a great job devs

[untold briar]

Oh it's https://discord.com/channels/1069381916492562582/1082775596767117312 channel maybe not visible in your default Discord channel list

[elfin sky]

Thanks bro

[open magnet]

have a look here: #📜┃rules

[supple frost]

Thanks, It wasn't selected in the channels and roles section. Now I see. Thanks

[fringe moss]

Hey there!! How do I combine clips? Can I do that? What is the max length of the clips? Thanks in advance!! Happy week ahead everyone!!!

[daring lava]

Hey YakNik 👋 here check out #⏩┃continuation it explains how to 🙂

[fringe moss]

... right on! Thnks David!!

[daring lava]

Happy to help 🙂

[fringe moss]

Happy Hump Day!!

[fringe phoenix]

is there a list of styles?

[open magnet]

have a look at https://suno.ai/docs/metatags (see #1151542418991239168 )

[fringe phoenix]

thank you

[limpid palm]

How download full length track? Only 30s.(((

[upbeat depot]

Can we expect to see much in the way of parameter adjustment or the ability to run this locally?

I imagine there would be a variety of possibilities and we are just getting one standardized algorithm for the most part

Sort of more akin to midjourney vs stable diffusion

[graceful condor]

/chirp

[tall orbit]

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.

[twilit smelt]

\

[prisma ermine]

Any plans for open sourcing chirp?

[open magnet]

have a look at #⏩┃continuation

[open magnet]

definitely looking into exposing parameters for the pro versions. locally probably not for the near future unfortunately

[upbeat depot]

i would happily pay more for a monthly subscription if enough parameters were exposed

[open magnet]

yeah biggest reason to start without was to keep things a bit simpler. most people don't really know what temperature etc is

[upbeat depot]

Im just still just so impressed by the quality that I keep having moments of surely thats just some layered premade loops where the vocals are the secret sauce. I've been producing music for over 10 years and this just keeps blowing my mind

[upbeat depot]

💯

[supple frost]

Temperature?

[graceful cliff]

Working on it

[supple frost]

I was trying to figure out what "temperature" would refer to in this capacity..

[serene plover]

"temperature" in AI is nerdspeak for "randomness"

[broken lotus]

it's a term from 'GPT' large language models – whether they are set 'low' which gives narrower more predictable results, or whether it's 'high' which gives more random results

[supple frost]

And here I was, thinking it could refer somehow to timber

xD

[broken lotus]

that would be timberature

[graceful cliff]

Sth will come out soon. Stay tuned.

[supple frost]

naw warriez. my interest was solely because, as a musician, it's really interesting to see what "controls" such a tool might have

I don't even know if it is possible to ell an AI what key, BPM, mode, structure, etc. a song should have

ahah

it's pure curiosity and interest

[graceful cliff]

I understand -- good to have knobs to play around

Won't be like your traditional control though -- what I mean is it won't be like key, beat, etc, but it will generate different feelings of outputs with the same input. Even though each generation is unique you can still tell the qualitative difference.

[supple frost]

Understood.. So in a way, like with AI imaging it will be a matter to understand what "language" they are more "receptive" to

[graceful cliff]

yeah it's like more qualitative than quantitative (hopefully that will improve over time)

[glacial stratus]

Will Chirp become open source?

[lucid pulsar]

Hello has someone created a private chat on discord with the Suno Bot? is it possible like Midjourney?

[daring lava]

Here you go 🙂

[bold locust]

it suddenly turned to rap... my music style was pop, how can I revert it back to pop after some rap?

[gentle rose]

Im loving Sunos output. Is kind of frustrating though not being able to choose only the desired generations to be stitched together. It would also be great if the generations were set in bars (2bars, 4bars, 8bars...) in music we count bars and not time.

[light ether]

if you want to keep forcing pop, you can try continuing to specifying genre in your continuations

[idle dew]

┬─┬ノ( º _ ºノ)

[graceful condor]

@light ether
There should be some kind of repository of genres. For example, which genres are working and which aren't.
I tried to get the music style of cartoon jingles like Gummy Bears, Ducktales, etc. I tried
"cartoon intro"
"disney"
"upbeat jingle"
Etc

None of those seem to work, like at all.

[untold briar]

Also don't forget you can browse all the public Chirp channels to see what kind of genre prompts are working well and what they sound like, as well as literally continuing your own clips from any other public chirp Clip. So you don't have to use credits yourself to test a prompt if you can find other examples in public. And literally continuing those samples to get an even closer match to that style.

[bold locust]

I never changed it to rap myself, the music style was pop, how ever I liked its transition from pop to rap for a few lines and then when I want it to keep singing it pop, it still sing rap. even though my genre selection is pop

[graceful condor]

Hello, How to generate a song in another language?

[open magnet]

just use lyrics in that language

[graceful condor]

show me the best songs yall have ever made using suno

[light ether]

check out #🐣┃suno-showcase some very good ones in there

[graceful condor]

oooo alright

[upbeat depot]

Is there the potential to input a melody to guide the output?

I know not currently

[open magnet]

not in the current version, but it's something we are considering. what's your use case out of curiosity?

[upbeat depot]

But is it feasible

[open magnet]

ya for sure

[upbeat depot]

Im a musician / producer

And love any tool that helps me get my ideas out

This is more a sketchboard for me at the moment

[open magnet]

what would be the best and most simple way of conditioning on a melody for you? uploading a track to get the melody from, uploading a midi? inputting chords in the lyrics field?

[upbeat depot]

And I've always envisioned making parodies of other genres or wishing I could make this or that style by my personal approach when it comes to it is not so board.

Picking chords by writing them in the lyrics field would be a Godsend

And the ability to say input a recording of me humming or singing

[broken lotus]

for musicians probably midi.
for instagrammers probably a few seconds of them singing

[upbeat depot]

And then transform the contours of that into the generation

Im a musician and would rather input audio than midi

So much more dynamic and detail

[open magnet]

and so when you pick it up from a midi, would you wanna keep the time domain or just the progression?

[upbeat depot]

Midi needs modulation to have the same effect. Tho it would be cool too.

[broken lotus]

yeah, i guess it depends if you want the ai to come up with the sound, or accompany your sound

[open magnet]

or even from singing/whistling etc would you wanna keep the tempo or just the progression?

[upbeat depot]

I think the time domain would be ideal to keep intact as possible. But I understand the need to quantize

[broken lotus]

both, but fixed / variation

[open magnet]

right so there is also the extreme of "overpainting" wher you upload audio and we generate lyrics that fit to the background audio

[upbeat depot]

Currently the audio is all purely generated yes?

[open magnet]

yea

[upbeat depot]

Including drums

Etc

Yeah

[open magnet]

yea

[broken lotus]

omg, singing lyrics on an existing track..., i didn't think of that...

[upbeat depot]

I think all these ideas are amazing

[open magnet]

haha yeah, but each of them is different a probably quite a bit of work

[broken lotus]

or the reverse

[open magnet]

we are still very small so we have to pick our battles

[upbeat depot]

Mostly for me I'd just be happy as hell to sing 4 bars of a melodic idea and turn it into jazz along with harmonies etc

Absolutely

Would love to give input whenever needed. I've spent way more hours using this than I anticipated ahah

[broken lotus]

or replace voice with violin...

[open magnet]

ok so how would you two rate the following options:

1. allow chords to be entered in the lyrics field to guide progression
2. upload any audio file to pick up rough melody from on keep during generation
3. add vocals to an existing backing track
4. add backing to an existing vocals track

[upbeat depot]

Like take an original track and swap the sounds? Or just ask for a violin solo

2, 1, 3and4 same tier

As in
2
1
3/4

They all get me excited but this is my current mood

I'd use the crap out of all of them tho

[broken lotus]

1, 2, 3/4

similar to above... 😅

[open magnet]

haha ok sounds great

[upbeat depot]

😍😍

[open magnet]

thanks for the feedback, much apprecitated. we're still learning a lot about music and what's possible ourselves

[runic dock]

It will be great if user is able to input a melody by uploading a song or puting chords in the lyrics

[runic dock]

Inputing chords directly in the lyrics

[graceful condor]

@light ether When I try to create an instrumental intro to a song (without lyrics), it does it very well - but it doesn't fill up all 40 seconds. Any trick I should know about to make sure it fills up all 40 seconds with instrumentals?

I'm making a song about a musician friend of mine, who plays guitar (with my brother on drums) in the band Baja Bug. He's the one with the black sunglasses. He had a hilarious experience with mushrooms in Rotterdam, which I want to write a song about. 😄
https://youtu.be/FKGBXQdMmrc

They are in studio right now. I thought, why not mess him up by having an AI write a song about his experience?

Suggestion: It would be nice if the "metadata" of the Chat-GPT request was stored in each iteration or generation. It can help with the workflow tremendously.

Basically, the text you wrote for Chat-GPT would be shown in the bot's post.

[light ether]

Good suggestion!

[light ether]

It's not super easy right now unfortunately. Some work on our end to do

[graceful condor]

No worries

You have done excellent work already 🙂

I am also glad that you are focusing on the "musical aesthetic" primarily, and letting the "bitrate" be something for later. It makes sense in terms of prioritization to do it like that, even if early version songs will sound a bit low-res.

Midjourney did the same consideration, and they were right to do so.

Even though they have truly neglected the Language Model aspect.

[upbeat plume]

api

[graceful condor]

I'm not getting 40 second songs, so it's not working

[broken lotus]

I've noticed my content prompt can cut it short.
like if I only have 2 lines of lyrics, or a single [prompt], it might do those and end the clip.

(not saying you did anything wrong, just trying to gather and share tips that work)

[daring lava]

Could you post a screenshot of your creation? Maybe I can help

[prime abyss]

Hey, did it crash or smth? 😬

[light ether]

Discord itself is a little sad

[prime abyss]

Ooooh, i see, thanks

[light ether]

Will monitor though

[light ether]

Usually these settle down somewhat quickly, if /info returns it should be back up

[light ether]

Yeah, that's definitely accurate behavior right now

[prime abyss]

Alright ^^

[static wigeon]

we can't choose male or female on chirp ? there are a solution ?

[broken lotus]

I add ... "with female singer" in the genre description. Seems to usually work.

[static wigeon]

thanks !

[open magnet]

yeah not a hundred percent reliably but tends to work

[spare pulsar]

Attending an AI UX hackathon today and wanted to play with making a visual UI for stitching tracks together from a tree of generations -- been doing this a lot, and it's easy to get lost in the chat thread. Is there a hidden API endpoint I could use or would I be better off scripting something that calls /chirp via Discord?

[lilac trail]

is there a way to see what the style of music was/the music prompt was on a previously generated music clip

[spare pulsar]

@lilac trail You can see it in the video file, but if the style was longer than a few characters it may be cropped. You can search for the lyrics to find when it was originally generated it and read the style from there.

[lilac trail]

thanks

[light ether]

Hey! I actually prototyped something similar internally, sort of like a Figma UI. Sorry, no hidden API. best bet is the latter, or to fake it by doing a bunch of gens up front

[edgy elk]

Does anyone know, how to change/set stress in a word for Chirp ?

[open magnet]

interesting idea, haven't thought much about it yet. maybe try capitalizing the word? or elongating a syllable? although maybe then you get someone shouting... no idea 😆

[worn loom]

how do I add the suno bot to a private channel?

[broken lotus]

You can direct message the bot, but not add it to your own discord atm

[open magnet]

actually i beliiieve that limit was just lifted, so let me know if adding to your own server still fails

[shrewd anvil]

Hello everyone ! Is there a way to just generate a music with no voice in chirp ? Like I did that, but by random, so if there is a solution, it will be great 😀

[untold briar]

You can try putting things like [instrumental] in the lyrics field - instead of actual lyrics. There's some other tags that people often use but off hand I can't recall what was working best. If you check the public channels you can find other examples for sure. Overall it is somewhat less reliable than regular songs with lyrics. Sometimes it's kind of a miss and you can just get a very short clip, but other times you can get a full clip of pure music.

[shrewd anvil]

Ok thank you a lot ! I hadn't thought of using metatags. I'll check the list, it might be useful.

[surreal basin]

under style, can you write a band? how many keyword can suno handle ?

[broken lotus]

They've said there is no training to connect band names and musician names to music styles.
Try describing the music and genre in the description.

It's a Large Language Model so use ordinary language. In my tests the more you load up with descriptions the more likely to get a clip with broken audio or no sound at all.

I have used both [End] and [Fade Out] successfully, but they seem more reliable on their own... which means spending a continue just for the [Fade Out] prompt, or for a few seconds of music extra just to end the song.

using the prompts in a chirp with lyrics seems to sometimes conflict (is it singing the lyrics, or ending? might be a conflicting instruction.)

[runic dock]

/chirp

[tall orbit]

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.

[upbeat depot]

What is the base model the music is generated with? Is it 2 separate models, one for vocals and 1 for backing track?

And what is the dataset(s) you guys used?

[eternal nacelle]

If they wanted to disclose that, they already would have.

[teal lance]

I seem to mostly get results with very loud background music and vocals you can barely hear. Anyway to get around that?

[broken lotus]

maybe add 'upfront vocals' to the description?

[pastel tapir]

I'm confident its a single model. There would be a ton of alignment issues with 2 models

[tall orbit]

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.

[knotty kite]

What’s the trick to do bass drops for edm, [drop] is recognized?

I tried just doing lyrics like - - - - - dashes but that didn’t work a few tries now

How does genre work when continuing? I’ve tried switching the style up and it doesn’t seem to alter it much if at all

[snow cliff]

When I continue a song it seems to not give me the same style - it’s like a complete different version - any way to command it to stay the same so I can have a complete song ?

[open magnet]

Usually it will do that automatically. What can happen is that one clip ‘ends’ and then the model doesn’t continue it but instead starts a new one, kinda like gpt giving another question instead of an answer in a q&a. Should be v rare, do you see it consistently with different clips or just with a specific one?

[open magnet]

Hm not sure, would be cool to pseudo label some stuff during training. Do you have examples of what the correct notation for these things would be? Haven’t really seen things like that in a lyrics annotation format

[broken lotus]

I think [Build] and then [Break] worked for me 1 time I tried....
(but it wasn't EDM)

[teal lance]

Is there documentation somewhere on the tags we can use? I did not see anything under Doc's.

[daring lava]

There is some info about meta tags here... https://suno-ai.notion.site/Using-Metatags-34944efe09ec41d693e314f13af44695

[broken lotus]

(I think they are being cautious about implying that metatags are actual 'controls'... as opposed to the randomness of prompting an AI)

[teal lance]

Anyone have a list for the tags like [verse] [chrous] [bridge]. those actually worked for me

[open magnet]

[interlude] [intro] [outro] i think are good as well

[broken lotus]

[Fade Out] works..., often.

[teal lance]

[Fade Out] will that work so the very end doesnt get cut and you have to burn a credit on continuation?

[broken lotus]

[Outro] feels like an ending but you will need to fade out in post edit software...

[End] works when it feels like it, lol

[Fade Out] often gets me an actual fade to silence, but if there are other prompts or lyrics on that 'card' it seems less reliable

[teal lance]

ty

[wind nacelle]

hello, is there an option or button to put my chirp continuations all together or do I need to edit that by myself?

[daring lava]

There most certainly is 🙂 you can read more about that here #⏩┃continuation

[wind nacelle]

thanks :D

[daring lava]

Happy to help 🙂

[runic dock]

Where can we find all these control metatags?. Not in metatags list

[daring lava]

A lot of uncharted territory here, friend. Will be working to compile what we all learn along the way so as to better define and teach what can be consistently performed with /chirp

[broken lotus]

I don't think there is anything official...

un-officially, I'm guessing people are compiling lists of what work for them.... When someone like David or georg 👆 gives advice, I search that name and see what other advice they might have given recently...

And I trolled through other people's chirps where I can read their prompts and descriptions....

I searched interweb for 'parts of song' 'vocal performance terms' 'music genre'....

It's a large language model + random..., so talk in plain language, but keep descriptions short.

I put my content [Prompts] is square brackets like
[Electric Guitar Solo]
, but
Verse:
also worked without the brackets... and one time it sang
[Call and Response between Percussion and Bass] 🎶 lol

Just try stuff. even with 'perfect' prompts, the AI does what it wants half the time. It is early-days....

There is also (old) info on Suno website...

These would not be [Prompts] now, these would go in the genre description without the brackets....

[open magnet]

yeah either works. the genre is just automatically put input brackets and prepended to the prompt (and slightly boosted). so a tad better to put it into the genre/style field, but brackets in lyrics should also work with style empty

[dry aspen]

Hey @open magnet, loving Chirp thus far and have a handful of questions about it:

• Will Chirp be later released as a repo on github, like Bark?
• In what ways does Chirp differ from Bark?
• Is there currently, or plans to add, a way to view/save/apply speakers generated for a song, so that a voice that's liked can be re-used for other songs?
• When does the monthly reset occur? 1st of the month or a month from first using the bot?
• In terms of custom plans, would there be any possibility of a plan that allows for unlimited generations, though slow/queued, similar to how Midjourney allows?

[runic light]

/chirp

[tall orbit]

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.

[summer dock]

/chirp

[tall orbit]

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.

[tawny forge]

I was just testing Chirp out with different lyrics and voicelines and I realized that sometimes it cuts off without finishing the rest of the lyrics , does it have something to do with a time limit of -40 sec per clip ?

[untold briar]

If you prompt too many lyrics than sometimes it simply doesn't finish them in 40s, yes. That's why they recommend not using too many lines all at once. Though like many AI models sometimes it just decides to cut off lyrics even when there is time remaining - an issue that should improve as the model improves over time.

[pearl surge]

#🐣┃how-to-chirp make a Morgan wallen type country beat

/chirp

[tall orbit]

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.

[tawny forge]

why not make the bot take half the lyric and use it and then run the other half as a separate task and merge it at the end to fix this issue ( which could also be used to make it process double the lyrics limit )

[untold briar]

That's is essetially how it works now, but it relies on the user to decide where to put the lyrics in each segment rather than being a fixed rule. So you can either use smaller amount of lyrics or start the next section from where it left off. Details: #⏩┃continuation message

So the limit is actually quite a bit more than double, since you also continue again from the section section, and so on..

[tawny forge]

that is very useful , good job btw on this project I just found out about it 2 hrs ago and tbh you are doing very well on it

[eternal nacelle]

It's all Greek to me, but this seems like something Suno techies might be interested in, hot off the presses! https://huggingface.co/papers/2309.08773 Something about increasing "controllabilty" of text-to-audio.

[shy minnow]

hey how can i add chirp to a direct messagw?

[dry aspen]

PC: Right click on "Suno Chirp Bot" on the member list, then click "Message", you can then DM commands to the bot directly this way.

Mobile: Similar to the above - Tap the icon which looks like a person to open the member list, tap on "Suno Chirp Bot", then tap message to send the bot commands in DMs.

[tall orbit]

The /chirp command is not available in this channel

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

[broken lotus]

"hypnotizing male voice in ASMR"
vs
"melodic male voice, a cappella"

[unique dome]

Hey @shy minnow, check out https://suno.ai/docs/dms for instructions!

TL;DR start a DM with @versed valley then use the /chirp command just like you would in the public channels!

[old spoke]

I made 4 songs in Suno and my credit ran out, I wonder why?

[daring lava]

A trial membership allows for 25 generations. That can include either initial or continuation generations

[light ether]

Sorry, are you getting an error? Your account looks OK to me..

[old spoke]

I only had 4 songs done and then I started getting this message

[carmine minnow]

it's flagged. maybe profanity? I think I've gotten it for F***

[old spoke]

?

[light ether]

I see, it is getting auto-moderated. Try different lyrics

[old spoke]

problem solved thank you

[light heart]

How big is the chirp model and which models is it made out of
Bark is 3 X transformers and 80m params

[grim dew]

I didn't know you need credits. How do you get credits?

[runic dock]

Got it. Thanks. Will try this

[runic dock]

Yes . Thanks

[sand brook]

/chrip

[broken lotus]

Sadly I don't think Chirp knows all.... 🥲

❌ Theremin

[light ether]

Ha. That's a good test!

[broken lotus]

I also haven't gotten 'Synth' to work as an instrument... 🤔

'Synth Pop' works... (usually give a vocoder voice)

'Synth Arpeggio' 'Synth power chords' (etc) doesn't seem to work...

Any ideas for a (retro) synthesizer-y sound?

[open magnet]

loll, i'll think on it

i think a bunch of these might magically work if the model gets larger since it requires quite some text knowledge (which compared to audio of course less model capacity is spent on)

[tribal vapor]

how can i recreate chirp on my local machine?

[tribal vapor]

how to create jingles

[gritty bone]

/chirp

[tall orbit]

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.

[rotund trail]

why i cant generate ? cant see generate rooms

[graceful cliff]

see #🐣┃how-to-chirp

[rotund trail]

thnx

[rotund trail]

i got 40 sec sound

is there any way to make this one longer ?

i just get 2 3 second more i think

[untold briar]

40s is the max for a single generation, you have to click the arrow (see: <#⏩┃continuation message>) and combine two clips into one to go further.

[quartz marsh]

Hey team, is there any known issue with generating lyrics? I have attemped to add lyrics to numerous generations with nothing working. I just get a continuation of the music (which I like).

[daring lava]

Hey @quartz marsh could you please provide a screenshot of what you are trying to create? Prompt included

[quartz marsh]

I am continuing from a generated track and then trying to add lyrics

The initial generated track was an instrumental

so I am simply just adding lyrics under [verse]

strange, it is now working

[daring lava]

Great to hear!

[quartz marsh]

I also tried with [vocal] didn't work

without any [] and it partially worked

[broken lotus]

I've seen where chirp has difficulty changing tracks...
Like going from lyrics to instrumental and vice versa

might help to add a vocal style to the initial genre prompt and then put [Instrumental] in the first chirp....

but yeah, I have seen this and not sure of a reliable solution...

And then when the voice comes it might not be the way you'd described...

[open magnet]

yeah that makes sense. history is obv a big bias for a gpt style model. like depending on how 'instrumental' the track seems it might decide that it's more likely to continue that rather than add lyrics...

[broken lotus]

If the genre prompt is 'weighted a little bit more' that the in-line prompts and lyrics...
... how does the history rank...?

It seems the strongest, since it will stick with it's own song and ignore your prompts...

[quartz marsh]

interesting, so adding a vocal style to the initial intrumental prompt might actually help?

[broken lotus]

ehhh... 🤷‍♀️ I don't wanna say it is reliable...

[open magnet]

yeah technically both is history, but text descriptions/lyrics will always be more noisy in the train data than unsupervised audio history. so the model kinda learns that audio history is always veeery strong

one way to patch this would be: right now a continuation uses the last 10s of the previous clip as the start of the prediction. we could reduce that, in which case it would be more likely to follow the text prompt. would also help things like genre changes. downside is that it would be less coherent to previous generation. like if there was singing and it stopped a couple of seconds before the end of the clip then a continuation might make up a new singer or something

[quartz marsh]

is not possible to create a loop and then match vocals to that loop?

[broken lotus]

the 10sec history feels good right now..., but obv I have nothing to compare to.

I sometimes wish I could assign a history from a previous clip in the sequence... to refresh the original quality or return to chorus that hasn't strayed off

[quartz marsh]

then I suppose you could structure the music generation to bars as opposed to time

[open magnet]

yeah been playing with that as well, like take 5s of the original clip with best quality vocals, then 5s or whatever of the previous one to get good continuation

[broken lotus]

Some of the devs mentioned 'over-painting'... where we might be able to do that sort of thing.

(It did not sound like a thing that exists, just speculation for the future)

[open magnet]

might work, but also a bit of a hack in the sense that hopefully next model version will get better with all that anyways. right now model generations are still coming out v quickly for us luckily

[open magnet]

yeah we haven't had time to implement it yet, but in principle shouldn't be difficult

[quartz marsh]

I like some of the features from Stableaudio. Have you guys tried it yet? It is totally different to Suno. Well at least, it feels that way.

[open magnet]

ya totally different. more similar to musicGen

to me, stereo is the nicest addition

the rest feels very very similar to musicGen

but if you have interesting findings in what's nice about it i'd love to hear

[broken lotus]

I was going to say 'stereo', so I could be nice. lol

[open magnet]

hehe yup

already working on it...

[quartz marsh]

I feel like with Stableaudio I have more control, but it requires more effort, I do find it difficult to make songs, but I can make 'music'. With Suno it is way easier to make songs but it can be so random. 5 generations in, and I feel like I am getting closer, then the next 5 are not what I am looking for.

[open magnet]

interesting, yeah probably a more constrained problem

[untold briar]

My personal niche feature request that is admittedly a pretty silly use of resources is letting songs be continued in both forwards ⏩ and backwards ⏪ directions - from any clip/segment. Thinking about the non linear music that would come out of it makes head hurt in a way I super vibe with.

[quartz marsh]

was thinking more about this, rather than having 5s of generated music I would prefer 4 bars at 100bpm for example, if that is 4s or 8s should be irrelevant. If you know what I mean. I would like more control over the structure of music

with Stableaudio you can add bpm but it doesn't allow you generate the number of bars

[open magnet]

interesting ya..

[light ether]

We can probably figure this one out!

[cunning bluff]

Всё не то, всё не так, ты мой друг, я твой враг
Как же так всё у нас с тобою?
Был апрель, и в любви мы клялись, но, увы
Пролетел жёлтый лист по бульварам Москвы
Третье сентября - день прощанья
День, когда горят костры рябин
Как костры, горят обещания
В день, когда я совсем один
Я календарь переверну - и снова третье сентября
На фото я твоё взгляну - и снова третье сентября
Но почему, но почему расстаться всё же нам пришлось?
Ведь было всё у нас всерьёз второго сентября
Но почему, но почему расстаться всё же нам пришлось?
Ведь было всё у нас всерьёз второго сентября
Журавлей белый клин, твоя дочь и мой сын
Все хотят теплоты и ласки
Мы в любовь, как в игру на холодном ветру
Поиграли с тобой, но пришёл сам собой
Третье сентября - день прощанья
День, когда горят костры рябин
Как костры, горят обещания
В день, когда я совсем один
Я календарь переверну - и снова третье сентября
На фото я твоё взгляну - и снова третье сентября
Но почему, но почему расстаться всё же нам пришлось?
Ведь было всё у нас всерьёз второго сентября
Но почему, но почему расстаться всё же нам пришлось?
Ведь было всё у нас всерьёз второго сентября
Я календарь переверну - и снова третье сентября
На фото я твоё взгляну - и снова третье сентября
Но почему, но почему расстаться всё же нам пришлось?
Ведь было всё у нас всерьёз второго сентября
Но почему, но почему расстаться всё же нам пришлось?
Ведь было всё у нас всерьёз второго сентября

[daring lava]

Привет! Если вы пытаетесь использовать бот /chirp, проверьте это. Лучше будет объяснить, как #🐣┃how-to-chirp

[oblique hornet]

/chirp

[tall orbit]

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.

[summer dock]

Hi all! I am a little dense when it comes to translating legal jargon and I wanted to know if anyone could clarify the commercial usage terms? Is there restrictions with commercial usage? Any information would be greatly appreciated!

[broken lotus]

Is there a specific part that troubles you...? or is it the general 'legal-speak'...?

[graceful condor]

türkce sarki nasil yapilir

[summer dock]

More just the general 'legal-speak'...I'm a songwriter and I'm curious as to how I can integrate this into my songs while still complying with any legal restrictions.

[open magnet]

don't quote me but rough tldr: with free plan you can use non-commercially, as well as basic social posts. and with Pro plan you own it and you can use it commercially

[summer dock]

Thank you!

[broken lotus]

ok, I will summarize but:

I AM NOT A LAWYER™

(the important bits)

• the legalese makes Suno the 'owner' of the song (transferable to any future company that buys Suno).
• Suno can use the songs to improve and promote Suno (Suno cannot steal the music to grab fame and royalties)
• Suno ≠ Discord. Discord ≠ Suno.
• Free subscribers can use non-commercially with attribution to Suno.
• Paid subscribers get their songs sub-licensed back to them for commercial use.

[summer dock]

This is greatly appreciated! Thank you so much!

[unique dome]

More details in the General Commerical Terms section of our Subscriptions page:

https://suno.ai/docs/subscriptions

[arctic jetty]

/chirp

[tall orbit]

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.

[meager flower]

how does it work?

[daring lava]

#🐣┃how-to-chirp

[autumn olive]

hi

[wraith lagoon]

how to make a solo guitar part?

[compact ibex]

How to tweak chirp so that it adds some outro part of a song, or does not cut off so suddenly as sometimes it even cuts off the alst few lyrics?

[sinful abyss]

Is it possible to get timing information? What words sound at what timestamp. So it's possible to create dynamic captions or karaoke

[open magnet]

You could end with [interlude] for example or continue with that

[open magnet]

Not right now, but would be a cool feature to have. For now you could try whisper

[open magnet]

Leave lyrics empty and type “guitar solo” or something like that into the style?

[tall orbit]

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.

[dapper imp]

/chirp guitar solo

[daring lava]

Hey 👋 check out #🐣┃how-to-chirp

[burnt bison]

#chirp

#🐣┃chirp-beta-1

[scarlet delta]

/chirp

[tall orbit]

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.

[wanton kestrel]

Is there maybe a plan to point out which words have caused rejection?

is there a list of banned words?

Is there a way of adding a sound effect to the lyrics? Instead of a word a sound effect maybe

can I ask is the word Triggered banned> Im so not sure which of my lyrics are being flagged> they all pretty safe>

[open magnet]

unfortunately not, it's model based.

[gilded siren]

Could I get any advice regarding ending the song? I have tried [coda] however I have not had much success with this

[graceful cliff]

try [outro] [end] [fade out]

[gilded siren]

Thanks I will give these a go

[untold briar]

Some people reported [song ends] works as well. And all of those typically work best with only a small amount of lyrics, or no lyrics in front. But it can be pretty RNG hit or miss, like a lot of AI stuff.

If you just want a fade out while the music/song is still going, that is one thing easily done in any audio editor. So if nothing works that's always a fallback to end.

[gilded siren]

Very true with the edited fade. Thanks!

[tulip abyss]

/chirp

[tall orbit]

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.

[tame axle]

Hi

[tight nebula]

how do i change the language?

[royal salmon]

Why does it chop the end off so often?

Also, I'm getting quite a few videos with no lyrics, just the music. Any ideas?

Thank you 🙂

Oh - and I should have started with: Amazing use of AI! And to think this is only going to get better!! 🙂

[lapis nimbus]

girlfriend love

In the moonlight, our hearts entwine,
Your smile, darling, is oh so divine.
With you, every moment feels so right,
Underneath the stars, in your arms, I'll spend the night.

[quick raptor]

¯_(ツ)_/¯

/chirup

[zinc palm]

/chirup

[daring lava]

Could you post some screenshots of what you are submitting? Maybe I can help

[open magnet]

yeah, working on improving the cutoff for next version. in the meantime try continuing with lyrics that start riight where it left off then it should continue seamlessly.
as for instrumentals, try adding 'vocals' to the style. sometimes when the model is convinced the style of the music is more backgroundy it will decide to ignore the lyrics otherwise

[broken lotus]

fwiw, instrumental cut-off works well. 👍 easy to edit, usually.

(agree the lyrics cut-off can be awkward)

[wild timber]

jko

[royal salmon]

Awesome, thank you. I figured it was just new service issues- just wanted to make sure 🙂

[sinful abyss]

Could you explain me meta tags like [intro] [outro] [verse] etc - it's dictionary or it's goes directly to the model? So any chance that it will react to my own instructions like [speed up], [more bass] etc?

[inland grove]

for some lyrics i dont need background music or say just mild one, how to do it?

[open magnet]

add "vocals only" or "acapella" to the description?

[inland grove]

I will try thanks, previously i tried "no music"

but didnt worked

[neat breach]

I just paid y'all 10 bucks, ytf can I not use the service?

Lemme guess, "it'll kick in soon!" Which means I'll have to scroll through 35K messages to get back to the project I was putting together? LOVELY.

[broken lotus]

maybe search your own name and it will narrow down the 35K messages you need to scroll through?

Also can type /info to find out your current remaining tokens...

[open magnet]

no it should be instantaneous. do you have 2 accounts maybe and web was logged into a different one? a few people had that. and yes, search your name or just DM the bot directly or make your own server and invite the bot

[open magnet]

haha, yeah i suppose "no music" is a bit ambiguous 🙂

[open magnet]

it goes directly into the model, so edit instructions like "more" won't work. however general instructinos like "heavy bass" should work

[golden nexus]

/chirp

[tall orbit]

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.

[graceful condor]

@open magnet what version will you call Chirp right now? 🙂

I think it would be nice to see some alphanumerical version number in the announcements. Even if it's all very wishy-washy vague assessment of where you are in development.

Which - I understand would be very difficult to gauge.

[open magnet]

haha ya fair. technically last model was an alpha and this is a beta

that's why we did v0 and v1

we'll try to make things a bit cleaner going forward. right now like u said things are just still changing so quickly that it's hard to even roadmap things well

[somber garden]

I've figured out that there's a difference between verse and chorus, but do you have like segue or sabi?

[broken lotus]

I've seen [Pre-chorus] and [bridge]
(not sure how reliable they are)

[royal salmon]

/chirp

[tall orbit]

The /chirp command is not available in this channel

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

[broken lotus]

What's the longest song you've done? 😎

[graceful cliff]

Just wait -- it will get longer 😄

But my current v1 is like 2mins

[next pelican]

I ran out of credits on the free plan and upgraded my plan to the premium but it is still saying I am out of credits? What should I do?

[open magnet]

Can you send a quick email to us? Probably you have two accounts and the wrong one got was logged into the web app and got credited

[grizzled slate]

I just tried Chirp, and it didn't do a bad job with the lyrics I used. However, the start seems to be a little off as an introduction. Is there any way to modify and regenerate the intro and outros etc?

[untold briar]

A feature like that has been discussed by the devs as something they want to add eventually - essentially audio inpainting/outpainting - but not yet. But in the case of the outro you can go back to the previous segment and keep trying again with a new continue. But you can't tweak just a few seconds of one segment, or generate in the opposite direction back in time.

Intro and Outro 'guidance' should be better in the not too distant version as well. As in the ability to specific in your song prompt that the song should have an intro or the song should wrap up, in a more reliable way. (You can do it even now, with [intro] and [outro], but it is more hit or miss)

[brazen iron]

/chirp

[tall orbit]

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.

[fast coyote]

I think that it can generate outros more reliably if you also add it to the style (for example "edm outro"). I managed to get it to produce outros semi consistently by doing this.

[fading current]

Are there times when it just won't finish? I submitted a chirp and it says it's working for ~10 minutes and then "the application did not respond"

[untold briar]

Yeah, you probably just have to try again, that's too long.

[lean junco]

I'm trying to replicate the delay effect of this song https://youtu.be/cQKGUgOfD8U?si=15OG9RI6z2qCQzZL&t=143 but uhh... well the lyrics are flagged, which I guess makes sense

[unreal tusk]

/chirp write a country song

[grizzled slate]

Many thanks for the reply! Suno is something I'll definitely be following. I'm an old retired guy with a few problems that leave me a housebound, wheelchair-using deaf and daft nut! Dementia is a bummer! I did a song-a-day challenge to raise money for Dementia Charity and they asked me to do another. I'd like to do some original songs and think Suno could be the magic to give me great starting point!

[grizzled slate]

Oh! I'll have to give that a go! It seems like how you prompt and what you prompt makes a difference. I'm just checking out the user section to gain insight into using Suno. A few questions: I'll be trying out lots of variations of prompts and formats to test the results. I'd like to share any positive results... but where do I do this, please? Often the results seem to be Rap orientated but I suppose that has to do with the text syntax. Is this something you're looking at improving? Does each generation and regeneration incur credit costs? Can I stipulate voice options? When is the next update/alpha expected? Many thanks and keep up the great job you are doing! Music for all, not just the few! Brilliant!

[grizzled slate]

It seems strange to call you The Table, but there you go! Thanks for the tip about the outro idea. It does seem that the prompt you use and how you construct it makes a difference. I read somewhere that using [ ] seems to help and avoid using commas and full stops. I'm sure there must be other things users are finding. I wish there was some easy way we could compile these somewhere on Discord where everyone could benefit and add to. Many thanks!

[open magnet]

Thanks feedback here is greatly appreciated. We only launched the beta very recently so lots of things to improve 🙂 you are right that in general the genre is picked up from the style of the lyrics however in the newest version you can describe the style of the song, so it should override whatever is picked up from the lyrics

[open magnet]

Yeah brackets can be used for descriptions such as [Chorus] [Verse]. Technically the ‘style’ field also just gets put into brackets and prepended to the text input.

[broken lotus]

I can't listen to that crap generic 'death metal' voice anymore 😩

[foggy trellis]

/chirp

[tall orbit]

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.

[regal crow]

how create a private Suno Chirp

[graceful cliff]

See #🐣┃how-to-chirp

You can DM

[unique dome]

Here's the bot:

@versed valley

And step-by-step instructions here if needed:

https://suno.ai/docs/dms

[stone dagger]

Does anybody have a nice and easy tool to remove a bit of noise from the Chirp clips? Online or offline...

[copper spruce]

Can chirp bot integrate with node js server ? I mean can I pos request for generating chirp bot from my node js server ?

[whole quartz]

doesnt seem to be an option for that in their docs unfortunately, although that would be great: https://suno-ai.notion.site/Suno-Docs-38e5ba5856d249a89dcea31655f4fb74

[tribal cargo]

/chirp

[tall orbit]

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.

[charred lake]

Hey my last two renders were remixes of this, but they haven't come out yet - wondering if they are in queue? #🐣┃chirp-beta-1 message

[digital flare]

/chirp

[tall orbit]

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.

[fast coyote]

Is stereo audio planned for the future?

[lusty compass]

Why as I'm building the song. The voice gets unclear as it develops? Rock vocals for an example? I'm using way too much credits trying to get a clear voice like the begging of the song?

[open magnet]

Ya definitely on the roadmap

[open magnet]

Unfortunately the context window for llm is a bit limited. Right now it’s 40s for chirp so eventually it starts drifting

[open magnet]

Maybe a worker temporarily crashed. Just resubmit, won’t count towards credits

[fast coyote]

Awesome! This alone will make the sound quality appear so much better

[lusty compass]

I can't really complete the song

[open magnet]

Sorry 😦 usually you can get a couple of minutes of coherence but is hard. Next version should have much longer context so the should help hugely for this kind of drift

Maybe you could continue of of the earlier clips to get an ending and then find a way to Stich the clips

[lusty compass]

I tried that but the music song is completely different. Doesn't blend well

I bought a 1000 chirps and already used half trying to fix 3 songs

We need 5000 chips so we can not stress about excessive usage trying to fix these songs.

[broken lotus]

when you say 'next version'.... 😋 😃

are we talking weeks away? months away?

Nosy busybodies want to know!

[graceful cliff]

Statistically the right answer is usually around the geometric mean of the lower and upper bound of your guesses 😄

[open magnet]

Loll

[charred lake]

Yooo can I get help with a flagged prompt

I promise it's not nefarious!

Also, can I get hepl maintaining the same lead singer voice in this bit? She is AMAZING but every continuation I've tried it's a meh sounding dude #🐣┃chirp-beta-1 message

[graceful cliff]

See #🐣┃chirp-beta-1 message, or pieced together: #😍┃v1-songs message

[stone dagger]

Given the limited context window and natural drifting, as mentioned above, I usually do at most 2 generations after the first one, but I do it multiple times. Then I get the best clips and stitch them together with a DAW software (actually using BandLab Studio, online), to obtain a song that sounds real and as most consistent as possible. I got some very good results with this technique.

[stone dagger]

now, I don't want to promote myself here, but I posted a couple of them in the chirp-showcase channel, if you want to listen to the results. You can hear they are pretty consistent, but once you know what I wrote above, you understand how it is made

[ancient scaffold]

/chirp

[tall orbit]

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.

[knotty kite]

As a subscriber, If we don’t use all our credits do they get stacked with our credits for next month or do they disappear?

[proven kiln]

Hi. is there a way to remove the chirps you get but do not like?

[frozen harbor]

Is there a premium version of this software that can do longer sets of lyrics?

[low idol]

What are the limits? 😮

[gusty granite]

/chirp

[tall orbit]

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.

[daring lava]

The credits that come with your monthly subscription do not roll over from month to month. Any credits purchased outside of the plan will

[daring lava]

Not at this time but you can use continuations to make longer songs. Check out #⏩┃continuation

[graceful condor]

Can we send messages to bot privately?

It would be better If we could privately request song

[broken lotus]

just dm the bot.
then you can use the same instruction privately

[graceful condor]

lol mb

thanks for informing

[spark blade]

How to retrace all of our chirps without scrolling up and down?

[winter vale]

how do we dm the bot?

[daring lava]

@winter vale

[daring lava]

You can use the search bar. Click on Mentions, begin typing your user name and select it when it pops up and it should show you a list of all of your generations

[ornate raptor]

/chirp

[tall orbit]

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.

[wide crown]

Is it possible just to generate music without the singing?

[graceful cliff]

yes try to prompt without text or just do ‘[piano intro]’

[mild ivy]

so, how do i make the bot follow a more specific style, like can i give him references, mix rithms and such?

[fallen wraith]

/chirp

[tall orbit]

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.

[fresh cargo]

/chirp

[tall orbit]

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.

[open magnet]

just text descriptions are supported for now

[charred fog]

/chirp command

[tight hamlet]

/chirp

[tall orbit]

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.

[still jetty]

#chirp

[tranquil epoch]

/chirp

[tall orbit]

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.

[frigid nest]

/Chirp

[tall orbit]

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.

[charred lake]

I ran out of credits after 2 days.... i sifted through quite a few 3-4 second generations and other ones that were just glitched out sounding. Any way I could get supplemental credits?

[fleet dagger]

.

[royal girder]

/Chirp

[tall orbit]

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.

[short perch]

/Chirp

[tall orbit]

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.

[tranquil fossil]

hi guys why chirp bot can no longer give good results why!!!

[open magnet]

Nothing changed, probably unlucky with the chosen lyrics?

[wide wing]

how do you get the full song? it only plays 40 seconds of it.

[daring lava]

Hey 🙂 check out #⏩┃continuation

[wide wing]

@daring lava One other question, is there a list of [parts] that are available to use (eg. [Verse], [Chorus], etc.) thanks

[unique dome]

It’s far from exhaustive, but here’s a list of the types of metatags that tend to work well:

https://suno.ai/docs/metatags

You can also try stuff like [Intro] and [Outro], but there’s no fixed list

[wide wing]

awesome, thank you.

[wide wing]

It seems to recognize some music genres that aren't listed either, like retrowave, synthwave, dubstep

[stone dagger]

Anyone managed to find a workflow to clean those generations with too much noise?

[stone dagger]

I would LOVE a quick and dirty solution to just remove a bit of noise for free and kinda instantly, but I seriously couldn't find any satisfying option...

Anyone from Suno itself has any viable solution? Anyone tried anything? I would actually even pay if it is good...

[red mauve]

hello?

is it possible to invite the CHIRP bot to my page?

i pay, so this would be great 😉

[obtuse notch]

guys

can i make a breakdown?

somehow

#🐣┃chirp-beta-1 message

the first one was really good

[daring lava]

This is something the team is actively working to accomplish. A lot of great things to come. I personally don’t solicit third party stuff, so I wouldn’t be much help…. But a quick search on google/bing/etc could turn up some results I bet

[daring lava]

Hey 👋 here’s how to add to your own server 🙂

Just click on the Suno Chirp Bot and click on the server button and follow the steps above (add /chirp bot)

[sour pawn]

Hey can I ask a quick, kinda dumb question?@daring lava

[daring lava]

Sure thing, what’s up?

[sour pawn]

So I notice some prompts end up getting flagged. I am not using curse words, or over graphic detail, but it still get's flagged, is there a public list of banned words? I saw someone's song the other day that was pretty much straight up icp and has pretty much your standard gangsta rap lryics, so just wondering.

[daring lava]

It could be that you are including too many words in your prompt. Do you mind screenshotting and showing an example?

[sour pawn]

Hmm, you know that does make sense, I have been using the chat gpt too much, and maybe it's becoming a bit too much for it, umm hmm I just got done a song. Here i will make a new one after I hear it, and I will make a new one. I have sorta..been on a strong binge of making music lol

[daring lava]

Try looking through the material posted in #1151542418991239168 there is a lot of good information in there. It is suggested to include between 4-8 lines for a prompt to get best result. The included ChatGPT prompt creator should already be doing that unless a user instructs it to do otherwise.

That being saud, I've seen hiphop/rap be able to carry a bit more lyrical "weight" so to speak, because of how quickly it churns through the lyrics. Either way, I would play with your line count and play/test around

[sour pawn]

I got you, thank you. Back to making fun evil songs for my project.

[daring lava]

Welcome 👋 Best place to showcase what you create is #🐣┃suno-showcase 🙂 the whole team is active here and regularly float through the channels to check out all the cool stuff people make

[neat nexus]

Is there a way to loop back a previous chorus rather than hoping to recreate something similar through chance?

[sour burrow]

wheres Suno chrip bot

[brazen sleet]

/chirp

[tall orbit]

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.

[daring lava]

Hey BaharSpring 👋 check out #🐣┃how-to-chirp

[neat nexus]

I get that you can keep regenerating, but what are the limits of the system itself?

How long can a single song max out?

How long can a single prompt generate in playback?

For instance, assuming no upper limit on single songs, what would prohibit me from creating a radio show, complete with DJ by prompting over time for a thirty minute segment?

I ask about the single prompt generation playback time because I notice that Chrip tends to try to speed up its speech to stay within thirty seconds.

[untold briar]

There isn't a theoretical limit on continues but like a lot of AI models errors tends to accumulate over time -- the model ends up continuing problems and so practically speaking it is hard to get great quality past a a few minutes. So if you wanted to make 30 minute single audio you would almost certainly get better quality be faking it being actually only using continues, starting fresh in places instead.

That said, sometimes if you just keep going you can push past a problem and it recovers. Also while there isn't a theoretical limit there may be a limit on the bot audio length for technical reasons or Discord reasons, I'm not actually sure if there is.

[neat nexus]

Cool. I see what you mean though. Things could quit blending properly.

Might make a fun experiment though.

[untold briar]

Sometimes it fails in a super fun way though, this is probably my favorite

The cats make Chirp absolute lose its marbles for awhile there, thought it starts to recover by the end. Still has a lot of extra noise leftover from the meows though.

[neat nexus]

Oh wow....lulz...

I ain't even mad, that's awsome.

[untold briar]

Yeah I'm just speculating but I think the words got turned into something like a musical rhythm, more like a drum-beat or guitar strum, in the model instead of lyrics there, so it just coudl not stop, lol

[neat nexus]

Yeah, but it's the kind of thing I wish I could watch people's faces when they hear it for the first time.

Makes me wonder...has anyone tried to recreate The Jones Laughing Record (Introducing The Flight Of The Bumble Bee)?

[neat nexus]

So that inspired me and I tried my hand at one...it kind of works a bit?

[open magnet]

very cool

[barren locust]

Hey people, can somebody explain top me how can I delete my chirps? Not my own posts, but the resulting songs from the bot. I guess I need to contact staff, but who exactly?

[stone dagger]

Thank you very much for your kind answer. No luck on the internet, I tried this and that and nothing works as I want. I guess I'll just wait for the next release of Suno. Anyway good to know that new features are coming. I had no doubt, but still it's good to know!

[neat nexus]

(Possibly) dumb question: Does submitting changes to the genre type switch genres or blend with the previously submitted genre?

For instance, do I need to always input the genre style each time or can I start with one, shift it by adding another but still get that first musical style present throughout the entire production?

If I say start off with "new age" as my genre and do an intro, then switch the second one to Gregorian Chant will it remember the new age part of the beginning?

[daring lava]

Feel free to dm me, I’ll poke around and see what I can find that might could help 🙂

[open magnet]

continuations keep ~10s of the previous generation. so if your genre is wildly different the model will probably respect the audio history more than that new genre prompt. however if similar it will probably slowly drift or something?

[neat nexus]

I see. So it is biased in favor of the previous type but will roll the dice when continuing with new stuff?

[daring lava]

Not a dumb question, a very interesting and reasonable question if you ask me. I have tried this a little here and there (rock->hip hop, pop->edm) As georg said, it really picks up from where the latter left off. That being said, results can bleed off and taper into something new. Drastic changes are a bit difficult, but even then, not impossible. A lot of figuring out in this new landscape though. Unchartered territory

[neat nexus]

Yeah. Definitely.

Basically I'm trying to see if I can mangle something slightly Enigma sounding.

(I say mangle because I'm not actually expcting to get there and hope instead to come up with something new but that is my starting point.)

[neat nexus]

Also when it keeps that ~10 does it sometimes lose stuff repeated in that section? Because it seemed to me that it did a bit there in some of my stuff.

[neat nexus]

On another note:

Do we get charged when an error occurs?

Weird. It gave me two errors in a row and then went back and actually generated one of the ones which errored.

[daring lava]

No, errors do not cost credits

[surreal venture]

I don't understand how to make a longer clip? I mean, 40 sec. is not enough for me. Also, how can i choose a vocal style, like "growl" in chorus

[open magnet]

not sure i understand. the model will only last 10s of the previous generation, so anything before that it will have no idea

[peak heart]

Can't we make the duration longer? Like 2.5 - 3 minutes.

[eternal bough]

1 minute would be great but often only 20-30 seconds are made if the song text is not long enough.

or is there a trick for this?

[graceful cliff]

try #⏩┃continuation for now

[graceful cliff]

you can try to add [intro] [outro] [instrumental] etc before / during / after the lyrics

[peak heart]

thanks

[sweet stone]

can we lower the voice volume?

[neat nexus]

I've done several attempts now to complete a chorus and keep getting back two second sections. I have credits so I don't know what is going on.

This is crap.

I've gotten five of these in a row now. Is something down?

Generation time was short on most of those too.

I typically get anywhere from a minute to a minute and a half for generation. These have been taking like nineteen seconds to make.

Proof that I should have plenty of credits to continue.

I hate to lose the momentum I felt when building this but these are crap.

The two seconds I get sounds like it could be good but never finishes.

[abstract current]

Hi everyone, after I've made something, how do I save it in order to play it somewhere else?

[neat nexus]

I guess I'm just going to have to go to bed and hope this gets looked at in the morning.

[neat nexus]

If you've already hearted it to compile the whole song, then you should be able to download it like any other video.

When you hover over it with the mouse it should show a button like the one in the image.

[abstract current]

thanks, Hepcat

[neat nexus]

Glad I could help. 🙂

[abstract current]

I selected the type of song and entered the words writer that prepares the file. Searching by my username. But can't find the file? Where is he hiding??

[graceful condor]

Hej

[abstract owl]

How its work

?

[daring lava]

Hey Hepcat! Sorry you’re bumping into this. Could you possibly email [e-mail] at your convenience. Please include the specifics of what is happening with examples.

[daring lava]

Hey 🙂 check out #🐣┃how-to-chirp

[neat nexus]

Have sent email. Thanks.

[daring lava]

Very welcome, thank you for your patience with this

[neat nexus]

Followed up with a second email, including a new screenshot and the two chirps.

[unique dome]

Received. Thanks, @neat nexus. We're looking into this, and we'll get back to you shortly!

[neat nexus]

Thank you for the fast response.

[neat nexus]

....

Is it possible to stitch it up and still come back to a work in progress to continue where you left off?

[light ether]

Hey, I deleted a few for you but right lcick on the message and click "Apps > delete chirp" (only works for yours)

[barren locust]

Thank you very much!

[neat nexus]

Possibly dumb question: Are meta tags case sensitive?

[summer lily]

Wish they increase the amount of free credits. 25 is really bit 😦

[marble condor]

hello, is there a list of music genres available? I ask because when I tried heavy metal is does not distinguish from speed metal?

[open magnet]

we case augment, so there could be differences but technically anything should work

[open magnet]

there is no well-defined list of genres since the model just learns from a bunch of different stuff, so not sure which sub categories work well and how to best describe them

[open magnet]

GPUs $$$ 😦

[neat nexus]

Is [outro] just broken then?

[open magnet]

hmm shouldn't be...

[neat nexus]

[Intro] and [Instrumental] I've had good luck with, [Outro] not so much.

[open magnet]

but i think technically an outro can still be pretty long...

[neat nexus]

I mean that it just...ends.

I dunno, I guess I expected a fade out or something?

Oh I do want to mention that some of the issues I was having with last night were clearly related to the length of my song--I had no idea I was doing an eight minute song. So thanks for the help and I apologize.

[open magnet]

haha yeah.. we gotta find a way to make the continuation stuff more obvious... discord is a bit limiting here unfortunately..

lemme try outro quickly

hm good point ya i see what you mean

i think this is a problem with when we allow it to end...

will see if we can do an easy fix but next model we are planning a bit more timing control in general

so hopefully won't be an issue anymore

[neat nexus]

Good to hear that.

[sinful abyss]

is it possible to get constant bpm for long track? now it's floating a bit

[daring lava]

That’s a great idea, mind posting it in #💡┃suggestions-web ?

[sinful abyss]

sure

[daring lava]

Thank you 🙂

[graceful condor]

/chirp

[tall orbit]

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.

[daring lava]

Hey hoha 👋 Welcome to Suno! Please check out #🐣┃how-to-chirp

[neat nexus]

So I tried to stretch into something new again, this time inspired by a mention on a YouTube video about beats. So I prompted Sixties Soul, funk for the style and did an [intro] to get the music started.

But ah....where did these words come from?

I admit, I'm hard of hearing so as much as I might like some sounds, I know that I can't always hear or understand those sounds. So I have no idea what is being said here.

[neat nexus]

What is it even doing?

How is this instrumental?

[open magnet]

😂

try 'instrumental, sixties soul, funk' into the style and keep text empty

[neat nexus]

I'll do that next. Just tried with removing funk and still got vocals. My guess is something about "sixties soul" is causing it.

[untold briar]

Well human voice is technically considered an instrument in music I think so... 😀

[neat nexus]

That worked!

Now I just hope this doesn't break things when I attempt to create lyrics. Thanks.

[neat nexus]

https://tenor.com/view/technically-correct-the-best-kind-of-correct-gif-17815153

[broken lotus]

try [Fade Out]

[neat nexus]

Is that a Chirp Metatag?

I mean I'll try it, but I didn't know it was a thing.

[broken lotus]

yes

use it on it's own card (often works).

Also [End] sometimes gets an actual end.

[Outro] seems to create the outro but you'll need to fade out in editing software

[neat nexus]

OK, that actually makes sense. I can see that being a feature even. Some people would want to control that manually for themselves.

[neat nexus]

Thank you for this suggestion! It worked well!

[limpid storm]

hows chirp api looking

[neat nexus]

Hey quick question before I go ahead and pull the trigger.

How do extra credits work? Do they roll over month to month? If I still have some and a new month begins and my automatic subscription kicks in, does the system go for the subscription first or the extra credits first?

(Sorry for double posting here and in #general)

[open magnet]

They are independent. The monthly ones refresh but the extra ones stay forever ontop of the monthly ones

[neat nexus]

That's good to know at least. At my current rate of burn, I suspect the question may turn out to be strictly accademic any way. 😅

[neat nexus]

I'm getting bit by a flag warning as well:

In the candlelight's glow our eyes they meet Sultry smiles exchanged hearts skipping a beat Hips start to sway as we grind and rub Lost in the moment feeling the love

Are sexy songs not allowed?

is this an actual expected limitation or a bug?

As expected, it was as we grind and rub that was the problem.

So no sexy time songs?

[crystal tendon]

@neat nexus The duration needs to be longer 😦 I could make this song perfect, but the duration is so short that it is difficult to progress.

[untold briar]

Important people on working on it...

[untold briar]

Moderation is fairly primitive at the moment but improving. This seems silly, I agree. I'm tempted to throw a 'steamy' (but extremely indirectly written) scene from a shakespeare play or a victorian novel as lyrics and see if even that gets flagged 😄

[broken lotus]

😩 how will I finish my Opera about Michelangelo who grinds and rubs a marble statue to polished perfection during the Italian Renaissance...?

I mean, the statue is a 17ft tall nekkid hunk... 😋 and there are at least 4 arias about rubbing and grinding on his... various parts. 🫣

What have you got against the Italian Renaissance, Discord???

[neat nexus]

lulz

[elfin coral]

lulz

[neat nexus]

Is it a Discord limit, a ChatGPT limit, or an internal limit imposed by the developers who aren't really interested in a bleepfest full of bleep, bleeping bleepers? 🤣

If it is the developers, that's fair enough--it's there thing and they have a right to choose how it's used.

I'm just wondering.

[open magnet]

yeah it's a tough call. we're obv aware that prob a good quarter of all existing music wouldn't pass our filters, but if we relax them... aehem well, we tried that...

[neat nexus]

Oh? You did?

....

....oh God. Yes, you probably did and I can already guess what probably happened. My deepest sympathies.

[neat nexus]

Any suggestions on how to get a song to return to an earlier voice? I'm stuck in a messy chorus tone when I'm trying to get back to the beginning vocals for the verse. It's fighting me.

Sixteen times I tried the same piece. I've edited. I've tried instrumentals, genre changes...it's just fighting me.

Whew...I tried an entirely different genre and that seems to have done it finally. I need to see how the next piece bolts on to be sure.

[open magnet]

yeah that's probably difficult considering the 40s total context

verse-chorus-verse pattern sometimes goes back to original voice but only if the verses are separated by less than 40s

[neat nexus]

40 seconds?

[rare junco]

Not sure if this is the right channel to ask, or where to direct my question to.
It looks like I've lost my bonus credits from last month (the ones in celebration of 500k chirps, see announcement channel) that I hadn't even used. Is this intentional? 😢

I expected the regular credits to get reset, but I figured the unused bonus credits would carry over.

At the end of the day, not a huge deal either way because I've not put a single cent towards this amazing project

[neat nexus]

Did someone put the volume at half all of a sudden?

[light ether]

We'll fix this one soon

[graceful condor]

Kann man auch längere Songs erstellen lassen ?

[daring lava]

Oh ja, auf jeden Fall. Bitte schauen Sie sich das an... https://discord.com/channels/1069381916492562582/1149508588470022164 es wird Ihnen helfen, zu erklären, wie Sie eine Generation verlängern können, um sie länger zu machen

[brittle wadi]

Hi where can i start to produce a song with a prompt?

[neat nexus]

/chirp

[tall orbit]

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.

[daring lava]

Hey T1nvest 👋 please check out #🐣┃how-to-chirp

[hollow delta]

@light ether I also lost my extra credits at the end of the month

[light ether]

@hollow delta @rare junco this should be fixed now - ping me if it's not!

[rare junco]

yup i got mine back, thanks for the quick fix 👍

[neat nexus]

....wait...

All our songs are mono?

[light ether]

Yeah currently, working on stereo

[neat nexus]

Huh.

AM was better than I thought, eh?

[light ether]

I am certainly surprised by how decent mono can sound, but we are definitely working hard on stereo goodness

[neat nexus]

And I'll be looking forward to it when it comes. 😉

[long drift]

My credits didn't come back either. Is there anything I need to do?

[hollow delta]

@light ether Thank you, I got back.

[old vine]

/chirp

[tall orbit]

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.

[old vine]

/chirp

[daring lava]

Hey linly:) please check out #🐣┃how-to-chirp

[graceful condor]

how to join all the separated songs?

[polar arch]

Heart react the final one

[broken lotus]

I heart react you.

(this is going in a song)

[neat nexus]

This is the way.

Ever since I started using this I've begun to see lyrics everywhere.

[light ether]

I suspect you may have used the extra ones last month. But it's a new month so you should have more montly creidts now (try /info)

[neat nexus]

Can I get advice here on how to progress a song or is that what #💬┃general-chat is for?

I was inspired by something that a different person put up that they did. So I tried my hand at it and failed and failed again and again. I tried it again today and as a joke put James Blunt in as a genre and it seemed to work so I went with it. There are two issues I'm having.

• I've noticed that the music can get a bit repetitive so I've been trying to mitigate that with instructions on music in the lyrics area. Some of the time it seems to be working but since I'm not actually musical myself I could use some recommendations on what to say to goose this along.

• I've lost my lead singer. I tried to keep it simple and under the time limit but you can noticably hear the vocals shifting as the song goes on. I could use some advice on how to goose him back into singing properly. It seems to have hot during the chorus (naturally.)

Here it is in progress:

[untold briar]

Keeping the lead singer is tricky and somewhat a matter of luck. You can try calling back to a phrase or entire line of lyrics at regular intervals, sometimes that can sort of preserve it a bit. Should be better in future versions as well. When you continue you usually don't need to put the music prompt in, but you can also put it in the continue and sort of makes the style stronger -- which *sometimes * can mean preserving it better, though it can also make it just weirder -- or you can even tweak it a bit. (By default doing this often makes it a bit overly strong, but can be worth a shot.)

[neat nexus]

I thought about a call back. I should give it a try and hopefully it will read as an adlib.

Hmmm... that's another suggestion idea...

I'm going to give it a try.

[untold briar]

You can also fake a longer song and put it together yourself in an editor. So you use the same single-clip that has a clean ending, and generate multiple verses from that one, rather than going continue -> continue -> continue in a line. So it has less time to drift. Not ideal of course.

[broken lotus]

out-of-box suggestion might be to pick a 'seed' clip (after long intro), then introduce a repeating [refrain] lyric, even better if it's not a full-length clip but just a short 2 or 4 lines, then go into the chorus or verse, and restart back at the short 'seed' clip to generate the various other parts of the song...

then of course use audio software to edit all the clips together....

Jonathan beat me to it.

[untold briar]

It's hard to blend some music styles, but some are easier to edit like that. This is a tool I've used before, it's not quite as fancy as it first appears (it's pretty simple under the hood I think) but can smooth out segment joins https://mofi.loud.red/

[neat nexus]

Yeah. Honestly? I've been avoiding trying to edit stuff together as it feels like cheating to me. But I do see the appeal.

Any suggestions on instrumentation?

[broken lotus]

in general?

[broken lotus]

and yeah... it feels like cheating....
but the AI won't mind...

[neat nexus]

Yeah. Just something to toss in so it's not strumming completely the same chords over and over and over like I notice happens.

I tried [Instrumental] but it seems like it's missing directions to get it to introduce elements.

[broken lotus]

I tried 12-bar and 16-bar... but it ignored me.

[untold briar]

I haven't tried this recently but in earlier Chirp I found switching music terms in the prompt between segments. Soft/Hard, fast/slow, minor/major, ascending/descending, etc

[broken lotus]

[Guitar solo]
[Percussion Breakdown]
[Bass and Percussion Interlude]
...

[untold briar]

For pure instrumental even prompting opposites as a single prompt was kind of cool

[neat nexus]

This is the kind of thing I was looking for with this! Thanks!

[untold briar]

The danger of instrumentals is you can lose your singer, so don't use it at the end

[broken lotus]

I had some things work surprisingly.... I'll go back through my prompts, see what I can find....

[untold briar]

If cost isn't any issue you can just RNG it until the singer sticks of course. Tried and true.

[neat nexus]

I mean that's what I was doing before with fairly...ok..results. But after the sixteenth time you start asking if there is a better way.

And while I still have those extra credits, it's a loooong way from the end of the month.

[graceful condor]

Hey guys, you know how to do a song like Linkin Park with 2 singers and one singing, one rapping?

[neat nexus]

Editing.

...

I'm being snarky because I just "cheated" by editing a track of my own.

You might be able to generate [singer] and [rap] metatags and it understand it, but you might also be beating your head on a hard surface trying to make it work.

Editing (cheating) might just work better and easier.

[neat nexus]

So uh....

How long should I wait to stitch a clip together before concluding something messed up somewhere?

[untold briar]

Not too long usually. Is it super long? Hearts don't cost credits so you can just try again whenever.

[neat nexus]

It wasn't really. It just hung and I didn't know if something was wrong or if I would be making it worse by telling it to stitch it again or not.

I grabbed it now. Thanks.

[neat nexus]

I posted that one in showcase as well now.

[broken oxide]

how to check your credits

[broken oxide]

thanks

[light ether]

May hang once in a while, feel free to click again if it looks stuck (no credits associated with that either)

[neat nexus]

Will do next time. Thanks.

[viral lotus]

@untold briarI don't suppose Chirp is available for download onto PC?

[untold briar]

Discord only, alas. Maybe some day the devs did say, but way too early to even think about, just getting off the ground really. (Assuming you mean open source. I guess there could be a desktop-interface version that just doesn't use Discord.)

[untold briar]

late night curiosity on chord prompts, checked 5 random Chirp samples. The ordering of chords was never correct, but the selection of chords that showed up might have a weak but nonzero correlation to chords in the style prompt. Being extremely generous and just checking if the note/letter of the chord matches - not the precise chord - results were: 1 totally wrong, 1 with 2/3 matches, 2 with 2/4, and this one might be a 4 for 4. Ordering seemed to have zero correlation, and in practice it matters a lot WHICH C chord you use (I didn't even check major versus minor) so it's probably not super useful even if there is a connection, but it could be better than chance.

Didn't check any chords in the lyrics field, just style. I was kind of hoping it'd be obviously doing absolutely nothing or clearly doing something, but it's ambiguous enough that you'd have to look at a lot of samples to know for sure, which I'll leave to the more musically inclined to investigate.

[viral lotus]

Thank you, I just meant a desktop version.

[viral panther]

Hi there! Does anybody has a clue how can create "instrumental themes" with chirp? Let's say having an initial beat, then make variations out of it (same style, different dynamics, tempo, adding sg to it or remove sg..)

[broken lotus]

what is 'sg'?

[viral panther]

"something" ... sorry

basically i'd like to prompt something like this:
For the first 4 sec “Soft instrumental music begins, resembling the early morning calm”
then for 12 sec “Music gradually picks up pace.”
then for 36 sec “Transition to rhythmic beats as fashion showcase begins.”
then for 24 sec “Music becomes softer, taking a reflective tone.”
then for 14 sec “Music begins its crescendo, building towards the climax.”

[broken lotus]

I have tried using multi-prompts on a single card... it is not as reliable as using prompts on separate Continuations....

the basic instructions are to add you prompt in square brackets down in the field where the lyrics go.

so

[Instrumental]

as the only text in the lyrics field should return a music-only chirp

the other step-by-step instructions are less reliable, maybe not reliable at all....

read some other people's prompts and final videos, to see how they are using prompts (and comparing to if it worked or not)

[graceful condor]

been waiting for like 12hrs to get my full song. its made from around 20 chirps i suppose, from different time stamps. did try few more times but no luck

[broken lotus]

If you have some time..., could you go back a few generations and see from where in the chain it CAN be completed...?

I wonder if it's a limit on the number of continuations, or a file-size limit for the final video, or.... something else....

(I've never made a chain that long)

[neat nexus]

This is a segment from the chorus of a song I did a few days ago.

How does one get that type of singing to work on command? What are some ways to prime the pump or to indicate to the AI that we're looking for voices of this nature?

Another example:

These are vocals using effects. How do I get those effects?

(This is not an accusation) but there feels a bit of a gacha type process here. I try one thing, pull the lever, wait. I like it then I go on and pull the next lever. I don't like it, I start again and try to get something I like. It's all throwing the dice. What am I missing? Is it all RNG and I just have to live with it?

Or are there methods that I'm just not seeing to get the spin and effect I'm seeking when doing stuff? Even something that gets me in the right ballpark would be better than this current method where I throw stuff accross the room in hopes of hitting anything at all.

[broken lotus]

you never said WHAT type of singing....

guessing from the 2 examples you want clear and melodic...?

and obv can't see what you prompted...

[neat nexus]

lemme look and see if I can find the sections again.

[broken lotus]

or maybe say what kind of vocals you keep getting that are NOT the examples?

[graceful condor]

thats some great suggestion, thanks

[neat nexus]

It's one of those deals where you know it when you hear it.

The way those voices have an intensity or echo to them.

[broken lotus]

like "reverb"?

[neat nexus]

I'm looking for that melody in a chorus for the 'groove together' chorus. The other one is just intriguing me because of the reverb voice. Yeah. But I'm really not sure how to force that to happen on demand.

(I'm really not educated musically. Not sure if this is a positive or a negative with this stuff.)

[broken lotus]

😉 understood..., just want to be sure we are talking about the same thing...

[neat nexus]

Yeah.

The groove together thing is really what's stuck in my head at the moment.

It has that whole slow dance vibe that I want to tap into again.

[broken lotus]

is it the way his voice goes up, and emphasized the word groove?

[neat nexus]

The other clip just feeds into my Daft Punk addiction.

[neat nexus]

Yeah! That!

It adds an intensity.

[gritty cradle]

Hello please help me in making hindi rap song

[broken lotus]

I think the standard answer is just type your lyrics in that language.

there are some language discussion threads in https://discord.com/channels/1069381916492562582/1101325881059840040

[broken lotus]

one thing I tried that seemed to work is to type the emphasis word in ALL CAPS, the singer almost always went UP several notes on that word....

I have also tried [Melodic Singer] (or [Melodic Singing]...?) to get the voice to move around a bit...

[neat nexus]

I'm trying the all caps thing with that older segment again to see what it does.

[neat nexus]

He's more concerned about getting a consistent beat and the clips not being long enough. He DM'd me.

Unfortunately while I enjoyed some hip hop in the past most rap is not my thing.

This was test one, just doing the caps.

Tried again, throwing in there the reverb with the instrumental, changed type to melodic singer.

OK!

Now we're getting somewhere...

[broken lotus]

it's like 50/50....

maybe it just likes the melody it created..., on GROOVE and sticking with it (ignoring BESIDES)

[neat nexus]

The first one got seriously close!

It did seem to be adlibbing a lot in the beginning part though.

[broken lotus]

😅 🤷‍♀️

[neat nexus]

And I think it's beat that needs to be emphasized, not besides more I listen to it.

[broken lotus]

it's singing 'gonna' like it's 1 syllable...

[neat nexus]

So huh... [reverb] works with [Chorus] as a metatag, doing something but not sure what. Melodic singer may have helped as genre type.

[broken lotus]

is maybe skewing GROOVE to an earlier beat

[neat nexus]

Here's the previous clip:

[broken lotus]

Well... I wish ad-libbing was more controllable....
this #🐣┃suno-showcase message is whole song without lyrics (almost) – but adding prompts like [VERSE] [CHORUS] without text, and the song is great!

[neat nexus]

Yeah I got lucky with the wawa wawa wawa, it seemed to have influenced the music since it wasn't actually vocalized.

[broken lotus]

There's a bigger influence from the lyrics than I expected....

(in general)

[neat nexus]

Wow that was great,

[broken lotus]

Riiiiite?
Like why am I sitting here trying to come up with WORDS when the AI can just make jibberish sounds that are perfect?

[gritty cradle]

please help me in making my rap song 😭

i have the lyrics

[broken lotus]

so what is going wrong when you try it?

[gritty cradle]

nothing is geeting correct

if beeat is good then kyrics prblm

and if lyrics said is correct then beat is not proper

[neat nexus]

lulz. They both have their place.

[neat nexus]

So you keep saying but you don't be showing..,

[gritty cradle]

tryied in geeting in hindi also but not worked looks like sppech

[broken lotus]

It is random 🎲 ... and AI 🤖 is stupid....

[gritty cradle]

showing means ?

but help me please

😭 😭 😭

can you please try it

[broken lotus]

showing means, we cannot drive your car if we cannot see your road.

[gritty cradle]

if am telling the road and ready to drive also but not getting the destiny

[broken lotus]

and also you must accept the AI 🤖 is also guessing at what you want vs what might sound good vs what it knows

It's AI, not an actual person on fiverr who can understand instructions with a human brain.

I suggest you look at OTHER people's prompts, and listen to their results...
Search this discord for 'rap' there are many many many results

when you hear a good one, look at how they wrote their prompts

[neat nexus]

I'm tossing together some unfinished doggeral to see if I can show you an example of "showing" your work...

(As if the above workshop wasn't already an example of showing your work... 🙄)

[broken lotus]

If AI could actually DO whet people tell it, word-for-word, we'd all be dead already.

[molten yacht]

Hello fellas

[broken lotus]

Hi Lady.

[molten yacht]

Wants to know about how to made a 3 minutes songs?

[broken lotus]

read the instructions here: https://discord.com/channels/1069381916492562582/1149508588470022164

[molten yacht]

Thanks a million!

[neat nexus]

You start off with an [Intro] (because nobody wants something that just starts with a verse, usually.)

You tell it what kind of genre or style you want the music. Here I chose trap house and added a metatag to style the tone further with [dark chords] to see what it would do. I liked the one and not the other so I picked the one I liked and hit continue.

Then I told it to use an [Instrumental] flute with the [flute] metatag threw in my [Verse] and put in my lyrics under that.

We gonn'
dial the sensity
feel my intensity
look at the density
Who would
and how could
you excite
and ignite
my immensity

like a diss
you blow a kiss
it hits
I twitch

Those aren't really rap lyrics but I figured it would work in a pinch.

Then it's a matter of continuing as you go.

Oh! for the second segment I chose dirty rap as the genre to supplement the trap house. You don't need or want to do this every time, but it can help tweak things a bit.

@gritty cradle, does that help any?

[gritty cradle]

trying

i am trying with intro ony but not geeting good start

कुछ करके दिखाना है, अब कल तो एक बहाना है
कुछ करके दिखाना है, अब कल तो एक बहाना है
कुछ करके दिखाना है, अब कल तो एक बहाना है
अब कल तो एक बहाना है, कुछ करके दिखाना है
इतिहास बनाना है, पसीने से नहाना है
सबको दिखाना है, उंगलियों पर नचाना है
करके मेहनत, बाताना है, जलने वालों को जलाना है

here is my intro

[broken lotus]

ok, well ... you are saying 'intro ONLY' but then you add a bunch of lyrics, so that is not an 'ONLY'....

the suggestion was that you use your 1st chirp with NO LYRICS, instead put the description of your genre in the top text field...

then where the lyrics go (middle text field) use:

[Intro]

generate that, you are listening for a beat/rhythm that you like... it may be shorter than a regular chirp (that has lots of lyrics)....

Once you hear an [Intro] chirp that has music you can live with...
hit the Continue ⏩ button that matches the clip (1 or 2)

then add your lyrics on the Continued chirp:

कुछ करके दिखाना है, अब कल तो एक बहाना है
कुछ करके दिखाना है, अब कल तो एक बहाना है
कुछ करके दिखाना है, अब कल तो एक बहाना है
अब कल तो एक बहाना है, कुछ करके दिखाना है
इतिहास बनाना है, पसीने से नहाना है
सबको दिखाना है, उंगलियों पर नचाना है
करके मेहनत, बाताना है, जलने वालों को जलाना है

[gritty cradle]

okay i will try with that

[neat nexus]

Oh so having stated the following:

• I am working blind with just Google Translate to give me a clue (and we all know how that is,)

• I am spending my own credits here

• I freely acknowledge my ignorance

• I know I misspelled [Instrumental] go bite me.

Here is what I was able to do:

I don't know what it means but it don't sound too bad.

For reference.

[gritty cradle]

yes

but the lyrics and meaning of song is good

[neat nexus]

Well now that you see some of the structure, have fun with it and keep trying to make your vision come to reality.

[gritty cradle]

i am trying a lot as my song says 😂

dude but not geeting little bit closer also

[broken lotus]

definitely take a break and listen to some other people's results.... to hear what worked and what didn't.

it's ok to steal their [prompt] ideas, when they are good.

[neat nexus]

Yeah, it's a lot like the early days of the web when people built their own websites by looking at the code and copying segements to use in their own pages to help learn how to format things.

[broken lotus]

and ALWAYS speel [Instrumental] correctly.

[neat nexus]

😂 🤦‍♂️

[gritty cradle]

i tried to find promot realted to rap but most of them are regarding melody only '

[neat nexus]

It helps to know the different styles and genres.

Wikipedia can probably help with that.

https://en.wikipedia.org/wiki/List_of_hip_hop_genres

By the way, @broken lotus your avatar reminds me of someone. Do you have a YouTube channel by any chance?

[broken lotus]

mmmmmmaybeeeeeeee

If you hate that channel then no...

[neat nexus]

I think I saw your work when looking for tips and tricks for Chirp. If that's you, excellent work. Definitely subscribed.

[gritty cradle]

okay i got beat for intro

[broken lotus]

Oh well then is definitely me and thank you!

[gritty cradle]

now tell me how to put my lyris into it

[broken lotus]

so you see the Continue ⏩ buttons?

pick the one you like best, hit the Continue ⏩ button for that clip.

an new chirp window will pop-up....
you have option of adding the same genre description in the top text area (or change the description if you like)

and add your lyrics in the big (middle) text box

wait for 2 responses from AI...

pick the one you like best, and Continue ⏩ again...

if neither is good, you can try regenerate 🔁

each time you ⏩ or 🔁 ... it will cost you 10 tokens.

[gritty cradle]

okay let me try it also where i can dw full song after continuing it

[broken lotus]

Once you get to a point where you want song to end, hit the ❤️

this will stich all the clips in the chain into 1 seamless song/video...

similar to using [Intro] in the first chirp, you may want to add one final Continue ⏩ , and use a prompt to end the song...

some ending prompts that (mostly) work:

[End]

or

[Fade Out]

(just use 1, with no lyrics, like you did with [Intro]

There is no token fee for ❤️

So you can ❤️ as you go along... to check how the flow is going.

[neat nexus]

@broken lotus, do you mind if I DM you?

[topaz willow]

didn't know about the token fee for the heart... good to know!

[gritty cradle]

i am not geeting the end

guide me to get better end for the rap

[neat nexus]

When you mean not getting the end. What do you mean exactly?

[gritty cradle]

song ending

the beat looks lkike we are in the middle of the song

[neat nexus]

I see.

OK, there are two things to do here

• [Outro]
• [Fade out]

I personally like them both. You can use them in that order and select which result you like.

This is also a good place to do an [Instrumental] before the [Outro] if you'd like. Up to you.

May take more than one attempt to get it to actually fade too, so you may need to play with it.

[gritty cradle]

so need to use befor lyrics

or in genere

[broken lotus]

no, use where you'd put lyrics...

[broken lotus]

so when we say 'genre' that is the overall song, like "Funk with Male rapper" – this goes in the top box that says 'genre description'.

when we say 'prompt' that means instead of lyrics, use the word in square brackets like [Intro] in the same middle box you usually put lyrics..

you can use both a prompt and some lyrics in the same box, but it becomes less reliable, like the AI sometimes does 1 or the other, sometimes both, sometimes neither....

I really think you could learn better by observing the many examples in this channel: https://discord.com/channels/1069381916492562582/1083129341870878780

it is easier to learn by observing other people's success and fails, and hearing what is a 'realistic expectation' of AI technology....

when you use a bunch of cry emoji, that does not actually tell us anything. So it is a little frustrating, but we will help you, but you will actually learn faster by listening to a lot of other people's chirps and watching the prompts show on the screen as music plays.

[orchid summit]

/chirp

[tall orbit]

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.

[rare junco]

does anyone know the prompt that is sent to chatgpt when filling in the bottom box with the /chirp command?

been trying various things myself along the lines of

"write a song about [topic] in the style of [genre]. one verse, one chorus."

but nothing ive gotten was quite as good or concise as what you get via the bot

[light ether]

we tweak it from time to time but if you want concision, ask gpt for four line verse and four line chorus (we also were trying 2-3 lines)

[summer dune]

/chirp

[tall orbit]

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.

[neat nexus]

Not sure I'm entirely too happy with this one. It just kept wanting to die. On the other hand it's a fun little ditty

Here's the original so you can see the changes.

I mean, sure as an experiment to see what I can get for more music in a song, it was a success but ideally something like this would have gone on quite a bit longer.

The crowd stuff as you can see was entirely Chirp's idea.

[graceful condor]

does anyone know when/if this will become open source, and if there are any alternatives that are already open sourced?

i couldnt find a single one that can generate singing voice from text other than bark and thats.... well.... not the best

[rugged carbon]

/chirp

[tall orbit]

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.

[untold briar]

Because of the way many of these new AI models work, figuring out how to prompt/guide/control tends to be a matter of discovery. It should get better in the future, but right now, it's much like the Stable Diffusion community learning process developing prompting techniques, over time. And there are for sure more ways of guiding it coming, if you search dev comments, many options being explored.

[untold briar]

Unconditional output is one thing tone-deaf Bark is still better at than chirp, in that you get actual words and phrases and sometimes even sentences, and sometimes even things that sort of makes sense. I hope someday a Super-Chirp can do that and be as good music, I love unconditional. And I swear later chirp is better at unconditional than earlier chirp was, actually. Some day Chirp will be writing joke punchlines...

[crystal tapir]

How do I tell chirp it’s the end of the song?