📚┃suno-school

[shell jungle]

matts server is really hyped, (mattvidpro)

[woven osprey]

Brilliant news and thanks for responding so quickly.

[fierce granite]

I don't really get it.. it's pretty cool, but I end up with a half created song.. what do I do with it then?

[open magnet]

you can continue it with any lyrics you want

web UI doesn't support that yet but it will next week

same as how it works in discord

[fierce granite]

I am new, I started with the web.. got some really nice songs. But didn't know what to do after that

Now I tried discord.. created a song. Nothing as good unfortunately, but clicked to Continue on one result, and it basically created a whole new song

Still a partial song, like the others. So am don't know what I'm mising

[neat nexus]

Your choices are to:

• ....wait impatiently for the full capablities to be released next week

• ...play with the current generation model on Discord while waiting impatiently for v2 to be fullly released next week

Either way you're going to have to wait.

[neat nexus]

#⏩┃continuation might be the answer you seek.

Note: when it comes to impatiently waiting, that's something we're all doing, 😭

[fierce granite]

Oh ok.. I tried paying, thinking maybe that was why.. but results were the same. So for now we're just waiting for that next level.

[open magnet]

yeah sorry, some time mid next week is the plan, will be just a few days for feature parity

anything you create now you will be able to continue later once enabled

[shell jungle]

i am hyped for when v2 comes out, a bunch of people love the new beta model but the site does not have the "continuation" thing

the week ends tomorrow👀

[neat nexus]

Time zone shenanigans?

take me to your time machine

[celest goblet]

i have credits but appear like manage account.... anyone with this problem too?

[light ether]

sorry, that really means later the week starting tomorrow 🙂

[copper flame]

Hey guys, will it be possible to be able to get a seed so we can recreate different lyrics with same artist and style ? This is major for me, id like to buy it after trying out but without seed it kinda is shooting blind , lot of great tracks but once you hit a gem, you cant make more of the same cause its kinda random seed

[untold briar]

That isn't a feature that unlocks if you buy Pro now or anything. Pro works the same as Free in terms of functionality right now. But features like that have been suggested so maybe in the future.

That said right now you can kind of sort of sometimes approximate that by using a shared starting song segment, then continuing it in different directions with different sets of lyrics, creating a kind of album of songs vibe.

[blissful raven]

I predict Weds for V2
https://cdn.discordapp.com/emojis/1142215257553637456.gif?size=96

[shell jungle]

🔥 Yay!!!!!! i can't wait!

[light ether]

sorry, that wasn't clear, I mean sometime this Thanksgiving week, not tomorrow. but we are working hard on it!

[teal raft]

Hope you guys aren't working on Thanksgiving, take a break and eat too much

[tame berry]

Another new track in #DeepHouse was been released just now. I'd appreciate your support 🙏
https://soundcloud.com/djanthy/hot-inside?si=5ee7bf05c8e341dca66f1b8676fc6228&utm_source=clipboard&utm_medium=text&utm_campaign=social_sharing

[shell jungle]

made some really cool v2 songs, go check it out #🐣┃suno-showcase

[kind coral]

Sono got seed number or not? If I like the random style and voice can I use it back?

#🐣┃how-to-chirp Sono got seed number or not? If I like the random style and voice can I use it back?

[wicked canyon]

Hello all! I bought the web version and have a question: I'm generating songs in polish and when I want to use a specific style, can I write the style name in english and the theme of the song in polish? Because some of the music styles don't translate well into polish. Tried doing it once and it did the style, but it kept all the song lyrics in english. Is there some sort of a trick for prompting this kind of request?

[karmic rampart]

I love this tool! Impressive! Regarding the following poll on 10/18:
"What is your ideal process for writing lyrics in Chirp?

1. Write/bring my own
2. ..."
   I'm curious if a way exists yet to paste in my lyrics, specify a style, and get a song back?

[broken lotus]

this works on Discord... there are separate text boxes for genre description and lyrics...

the website is getting the same functions. Devs are building the UI this weekend, hopefully public in a few days.

[inner steeple]

v2 isn't also always better right now

when it works, it works nice

but it seems to have higher probability to just not do vocals at all, or do them at really low volume so that you can't hear them

this is really common right now, things like these happen on current discord's model too but not in this scale

[broken lotus]

Are certain genres more prone...?
I feel like 'pop' is usually a loud female singer...
other genres (rock) (metal) get lyrics that are 'undermixed'

[inner steeple]

could be genre specific for sure because you can't really enforce the genre on the web version right now

so when it puts some random genre there in style of what you ask for, it may cause these issues

there's always been mixes of genres that just confuse this and it can't figure out what to do with it

[broken lotus]

have you tried prompting for a louder voice...?
(not that it would be reliable on web rn)

something like 'hot vocal' or 'upfront vocal'...

[inner steeple]

last ones that created just music were genres "jazz hard bop" and "indie goth uplifting edm" where in latter I'm pretty sure it was the goth that broke it

one may like this if one likes jazz

I don't really 😄

just trying random things here

[broken lotus]

I am always prompting 'sultry vocal' because it makes the voice quieter... but it's practically a compulsive habit by now..., not really sure how it's working because I don't do anything else...

[inner steeple]

oh.. also typo there

I really like "suomipop" tone with this, it just means finnish pop, I dunno if it's a placebo effect but I keep using it as it usually gives that softer tone to things.. but I can't use it with website v2 as chatgpt then interprets I want the lyrics in finnish too (which isn't really the point if I describe the song in english)

[broken lotus]

suomipop
(from Discord)

these are great!
Should I be looking for Finnish bands in rl?

[frigid talon]

/chirp красивая девочка Агафья, русская балерина, умничка Агуша

[light ether]

Unrelated but that particular bird image is awesome

[broken lotus]

Hi..., try just
/chirp
in one of the beta channels https://discord.com/channels/1069381916492562582/1149728729740882080
a pop-up window allows you to add the prompt

[inner steeple]

yeah that suomipop also mixes well with other pop genres and edm

[shrewd spade]

On the new version’s website, I didn’t realize the “My Clips” page only kept up with 10 pages of generated clips, and I’ve lost access to a beautiful song that got me hooked on Suno. Is there a way to access those generated clips that don’t appear in that section anymore? Or download the whole of my generated clips altogether? Thanks.

[light ether]

we're gonna improve this paginator, but cna you see this at the bottom of the page?

misread

[shrewd spade]

Yes it stops progressing at page 10

[light ether]

that's a bug

standby 🙂

[shrewd spade]

Ah 👍

[light ether]

thanks forusing it a lot and discovering 😛

[shrewd spade]

Yeah no problem, 10 is plenty i just would have saved that one song

[light ether]

it'll be around somewhere, just need to fix that frontend issue

[shrewd spade]

ok thanks let me know

[placid gorge]

Just a thought Double < Most recent Double > Oldest Normal > and Normal < One page in that direction

[grand leaf]

Hello everyone, the online app is is amazing, there's just one problem. I made probably the best song I've made yet, but there are no options to continue it, I've tried everything, can anyone help me out

[light ether]

Continue coming to the web this week - just make to favorite it!

[light ether]

Page issue is fixed now

[grand leaf]

Yeah it's definitely a favorite, you are saying this week they'll show us how

?

[light ether]

Yep, we are going to add continuing workflow on the web

[grand leaf]

Ok what do I do, I know it's a stupid question, but the options you get on discord just isn't available. Like this song really resignates with me brother

[light ether]

Sorry that was a response to a different question. Just hold on until later this week and the web app will update to have similar options to discord

[grand leaf]

All good thank you my friend

[shrewd spade]

Awesome thank you. This is a very powerful tool, and from a long-time musician who works with samples, this is pretty revolutionary

[pine agate]

does that include adding our own lyrics? It's a bit sad when the web app includes parts of the prompt in the lyrics that I didn't intend for it to do. Tangential: can chirps be purely instrumental or does the model only support voiced generations?

[untold briar]

Yeah to both

[teal raft]

The pure music is what I'm looking forward to in the web UI. The lyrics are fantastic don't give me wrong, and wildly impressive.

[untold briar]

You can get it now but you have fight a bit with ChatGPT to get the right words in the style of music field. It thinks it can't do it so you have be a bit indirect, and it's hit or miss, so I wouldn't recommend trying too hard right now since the misses will still burn credits

[untold briar]

example looool

[teal raft]

Oh yeah I know and I'm not too worried about burning credits, the amount of entertainment I'm getting out of just exploring this thing is easily worth it.
I can somewhat reliably get it to output just music by telling "instrumental no lyrics" since the model interprets it literally just like how sometimes it will literally describe the genre

Like this:

Those sure are words

The interaction between whatever models are working behind the scenes is ultra interesting

[untold briar]

Yeah intrumental in genre field can be enough. Another thing you can do it get some kind of text that fills space up in the lyrics field that it tends not to sing. Examples would be unicode upside down text, ascii space characters, line breaks, empty parenthesis, or the words (instrumental) repeated.

[teal raft]

Oh rad, that rules

[untold briar]

But fairly often ChatGPT decides to do something like this instead with similar command. Which can work out amazingly, 20s onward is what we call in the field a BLANK SPACE BANGER

[teal raft]

lmao this rules

heh, sometimes you can get it to not say something specific just by going "NO X" but sometimes it sings about 'No x'

[untold briar]

Yeah. It's sort of fun sometimes because those tend to make for more creative and weirder lyrics than it normally would

[vivid fjord]

/chrips

/chrip

[neat nexus]

The command is /chirp and only works in the #chirp-beta-X channels.

[broken lotus]

BLANK SPACE ... BLANK SPACE ... BLANK SPACE

[plush wyvern]

Is there a way to delete chirps we dislike?

To keep things organized?

[untold briar]

Not yet, hopefully future feature though

[plush wyvern]

Thanks

[random minnow]

The vocals are very faint, but I'm assuming there isn't much we can do about that right now, is there?

[broken lotus]

you could run it through something like UltimateVocalRemover (free)..., demux the voice from the music, add both new tracks to a DAW (garageband, et al) and raise the volume on the vocal track.

since you wouldn't be doing any real edits, it should be a near-painless process...

it's a cute song... worth the effort.

[random minnow]

New to all of this, so thanks for the tips and recommendations! Appreciate that.

[teal bane]

hello. i subscribed just several hours agao but it still says i need to buy more credits. anybody knows what's needed to do here?

[inner steeple]

I feel that that faint vocals won't really detach well with UVR

[old wing]

I was searching everywhere in the docs/chats/youtube videos/etc for a way to delete and clean everything up. Thank you for sharing this info!

[inner steeple]

I guess you could just make your own discord group, invite the bot there and then moderate, delete the bot's chirps around the channels as you wish

I've been meaning to do this myself, to keep things more organized

hmmm

actually you can't move messages around

but you can delete them (as long as you have rights to do that on your own group)

could still make bunch of category channels and use those for generations to keep things sorted in some way

[old wing]

Thank you for trying, and sharing the results! That is extremely helpful!

[untold briar]

Yes, that's a nice alternative since unlike DMs you can in theory even invite people to your Discord and build on each others generations still

[inner steeple]

yeah I just setup personal group for this

I may have lifted that one chirp image to it's logo

I do plan to explore this more extensively soon, planning to get that premium sub once the v2 is on discord, and having some way to split the content would be nice

right now my chirp stats are rookie numbers

I guess at least half of those are from original v0

[teal raft]

I paid for a sub right away because I knew I'd do it anyway, you get on a tear and you do not want to think about limited generations

[inner steeple]

hmmm, I noticed #😍┃v1-songs mentions #bot-sing-alpha

is that bot sing alpha some old chirp channel?

[untold briar]

There's been 3 versions. V1: English Only (sort of), no style/genre prompt, only lyrics. V2: Multilingual and style/genre prompt V3: Current Web Version, longer context, stereo, higher quality

[inner steeple]

yeah, except Suno labeled them v0, v1 and v2

1-3 would be easier to explain to people though, v0 made more sense when you thought about it as alpha/beta, but it's still labelled as beta today

ah

[white vapor]

Is there a way to get chirp to carry a note in the lyrics or do we just have to hope it'll do it? I tried using hyphens (--) or tildes (~~) but it doesn't seem to work all the time (probably coincidence when it does)

Or how to make it pause between verses for a guitar solo or instrumental piece

[placid gorge]

Dunno how but im at 131 already?

Also yes my credits do reset 7 pm for some reason

[broken lotus]

just keep adding vowels:

I waaaaaant yoooooo

if you want melismatic notes ad the hyphens

I waa-aa-aa-AA-aa-aant yoo-OO-oo-oo

the capitalized ones will usually inflect up, like an accented note

[placid gorge]

At least if im reading correctly it says 50/50 on usage if that means no than i guess its 10 PM

[inner steeple]

I think the earlier mentioned 3AM UTC is probably accurate

it's either that or 4AM UTC

so, should reset in 2 and half hours?

[white vapor]

Thanks, I'll give that a shot!

[neat nexus]

Have you tried logging in and out of Discord and your Google account then logging into one or the other to see if this makes a difference? Otherwise you will need to email [e-mail]

[jovial plaza]

It's not possible for me to cancel my subscription

It seems to be loading forever when clicking on cancel

[runic hinge]

Is it possible to combine the melody of an existing song but put lyrics in another language?

[broken lotus]

hi, please email [e-mail]
if the subscription status hasn't updated for you.

[broken lotus]

by 'existing' do you mean a song made with Chirp, or the song of a recording artist?

(If Chirp, there has been discussion how new features might work, and you can add your thoughts to https://discord.com/channels/1069381916492562582/1154587937611190354)

[normal grove]

What is the sample rate of the tracks created? Sounds like 22kHz, is it possible in the future to increase this, or is this simply not possible with the speed of current GPUs, assuming it's GPUs doing the work.

[placid gorge]

I mean theres still a cap at 100 :]

[jade mist]

Some questions:

1. Will you continue on expanding the duration beyond 80 seconds? It's annoying when songs with long pauses between verses cut off because the limit.

[placid gorge]

Continuation for web should be up by the end of this week

[jade mist]

Sorry about those separate messages. I forgot pressing enter will send them so...

Sorry... Will I be able to enter my own lyrics and will you expand the duration?

[placid gorge]

Yes both should come over to web really soon™️

[broken lotus]

all the Discord features are coming to web this week (Continue, separate lyric and prompt, etc)

and v2 is also coming to Discord so they will have same features and quality.

I have not yet heard how long v2 Continues will be.

[unique dome]

Yes, this was a mistake / remnant of the past. Just updated — thanks!

[normal grove]

Just out of curiosity, why aren't you allowed to generate a song with swear words, but most rap music in real life has it?

[vast dew]

issue is probably less likely the chirp model itself but the censoring of OpenAI's GPT Models.

[normal grove]

So it isn't against the terms of service?

I thought I saw something in there about the types of songs, but I don't recall seeing about swearing

[vast dew]

at least Discord has these rules to not use profanities. so you should not use them in chirps since they are posted on discord in public. Not sure about chirps in private.

[placid gorge]

Discord>GPT>Manual/Suno for filtering iirc

[normal grove]

Yeah I wouldn't expect people to release to the public that sort of thing.

[vast dew]

i thought the question was more about the technical side why it its not allowed. 🙈
And there, the chirp model most likely can swear as much as it wants. its just the song-text generation is censored by OpenAI.

[blissful raven]

We need LLM uncensored 👻

[vast dew]

or just try without swear words. Just be friendly to others. 🫶

[normal grove]

It is a shame that most LLMs are sensored, I think they should have an age verification and then it can be opened after that. Sensored LLMs do feel like they treat you like a child.

I generally don't have a want or need to use LLMs in a bad way, but I feel that it should be my choice. In the same was if I created a song with swear words in them, it wouldn't be to hurt others, just for personal amusement, or style, sometimes preference depending on the song type.

[blissful raven]

In the future AI will Be Uncensored

[vast dew]

the worst censoring i have seen so far is by microsoft. For example in the Grounded game, where you can't even write most of the enemies names, because some part of that name is probably some swear word in some language of the world that is spoken by maybe 5 people total.

[normal grove]

Yeah I don't get it, especially if you're playing the game yourself or with a bunch of friends. I also can't believe I've just spelt censored wrong about 8 times, I even know how to spell it, haha.

[broken lotus]

THIS 👇

I've learned there are like 4 steps and each has a different censor attached...

discord is just bad words (and really long 'spam')...

but GPT will censor your whole concept if it doesn't like it...

I used 'aroused' instead of 'attracted' and GPT said no no no no...

it's not a bad word, just a shade more sexual...

also one time used 'girl' instead of 'woman'.., daughter instead of wife..., it imagines the worst whatever you type...

[normal grove]

Yeah see I think it takes it too far and needs to be toned down on occasions

[long tapir]

I have submitted music to some professionals, 4 out of 5 of them said the same thing, the vocals are JANK. Is there anything which can be done to improve the vocals coming out of SUNO?

[neat nexus]

Not saying you're wrong here...

...but are we sure that professional bias is not involved?
It's hard to get a man to agree with something that directly has the potential to affect his pocketbook most of the time.

[placid gorge]

tru

[long tapir]

the vocals are JANK imo.... not good on the ears so I agree with them, not professional level yet

[placid gorge]

They arent 100% Jank they can be at times but its not an Always the case type thing. But thats just ai for you, things will get better eventually however.

[wintry grail]

I have gotten some awesome vocals and screams, so I strongly disagree

[broken lotus]

NAS had a singing career, so...

😂

[neat nexus]

Also something something enemy of the perfect is the good enough...something something...

[wintry grail]

@long tapir it sounds like you submitted something to them and mentioned it was AI, whereas you should've given something with great AI vocals and just ask their opinion that way

people are biased either way when you mention it's AI, it's "fake" to them

[broken lotus]

also fun fact: Van Gogh was a really great painter, much better than StyleGAN.

But AI draws in seconds so... ⚖️
not really equivalent at all...

Does Midjourney art hang at the Met? no.
Can I prompt dead painters at the met to make me an art in 10sec?

also no.

[wintry grail]

also, standard Suno bot files are mono, personally i go out of my way to extract layers and widen the stereofield, it ends up much more pro sounding

Izotope RX10 ftw btw

[broken lotus]

(just wait for v2... it won't replace Celene Dion anytime soon, but better than your average karaoke)

[wintry grail]

yes, i am waiting very impatiently 🙂

@broken lotus your MJ comparison is very valid, I have been into AI image creation from the start, and I've witnessed the "resistance" first hand... to me it's all bs, just people who are afraid of change... look at it now, photoreality is already here, I think people will get over it, it just takes them time to adjust (never mind the stubborn nay sayers)

[normal grove]

I use all of those things, Midjourney, Stable Diffusion now this, I personally love it all.

[wintry grail]

GPT+Dalle is my go-to atm, don't even need to make prompts anymore and the results are lush (or can be, at least)

[broken lotus]

Also, did live singers go extinct somehow?

can you get a live singer for $10/mo...?

Can we just live in reality and enjoy the moment?

[normal grove]

Ah yes, ChatGPT and Dall-E 3 too, I also do those too lol

[wintry grail]

i have been producing for over 20 years, to have a singer at no cost is a no-brainer to me

and yeah all this AI art does not invalidate the actual art of doing everything manually

but I'll be damned if i'm not jumping on it due to the sheer convience of it all

[normal grove]

No exactly, I still admire artists and think they're amazingly skilled, I wish I could draw, paint like them.

But because I can't using AI is a great alternative.

[wintry grail]

for sure, it's a wonderful method of expression

[broken lotus]

I feel like replacing the AI voice is easier than replacing any other part of the music...
That's a normal thing of music production. You don't walk in with your bedroom demos and the producer supposed to make you sound like Pavarotti by adjusting the faders....

Hire a singer. If that's holding you back somehow. Do the extra work. You're still ahead... way way ahead.

[wintry grail]

well, actually replacing it will be a matter of getting AI results in stems, because seperation after the fact isn't an exact science, some of the vocals tend to bleed over in the other stems

[broken lotus]

yes, and Chirp sometimes likes those soft-tronic-sounding voices... Too hard to lift those out with demux...

[wintry grail]

I am making mainly electronic stuff, so repetitive vocals here and there are a great asset in that context, I've not actually been trying to make complete lyrics

but the screams, damn, those are gold

very emotive

i'm sitting on 4 AI albums/ep's right now, I will be editing/finishing/releasing them on Bandcamp every week starting this week

definitely not Jim Morrisson inspired, i wouldn't dare 😎

[broken lotus]

all my stuff is whispery voices, lol

(realizing I can probably whisper into a microphone lolol)

[wintry grail]

one more, for fun

from the non-electronic album

[broken lotus]

I kinda love when the AI is singing off-key bedroom pop...

Oh! did you put that echo on the voice...?

[wintry grail]

yeah 🙂

delay/reverb does wonders to the stereofield

[broken lotus]

😮 how'd you get it to speed up?

[wintry grail]

that's Suno 🙂

[placid gorge]

Meanwhile me waiting for route note to review and post some stuff [All singles posted over the course of last week 3 longer ones and than some one-offs] Also plan to make actual music at some point in the future. (If im reading the room right this should fit here) also have one of zed One-Offs

[neat nexus]

I just keep running into the chorus as a life altering event in my stuff. I usually dread the chorus as I know it's going to be an almost impassible minefield and that I often will need to butcher (edit) the audio to try to make a complete song unless I want to play a cosmic roll the dice game at trying to resumer just the right vocals and music,

[placid gorge]

Maybe try continuing from pre chorus and have the next clip be chorus and than verse?

[neat nexus]

Not sure what you mean there.

This song doesn't have any pre chorus so that doesn't really work.

Doing a traditional song right now for a project I've been working on for most of the month.

[placid gorge]

Clip A: [Verse], [Optional prechorus]
Clip B: [Chorus], [Verse]
Clip C: whatever comes next

[neat nexus]

Right but for a song like say...Jingle Bells. There is no prechorus. Just a refrain.

Makes it a bit harder.

[placid gorge]

Than just have the first clip be just a Verse continue you it for the refrain than go back to the first clip continue it into the Chorus and whatever else. Than download the clips and assemble them correctly combine them and bam. [Probably odd wording]

[neat nexus]

Yeah, that'd be the "butchery" of the audio I mentioned. I do it. Sometimes I think I do a good job of it even. But I'm generally much happier with my lot in life when I can manage a Chirp that is beginning to end continuous. I don't know why. It doesn't even make sense. Editing is a fact of life, but that's how I feel, rational or not.

[placid gorge]

👍

[neat nexus]

But yeah now that I'm thinking about it, it is entirely possible I am not loading enough lyrics into the clips.

One of the things I did early on was limit how much I did in each section and format for smooth vocals. It's possible that's getting in my way now and I need to put more segments together? Bit there's only so much you can shove into the old v1 Chirps before it starts singing too quickly to beat the clock.

[neat nexus]

I'm going to tinker a bit and see if I can make anything good happen. Thanks.

[placid gorge]

Or, too soon

[neat nexus]

Definitely! I know it irritates some people but I've gotten a lot shorter intros lately and that doesn't exactly fill me with joy as I like longer intros, depending on the style.

[placid gorge]

Same, and Longer solos and may be weird but also longer outros which I never seem to get

[neat nexus]

Forgive me but as one of those fortunate few who seem to constantly end up with the outro that never ends, (or worse--Chirp deciding it's done with my sh-- and doing a fade out on me) I have no sympathy. 😂

You should ask wetcircuit to show you her endless piano solo..

[normal grove]

Why do I keep getting try a different prompt? There is nothing bad in the prompt.

[neat nexus]

https://tenor.com/view/gasp-surprise-fake-shock-shook-gif-11845590

What are you prompting?

[broken lotus]

it's a generic 'error' message so we have to guess what happened.

if the prompt mentions a recording artist by name, GPT MIGHT reject on the grounds of copyright 😩 (not because it actually generated anything but it is censoring itself...

[normal grove]

After seeing the thing just now about the piano solo, I asked for A piano solo [instrumental]

[broken lotus]

'Jazz Piano' was the prompt that wouldn't shut up... ever. lolol

[normal grove]

Ah, maybe because I didn't add a style

I'll try that haha.

Actually got a song, not bad lol

[placid gorge]

lol

[neat nexus]

Found it!

#🐣┃suno-showcase message

[placid gorge]

My Luck

[neat nexus]

https://tenor.com/view/rock-head-bang-metal-rock-yeah-gif-15728468

lulz

All you can do is try again.

...

Starting to think this version of Jingle Bells is cursed. If I don't get something that works soon I'm going to post the remains here for dissection and move on.

[normal grove]

Just listening, brilliant!

[placid gorge]

I have wasted 40 credits trying again lol

[neat nexus]

I've been at this since 9:16 AM CST. It's now 11:40 AM CST...

[placid gorge]

zamn

[neat nexus]

And really, has anyone really read the original lyrics? Kind of a messed up song...

Verse Two:

A day or two ago, I thought I'd take a ride
And soon miss fanny bright, was seated by my side
The horse was lean and lank, misfortune seemed his lot
He got into a drifted bank and we got upset

Verse Three:

A day or two ago, the story I must tell
I went out on the snow, and on my back I fell
A gent was riding by, in a one-horse open sleigh
He laughed as there I sprawling lie but quickly drove away

Verse Two is basically about a sick horse, Three is about laughing at someone falling

[placid gorge]

I mean it was created after the great depression and during the panic of 1857

[neat nexus]

This version is cursed.

I may be able to butcher the pieces together but I just don't feel like I'm keeping my vocalist. And that makes me sad because I really liked the vocalls at work here. But the timing is just off to me as I go along.

For clarity I would have edited the first clip into the second somewhere about 1:04 in the second clip.

[broken lotus]

let's face it. Victorians were terrible people. 🤣
A Christmas Carol was basically a documentary.

[neat nexus]

LOL, you're not wrong, sadly.

Any way this is how I would have edited it if I were going to try to continue it. It just doesn't feel right to me though...

That bad?

[broken lotus]

those high notes... 😞 ... v1

[neat nexus]

Yeah well there's nothing else available...

🤷‍♂️

[broken lotus]

i know... I know...
It's not a solution to complain.

but it's a cute song!

[neat nexus]

I'm just trying to complete my three songs I need.

[broken lotus]

like the chorus!

[neat nexus]

Should I head back to the mines and try to complete it then?

I just don't know how to get that really smooth vocal from the beginning again. I have tried to go back up the chain. It fights me. I get full 29 second clips of instrumentals.

[icy nacelle]

/chirp

[flint haven]

The /chirp command is not available in this channel

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

[broken lotus]

Bet the house! Keep rollin! No WHAMMIES!

[neat nexus]

See, you get me...

😄

https://tenor.com/view/throwing-dice-quavo-trouble-popped-dice-roll-gif-17702925

[wary sierra]

Hey everyone. What's the best way to continue the song? When I'm pressing the "Continue 1" button it asks me again about genre and lyrics, but how to keep track the same and just continue it?

[placid gorge]

Leave the Genre part blank and insert the lyrics

[wary sierra]

yeah, I've tried, but then the lyrics I made doesnt fit 30 seconds limit (still 8 lines) and the track is completely different

[broken lotus]

Continue listens to last 10sec of the chirp...

your voice and beat and melody need to be in that last 10sec, or Chirp will make up something new

(rather Continue is only continuing that last 10sec)

[wary sierra]

thank you, it's way more clear now

so, for consistency I should just generate 10-seconds fragments one-by-one, right?

[broken lotus]

well... that's one strategy... I have tried 😓

but there are prompts that can nudge the bot... and sometimes I cycle the genre description around so it is 'same but different'

it's a short window but chirp is pretty good at understanding that the song needs 16 bars, and will keep steady until it feels it's time to ad lib... that's totally random so...

usually the 10sec is enough for it to pick up on the song's pace... some genres are more repetitive than others...

[neat nexus]

Rather than keep beating my head against the wall I'm going to skip the slip and fall section.

Wikipedia says it's usually skipped. And I can see why! It's such an akward verse! Chirp tries to rap it half the time...

[broken lotus]

you can kinda force the bot to give very short clips, feeding it only 1 or 2 lines at a time...

it will take more tokens to complete the song...

[cold gull]

Could this also be an issue with the new v2 once it gets the continuation function?

[broken lotus]

Might end up with a different strategy... start length is twice as long... continues also might be twice as long...

less worry about consistency if each single chirp holds more of the song...

🤔 could maybe hit 3min in just 3 chirps... 🥹

that would actually make full-length songs like HALF the tokens it takes with v1....

(I don't know how long they will be, just speculating....)

[neat nexus]

See what I mean..it just starts doing 90s era Eurodance "rap" for no reason...

[broken lotus]

you say that like it's a bad thing...

[neat nexus]

If it was something I was aiming for I'd not have problems with it but that's not what I'm trying to pursue right now.

[neat nexus]

Named appropriately as far as I'm concerned. I do not like it. I was unable to recapture the voice in my opinion and I am disappointed with this result. But it's done at least.

[dawn depot]

I chirped a crazy sounding song on the suno website but I can't find the continue button to add more to the song. any ideas or help?

[neat nexus]

At this time the webapp is a functional but limited demo of the new v2 Chirp model.

It does not (yet) have the ability to continue.

There is supposed to be a release of a more fully available and featured model during the week.

Soon™️

[graceful condor]

How are you guys able to pay for running suno at this point? Do you have investors or is this a research project? Do you only use chat-gpt 4 for the lyrics? What is generating the music?

[low hare]

suno will be perfect when they fix the occasional inaudible vocals

[pliant cave]

I can't emphasize this point enough. I have found this bias to be far-reaching and present in pretty much every aspect of what is considered "AI". As soon as they hear it, their opinion is biased immediately to look for the negative. I've done stand-up comedy written and voiced by AI and immediately have a negative reaction if I say it's AI, yet the same material if I say "listen to my crazy neighbor Larry going on one of his stand-up comedy rants", it's met with an extremely positive reaction.

[spare lintel]

is it possible to make songs with no singing?

[pliant cave]

As an example, consider this track, which was generated by the long-context stereo model (which I believe we are calling v2) and demuxed and remastered. If you play this in your car for pretty much anyone, even music industry people, I highly doubt they would describe the vocals as "jank". In my opinion, it's nearly indistinguishable from what you would hear on the radio today from actual artists.

(Side note: The chorus went off the rails on this, sadly. If anyone at Suno feels like attempting a continuation of any kind, please feel free. It was a song about "I miss the 90s".) 😅

[shell jungle]

really hyped for when suno v2 comes out! i know a bunch of people wanting to finish their songs they made with beta

[teal raft]

Yeah a lot of what I'm getting to come out of this IS music, flatly, and isn't weird

It has tells if you look, or what people not paying attention to this stuff would describe as weird but like, it's just music

[lusty grotto]

Literally half of my last 6 songs were missing the lyrics. No controversial words, all G rated.

[untold briar]

If the song generates audio of any kind, it wasn't blocked for rating or anything like that. The stuff that is moderated by AI for PG-13-ness is stopped before it makes a song at all and doesn't use up a credit.

[teal raft]

At least with V2 you can still like, hear? The lyrics but it's either ultra faint or the system things the music its playing you IS the lyrics

But you don't hear it since we're people and know what music is and what lyrics are, these LLM systems don't, not really

But I'm no expert so I could be very wrong

[lusty grotto]

So it's just a bug? It generated lyrics but sung none of them (web app)

[untold briar]

Yeah it's a bug in the same sense that ChatGPT completely forgetting to answer your questions and going off on some random topic instead is a bug.

[light ether]

Yeah, we're working on making this happen less

[lusty grotto]

Occasionally delusional song lyrics could be hilarious 😆

[light ether]

That certainly happens from time to time 😛

[teal raft]

I did get a thing today where (on the website) the second song was singing different lyrics

The first one was normal, but the second one was ???

Same subject just absolutely not the same

[warm flicker]

How did you came up with no digital glitches in the sound? Is this what you call “demux”?

[quaint rain]

||https://www.youtube.com/watch?v=tfmmyD1EyYg||

[untold briar]

Just one link in chirp-showcase please

[serene current]

hi all! 2 questions please:

1. i can't merge my account on discord and on the site. I payed for the site but here i have free (50 credit)
2. there is a function on site to continue the song? ( see there is one here)
   Thanks!

[quaint rain]

yep

[untold briar]

For merging email [e-mail]. For 2. the feature is coming soon maybe even this week

[tight frost]

It seems that v2 emphasis the voice while v1 emphasis the music. Is that right or it's just my confirmation bias in random prompts ?

[cold gull]

I feel like i read about this on here, but want to confirm if there is gonna be a sound upgrade in the close future?
Also, if i buy the subscription and start generating today, will i be able to continue those once the continue function comes to browser or only continue the most recent one you generated?

I am wondering because i was thinking of buying and generating on the site but dont want to suddenly realize (IF this is the case)that i cant continue what i generated or that quality to audio is suddenly upgraded, making my generations "outdated" in a sense.

[untold briar]

You will be able to continue web clips you generate now, soon in the newer higher model. I don't believe you'll be able to continue lower-quality Chirp V1 Discord clips with the Chirp V2 (which would have to upgrade them into stereo, etc) but if possible I think the devs do want to enable that eventually. But won't be in at the start for sure.

[cold gull]

Thanks for replying! when it comes to quality, i meant that people were talking about quality of v2 bitrate. I feel like I remember there was a discussion about possible future bitrate quality bump but i might be wrong. but if im not, the thing i meant is that if there was talk about such a bump happening soon, then it might be worth waiting instead

[untold briar]

As far I know the audio bitrate won't be changing from what is on the current website. I think if they fix or upgrade that though, they could actually upgrade old songs. Just my personal guess assuming it works like Bark, where you don't really save the audio as a pure wav, you can just save the Bark outputs raw and turn them into audio again later. (But most of the audio quality problems are not from the codec, I think, more intrinsic to the AI.)

[cold gull]

Ok 👍 Thanks for the info!

[odd fable]

Hi, how do i induce a command with which chirp will make a male voice sing

I want these lyrics to be sung in a jazz blues style with a male voice (preferably deep voice)

(Verse 1)
In the cyber city where the hashtags play,
GIGA CHAD MEDIA, lighting up the digital day.
Social blues, we weave a savvy tale,
Boosting your brand, setting sail.

(Chorus)
GIGA CHAD MEDIA, in the marketing haze,
Bluesy rhythms, turning clicks into craze.
From tweets to trends, our expertise displayed,
Your online presence, in the blues, will be praised.

[broken lotus]

'sultry male vocal' is probably my most overused prompt.

[odd fable]

Alright, thanks

[odd fable]

Do i have to put that in the style section or in the lyric section in brackets() ?

[broken lotus]

I put it in the genre description

[odd fable]

Alright thanks

[normal grove]

Is the system seed based? Do we know of any cases where the songs have been the same? I know it says in the terms of service it is a possibility, but have you seen any cases?

[tight frost]

Hey guys, I just spent several hours trying to get ChatGPT to write a practical guide out of this discord room, and it's a complete failure. I wanted to get advanced tips to generate the best songs possible as I want to create a full album with ≈20 songs. Is there any power user who would like to write a guide ?

[broken lotus]

Chirp is upgrading to a new model with double the generation time.... I think a lot of our 'rules' might need adjusting in the next weeks....

[grave violet]

I have a problem: i've created 20 sec song and i want chirp to continue it, so i presses the continue button add new lines of words but I have another voice so i cannot continue the song

any help would be greatly appreciated

[tight frost]

Have you tried v1 rules on v2 and they don't produce the same result ?

[broken lotus]

Continue only listens the last 10 sec of the chirp (v2 may be longer)

the voice, beats, instruments you want to carry over to the Continue needs to be heard in that last 10sec.

[grave violet]

what you say don't work unfortunately and doesn't help
look:
this is the song i try to continue
#🐣┃chirp-beta-7 message

(i press continue button and make a new 4 lines of lyrics)

and I get the different song (although the part where you say last 10 seconds is okay in this
but continuation of my song goes in another voice)
here the result:#🐣┃chirp-beta-7 message

Would be really great you or someone could help because it seems a basic thing to do: users don't want a short 20 sec song, they want a standart song .

[broken lotus]

there's a new 1-prompt feature that was hard to get use to... website getting the lyric/genre split text as soon as it's ready...

from what I can tell, the genre should be more or less same (with the genre text), but the 1-prompt 'interpreted' the genre so it felt completely different...

I think the lyrics length will need to change from what we got use to...

some genres I would have stayed away from (because v1 would start 'ringing' after a few continues) are probably do-able now... I plan to do a lot more ambient/slow music because the quiet parts sound cleaner... and longer chirps means the continuity will be better overall, so I can make longer songs that are still the same song by the end...

I think everyone is going to probably adjust on the new model (not night and day adjust, but all our secret rules are gonna need to be reviewed, lol)

[broken lotus]

Is this the continuation?
#🐣┃chirp-beta-7 message

I hear the female voice from the original chirp (clip-1)

I hear her in both of the new clips...

[jade mist]

Is it possible for the model to for example remember how the chorus goes when continuing with V2?

[tight frost]

Does the model learn from the likes ?

[broken lotus]

each chirp is twice as long... it should (generally) be easier to get to the next chorus in 1 chirp (less opportunity to change course when continuing).

Tests I've heard do sound better (like 1 song) but we'll know for sure when the feature is public and we can all try it...

[broken lotus]

the 🤍 on the website looks like a favorites feature

the ❤️ on Discord is how we stitch all the continued chirps into 1 seamless song

I heard devs say an actual 'likes' system is coming that would actually train the bot... idk any details.... there are a few new features coming (later) that have been hinted without a lot of specifics or eta 😅

[shell jungle]

@flint yarrow 👀

someone leaked what it does

really hyped

[neat nexus]

Leaked?

https://tenor.com/view/sadhotgirl-huh-dog-turn-head-dog-huh-what-did-you-say-gif-22054052

[sturdy flax]

Huge for education

[broken lotus]

😋 original post is here: #💬┃general-chat message

[teal raft]

woah what the heck you didn't even blank out anyhting either u must have the inside scoop

Kidding, of course

[broken lotus]

🕵️ 💦

telephone, telegraph, tella woman.

[shell jungle]

i was blocking it out bc i didnt want to get anyone banned\

also the new suno bot thing is off of command list, i guess they found out that we know, this is the earliest photo of the bot on command list

and its gone

[broken lotus]

👀

[teal raft]

Considering how long ago the post was in #general

I'm assuming it's the v2 we're all playing with on the web

[shell jungle]

yea

[broken lotus]

😞

[shell jungle]

they are giving us v2 on discord this week

v2 is the beta model on the site

on discord we are using v1

[teal raft]

gimmie the UI

yeah I haven't used discord for it at all, I'd, rather not

[shell jungle]

https://app.suno.ai/

the site

[teal raft]

It isn't there

I know I'm using it

[light ether]

lol no, @flint yarrow is just a test version of @versed valley 🙂

[inner steeple]

I'm not feeling the 1 prompt thing.. I mean. it can give new things to explore but I'd want to at least enforce some part of it on the genre

like, if I ask for female vocalist, this will get just removed by the website gen

in general, v2 is huge improvement but I keep getting a lot of artifacts where it tries to do polyphony but for only few words here and there

it wasn't that common on earlier versions

[shell jungle]

okk

[teal bane]

hello. anybody know how do I complete a song?

[mortal kelp]

how to stitch the parts of song together?

[neat nexus]

#⏩┃continuation

[neat nexus]

#⏩┃continuation

[thin flint]

Website or Discord is fine with me, but I'd rather use the web app I think. But I'd like a way to do custom lyrics and complete songs with V2. The web app feels like Chirp V0 in functionality.

[teal raft]

yeah that's the parity they are talking about

End result is supposed to be V2 on both and the same functionality on both

[graceful cliff]

It will come out soon

but parity violation happens naturally in nature, when the force is weak 😄

[untold briar]

As a pure UI Discord has a lot of limitations, but on the other hand you might imagine ways discord has the potential for something analogous to 'jamming' - a cooperative/social music generation experience.

[jade mist]

I really hope the custom lyrics and continuing comes this week as they kind of promised. I mean it's already Wednesday, at least in my country...

[proven tundra]

Can you train or finetune "Bark" use Custom data ?

[random minnow]

I don't see any continuation in the web version. Am I looking in the wrong place? Does it not exist on the web version? Thanks.

[untold briar]

Not there yet, coming soon. Devs said they hope for this week

[random minnow]

Ah, okay, thanks. Will it retro back to all existing chirps?

[untold briar]

You should be able to continue web chirps (Chirp V2 chirps) when the feature is added. I don't think you'll be able to continue V1 (current Discord) with V2, though maybe eventually.

[random minnow]

Appreciate the info. I have one chirp in particular that I REALLY want to continue. Thanks!

[graceful condor]

I have a question, I went on the suno website and made a song, are we allowed to further edit the song?

[placid gorge]

iirc yes

[graceful condor]

thanks!

[shell jungle]

i have a question. when the music is being made are the vocals and music split or all the music and singing all in one file? cuz i was gonna say you could find a way to enhance the voices then add them to the audio.

[graceful condor]

That's a good question, wud also be interested in the answer to this!

[normal grove]

I third that

[open magnet]

all in one model

[normal grove]

I mean it's amazing anyway, but sometimes the lyrics can be quiet on some songs.

[open magnet]

that's why things sound a lot less 'artifical' or canned

yeah that's the downside that the model tries to get the best overall sound, and doesn't pay suuuper special attention to vocals quality sometimes

[normal grove]

Sometimes the track is a great one too, I'm sure that'll improve overtime, though an amazing product regardless, it blows my mind how you've got it to work like you have.

[open magnet]

more scalable in the long run though in my opinion. would rather have a model that can do verything and can be easily grown rather than a solution where you splice a bunch of stuff together

[normal grove]

Is it seed based? Be great if you had a great track and you could repeat the seed to continue it with different lyrics

[open magnet]

not really but we will probably do something where you can continue with timestamps, like: from this track use from 3s to 7s and continue with these lyrics

[open magnet]

would that get you what you are looking for?

[teal raft]

Sounds good to me, tbh

[normal grove]

Yes that sounds great, I've had a couple of songs end before the end of the beat, meaning you couldn't add on to it without it sounding wrong, is there any way around that?

[teal raft]

It sounds like what georg is suggesting is something like Control Net for stable diffusion, in that you splice a section from an already-created song and tell the model "Like this" so it generates new audio from that template

It wouldn't be a seed in a mathematical, traditional sense in so much as how AI stuff works, but a 'seed' in that it's a structure for the model to glom on to so it has a place to start

Do it right and there shouldn't be a seam

[normal grove]

If that works then that'll be excellent.

[normal grove]

Sorry not if it works, of course it'll work, I just wondered how it would be done, thank you.

[jolly steppe]

I just paid for the Pro plan, but I keep getting the no more credits message. Does it take time to propogate or something?

[light ether]

We may need to link your accs manually - please email [e-mail] and someone will help out

[jolly steppe]

I'm in a bit of a hurry, do they respond quickly?

[placid gorge]

From my knowledge they respond pretty quickly maybe like 5-30 mins unsure i Just know its not gonna take several hours

[light ether]

depends on time of day, rn is probably quick but < 24hr would say 🙂

[placid gorge]

tru

[jolly steppe]

Annnnnd we're LIVE! LOL - Thanks all.

[ivory axle]

Hi all! Can you tell me please, how to increase limit (40 sec) and get full version of the song (3-4 min)? I can subscribe if necessary

[neat nexus]

#⏩┃continuation

[ivory axle]

thanks for the answer!

[shell lagoon]

2 months have passed and still no solution to out of tune vocals? like 90% of my new chirps are unusable

[placid gorge]

?

[shell lagoon]

I feel like it just gets harder to get good results. 2-3 months ago I could make a 3+ minute song with about 30-40 generations. now I get 2 minute ones of about 60+ generations. also original vocals are lost much faster now. very hard to get consistent results compared to few months ago

Am I the only ones getting worse results now? used to get good choices felt overwhelmed by options, now generation after the next bad results, cracking vocals. 11 tries so far today (22 results), all bad, almost all of them out of tune. feel like things change for the worse over time

[placid gorge]

@light ether any insight? ^

[light ether]

discord or web?

(it could be your standards / ear have gone up or gotten better as well, this happens to us)

[shell lagoon]

discord. web is , idk, people use it ? I dont get the point of it. not my standards, results getting much worse. noticed it last months where it was really hard to get consistancy. now I can't even get a "good start"

[light ether]

we havent changed anything with the underlying discord model for a while 😦 That said we are working on getting a new, better model onto Discord soon (a preview of that model is running on the web version rn.)

sorry!! dont really know what to tell you. I do notice that my standards have gotten way higher as I work with Chirp more. So I also feel a similar feeling

That said, please try out the create workflow https://app.suno.ai (its a different workflow for this moment) to get a sense of the new model, i hope its a better listening experience

[shell lagoon]

tried the web. I hope it will be usable with own lyrics and will allow "stuff" happening between verses in one take. for example instrumental, guitar solo (something I'm really having fun with, adds so much depth to a song)

but now it also takes time to get, I can write guitar solo, get some generic tune, then 2 parts later (continue x2) suddenly get a guitar solo een though I didnt write it

[light ether]

soon!

[digital jungle]

chirp offline?

[neat nexus]

Appears to be working for me.

Anyone else having issues?

[placid gorge]

Couldve been a Region based issue but works for me

[blissful raven]

Anyone figured out how to duet yet?

[neat nexus]

Someone did, but I don't think they shared how to do it consistently. I'd love to know the secret if there is one.

[broken lotus]

duet example.

they will take turns singing, but when it comes to the both singing together Chip seemed to pick one voice.

[blissful raven]

So you have to make a Name for the character?

Example John:[lyrics] Sarah[Lyrics]

[gilded folio]

In love with this so far.... wondering if there is some sort of cheat sheet reference for prompt building?... like for example if I want to fill some time with just music and no lyrics... is there a prompt for that?

[broken lotus]

some examples here: https://discord.com/channels/1069381916492562582/1166166196467548280

[broken lotus]

this works on Discord rn, not the website yet...

[gilded folio]

The website seems to have very limited character entry. Am I doing something incorrectly?

[neat nexus]

It's a limited tech demo of the v2 Chirp.

A much fuller featured version is expected to be released very soon now.

https://tenor.com/view/soon-gif-19121692

[gilded folio]

ah. Exciting. Just found the app today. Very cool.

[thin flint]

Is that from web V2? Curious of all the creative prompts people are using with it, never thought of trying that.

[thin flint]

Oh, does the website currently remember the context between generations? I think it just makes a new prompt without being aware of the other generations so when you write continue x2 it's just luck or placebo effect.

[teal raft]

So!
I'm trying to continue a song in V2 on the brand-new updated web app, but it seems to be cutting of the first line?

Like the words are there, it's genreated them as well

But the audio continues from the second line and not the first

I'm just trying to generate a chorus alone because it seems like when you continue a song it's limted to a solid minute vs. 1:20

I wonder if it's because I am trying to go directly to [Chorus]

[light ether]

is it skpping the first line?

[teal raft]

Yeah

It starts at "Can't resist"

Yeah

Ok it's becuase [chorus]

If I cut that out and just have the lyrics without [Chorus] it reads the first line

Though now the music seems off a line? It's singing the lyrics but out of step with the original song

Eh maybe that's just the way it generates

[light ether]

ah, sorry, we are gonna keep iterating on continuation so you can have more fine grained control - definitely some things continue better than others right now

[teal raft]

No worries, it's a toy. A fantastic toy but like, cmon, this is all brand new tech

[vital elbow]

Is there some kind of filter, it told me I couldn't do a prompt, but there wasn't any bad words in it

[teal raft]

Sometimes it just does that, I've made a habit to copy whatever prompt before I enter it

[frank isle]

With Chirp v2, I've noticed that verses can be repeated after they're all spoken

[clever depot]

so the new web app now allows manually setting the style of music, that's great, because before it never used "atmospheric black metal" as the style even with the prompt asking for "atmospheric black metal". but now, if I use "atmospheric black metal" as the "Style of Music, I always get a song that sounds nothing like atmospheric black metal at all

is there just no atmospheric black metal in the training data?

I mean, you don't need to be an expert in metal music to see that this sounds totally not what anyone would imagine from "atmospheric black metal", it sounds way too happy 😅

[thin flint]

I can't seem to play this Turkey Time song. None of the play options work. What's up with that?

[light ether]

sounds like a bug, sorry... what browser are you on / mind sending a screenshot?

[untold briar]

When you first navigate to the page, there is a play button at the bottom that does nothing. But if click the song then it works. I guess the first play button is supposed to play that whole 'album' if it were a list of songs? Actually there are two click through dead buttons. 1: After you click Turkey left nav ( This one is supposed to not work, but it maybe confusing I think. Probably less confusing when the right page has more than one choice in future. ) 2: After you click Click Turkey thumbnail.Then you click the song itself and it works

[light ether]

Boo. Let us try to fix this quickly

[thin flint]

Whether I click on the song or am just on the main explorer page, it still does not work for me. Chrome here.

[light ether]

Thanks, can you double click on the song? That Blue play button not working is a bug though

[frank isle]

I'm trying to make Chirp speak fast like an intense rock song. How can I do that?

[vital elbow]

On the new version, when continueing, is it best to only add new lyrics, or include the full lyrics (if it was previously cut short)

[light ether]

I've seen success with both, I hate to say "whatever starts to work best for you" but....

[maiden forum]

Hi, i did a lot of experiments with suno and its amazing

the new webapp is also turning out great results, but has problems with foreign languages

before i was even able to get songs in ancient egyptian language

i mean with the discord server

[thin flint]

Was this Turkey Time song made with a bunch of continuations? Looking at the prompt it looks as though you can put a whole song structure in it? Because that was a full song with guitar solos and a nice outro and everything.

[light ether]

Ya, I think maybe 5 - if you download the video those are the cuts

Probably took 10 -15 Chirps total to get those 5?

[clever depot]

every time I try to continue from an existing clip in the web app, I get this:

using the same prompt that the original clip was generated with

[teal raft]

Seems to be service related

I get it too, I just keep trying

[clever depot]

does it work sometimes then if you just try often enough?

[teal raft]

For me it does yeah

[light ether]

can you dm me your prompt?

[frank isle]

The style doesn't seem to be working well in another language. I put heavy rock and it didn't give me anything close to rock

[inner steeple]

how do you get the whole song through website version?

[clever depot]

I can write it here, one I used was atmospheric black metal instrumental intense without any lyrics

[inner steeple]

I don't see obvious option for this

[light ether]

It's under the ... menu on a clip "Get Full History"

[frank isle]

Oh yeah, is there a way to make it, like, add a specific instrument in the middle of the song at a certain part?

[inner steeple]

so only for v2 generated?

I tried looking about the old continuations and didn't see it there

and I haven't yet continued with v2

[light ether]

I see, we'll add that probably next week

This is for old v1 clips from Discord right?

[inner steeple]

I might be looking in wrong place too

yeah, ones generated via discord, but that show on webapp

of course I can get the full clips on discord

[light ether]

Ya, we need to hook that up, so that's why the option is not currently there on web, sorry

[inner steeple]

no worries, just checking I didn't miss anything

is the chatgpt prompting secret sauce or is the current template known thing?

[clever depot]

I had it successfully continue now after trying a few more times, but does it not result in a song with the length of clip1+clip2? I only get a new short clip without the old clip at the beginning?

[inner steeple]

I do like some of the generations made with the non-custom prompt but it's obviously bad way to iterate on the lyrics

but I do get better lyrics overal with the suno prompting than what I get if I ask chatgpt myself

so I'm guessing there's good template

[light ether]

if you want to get whole clip, you can click the ... on a clip and click "get full history". we'll probably make this more intuitive in QoL update

[clever depot]

ah, thanks!

I think it should just directly put them together by default

[inner steeple]

you don't always want that

[light ether]

we might make that a setting or somthing

[inner steeple]

it's also easier to review new gens when you don't have to play the start again

of course the transition is harder to judge that way

[untold briar]

Yeah "history" makes sense technically but it's kind of a confusing term

[light ether]

we're hoping to have a more dedicated song builder mode at some point

[inner steeple]

"get whole song" might be easier to understand

[light ether]

feed is definitely v0

[clever depot]

is the continuation of a clip somehow much lower quality than the original clip?

it sounds like the original clip is in stereo and the continuation is in mono

[teal raft]

Yeah actually

I'm getting that too

[thin flint]

I reused the Turkey Time prompt in order to make it into a new genre or whatever, and I hit generate and it's just not generating. No errors or anything that I can see offhand.

[teal raft]

You're right @clever depot the continued portion of the songs are in mono

[clever depot]

this is a super repetitive song, but that's great to see the change in quality in the middle where the continuation starts

[inner steeple]

wow

that really is

[teal raft]

Right at 1:21

I'm in headphones and it's way obvious

[inner steeple]

it does seem to keep the style intact tho 😁

[clever depot]

yeah the style stays nicely

[inner steeple]

it's pretty obvious on my speakers too

[untold briar]

Yeah there is some kind of stereo wonkiness in continues -- not always though *sometimes * mine seem great. A few rough edges hopefully get sorted soon

[teal raft]

Is that an old V2 or did you make that new after the update?

[clever depot]

the one I just posted I just made a few minutes ago in the web app, both original clip and continuation

[teal raft]

I've been exsluviely messing with old cretions

ah dang so yeah that part doesn't matter then

[inner steeple]

yeah but is the original part you continued made now?

[clever depot]

original part is also just made a few minutes ago

[clever depot]

all 4 continuations that I had succeed so far were mono

[teal raft]

Woof yeah I thought something was off but like, pushed it out of my head. Now it's really obvious and I uh, don't want to burn any more credits

Well the mp3s of JUST the continuation still think they are in stereo

just the left and right channels are the exact same

[light ether]

cc @clever depot looking into it, sorry

[maiden forum]

yeah when i try to continue with new text, it just makes the last half of it or so

will burn more credits to see if i can find a solution

[teal raft]

Wow yeah you better be sorry! Slacking off like this on US Thanksgiving!!
Seriously though thanks for just being here in the chat and responding

[maiden forum]

ah its thanksgiving in the US 😉

happy hollidays

[inner steeple]

yeah we don't do turkey time here

[teal raft]

I'm going to eat so much middling food because my step mother in law can't cook for shit helll yeah

[light ether]

ya we knew turkey time would be a very niche thing

[teal raft]

This sounds like what I was hitting earlier when I was trying to make JUST the chorus. It would skip the first line in the section and start from the second, out of time with the music

[light ether]

oh boy, i may have found the issue (mono one)

[teal raft]

I'd say grats but the 'oh boy' means something

[maiden forum]

bit a fresh turkey it was a typo somewhere

[teal raft]

I'm guessing something to do with discord

or the V1 model and the continuation code or something like that

[maiden forum]

when coders start a sentence with "Oh boy" it means nothing good

trust me, i do software since 44 years

[inner steeple]

or.. it's something really dumb

[light ether]

ya this one is both nothing good and really dumb

standby

[teal raft]

rip

[maiden forum]

❤️

[inner steeple]

standing by

[light ether]

hoping to have a fix in ~15 min

[teal raft]

15 seconds wow ur so fast

[thin flint]

So the first part of this prompt worked alright. I'm getting an error with the bridge though, is it because of the [guitar solo] metatag? Just figured that might work, but that's the only reason I might be getting it...

[Verse]

Thunder rolls on the horizon line,
In the darkness, we all align
Iron chains, steel in our soul, we dine (we dine)
On the echoes of the battles we've outshined

[Chorus]
Metal nights, under the moon's prime,
Raise your fists, feel the rhythm, sublime.
As the stars ignite, our spirits climb,
Roaring loud, in unison, our hearts entwine (yeah!)
Metal nights, in the moon's prime

[Bridge]
Metal hearts beat as one,
Through the night, until the rising sun,
Echoes of the steel, never undone,
In the brotherhood of metal, we are one

[Guitar Solo]

[frank isle]

I'm trying to continue a song and it doesn't work

[clever depot]

it seems if you just try again often enough, then at some point it will work. but there is an issue at the moment that the continuation is too low quality

[frank isle]

Ah, is that so?

[clever depot]

that is so

[frank isle]

I assume the continuation having too low quality is getting fixed now?

[clever depot]

yes, it seems we can expect a fix in 15 minutes 😄

[frank isle]

Already? That is surprisingly quick

From my past experience of other stuff I have tried, I would have expected a few days minimum

[teal raft]

I suspect they might be OK with pushing this update right now, really late for the US, Thanksgiving

Probably minimizes the amount of voices screaming

[clever depot]

on that topic, better quality of voices screaming would be nice too 😂

[thin flint]

Yeah, I know about the quality bug, but I'd rather have that than an error about my prompt when my prompt seems fine, E.G. passes the filter. You may want to add a message about a prompt being too long for reusing. I tried to use the Turkey Time prompt and change styles, but the prompt comes with all the lyrics, and I guess I have to split it up. But I had to figure that out and the app didn't tell me, just rejected the prompt with no reason.

[frank isle]

On that topic, better quality of voices would be great too (especially for different languages)

[light ether]

Yeah, we need better error messages for sure. TY!

[frank isle]

But I get it

[light ether]

which languages thinking?

folks watching, react to this if you experienced continuation bug and want to quickly test a fix (credits included)

[frank isle]

I was about to say before you edited it

Then yeah, sure

[thin flint]

I think Suno is in pacific timezone, so it's only 11 PM there, not sure if everybody's in remote locations. I mean that'd still be outside typical business hours haha

[light ether]

@frank isle @thin flint @teal raft @clever depot added some to test with (refresh, if not DM me whatever you're logged into the web with)

[thin flint]

That moment when you're on Discord with a bunch of people at 2 AM (eastern) messing with and breaking software for on-the-fly fix deployment. Proper QA work at it's finest.

[teal raft]

Refresh worked for me. I assume then that we try to continue and see what happens?

[light ether]

yep. i just tried a few,s eemed ok to me

[teal raft]

roger roger giving it a shot

[teal raft]

It's fixed on my end

That's some good ol' full sounding music, I tell ya what

I downloaded the audio and played it in Winamp since I'm old as hell and tested it there

[clever depot]

both continuations I got at the moment have a length of 3 seconds with only silence in them 😅

[teal raft]

Winamp has a little slider you can whip back and forth

[light ether]

this can happen if the previous clip ends in silence, or also a bug

[clever depot]

yeah, it was quite quiet at the end

[light ether]

we working on a workflow where you dont get charged credits for silence...

(not hard...)

[light ether]

Thanks all for testing and fixing this dumb issue, Refresh again and should have some more credits (DM me if not)

[frank isle]

Still giving me errors trying to generate it

[teal raft]

Yeah it's fixed. Sounds great too. The join is at 1:14

[light ether]

can you DM me your prompt? This is probably something else like moderation now

[teal raft]

Whew that's pretty generous, thanks mc

[light ether]

It's much appreciated, y'all saved us a bunch of time here catching this early

[frank isle]

Would it be okay if you sent me a friend request so I can DM you properly?

[teal raft]

Glad to be of help.

[light ether]

All set

[thin flint]

Continuations work, this was just an odd one. It was supposed to do a bridge, but what it did was this weird sound with a second of hard guitar.

[teal raft]

Soundds pretty rad tho

[maiden forum]

whenever i use [female] i mostly get a male voice, any workaround?

[inner steeple]

put female on your genre

[maiden forum]

ok, will try

[inner steeple]

doesn't seem to work without custom prompt though

it can still occasionally get it wrong, it's not guaranteed

[clever depot]

I'm also still sometimes getting the error when trying to continue

[maiden forum]

will, try. just had luck with a song text in ancient persian looks good

[inner steeple]

I got some weird error the other day when I just had my credits depleted

[inner steeple]

webui still claimed I had 10 left but it was gone and it got confused (this was due to the weird double generation thing I reported the other day)

[light ether]

looking into this one

[maiden forum]

this is a soldiers song in latin

[light ether]

this is awesome

[thin flint]

Are there new metatags and such to use with V2? Turkey Time has stuff like [guitar, outro] but just got a few odd ones with [guitar solo] metatag, so just wondering what best to use as far as V2 goes.

[maiden forum]

well, its not what i wanted but better than expected

is there an updated metatags list somewhere?

[thin flint]

Ah, I see your continuation became mono. I guess it's not fully fixed yet.

[untold briar]

No updated list of tags yet. In my opinion a lot of the more specific tags that worked occasionally in Chirp V1 like [piano instrumental] work much more consistently in Chirp V2.

[thin flint]

Does that work better in genre or lyrics? In Turkey Time they used it to separate song sections, like instrumental and vocals in one song.

[untold briar]

In Lyrics, I mean. I do think "piano" probably also works better in the genre (as a way to indicate an instrument overall), but in this case I was talking about the fine-grained tags in lyrics

[maiden forum]

hope they work on a fix now

[clever depot]

I have the feeling that the order of the lyrics in the combined "Get Whole Song" are not actually correct

[maiden forum]

also a fresh metatag list with a few instructions would be rather helpfull

yes, its shuffeling the lyrics sometimes

[thin flint]

Specific genre metatag lists would also be helpful for the most common ones, like pop, country and metal.

[clever depot]

in the generated video, the lyrics are in the correct order.

[light ether]

how exciting

[light ether]

I buy it, looking into it

[clever depot]

I think it's a nice christmas song though 😄

[vital crypt]

/bark welcome welcome to sea sparkles corner

[maiden forum]

i just did it, but it also mixed the vocals

but since this language dies out like 2500 years ago i guess i wont get any complaints

[cold gull]

Is there any way to "steer" the song into ending with the continue button on the web app?

[maiden forum]

use the metatag [Outro]

[thin flint]

This song here ended without an Outro. The next continuation with an Outro tag didn't end it properly. Ehh, not bad for a first song with this new interface.

[light ether]

I think we will do a dedicated Outro button at some point

[maiden forum]

its still chaotic to the end

dont know why, but after 2 or 3 continuation it sometimes has problems

[light ether]

this is pretty epic

how many continuations is that one?

[maiden forum]

yes, exactly what i wanted

i had like 5 continueations with 2 complete failures

one time i got a commercial break in modern farsi

[teal raft]

I'm sure this is just inherent to how this all works but yeah if you continue off a continuation you start to get a drift away from the original audio. It's cool though, like a very slow slide remix

If you're trying to make a complete song it seems best to constantly feed off your original audio

[thin flint]

Mine had an estimate of four, hard to count them as a screen reader user but yeah. Would be nice of an indication of how many clips there are when you tell the app to clip the whole song together.

[light ether]

NB: there's a bit of trade off between context and gen length

[clever depot]

I now tried to continue my festive santa song in a black metal style about evil santa 😄 it stays quite close to the original style that it continues from, but the subtle change to a different style is quite good I think

[light ether]

we'll probably do a slider wher you can say - get 40s continuation with a longer context

[maiden forum]

that would be cool

[teal raft]

yeah

[maiden forum]

much cooler would be buttons to put instruments in

[teal raft]

Now though the limitations of how you organize clips is going to become a burden with all these clips.

[light ether]

thats a whole nother research direction (we want this too)

[blissful raven]

Great Job Team V2 is awesome

[light ether]

thats part of QoL we're thinking about, if you have any desires/suggestions on organization as you gen more, DM me or post in #💡┃suggestions-web

[maiden forum]

suno is still the best, kits.ai is rather useless (for me)

what means QoL?

Quality of Lyrics?

[teal raft]

Quality of Life

It's a term for UI sorta stuff

[thin flint]

I don't hear the black metal in this.

[clever depot]

yeah it didn't manage to actually get all the way to black metal, but you surely hear the last 2 minutes are quite a bit darker than the first 3 minutes? 😄

[small kite]

New here and have a question, on the website it creates roughly 1 minute cuts and on here is it always just 30s? Tried the same lyrics and notes in both places and this one just smushes one minute of lyrics into 30s and makes it less ideal..

[light ether]

Welcome, Ya, we are updating the discord to match the website next week, just need to do some bot upgrades

[thin flint]

Sure. I wonder what genres you can switch to from country. I guess the aI doesn't see that a logical progression or something.

[maiden forum]

I tried german folkmusic with suno

the results been so bad they were good already

https://youtu.be/I2D_q-XjRtg?si=2avhtpwNM7t-SAqT

but serious now, i know a dude just doing music kinda this one

[clever depot]

I'm surprised it can do German so well

[small kite]

Awesome thanks for the heads up, overall super impressed!

[maiden forum]

yes, even arabic and farsi is rather well

i do music in diffrent languages

i also speak some, like 8 or so

[light ether]

we finally got Thai to work pretty well with this one ( I dont speak Thai)

[maiden forum]

ah cool

[clever depot]

and when can it do growls? 😄

[maiden forum]

however my tries in romanian were not the best

will check it out later today

[light ether]

how would this be written out in normally?

or is that more of a technique while doing a phrase

[clever depot]

there's no way how it's written out I think. a band that uses growly vocals usually just does the whole song in that vocal style

[maiden forum]

do you think we could do a song in klingon?

[light ether]

I see Fly typing 😛

[inner steeple]

would be nice to enforce intro too.. it's pretty hit and miss when it starts right away from the vocals now

[maiden forum]

i am on it

[clever depot]

have you tried just giving it klingon lyrics?

no one understands it anyways so I don't think it matters much if it mispronounces it 😄

[untold briar]

You can try really strange styles. Klingon, seems like a long shot, but who knows. Like look at `zombies zombies zombies' genre, it's PERFECT

[clever depot]

why can it do the "zombies zombies zombies" genre better than an actual genre like "atmospheric black metal"

[untold briar]

It may not be better, I'm sure I tried a few gens and picked the best.

[jade mist]

Great! Finally now the web app allows creating your custom lyrics and continuing! Very cool!

[untold briar]

One thing I'm not clear about in V2 is how to properly input multi-word genres like "black metal" is it "black metal, atmospheric" or just all spaces, or maybe you need to make a compound word. That may be one advantage to zombies.

[clever depot]

very good question, yeah. would be helpful to know how exactly it needs to be prompted

I think the biggest problem with "atmospheric black metal" is that it doesn't even know pure "black metal" well

[untold briar]

If it happens to be parsing it as "metal, guitar, heartfelt, black" like just another word, that could explain it, for example

You could try a more descriptive general terms. Maybe like: shrieking distorted fast etc

[clever depot]

hm yeah, I will try that. but it would really be way nicer if it would just know what "black metal" stands for

[light ether]

we'll try to put some FAQ together along with the discord release next week

[maiden forum]

i guess i crashed the server with a lovesong in klingon

[clever depot]

btw, I just recently added Klingon language support to the game I'm developing 😂

I think its fun if everything supports Klingon

[untold briar]

I can't find the clip but I got a pure hallucination, so basically "SimLish" just Chirp making up fake language, and it sounded so much like somebody casting a spell it was so eeries. But since it was a hallucination I can't even remember what the prompt wsa

[light ether]

something break?

[maiden forum]

no all good, just needed a little bit longer

the result is actualy good

any native speakers here?

[clever depot]

so that came out of that. I would say its indeed better than what just "black metal" generates

[untold briar]

Klingon is sort of a hybrid of a bunch of human languages, I'm not too surprised really. You can almost read it if you know the right set of languages. And a lot of the words are puns in like a billion other languages I think. So if you know 50+ languages, it might not look that strange

[light ether]

you got gpt'd tho

[clever depot]

what does that mean?

[light ether]

we gotta fix this genre popping into the lyrics... currently quite the meme

[clever depot]

ah. yeah, I just typed "black metal lyrics" to get the lyrics

[maiden forum]

yes, true, and gpt for lyrics sometimes realy realy sucks

[light ether]

ya, i would say this is an intentionally campy v0 of lyrics

hope to make it a lot more interesting

[clever depot]

the lyrics are already better than the audio though. its just a lot easier to generate good lyrics

[maiden forum]

ok folks, breakfast is calling

catch you later

[thin flint]

https://www.youtube.com/watch?v=TV27s4EetPk

[lavish stone]

🦟☄️💥🌼

[cold gull]

Weird? It made a song of gpt's output so its singing that it didnt find lyrics (never asked to copy a style or lyrics)?

[jade mist]

I think if you make a song builder, it would be good to keep this classical method as some people may prefer it.

[rose galleon]

do we have an updated list of metatags we can use with v2?

[graceful condor]

using app, sounds great! but i cant get a whole song?

[broken lotus]

click on the '...' and pick CONTINUE FROM THIS CLIP

[graceful condor]

i did! and.. nothing 😉

or else - where should i look for the full song?

[placid gorge]

and thats where.. refreshed page 5 times now doesnt exist [NVM thats just for the V1 clips okay..] | Also cant use ChatGPT page as a continue source.

[broken lotus]

hit the CONTINUE FROM THIS CLIP (on a v2 clip only?)
and you should be seeing a new window CREATE (CUSTOM) with a little thumbnail of that clip, and a new prompt/style boxes. Add lyrics or switch up the genre style and hit GENERATE (costs 10 tokens)

on continue clips a new option shows up : GET WHOLE SONG
this will stich all the clips into 1 track (no tokens, free)

[deft raft]

I click Get Whole Song but where is the stitched song? I still only see each part individually

[untold briar]

It shows up as a new clip

[broken lotus]

(it's a new website, remember these issues so we can put them in suggestion/bugs)

the difference is the NEW CLIP will only have 1 play/like (because it's not 2 gens)
and if you have lyrics they will allllll be there....

it should look different....

(would be great if it was a different color background or stood out in some way)

[deft raft]

For some reason it's not creating a new clip for me. I'll try different clips later.

[untold briar]

Interested in hearing prompting techniques to rescue a longer song where the vocals are fading, or if it's instrumental, the song is stuck in a 'end of song' mode. Maybe prompting intros again, or numbered verses...

[broken lotus]

I have sometimes seen a delay (or complete drop) when stitching together... and later it worked, or later I refreshed and a couple were there....

It may be growing pains, may be a bug, maybe a crowded server....

Try again later 😓
(but also give feedback to bugs/suggestions where it's needed)

[normal grove]

@broken lotus I've sent you a DM

[broken lotus]

😉

[normal grove]

Is there a problem with getting the whole song, I click the button and it isn't doing anything, I've tried refreshing and it's still not there?

[random minnow]

Having same issue. It worked fine earlier this morning, but now I click the whole song button and get nothing either.

[normal grove]

I have a feeling it could be the servers are under heavy load.

[vast dew]

https://tenor.com/view/server-is-fine-burn-fire-spongebob-blow-gif-14178373

[broken lotus]

Getting the complete song is free (no tokens) so you won't be charged if suddenly 10 show up in your acct 4hrs later... 😅

[normal grove]

Oh, I hope it doesn't remember how many times you try haha.

I know it's not to everyones taste, but you honestly wouldn't know it was AI. Unless there are some very very over worked staff members 😂

[graceful cliff]

There are some very very over worked staff members 😄

[pallid seal]

How can I get songs similar to this second one
#🐣┃chirp-beta-1 message

& how can I make it bigger

With the similar sound

[graceful cliff]

#⏩┃continuation

[pallid seal]

It didn't work

Btw I don't have my own content

I asked chatgpt to continue

[graceful cliff]

are you on web app or discord

[pallid seal]

discord

[graceful cliff]

it should work...if try again doesn't help, then...I don't have any suggestions 😦

[pallid seal]

Alright

[inner steeple]

where should the whole songs arrive at?

I'm not really getting any either

[storm spruce]

Is there a way to prompt the [Continue from:] thing to start the next verse sooner. It always seems to put it at the end not being able to finish the text or just does 1 minute of continuing the chorus/song from the previous part with no lyrics

[tight frost]

It seems that custom creation on the web rejects artist names. I would appreciate this being mentioned clearly in the error message. Same thing for lyrics censorship, I'd like to know which key phrase the model (GPT ?) refuses to process.

[jade mist]

How do I get access my older clips with V2? I've been using it for a week now and have created about 100 clips. How do I see all of them?

[untold briar]

You can find old clips on the My Clips page, and if they were web clips, you can continue them and customize the lyrics. The sorting and info on My Clips is barebones now, but they should be there if you scroll back. Page navigation is bottom right of screen.

[vast dew]

would love a volume control on the web version. 👀 (and have it remember the volume. Most websites do not remember the set volume for some reason)

[inner steeple]

I'm really digging the lyrics on the automatic gen, wish I could get chatgpt to output similar quality 😄

[broken lotus]

STEREOPHONIC

[pliant cave]

Jonathan, we can always count on you to create the most bizarre and unique prompts imaginable. Creativity maxed out. LOL

[simple anchor]

same here getting the whole song is not working

[inner steeple]

wait.. you have to like these?

maybe that's why it's not working for me

[broken lotus]

yeah, not able to get a song history for a while now... mc is looking into it...

(we might need to get these later... keep track of them somehow, it's easy to lose track)

[broken lotus]

no, it's just not working...
(I have tried several voodoo-magical things..., but alas it is just an annoying glitch, lol)

[inner steeple]

ah ok

yeah I don't think I can even quickly find the right clips

only know from last clips which one I want to be final

[broken lotus]

most of mine are instrumental... so it's just hundreds of birds in the same colorscheme...

😭

(am am thinking to screengrab the bird, maybe. at least I have some record of what I wanted)

[inner steeple]

I got bunch of accidential ones too

[naive jungle]

we have a list of genres / artist to craft on prompt?
love the website UI , congratz!

[inner steeple]

this was supposed to have vocals

[broken lotus]

artists/bands don't really work (except in generic ways, like the name is associated with a genre so you will get something in that genre...)

there are some guides here for Chirp v1
https://discord.com/channels/1069381916492562582/1151542418991239168

[naive jungle]

thnx!!

[digital jungle]

my credits aren't showing up on teh web app Chirp v2

[light ether]

sorry, refresh/logout/login - we just had a brief downtime

[digital jungle]

what's teh best way to get the style of music you want? I can't seem to find the right sound and each continue the vocals seem to go higher on a scale

[light ether]

what genres are you interested in?

[digital jungle]

Like if I wanted it to sound like Green DAy, I know you can't use Green day as a style but how would you describe that style of music?

[light ether]

perhaps - punk rock? alt punk rock, pop punk

[digital jungle]

web app working now but I can't seem to continue

I hit continue from the first version and nothing happens

[light ether]

Ah, it's a little different in the new web version. it selects the clip you continue into the form, then you have to submit the form with whatever new lyrics you want

[digital jungle]

do I leave the original lyrics or change to the next verse and chorus?

[broken lotus]

change to the next verse/chorus

(or it will sing those again)

[digital jungle]

love the stereo version!

[light ether]

nice to see you in chat!

[digital jungle]

and with the web version I don't have to scroll through everyone elses to find mine!

[simple anchor]

now it was working - thank you! 😉

[digital jungle]

LOVE the web version! Best chirp ever!

[blazing schooner]

/chirp

[flint haven]

The /chirp command is not available in this channel

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

[light ether]

Glad you got sorted with Continues!

[simple anchor]

my first test song with new version. First thoughts:

• much less drifting which might also a bit more boring depending on each taste
• better quality
• stereo
• easier to use because of new length

[light ether]

ya, really depends on taste, we'll add ways you can force drift

[inner steeple]

I'd love the two chirps you get to work so that if you continue, other would try to follow previous clip more stricly, and other would be more loose..

this way you'd still get those happy accidents

[simple anchor]

yes that is the best I got so far I really like that: https://youtu.be/7y5vA5CdDaQ

[untold briar]

chirp is pretty hard to confuse with crazy parenthesis now

[broken lotus]

it's giving you harmonies!

[untold briar]

In Bark and earlier Chirps using () and maybe >'s (like markdown indents) you can sometimes get crosstalk or cross singing, where people talk over each other, but here it just kind of sounds like a song yeah

[neat nexus]

Yeah I noticed I can't fake it out with an onomatopoeia sound and transfer it to an instrument anymore.

[untold briar]

really the lyrics can just be any random things formatted and it comes out amazing, which in some ways is a bit frustrating to my whims. at least i can reliably produce bad singer even Chirp v2. (not this clip, this is a\n example of the first thing where any set of characters in the text field just works)

[inner steeple]

how does the sub work, like, you get rolling credits for 30 days from moment you sub, or is it just reset when month changes?

like, if one got sub today, you'd have week left to spend those credits?

[jade mist]

How do I make the model to remember how certain parts of the song went? For example, if there is a verse and a chorus, I want them both to be sang in the same repeating melody, as in a real pop song for example. Another question: If the song cuts off before the end of lyrics, how do I continue them to the end?

[broken lotus]

monthly subscriptions will renew on the same date each month... based on when you subscribed

(no idea what happens if you subscribe on leapyear lol)

[broken lotus]

chirp listens to the end of the song you're continuing from... it continues based on what it heard.

If you get too far from the Chorus (or the voice), it won't be able to get them back (like a long instrumental in between)

[inner steeple]

can you somehow track this day

like does it tell which one it was

[broken lotus]

yes (just checked lol)
/info will tell you

[inner steeple]

ah nice

[jade mist]

I can't wait the time the track limit is like 3 minutes, that way there won't be a problem of Chirp not remembering how it went...

[fiery sinew]

new web version is nice. It even continues the old favorites. However the compression level is a bit too tight, and the vocal sound breaks often.

[untold briar]

Do you happen have a good example of where it falters you can post?

[jade mist]

How do I make an electronic party song start with only an acoustic guitar but then slowly build up, like Wake Me Up or Hey Brother?

[high magnet]

I noticed this morning that on the web version, my 2nd song version has been without vocals like 3 times in a row now, even though the 1st one does have them...? Dislike that, I do.

[inner steeple]

you can try generating the clip first with acoustic guitar genre etc and then swap the genre on continuation

[broken lotus]

try describing the vocals first in the prompt.
(like with image-gen, the first item seems to get more weight)

[inner steeple]

it may be tricky to get it do that, but in theory, it is possible

[broken chasm]

hello Admins and everyone else! please help: does anyone know if older chirps get deleted? I made a chirp yesterday and after generating 3-4 today, the older ones disappeared when I started using 'create custom' feature. Is there is a cache I can get this file back from? I have only been using Suno for a couple of days so my log file should be relatively light?

[neat nexus]

It's not in https://app.suno.ai/me ?

[broken lotus]

use the MY CLIPS link. there's pagination links at the bottom.

[thin flint]

That's a top quality song. Did you just make each continuation from the first clip?

[topaz willow]

If I have a "Continuation" Stuck how I can stop the loop because doesnt let me on the new version V2 web?

[broken lotus]

refresh the webpage?

or, not sure where is is 'stuck'

[digital jungle]

@thin flint yes with the new web based version

[topaz willow]

Thank works, Thanks!

[serene current]

Hi all. There is any kind of prompting guide? (voices, instruments, riffs...)

[serene current]

I'm doing and entire album on BItcoin 😄 😄 😄

[broken lotus]

here are some guides https://discord.com/channels/1069381916492562582/1151542418991239168

also search for your https://discord.com/channels/1069381916492562582/1154790975240155247, there's sometimes quirks that work really well.

[serene current]

thank you!

[north briar]

@light ether Bark is not now support Thai language now?,

[untold briar]

chirp does

[north briar]

How to make Bark support Thai as well?

[untold briar]

Unless it happened recently I haven't seen much success in adding entirely new languages to Bark, though it is possible in theory. But the specific training code and methods are not released so it would be a bit of science project for someone to do, there's not like a guide or anything.

[graceful condor]

There should be a STOP / CANCEL button for mistakenly pressing Generate, I think..

though im sure whatever hardware has started processing it might have already processed the tokens

in that case, it could be prorated..

but how can you prorate something that you never give to the customer, i guess

[tender dirge]

Hi, i encounter a bunch of errors and it used my credits, ended up with some duplicates and corrupted tracks that dont play, any help?

[graceful condor]

can you specify it to scream at certain points, etc?

what kind of modifiers exist?

[neat nexus]

It's a new model. You tell us.

[naive jungle]

you can CAPS WHEN YOU WANT TO YELL, or maybe do some (yell) at the end of the phrase, that works for me. take care of the phonetics ( maybe wrong word 😛 )

[graceful condor]

kewl

thanks

What would it take to run something like this locally?

[teal raft]

A rack of A100s probably, like anything else

if not more since this uses chatGPT

basically none of us are running this locally unless we have a server cluster of our own, it's one of those things

[graceful condor]

quantize!

lol jk

[simple anchor]

That new version is amazing👍 Great work!

[fierce granite]

WOW! New web version is amazing, now it let's me finish the initial great songs I already created.. super nice!

Next I'd love to be able to give some instruction, unless that's already possible? I.e: (fade), or maybe I love the music, but not the vocals, so I want to try different vocal styles, while leaving the music the same.

[jade mist]

I made a good song for Geometry Dash by accident! Chirp memorized the drop melody and made 5 of them!

[gentle portal]

Hello, has anybody tried death metal with Chirp? Something like this: https://www.youtube.com/watch?v=MwtVkPKx3RA

It's largely a jumbled mess but it catches ArchSpire's elements to a fair bit. even though it's not exactly musical.

[warped goblet]

i wonder the same since bark and tortoise works so well for local

[warped goblet]

local models are not that bad nor do they need racks of A100s

[teal raft]

Sure but the local models you and I can run on 4090s, or even two or three of them is beyond what suno.ai is using. It's using ChatGPT in addition to whatever is generating the music

I don't know what music takes to spin out of a machine, images are 'easy', choesive text with enough tokens is the hardest, where music is I don't know

[thin flint]

So When you continue a song in the web version, I notice that my lyrics from the previous continue aren't deleted. Should I leave them there, or will it get confused? I had a bridge section and part of a line was cut, is there a way to continue that into the Chorus?

Why on earth are my continuations being rejected? It says composing two songs and then the prompt goes away like wht the hell?

[mental maple]

im trying to make a song in rap, death metal . but the output has no audio lol i guess the fast paced energetic songs r the death of ai gen music rn as they take the most effort

[gentle portal]

Might be an issue with the dataset it's been trained on?

[supple pawn]

God, it's still going on? It's been 4 years! Amazing.

[shell lagoon]

The new web custom give really bad results to everyone? like the singing is mostly muted beneath the music , and it just try to stretch the verses too long. the first verse always gets to about 50 secs when in discord version its normal 20-35 secs. don't know, I can't get anything even OK with the new web costum

[sinful abyss]

The new version is really unstable 😦 Sometimes it ignores the lyrics and generates 80 seconds of music. Sometimes it adds a lot of silence at the end of the track and the next track becomes completely silent.

And sometimes it sings something completely different and not related to my lyrics

Also, encounter reduplication - in continue section it sings last lines of previous generation

[shell lagoon]

I tried metal, rock and rap in the new version and it was really bad. far inferior to the current discord version

[sinful abyss]

Oh, and a couple of times I came across that it sings meta tags 🙂

[south estuary]

I found it amazing, mate. Not sure if it is a matter of a tate or if you just were a bit unlucky with your generations.

[shell lagoon]

I guess some styles it just cant produce well, don't know what styles you tried. but I got 100% bad results with muted vocals in different metal, rock and rap styles I tried

[south estuary]

Oh, alright. Yes, maybe in some subgenres you may have different results. But I got some pretty decent metal.

For instance, would you say this is good?

[shell lagoon]

Ye this sounds good. what style did you write exactly?

[south estuary]

I think it was something like "female melodic heavy metal"

[sinful abyss]

Do you use metatags?

[south estuary]

I used [female] and [chorus] [verse] for parts.

[shell lagoon]

still the vocals sound muted a bit beneath the music and unclear, the same is in your song. I compare it to the discord version and it just muted

[south estuary]

I hear you. Maybe you can try separating the vocals with something like lalala.ai or other stem separation

and use something there

This is rap... rapping starts at 27 seconds

sorry, actually 20 seconds

I used "oldschool rap"

[shell lagoon]

The vocals just muted, its very clear. compare this to 90% of discord version generation and see

so maybe when the music is just beats you can hear the singing or rap more clearly although its muted, but with music playing, not just beats like rap or house, but rock, metal, whatever, its much clearer vocals are muted

[south estuary]

Oh, I hear you

The accent in my language though, is 500% better

Maybe still you can do something with stem separation

to improve it

[shell lagoon]

I really with this was good. been waiting long for all the features , but it just doesn't work for me at all. like not even one single usable generation. even the music itself is much inferior to the discord version. it was like this in the "older" web too, and didn't improve at all. music of the discord version is so much better

everything I get in the new web version sounds like an alpha version compared to discord

[broken lotus]

you may be at cross purposes

(some people aren't actually looking for solutions, they just want to be heard because they feel frustrated)

Healthy to agree to disagree and save your positive energy for people who are in a receptive frame of mind.

[shell lagoon]

I got the first usable generations and found out its 38 seconds only. so it's basically made like the discord version. just the longer ones are bad...

[south estuary]

lol, in a way you are right. I am still flabergasted at the output V2 is putting out.

[warm iris]

hi, i have issue with billing

[broken lotus]

hi, sorry about that...
email [e-mail]
they are USA time zones, usually respond right away, (but not positive it is 24hr)

include your login email, and your Discord name
(assuming it's about connecting website and Discord accts)

[neat nexus]

So I came across Mademoiselle from Armentières in the public domain sector and started playing with it. But I'm having a horrible time getting it to go along with me...

[male chorus] Mademoiselle from Armentières
[female vocalist] Parlez-vous?

[male chorus] Mademoiselle from Armentières
[female vocalist] Parlez-vous?

[male chorus] Mademoiselle from Armentières
She hasn't been kissed in 40 years
Hinky dinky parlez-vous

[male chorus] Mademoiselle from Armentières
[female vocalist] Parlez-vous?

[male chorus] Mademoiselle from Armentières
[female vocalist] Parlez-vous?

[male chorus]
Our top kick in Armentières
Broke the spell in 40 years
Hinky dinky parlez-vous

[male chorus] Mademoiselle from Armentières
[female vocalist] Parlez-vous?

[male chorus] Mademoiselle from Armentières
[female vocalist] Parlez-vous?

[male chorus]
You didn't have to know her long
To know the reason men go wrong
Hinky dinky parlez-vous

[male chorus] Mademoiselle from Armentières
[female vocalist] Parlez-vous?

[male chorus] Mademoiselle from Armentières
[female vocalist] Parlez-vous?

[broken lotus]

what is happening in the chiprs?
or what is not cooperating?

[broken lotus]

does that come up immediately, or after thinking a bit?

[neat nexus]

So I don't know if it's on a blocklist or if it really hates the structure I am trying to use.

It ponders it and then gives the message. Usually within a twenty seconds or less.

Usually less.

[broken lotus]

my theory is that it takes almost no time to reject if it's an actual content rejection from GPT...

longer if it is a 'technical' error (ie the prompt is too long, etc)

[neat nexus]

So it doesn't like the words.

[broken lotus]

(based on what I can observe from here)

[neat nexus]

Hmm...

I changed things around...

Mademoiselle from Armentieres
Parlay voo
Mademoiselle from Armentieres
Parlay voo
Mademoiselle from Armentieres
She hasn't been kissed in 40 years
Hinky dinky Parlay voo

Mademoiselle from Armentieres
Parlay voo
Mademoiselle from Armentieres
Parlay voo
Our top kick in Armentieres
Broke the spell in 40 years
Hinky dinky Parlay voo

Still almost immediate rejection.

I'm starting to think it's rejection by content.

[broken lotus]

I see what you want of course..., but I might try simpler formatting... (this would be more chirp-like...)

Mademoiselle from Armentières (Parlez-vous?)
Mademoiselle from Armentières (Parlez-vous?)
Mademoiselle from Armentières
She hasn't been kissed in 40 years
Hinky dinky parlez-vous...

Mademoiselle from Armentières (Parlez-vous?)
Mademoiselle from Armentières (Parlez-vous?)
Our top kick in Armentières
Broke the spell in 40 years
Hinky dinky parlez-vous...

🤔

let me try some...

errr.... what is our style?

[neat nexus]

Buggering 'ell...I'll almost bet money that it's getting hung up on it because of the inclusion of it in Peter Jackson's They Shall Not Grow Old. But it shouldn't matter the song itself is public domain.

[neat nexus]

military chant, tropical house, EDM

[broken lotus]

naturalich

[neat nexus]

The only time I got anything good I got this instrumental...It's so good...I really would like to see what I can make it do if it'll accept the prompt.

[broken lotus]

is 'hinky' a naughty word? lol

[neat nexus]

I wonder if it's a name filter?

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

I mean I suspect that if I did something with say...Walmart in it that might cause some eyebrows to rise.

[broken lotus]

replacing 'hinky dinky' with 'topsy turvy' works...

it reversed your soldiers and woman voices... LOLOL
but close...

[neat nexus]

Trying a substitution with K_nky P_nky

Apparently a word is blocked here...

Those with certain types of hair need not mention it.

I guess....

[broken lotus]

I suspect 'hinky' is a euphemism for sex...

in 1871