📚┃suno-school

[leaden comet]

Hi, hopefully this is the right channel for my question : I'm exploring the new Chirp web app and I was wondering if it's possible to sort posts by "Newest" instead of "Trending" in explore mode

If not, please consider implementing this function in the future

[fallow gull]

Oh wow, Explore tab finally changed. I thought Turkey Time would be at the top forever.

[untold briar]

I still miss Mistletoe and Missing You

[queen token]

are there any known tips/tricks to getting clear vocals without distortions?

[deft hollow]

Honestly, I'd recommend waiting for V3 for that

Look at #🐣┃suno-showcase

[daring lava]

If you’d like to share the specifics of your input (lyrics+style of music), I can try to help refine it

[heady hearth]

Suddenly unable to download audio or video from desktop site on Android. Tried refreshing - Nada. If this is a known bug are there any solutions? I'll keep troubleshooting.

This is as far as it gets

[daring lava]

Are you still having this issue?

[high pond]

hi @daring lava , when can we expect support of chirp v2 command?

[heady hearth]

@daring lava unfortunately yeah. Just did a phone update and cleared the browser storage inbetween tries to no avail.

[daring lava]

So sorry to hear you’re having this trouble. When you get a chance please reach out to [e-mail]. They’ll be able to give you the best advice on what to do.

[daring lava]

Not sure what you mean… could you be more specific please?

[lapis dragon]

Are there any ways for me to run this locally?

[untold briar]

I think they means Discord chirp will get V2. I don't know the answer, but at this point I wouldn't be surprised if maybe it skips V2 and goes straight to V3.

[heady hearth]

@daring lava cheers David. Thought It might have been my VPN desiring an update, which has caused issues with downloads in the past, but no. It's a little strange though, the SUNO drop down menu was offering the options of video or audio download directly and not through a further selection... I'll contact support👍

[daring lava]

Okay best of luck getting this squared away asap!

[heady hearth]

Thanks!

[high pond]

@daring lava I just want to know what chirp model version is used on discord bot, v1 or b2? In announcements I found that "v2 is available on the web app today". But it is not clear for me what version is on discord?

[daring lava]

You can add the Chirp bot to your Discord server. But that’s the extent of locally. I’d suggest checking out the browser version though first if you haven’t already

[lapis dragon]

Ah ok, thanks. I was just wondering if I could run it in python scripts like Bark

[daring lava]

V1 is on discord

[daring lava]

Oh yea, no that’s not a thing

[heady hearth]

@daring lava cripes. Once again wasting your time: secondary drop down decided to wake up after me hammering 'download' like a madman😅

[high pond]

@daring lava do you have plans to update discord?

[heady hearth]

That is to say: ISSUE RESOLVED😁

[untold briar]

There are plans, it's 100% on the "list of things to be done eventually" but I don't have news for you about how soon.

[daring lava]

Yay!!! Glad it’s worked out!

[heady hearth]

😊

[high pond]

@untold briar , I got you, thanks

@untold briar , another one question, maybe do you have some kind of rest API for chirp v2? or going to implement this one in future?

[untold briar]

The only thing I've heard from official sources is "An API is on our roadmap" so tldr some day...

[thin flint]

Very interesting end. Not exactly what I wanted but it was kinda successful, it did fade at least. https://app.suno.ai/song/f1236bd6-0e68-4a17-a220-879d44b4ec56

[thin flint]

Interesting sounds that came out of this one. https://app.suno.ai/song/68304ef9-7c46-4fc2-8608-b097154729bd

[jade mist]

It would be good if V3 would be added to the app after the Vday thing!

[thin flint]

We can only hope.

[honest dagger]

That would be 10/10 publicity move to be fair

[thin flint]

At least put up a beta site separate from V2. The web player with it had some playback glitches though.

[acoustic heron]

Hey.For some reason I cannot dm the bot - any idea what could be causing that issue?

[versed valley]

Welcome! For help, check out Suno Help

[broken lotus]

I have an existing (old) dm with /chirp it seems to be working.... have to hit ENTER twice to bring up the popup...

[obsidian lintel]

my old one works too

[acoustic heron]

A friend of mine also has an old chat that is working fine. But given the how to literally says "dm the bot" I was wondering why it would not work

[obsidian lintel]

he cant get that far it wont let him dm at all

[broken lotus]

🤔

[broken lotus]

mobile or web or app...?

[acoustic heron]

It just says "unknown integration" or "the application does not respond" - I'm using discord in a browser

I just tried to relog / clear the browser cache, still doesn't work

So if anyone has an idea, I'd appreciate any input. It works in the beta channel though

#🐣┃chirp-beta-1 message

[broken lotus]

I have alerted the devs..., I'm sorry I can't help troubleshoot..., seems my existing DM bypasses the issue....

Can I ask why you do not use the website?
The clips are 2x longer and in stereo, same credits....

[acoustic heron]

I just recently discovered this and wanted to explore it a bit tbh, that's all

But thank you for forwarding the issue

[broken lotus]

👍

[obsidian lintel]

i like the v2 stuff on the web site but it doesnt let me have detailed prompts there it fails if i use more than 4 words

[broken lotus]

there is a Custom Mode, switch it on in the Create page...
It has the 2 prompts like Discord: Style/genre and Lyrics, with a Lyrics Generator you can edit...

[marble breach]

Hi, i want know, How extend one music with 0 crédits?

[broken lotus]

you will need to wait a few more hours for the daily credits to refresh..., then you will get 5 more chances...

Have you Continued a clip before...? You know how it works?

[marble breach]

Yeah, i have continued a clip before

I don't know How It works

Look

I want more letters and time

1 minute is...... so small

[marble breach]

I want This music have 3 minutes

Or 4

[broken lotus]

so on the website... next to this clip, find the [...] button.
and choose Continue From This Clip...

turn on the Custom interface in the Create page... you will see the thumbnail of that clip loaded below the prompts...

you can edit 2nd verse and chorus..., and Generate a 'Part 2'... (max 60sec)

it will generate 2 more clips labeled 'Part 2', pick the better one and find it's [...] button, and Continue again, from 'Part 2' to make 'Part 3'... and so on until finished...

then on the last clip use [...] and choose Get Whole Song (free, no tokens)

[tender eagle]

so how do you even end a song, when it ends, it ignores the outro lyrics hah, but when it includes the lyrics, it doesn't END ><

[tender eagle]

wait so you don't have to make whole song and then continue :\ damn

[shadow timber]

It doesn’t really hurt anything if you do, get whole song doesn’t cost credits

[broken lotus]

That is not at all what I wrote. 🙄

[tender eagle]

That's exactly what you did.

[tender eagle]

You said you can continue from continuation 2 and it'll make continuation 3..

I was under the impression that I had to take my continuation, make the full song and continue from the full song. Which if what you said works I do not.

[neat nexus]

https://tenor.com/view/bobawooyo-dog-confused-dog-huh-dog-meme-shocked-dog-gif-16713203299056947073

Whaaa---

[tender eagle]

I didn't know you can continue and then it'll combine multiple eg. More than just first and continued

[tender eagle]

They said you can choose the best of two from the second generation and then choose continue from that for the third generation.

[broken lotus]

ahh, sorry to misunderstand...

it should work either way...
but normally I do not Get Whole Song at each step..., there is no charge if you do...

[tender eagle]

Origins -> part 2 -> part 3 -> whole song. Versus how I have been doing it original -> part 2 -> whole song -> part 3 -> whole song

[broken lotus]

can help to hear where the song is going...

[tender eagle]

Yeah it works except it uses tokens lol

But I mean I assume you can hear where it's going just by following your parts before you combine the whole of

Does it not take tokens to combine a whole song I thought I saw it deduct a few credits

[broken lotus]

it's not suppose to cost any credits...

[tender eagle]

Oh, hmm. Sometimes it takes a moment for the credit update. Maybe I was just seeing the continuation generation deduction when I made the whole song and noticed it after

I always assumed it did

[neat nexus]

You create your first clip. This will be in a set of two. You choose between those or try again until you're happy.

That will be your starting point. You then continue from this clip. You adjust the lyrics to be at the point the first clip you are using ends. You add more lyrics to fill in the time.

You generate. You get two clips. You pick the clip of the two that matches your previous liked and continued from clip. You do this until you reach the end of the song. Once you have a satisfactory ending you pick "get whole song" from the clip that ends best.

This creates the song.

Now it is possible to hear how a song stitches together as a whole from the clips you have generated already even if the song has not ended. This allows you to hear if it is hitting the flow you seek.

Generating new clips costs credits. Stitching the clips together does not and should not do so.

[tender eagle]

Oh, I assumed it used a small number of credits just simply because I assumed it was also using AI for the stitching. Because if you take part one and part 2 and put them side by side in for example audacity, they definitely don't run together well

I didn't know if they had a different model running that

[broken lotus]

also sometimes I think we have to refresh the page to see the accurate number of credits..., like the webpage will fake the numbers, not grab the real number all the time...

[neat nexus]

I blame browsers for that. Some browsers really lag on some refreshes.

[broken lotus]

yeah..., the stitch works pretty well... I think.

I tried to do it by hand and... it was weirder than I expected, had to carefully crossfade each clip and trim a tiny bit... 😓

thankfully the stitcher is free... 😅

[tender eagle]

Yeah you have to align the best on the 4th and 1st or it can throw it all off and then you have to make sure it doesn't have a beat that is not too distinct. The merge it does seems to almost flawlessly. Even sometimes when lyrics are cut off it'll just go and remove the bad spot

[low remnant]

I was having terrible luck with auto joins yesterday, on Get Whole Song. Thankfully they were all in the chorus, so I was able to copy from one of the other instances and smooth it out. But usually it joins so perfect that I don't touch it.

[tender eagle]

I was just having trouble getting it to end the damn song. It kept making even the sections I had marked as outro. Just keep going

[broken lotus]

I think it gets confused by [Outro] followed by more lyrics...
Like it's not sure what it's suppose to do so it just keeps going...

try [Outro] without lyrics...

but you also might need an [End] or a [Fade to End]...
these aren't always reliable...

SOmetimes it settles into a 'loop' and I think it's easier to just do a fade out in post-edit...

[tender eagle]

No I did several like that. I did a end, outro alone, outro with lyrics, it didn't matter. I finally got it to do it after like six tries, half the songs had no lyrics at all. Just 1 minute of music

It really did not want to end that song. It was also messing up really bad with sound effects. I put something in a (and it would ignore the entire lyrics and just be blank for a minute and 20

[low remnant]

It all seems to be shooting craps to me, but I had some luck with [Fade to Silence] yesterday. If you haven't gone there yet, might be worth a try.

[cold zenith]

told you that the other day ^^

[broken lotus]

I stole you advice and passed it off as my own, lol

[pliant cave]

The biggest issue with the current iteration of V3 that I've found, or at least the one exposed on the V-Day site, is the model not stopping at the proper point. It either loops back to an earlier part of the song or just goes off the rails and hallucinates. I assume the team is well aware since it seems to be a pretty common bug and probably one reason it's not quite ready for general release yet.

[vernal meteor]

Yeah pretty much anything you generate will flip out and hallucinate in the end

which is awesome but not particularly useful 😄

[tender eagle]

I'll try that one next time

[untold briar]

As far a problems, I think hallucinating-too-much-song will eventually be much more preferable than than ending-too-quick, since we will be able to use new features to choose the time song to continue from, so those clips will still be perfectly useful. But a ended-too-soon clip may be more awkward.

[fossil nymph]

that kinda sounds amazing. Tbh some of the best ideas I've gotten from hallucinations lol in terms of melodies and chord progressions

[fallow gull]

In my first generation, I added a [Build] before the chorus which caused the song to get more aggressive. Now when i continue, it wants to stay aggressive. Are there any tricks to get back to the style of the first verse?

[untold briar]

You can try using less-aggressive music style terms in the style field in the continued verse. I'm not sure what's best for the opposite of a [build] off the top of my head. How long as the aggressive section in seconds in the first clip? If it's longer than a minute it will especially hard to change course, and still hard close to a minute.

[fallow gull]

It is just the chorus of the first 1m clip, so it's about 20s.

[untold briar]

I'm surprised it doesn't return to normal at the next [verse] tag if it's only 20s, but the ways of the AI are mysterious. You could try [bridge] or other structure tags maybe.

[fallow gull]

The beginning was sung with little instrumentation, but it builds, then it switches to practically Deathcore for the chorus with epic drums.. Such things happen when using metal ballad in the style.

[untold briar]

On thing that may be worth a try is 'maximum spam power' style. That is, whatever style tags you think are most likely to calm things down, repeat them over and over again in the style prompt with commas. I would not even use 'metal' in the style just pure descriptions of slower more relaxed chill music in that genre, or just generally.

If nothing you do seems to change it at all though, you may have to give (at least for now, maybe with later features to control continue time it may be different) on the clip. A lot of what you get is determined purely by the previous clip so it can be the case that no prompt in a continue will change it, that is possible.

[fallow gull]

I'm biding my time until v3 anyway.

[deft girder]

anyone know which genres produce the highest quality?

[broken lotus]

RnB. It's all bass and very sparse treble used for percussion. The audio stay clean for a long time.

RnB studio recording, melodic bass, muted guitar, sultry female vocal

[pliant cave]

The developers themselves have mentioned 'EDM' and 'Bluegrass' as genres the model is particularly good at, at least for the v1 and v2 iterations.

[fallow gull]

I think you can randomly hit a good song in many genres. I personally find it very difficult to get decent Folk and Rock songs.

[eager topaz]

Quite impressed with VDay

[untold briar]

For folk try "indie folk" that's been one of my go to styles. https://app.suno.ai/song/995ef6c7-5825-4520-bfa2-c0bb0125163b/

[shadow timber]

I find adding “primitive” brings out interesting results in folk, rock, and blues

[untold briar]

I don't think I've tried primitive, interesting

[warm oak]

I used to get 6 or 7 decent songs a day, now I'm lucky if I get one a week. I'd love to wait for v3 but I paid for an annual so I keep at it. Tremendous waste of time and money right now. I find a lot of the vocals are bad. Like, really pitchy. The list of problems goes on but I'll leave it at that for now. Whine, whine, whine. First world problems.

[oblique prairie]

Does tunecore allow to distribute songs that are created using suno AI if I add credits

[broken lotus]

Please ask in only 1 channel.

#💬┃general-chat message

[oblique prairie]

Ok sorry my bad

[empty anvil]

I am not able to end the song in the second part of the song . Even after deleting the first part song lyrics it repeat the same deleted song😓

[charred lake]

Does anyone have a catch-all tutorial on mastering Suno outputs? I feel like i could put something together, but don't want to if its something that already exists. @vague girder Would you know if documentation like this exists for the Suno community?

[golden pasture]

I have not found anything like this,

[broken lotus]

I think it would be appreciated by many!

[neat nexus]

Unfortunately there isn't any at this time. Also were you trying for me or that other guy?

If someone were to put something together it'd be greatly apreciated. Might even be a project for someone on the unofficial community wiki. If something like that existed and I was made aware of it, I'd definitely be linking to it (and trying to follow it myself!)

[honest dagger]

Is the v-day web just staying open for the foreseeable?

[main breach]

I came in here to wonder what the issue was -- I was generating some fantastic stuff without issue before, now I feel like I'm wasting credits and getting a lot of "corrupt" kind of results (mangled vocals, completely wrong genre like generating polka music for some reason?), etc. I wonder what the issue is, because I've wasted quite a few credits at this point.

[tired ginkgo]

can anyone tell me how to use bark?

[warm oak]

I don't want to demean what Suno does. It is still miraculous at this point but it seems to have taken a step back recently. I have ten more months of subscription to play with. I'm sure it will become even more amazing in the near future. Right now there are too many growing pains

[main breach]

Yeah I can't demean it at all -- I think Suno is pretty awesome! And I can't wait to see where it goes. I just wish I wasn't generating variations of polka music 🥲

[warm oak]

Polka is an untapped musical style these days. The world needs more polka!

[twilit mason]

Like "driving Suno crazy" using "{}". That's not "occasional".

[wary plover]

Hello guys, was removed the download mp3 option from the free tier of suno webb app?

i've created one more song yesterday, and i can't download it, or ani other old song. the option to download the music (or video) is gone. ¿something appened?

[karmic tinsel]

I want to share an unlisted genre that is working well, and a language tweak. The genre is Fado which originated in Portugal and is still common. The language tweak has to do with Portuguese. When you specify Portuguese in Suno you will get the Brazilian dialect most likely because it has the largest amount of speakers. I found that specifying European Portuguese will result in Suno using the dialect spoken in Portugal which is awesome. I have not checked other Portuguese dialects yet such as Angolan.

[twilit mason]

In my case, the simple fact of using less colloquial Portuguese was enough for Suno to use the accent of the country Portugal.

Facto # fato

[karmic tinsel]

Oh, good idea. I did a Fado and specified just Portuguese and the result was Brazilian. Changing it up to European Portuguese did the trick for me.

[twilit mason]

There are some words that are spelled differently in Portugal and Brazil (facto # fato). Inserting them in the lyric will help.

In fact, I think they're still written differently.

[runic dock]

😀😀😀 You did it 😀 congrats . What is the title : Riff Rumble?

[twilit mason]

Nameless😁

[deft girder]

thanks! ill give it atry

[deft girder]

thanks for this too. ive had good luck with modern pop

seems to have trouble with rock the vocals and music start to blend into noise after too long

[slim moat]

if I already have lyrics and want to create a song with music, how to do that?

[broken lotus]

use the Custom switch in the Create page

[exotic fiber]

sounds like {} is a negative prompt activator, will make some tests!

Edit: nahh.. false alarm.. xD

[fierce orchid]

Which prompt do you recommend to create a softer verse of the song and without percussions?

[final hazel]

a cappella

[fierce orchid]

like this?
[Verse]
[a cappella]

[final hazel]

in music style

[fierce orchid]

Thanks, but it's not working, there's always percussion. Can I try something else?

[versed valley]

Welcome! For help, check out Suno Help

[wary plover]

Again guys, do you know if appened something with the donload mp3 feature?

[rose ore]

Now that is a good bend

[peak veldt]

@wary plover It looks like it might have moved... I found it under here now...

[rose ore]

Use "folk" in the genre prompt

Or folk singer songwriter

If you need to add drums later in the song, like on parts 2 or 3, or whatever, add drums to that, put backbeat next to verse [verse, with backbeat]

[exotic fiber]

Try something like "acoustic pop unplugged" on the style and take out also the instruments u dnt want, like no [Verse,no snares]

[rose ore]

unplugged will still get you drums, but the vocals will be really clear

hang on, I did this the other day, will look up my genre prompt, brb

[exotic fiber]

[Verse,no kicks] or [Verse,no drums]
Lyrics....
[Verse,Acoustic guitar solo]
Lyrics....
[Verse]
(Acoustic guitar and vocals solo)
Lyrics....

try it

[rose ore]

interesting

[exotic fiber]

doesnt work all the times but usually do the job

[rose ore]

So the way I got this one was with "upbeat energetic female, pop country, folk"

[exotic fiber]

also, we all have to keep in mind that as genres is important, the format you present the lyrics also adjust the genre.. for example, if you want to make a hipo hop, use a lyriucs with the beat of a hip hop music... if you aregoing to do pop, do it with pop.. egvery genre has a different style for music writing, use that your favor

[rose ore]

It depends. Most pop music, regardless of genre is written 3/4 or 4/4. So you've got some flexibility there. Suno tries to be context sensitive though, which can really mess you up if you're stretching genres.

But yeah, thinking about timing, so important

[final hazel]

It does work, but sometimes you have to do it a few times... Don't be affraid to burn some credits to get the result you want.

[exotic fiber]

also when u want to mix genres.. like a pop with a rap interlude

you can use like

[Chorus]
Lyrics adapted to pop singing

[BRIDGE, RAP INTERLUDE]
Lyrics adapted to rap

[VERSE]
Lyrics adaspted to pop

also in the style, add, pop, rap, pop rap,

[exotic fiber]

suno looks for patters in song the model has learned, so, if you make the model to look for sertains patters on styles or bands or singers u love, making the ingluence on how that type of music is written, you will get better results

we need to push the model to whatever we want to get... and thats with promptinmg

and how suno promptinmg works?

Lyrics and song struture

[rose ore]

and breaking it! 😄

[final hazel]

https://tenor.com/view/ragepc-destroy-anger-gif-11869186

[rose ore]

hehehehe

so much yes

[exotic fiber]

structure your song as aBad Bunny reggaeton hit, intro, buildup, pre chorus, chorus, verse, chorus, poutro... with the same amount of verse and words ion each verse, and i can assure you if yopu put reggaetonm in the style, you will get a great result

[rose ore]

nice

I haven't tried that

[exotic fiber]

yep, somethime people says: the app is working worst.. but, no its not.. is just we cant put text on it and hope everything sounds perfect... we need to guide in everyway the model to reproduce what we want (kind off)

[rose ore]

I'm frustrated with the last update that constricts the meter calculation to four lines

[exotic fiber]

nah,.. u can also play with that...

[rose ore]

Oh yeah, I have been

[exotic fiber]

look, im gonna pastre here an example of an structure of a song the way i do it

[rose ore]

go for it

Question: Is there a reliable way to get longer intro leads? [intro] does not work consistently on english language tracks

[final hazel]

[intro]
[short instrumental]
works for me most of the times

[exotic fiber]

Style:
Alternative Rock Rap, Intense, male vocals, male duet

Lyrics
[INTRO]
(Harmonies, layered and haunting)

[VERSE, Bryan sings]
In the echo of silence, Lynn Cole finds her voice (In the shadows)
Amidst the chaos, making music her choice (Rise above)
With Suno as her canvas, painting dreams so loud
Her heart beats a melody, piercing through the crowd

[PRE-CHORUS, buildup,Bryan sings]
(Whoa) On a journey, through the sound (Echoes calling)
(Whoa) With Suno, where her truth is found (Light in the falling)

[CHORUS, DROP, powerfull,Bryan sings]
She's breaking the silence, with every note she plays (Breaking the silence)
In the rhythm and the rhyme, she finds her way (Lynn's defiance)
With Suno's guidance, she crafts her songs
A symphony of freedom, where she belongs

[VERSE,Toby raps]
In the lab with Suno, Lynn's laying down the track
Finding beats, spitting truths, never looking back
It's more than music, it's a revolution
Every lyric, every chord, her own solution

[BRIDGE, Buildup,Toby raps]
She's not just dreaming, she's creating
With every barrier, she's breaking
Suno's the stage, and life's the script
With Bryan's melodies, she's equipped

[CHORUS, DROP,Bryan sings]
From the whispers in the dark, to the spotlight's glow (From the shadows)
Lynn Cole's journey, on Suno, a show (Her story unfolds)
With passion as her guide, and music as her friend
Her songs are her legacy, from start to end

[OUTRO, Bryan and Toby together]
(Bryan) With every chord, she's rising, reaching for the stars
(Toby) Lynn Cole, on Suno, breaking all the bars
(Bryan) In the harmony and flow, her spirit flies
(Together) On Suno, where her music never dies

[END SONG]

try with this, its not going to produce a perfect tyhing at firsdt, but it works

also, its a little bit overcharged because i tried to add a lot of things to show examples of what i do, but in just one track xD

also, if suno starts to do crazy things, remove some things from lyrics

like toby raps, bryan sings... etc

[rose ore]

oh cool, thank you

[rose ore]

also love you pfp by the way

[final hazel]

Thanks.. it's a 3D render of me 😄

[rose ore]

nice

[fresh narwhal]

some one a know why im stay in Subscription
Basic Plan (Free) i have added discort account to websie etc

[exotic fiber]

formthis u can add a 5th line to the verse or chorus, and sometimes the model will pick a song with that extra line, or, it will create it using something like a short buildup

also, adding [Pre-chorus] and 1 or 2 lines of lyrics before the chorus also works.. (the way i prefer because we have more control over the extra lines on that especific part, we can try to assing a buildup, a shor drop, or make it intense, or sad, or even weird...

[fierce orchid]

Thank you guys for the tips 🙂 actually I'm trying to recreate the softer and more melodic part of a bachata which is generally without percussions

[fossil urchin]

does the song title have any impact on the generation? i've been trying to generate noise stuff but it can be hard to wrangle the ai to do that

[lyric umbra]

Hi, In almost every song generation, there are strange cuts at the full minute mark, exactly at 1:00, 2:00, etc. Have you also experienced this? Do you have any suggestions for a workaround?

[broken lotus]

the title doesn't effect the generation, but it can trigger the censorbot.

[fossil urchin]

cool ty. i felt like it did but good to know

[broken lotus]

there is some bleed through between Style and Lyrics.

[fossil urchin]

yeah i know that lyrics affect it for sure

i have to not put any lyrics, just [instrumental] or a single line of gibberish to get what i want

[broken lotus]

I've been triggering hallucination songs with
[atmospheric vocalizations]
😃
potato-potato

[fossil urchin]

cool will keep that in mind

can you really call them hallucinations if it's exactly what you asked for

[shadow timber]

When was there an update? I was getting long verses last night

[rose ore]

I wanna say... sixish weeks ago?

It'll still do long verse

but they don't flow the same way

and it's frustrating

[shadow timber]

I probably didn’t notice since I joined shortly before that

[rose ore]

it's all good

[runic dock]

Hi all. It is a bit hard to do continuation when suno has stopped in the Middle of the chorus or in the Middle of a sentence. I use to start, in the continuation clip, by the remaining part of the chorus or the end of the sentence. It needs a bunch of trials to get something acceptable. Any advice ?

[full kettle]

Does anyone have any suggestions of encouraging it to continue with vocals when it gives a completely instrumental rendering that you like?

[cold zenith]

click continue

[full kettle]

If I put in lyrics for continue it still ends up being instrumental

[cold zenith]

hum put [break] and put the lyrics after

@broken lotus i can't find the answer on the FAQ , who's the composer ? ( when u want to put it your music on spotify etc, they ask for it )

[broken lotus]

you...., but a title like Producer might be more accurate?
(You don't have to fill in them all, it's just about their search engine... is the composer someone famous with other recordings... Like Beethoven...? no, so ok to leave blank)

[wary plover]

ok, thanks. I'll try to find it.

[wary plover]

noup. I cant found it. I've entered to the specific song page, and noup, i cant find any download option on one of the tree dots menus, or something similar. I just have the free tier of suno

[final hazel]

@wary plover can you post a screenshot of your menu then?

[wary plover]

sure. first, a screenshot of a specific song (part1) page. with closed menu

and a screenshot with the "tree dots menu" opened. tell me if the capture is wrong please. im blind, so i don't know if im take the correct screenshot 🙂

[fallow gull]

The last 2 options are "Download" and "Report"

[wary plover]

but not are grayed?
when i move with the keyboard through the menu options, the latest download dont appear. hum. i'll try using a OCR instead the screen reader to check if i can find the options

[exotic fiber]

I made this kind of a meme video, the song was made using Suno... Managed to make a brazilian sing in spanish and brazilian portuguese :p

https://www.youtube.com/watch?v=N92Z7mlOKYs

[wary plover]

thanks guys, i found the options. I've used a other navigation tools of the screen reader because yes, are on the menu, but after a separator that brocken the screen reader navigation. but ignoring that, all things are well. again, thanks for your answers!

[cold zenith]

Okay thakn

[cold zenith]

i tried to let it blank but they gives me no choice to fill it

[broken lotus]

https://app.suno.ai/song/e2347d1e-8cf5-4e78-a240-090506afc829

[wild lark]

Does anytype of laughter work in a song? Like a creepy laugh

[broken lotus]

if you can find a soundfx creepy laugh to mix in, you will save a bajillion credits trying to get the bot to deliver one you like, within a song you also like....

[pale night]

did this work? can you share the song?

[iron thicket]

I am using Suno AI since several month now. Since V2 i have noticed real improvments on sound quality, but not for every genre. Maybe explained by the songs used to train the LLM model. It seems to me that Folk genre is ("significantly") higher quality in general. But some genres like Nu Metal are very muddy and over compressed. In addition, i am not convinced by techno (barely usable at that stage) or EDM tracks generated so far (seems very 90's 80's, even when using 2000's reference in the prompt). What do you think ?

[meager mason]

i assume, there is some sort of LLM trying to map the "style" prompt input to a specific genre the AI has been trained for. If they used for example Spotify classificiation data, like i did to build a classifier, before it become a TOS entry that using data to train AI, there is much data being returned if you search for example for EDM... as they seem to threat genres hierachically (Spotify did at some time at least), when searching for genre:"house" i got also deep house, etc. and generally speaking, they also have completely empty genres on their db, no artists no songs to be found.... and sometimes songs even switch the genre, as the datasets centroidds per genres have had significant changes etc. So if you seek for example genre:"EDM" on Spotify, you might find different genres. I personally treat EDM not as a genre, its more like a genre-group, everything primarmly made for dancinng, being electronic - so i would - as oldschool-guy that started with Trance And House (which have been referred initially as Rave, even count House , Trance etc. all to EDM... but thats me, Spotify and Beatport etc. all treat the genres differently. For Beatport, they often simply take what u enter when u set up distribution fdor a new release, unless its completely off the genre u renter, they take it... progressive electro house for example u can tag EDM on Beatport, they dont change it, but if you tag it like Trance, they will change it, if they check it, as its too much off the classification - but they don't check all tracks, at least not, before u got some activity going on on your release. Same with datas sets, one might put a track into "tech house" or "tech-house" or "UK Tech House" (which is agenre on Spotify for example). It all depeends from where ur data is coming. Spotify had over 6000 different genres, last time i checked my data, some of them i couldnt classify ssongs into myselfd, they are obscure and really niche... but none the less,, until we get a full list of actual genres the AI really knows including their correct writing to use the genre_x helper listed in the docs, its hard to actually get really what we want on majority of results.

One thing that might help u @iron thicket : Be more specific genre wise, instead of EDM, use something more specific , maybe try genre names from Spotify and/or beatport. For m, what helped me to get more like that bouncy tech house im lookingf for, is putting my style prrompt into " like "bouncy tech house" - that worked well better than without "" - where it often got bnouncy songs that aren't tech house at all,. So instead of EDM, try maybe "Progressive Electro House" or maybe something like "Sky Room" (an othjer EDM Subgenre being defined by Spotify, to see if that actually works here)... just a few tips... add a precision, or maybe try to describe the style, instead of using EDM - so the LLM can infer what style u want better... or try genre_EDM instead of just EDM, as it infers the genre it dfoesnt necesserly take your input as genre, genre_YourGenreName actually should work as hint that this is the genre you like... maybe genre_EDM year 2020 or something.... ifd they have the release date on training data , or the year at least ^^

[exotic fiber]

on that example i have added thiings that i use on mines... you can start trying to use things from there and try what works for you.

[real crow]

Thanks for your insights on fine-tuning Suno AI with specific genres and style prompts. Really helpful!

I've noticed a significant improvement in sound quality over the past 5 months, especially in seamlessly chaining tracks together. By the way, I found something cool: adding [Duets] or [DUO] in prompts lets the AI alternate between two singers. Thought you might find this interesting too! Feel free to PM me with some tips

But it takes sometimes maybe 1 or 2 times of 20

[meager mason]

i wish there would be some information, about the prompting mechanics, like, if we can actually apply specific weighting to words in the style prompt - or if the weighting is part of the randomizing process to avoid having same output being sent to different users). Hard to tell the influence of a prompt change, as the results are so different on the same prompt input.

[pale night]

Agree with EDM and Techno, subpar outputs

[exotic fiber]

https://musicmap.info/

i love that web...

[meager mason]

i love it too,!

if ure more intrested for spotify's categorization, u might be intrested in this: https://everynoise.com .

its just static as of end of 2023... as the creator did quit his job with spotify and thereby has no internal data access anymore to keep this external site updated...

[meager mason]

i just tried edm... jeesus.... thats horrible output 😛 - i thought in my niche itrs hard to get some good results, but edm.... thats worst genre i tried so far...

[rose nexus]

I know when v3 is coming out

yall are gonna go nuts this week

just you wait!

its gonna be worth it

[eager topaz]

Can't wait!

[graceful cliff]

how do you know...I don't even know

[rose nexus]

v3 alpha is coming out this week

but don’t know the day of the week, might change date who knows

[graceful cliff]

🤔

well, not Monday 🙂

[rose nexus]

its not midnight yet

there is a chance 🤣

[lusty tree]

is it possible to do instrumental songs in chirp? Or does it have to have lyrics no matter what to work?

[limpid spade]

Try instrumental in the style or piano instrumental etc

[cold zenith]

I tried to clean my music, when i got my ears, the quality is good, but in my car, it's totally different. The sound is bad. Someone got an idea ?

[iron thicket]

Thank you @meager mason. Yes agree with your comment. Yes I used the term EDM in my post but to your point I was using more detailed style descriptions like the one you’ve mentioned when asking for track generation. At this stage, I am still not convinced by the songs produced by Suno on EDM/Techno/melodic techno/dark techno/Electro House ….so electronic music in general. Maybe some times to get idea for a loop but nothing more for this genre. As I said the results at least the one I’ve got were very very muddy or were very old school out dated for EDM. While I’ve got tremendous results and quality tracks with folk for example.
Maybe suno needs more learning data on EDM/techno

[meager mason]

i had some nice results using this combination as style prompt: tech house deep house bass house

[iron thicket]

Happy to listen to one if you do not mind to share here or in DM

[meager mason]

check your DM

[knotty kite]

poetic dibs😐

[neat lake]

Hi guys, using a screenreader and I can't sign on the webb app, since using my keyboard only says blank blank when moving the cursor, this wasn't the case before. Any tricks on how to locate this sign in dialogue?

Ok I managed to sign in somehow with some random key combination, but there is no download button for songs, hmmm

[austere violet]

not even the 3 dots for the menu thingy??

[neat lake]

Nope, all the other options are there, but for some reason not the download and it doesn't seem to be a link or button to click either

[jade mist]

I don’t believe V3 will come this week untill I see it

[rose nexus]

Just you wait 🙂

[austere violet]

https://tenor.com/view/staring-mario-the-super-mario-bros-movie-waiting-to-drop-checking-the-sink-gif-27210847

[jade mist]

You will get 100000000 euros if true😂

[austere violet]

https://tenor.com/view/dr-evil-austin-powers-mike-myers-gif-6548426555310453841

[neat lake]

hi friend, I am also using a screenreader and can't find it, can you tell me where you found it and what it's called now?

[shell jungle]

imagine if it was XDDD

[charred lake]

EDIT - I figured it out, yes it's legit

どんな暗闇も越えて、昇る太陽
The sun rises beyond any darkness

Can somebody here help me verify the authenticity of these Japanese lyrics? I asked GPT to create me a Japanese rap song about coming up, with the mainline repeated as typical per drill wrap

https://app.suno.ai/song/3efc5a8d-2144-443c-a6a3-18213b45ab3f

[fiery sinew]

I really really wait for V3. Technical quality is such a crap now.

[charred lake]

To each their own, I think it depends on prompting and being very oddly specific (w/ help of GPT) ^ I'm loving my result here, just remastered the track and the audio is amazing imo. Will be streaming on DSPs very soon, just submitted for release...

https://on.soundcloud.com/Jtysi

[half dragon]

Hi cool song... just wondering, those tags in the lyric... did you add it or Suno tags it for you? for example: [Intro - Tension-building shamisen melody into heavy 808s.]

[cold zenith]

I tried to clean my music, when i got my ears, the quality is good, but in my car, it's totally different. The sound is bad. An idea ?

[exotic fiber]

Exported audio files from Suno are at 128 kbps (bitrate) + the noise the app introduce into the audio... To have a clear audio, you must process the tracks... and there r some ways to do it. Splitting the track into stems, then process and remaster them individually, or remaking them on a DAW.

[cold zenith]

this is what i did, splitting, and as i said. The sound is good with ears, but not in my car. This is the problem i'm having.

[exotic fiber]

In simpler terms, think of it like this: A 128 kbps music file has less detail because it's been squeezed to be smaller. It's like a low-quality photo. A 320 kbps file keeps more of the music's original details, making it sound clearer and richer, like a high-quality photo.

Remastering a track at 128 kbps might clean up some noise and adjust levels, but it won't really improve the quality. It's because you can't add back details that were lost in the original compression to 128 kbps. It's a bit like trying to make a blurry photo clear—you can make it look a bit better, but you can't bring back the details that weren't captured in the first place.

and, to get a decent sound quality on your car, or any home sound sytem, you need at least a 320 kbps file

[shadow timber]

I really like this translator for working with Japanese lyrics since you can see the romanji and also the meaning of specific words and how they fit together https://www.romajidesu.com/translator

[eager topaz]

Vocals have the poorest quality with Suno. I have very few tracks that I would call "release ready" even with extensive editing and cleaning up. Things like missing words in vocals, poor endings, poor beginnings, etc. For me, Suno is fantastic at what it does, but at best, at present, produces tracks that are demo quality and great for home use.
Instrumental sounds such as yours produce an acceptable quality, mostly, but it depends heavily on the style.

[shadow timber]

Suno generations just spinning?

[torpid elbow]

I’m sometimes satisfied result on lyric songs. But instrumental song and edm w/o lyric songs are so letdown. My biggest concern, too monotonous it Any idea for solution

[eager topaz]

No solutions that I know. I experiment a lot and use a lot of credits.
Yesterday I wrote words for a song and created in Suno. The 1st chorus was totally beautiful, one of the best I've produced in Suno, but another 3 choruses were all different to that first. This is another problem with Suno. Verses and choruses sometimes are not the same backing and melody. I haven't yet found a fix for that other than with the chorus, rip to stems then copy and paste the best chorus.
Suno is still a lottery for me, but the successes are amazing and I am happy with that. 😀

[torpid elbow]

I tried stems split then copy and paste the best ones. It actually work but sometimes its weird beat and vibe too much different style with previous part. I used audacity. Btw Any suggestion for stems split and audio edit tool?

[eager topaz]

I use RIPX DAW Pro, which does a great job, but it really shows the low quality of some instrument tracks. RipX is so precise that you can take a very small piece of audio, say the letter "T"and copy and paste it. RipX DAW has a free trial if you want to try it. Not cheap to purchase, but worth it.

[meager hemlock]

UVR5. The best tool for splitting. RIPX may have more features, but UVR5 is free, and i think overall results in cleaner higher quality vocal/instrument splits as well as instrumental to 4 instrument splits.

[meager hemlock]

I agree mostly on this. However for vocals (not all but some are good enough) you can split the vocal stem, then clean it in a DAW (de-noise, de-crackle, etc.) (sometimes even autotune it) . after its clean put it through RVC, with a well trained RVC voice model it will output clean (cleaner*) vocal.

or just wait for Suno V3 😂

[eager topaz]

Let's hope it's soon and addresses all of the issues 😩

[crystal sapphire]

Can someone explain this? https://app.suno.ai/song/2a5aae1c-4bbd-4205-b656-612f9f0e374e

[exotic fiber]

https://tenor.com/view/roll-the-intro-ignace-aleya-insert-intro-play-the-intro-present-the-intro-gif-17789073

[jade mist]

It is wednesday and no V3. It’s easy to claim info as true and provide a fake image as we have seen before

[fathom sluice]

Sounds like the AI had their head stuck into other stuff and just decided to hallucinate up things, seen it happen a bunch of times, sometimes the hallucinations are actually nice.

[runic dock]

I have the same issue. Most of the time the chorus will be fine and identical but Verse become worse . You will spend lot of crédits trying to have the quality of the first Verse. I often give up after several trials. But in few cases it works perfectly

[neat lake]

Well, since everyone's talking about the issues they have with the results of their Suno productions, I find that often the songs are monotone, like the AI finds one pattern it likes and it will never switch anything up, never surprise me. This is what really sets it apart from a real artist. It doesn't do anything interesting most of the time, even in repetitive genres like trap it could honestly do more.

Recently it's keep generating instrumentals without vocals as well, though I don't prompt it it to do so

[full kettle]

https://www.youtube.com/watch?v=Ak0v-RqXkcI&pp=ygUHc3VubyB2Mw%3D%3D

Not sure if someone posted this already, but is this a legit sample?

[pine turtle]

Is there somewhere a list of keywords to incorporate into the lyrics to tell the model what we want?

[wet minnow]

yes here : https://www.notion.so/suno-ai-wiki/4e24aa838bf84fffaf383b1f9ed50748?v=755ec8954abd4e3f9f5c899950f0f315

look at "metatags"

[meager mason]

its funmy, the best reslts i always get if i icorporate created words into it -laugh-

[pine turtle]

Thanks, I'll put that info to use. I'm trying to make Suno create an anime opening j-pop song.

[half dragon]

Is it possible to tell the model to generate female or male voice specifically? so when we continue the song, we could have a duet sings separate verses...?

[broken lotus]

is is possible but not reliable..., it will soak up all your credits trying to get it right...

but as a novelty... try 'male-female duet' in the style...
and give each section a singer's name:

`[Mary] You say potato and I say potato

[John] You say tomato, and I say tomato

[Mary and John] Let's call the whole thing off!`

the voices will switch, but it's totally random if one is male and one is female, or if they stay with the right names...

the more ambitious the duet... the more options for it to randomly maess up....

[stray pebble]

when continuing in the middle of lyrics, do I write the rest of the lyrics or the whole lyrics again?

[broken lotus]

sometimes one, sometimes the other...

I (anecdotal) think it might have to do with how fast the words are sung... a slow verse can just start where it left off... fast lyrics might need a little lead-in of overlapping lyrics...

You might need to try it, hear what the continues are doing, then try it the other way...

Even with my theory I still guess wrong sometimes, and need to switch...

[vague light]

This site is incredibly helpful, thank you for posting this

To anyone that edits the site, if you want to you can also add Hardcore Punk to the genre list under Rock, I use that one quite often and it works really well for heavier punk songs.

[half dragon]

I think you can update it as well... Never tried but the instruction on the page says so - need to have a login

[warm oak]

C'mon V3! Saving my credits just for you.

[vague light]

Ah mkay, thanks

[jade mist]

I will not trust some random guy with ai leaker as username. He will get 100000000 dollars if it comes before friday.

[limpid spade]

I'd be surprised if that troll account wasn't suspended for rule 3.

[half dragon]

Say... I like the 1st generated clip, like the beat, the mood, but it ends erruptedly in the middle of the chorus... if I chose to continue... would my whole song result be affected by that? is it smart enough to finish that chorus when patching the parts together?

[limpid spade]

Mixed results. Some of mine have continued well, some repeat a line which you can edit in Audacity or whatever. Others miss it completely

[pine musk]

Hope one day, we might be able to have duets and ensembles, and maybe their specific vocal types (soprano, mezzo, contralto), genders, accents and ages could be saved.

Especially for us who want to do more ballad and broadway-styled works

[jade mist]

Can a person with actual info tell if ai news leaker is lying or not so we don’t get disappointed at the start of the next week?

[final hazel]

https://tenor.com/view/meerkat-looking-sneak-out-peeking-checking-out-gif-16965207

[crystal sapphire]

Sounds like average Suno audio quality I'm used to hearing, but I haven't seen anything about V3 having been released. That's probably V2.

[shadow timber]

I think it’s a Valentine’s Day song which makes it v3

[crystal sapphire]

Was it ever confirmed that V-Day was version 3? Otherwise, the song didn't sound any better than the audio quality and results I can get now, so I hope that song's not V3. if so, there doesn't seem to be much to look forward to.

[fallow gull]

Yeah, its in the JSON response:
"major_model_version": "v3",
"model_name": "chirp-v3-engine-d"

[final hazel]

https://tenor.com/view/you-see-this-peak-peaking-kevin-hart-tilt-shades-gif-26320836

[neat nexus]

https://tenor.com/view/mib-men-in-black-will-gif-20590125

[vague light]

So just out of curiosity, is the one used by the discord bot currently V2? Or is V2 only on the website at the moment and discord uses v1?

[shadow timber]

Downloads from vday are higher res, as well, and the sound quality is much cleaner. I think the YouTube video you linked to is compressing the sound.

[neat nexus]

Discord is firmly v1 at the moment. The website is firmly a v2.x and the valentine's day site is all there is to say about v3 at this present time.

That having been said, I will again remind people to finish any songs in v1 (the Discord version) because it will eventually be brought up to date and when that happens it will no longer be possible to stitch those old clips. Get 'em done folks!

[vague light]

Oh cool, thanks for the explanation

[half dragon]

About the upcoming V3... is there info anywhere to see what it'll bring or improve? I guess being new here... and I hear a lot about V3... new version is always better but I'm just curious 🙂

[jade mist]

Json doesn’t mean a thing. It can be 1 day or 1 month until V3

[wary plover]

Hello guys, a small question. ¿what is the best technike to tell suno continue a song part when it ended on the middle of the lirix?
Example. you write a song with intro, verse with a chorus, but the intro took a lot of seconds, so the first part ended at the first sentence of the chorus.
so, when you generate the part two, what is the best method to tell suno, finish the incomplete chorus before continue with the next thing that you will put on the lirix block?

[broken lotus]

I think it depends on how fast the words are being sung... (anecdotal)

when the lyrics are sung slow, it seems best to just continue where it left off
when the lyrics are medium or fast, seems better to give it a little lead in, repeat some lyrics that were already sung (1 or 2 lines, or even the whole block)

I still guess wrong sometimes and after hearing it fail, I try the other.

[wary plover]

o thanks. summary, try and fail 🙂

and when you add the missing lines, is necesary tell him that they are part of a chorus, or better don't do it because that can breack your chorus for future repeats?

[broken lotus]

I don't repeat the [chorus] tag..., because I think it encourages the bot to see it as a new section...

[broken lotus]

hearing if it is being sung fast or slow I guess right more than not.
After a while you will get the 'feel' of it, but it's not a hard science.

[wary plover]

okidoki. thanks for your answers 🙂 i'll put them on a try 🙂

[jade mist]

Will website get V3 before discord gets V2?

[final hazel]

I am just a user like you, but logic tells me yes.. Since discord is still V1

[low remnant]

My understanding from staying in a Holiday Inn Express is that the limitations on Discord being stuck at v1 have at least something to do with Discord's limitations and possibly not Suno's. I imagine for the devs working on v3 is much sexier than trying to convince the Discord API to accept higher quality or bigger files or whatever the issue is with v2+

[jade mist]

Let’s be honest, who cares if discord ever gets updated to V2, not me at least

[low remnant]

Although, now I'm picturing the devs having a Monday meeting and awarding the cone of shame to one the who failed to perform the previous week. "Ok, pat. As punishment, you have to work on the Discord integration this week. Do better next time."

[ivory terrace]

Hello, I'm having a problem, I'm creating a song but when it comes to part 4 or 5 I want to have the same melody as part 2 but I can't do that. Hope someone can help me with this!

[low remnant]

If the melody wasn't reused in pt 3 or 4, you probably can't do it. You might be able to continue part 2 and keep the same melody and splice it all together in an outside app but Suno by itself is unlikely to oblige you. It only listens back so far.

[ivory terrace]

Thank you for your answer! Do you plan to add this feature to Suno? I think it's necessary!

[low remnant]

LOL I'd love to, but I'm not associated with Suno. Just a user with a big mouth.

[final hazel]

https://tenor.com/view/big-mouth-gif-14411882

[limber iron]

Are there plans to make the voices on songs have less or no autotune?

[graceful condor]

/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.

[ivory terrace]

I have one more question, I accidentally had a song that entered Suno's top trending even though I didn't share it on social networking sites. Why is that?

[final hazel]

Did you make it on discord and made it with a free account?

[ivory terrace]

I use a premium web account for $30 a month

[final hazel]

But did you make it on discord? Or did you share it on discord maybe?

Which song is yours?

[ivory terrace]

Currently the song is still in the top trending with the name playing games

i didn't share it on discord or any other platform, the song was finished around 3am this morning and when i logged into suno in the afternoon, to my surprise it was on the top trending

[final hazel]

Yeah.. not sure it that is your song... I pesonally doubt that very very much.. that song is been trending way before yesterday

https://tenor.com/view/eating-popcorn-watching-movie-time-serious-focused-gif-16842872

[ivory terrace]

I created this song 1-2 days ago but haven't finished it yet

[final hazel]

Given the situation, it's best to contact Suno support directly at [e-mail]. As regular users, we don't have access to detailed information that could explain this occurrence. The support team would be better equipped to assist you with your query regarding your song's trending status.

[ivory terrace]

​
Oh thank you for your information, I thought you were a Suno support person😊

[final hazel]

Most of us hanging out are just friendly users... Staff is lurking too, but more working while we help each other...

[thin flint]

Probably not because the voices are synths.

[acoustic heron]

Any chance you got a response? I still can't interact with the bot via dm

[meager mason]

did it ever work via DM?

[acoustic heron]

It does for some ppl I know and I followed the how to, which is basically: message it

[meager mason]

sorry, i just was unsure but had the thought i read somewhere that DM doesnt work, only via channel(s)

The /chirp command does not work in any other channels or DMs.

in the "how to chirp"

[acoustic heron]

Strange, I know for a fact it works for a friend of mine. Maybe because he added the bot quite some time ago?

[shadow timber]

Are you sure he’s using Suno and not something else? I don’t think Suno has ever worked in dms

[willow wren]

Yes it works as described in how to Chirp: Welcome! Follow these steps to start making music with Chirp 🐣

1️⃣ Start a direct message (DM) with @versed valley

[pliant cave]

@light ether @open magnet Big moment, gentlemen, congrats on shipping! I see you took my suggestion for a dropdown for the Model version, or more likely it was already planned. 😅 Either way, great work!

My only technical question for later, since this is the technical channel; and strictly only if anyone on the team has the time or inclination to answer:

I see in the announcement "Faster generations." Does this refer to streaming being enabled, or is the model itself actually able to generate faster than version 2, despite it being a better, possibly bigger model?

[shell jungle]

i got some bad news... v3 alpha is out

[shell jungle]

v3 alpha is out!

[azure saffron]

Seems like true!

[meager mason]

loving it so far 🙂

[crystal sapphire]

Yeah, perhaps it's compression. But looks like the guessing on V3's release date ended today!! Looking forward to using it.

[scarlet mica]

How do I select v3?

[shell jungle]

pay money.

[balmy phoenix]

has suno given any information on what data they use to train? where does this data come from?

[tight frost]

I wanted to ask, is the suno team going to take advantage of the technology that Sora uses ?

[full kettle]

Wow V3 is amazing

[eager topaz]

I love the Instrumental button😀

[jade arrow]

Hey, I'm a pro member, but not seeing the V3 option anywhere. The UI is the same as before.

[jade arrow]

Yeah, I'm not seeing that in the create tab on the site.

[open magnet]

Is Sora gonna take advantage of the technology that Suno uses? 🤔

[open magnet]

maybe command shift r to refresh and clear cache?

[jade arrow]

Nope. I even cleared my cookies and history. I'm still not seeing it.

I have 2315 credits, and my subscription is renewed on Mar 11. It says I'm on the Pro plan. Not sure what's wrong.

Maybe it's something to do with my last payment failing? I charged it to a prepaid credit card, and it ran out of funds this month. I still got credits, but I wasn't charged. Just told to update my billing information.

I'll see if it's working on Tuesday, which is when I'll be able to put more money on it.

[jade arrow]

I noticed you were using Brave, so I tried to switch to that instead of Opera GX, and it still hasn't helped.

[final hazel]

Oh that shouldn't be the issue.. and I am just a normal user like you 🙂

[jade arrow]

Don't put yourself down. I'm sure you're a superstar in your own right.

[lavish badge]

doesn't this open screen recording? ctrl+shift+r should be it

[final hazel]

@open magnetdo you have more ideas?

[jolly karma]

Is there a way to determine the vocal, like if you would do songs with the same singer?

[jade arrow]

I've wondered that as well. Describing the sound of the vocalist using words like "Sultry" or "Gruff" seems to have an affect, but I can't seem to get consistent results.

[half dragon]

cmd+shift+r is a macos command... vs ctrl+shift+r

[final hazel]

not at the moment

[lavish badge]

oh yeah mac users exist

[jade arrow]

I pronounce cmd, commed, and ctrl, sitril.

[tight zephyr]

kudos for the team for the v3 update! Just prompting some stuff that was hard to do under v2. Classic rock with guitar solos. https://app.suno.ai/song/cdd7b955-0df8-4f44-8d1f-85f296bc4816/ Solo piano. https://app.suno.ai/song/d2f38646-e849-4e14-a6cd-d8ba9a387b52/ saxophone was almost impossible. https://app.suno.ai/song/5278f826-7151-48be-b683-bb1d6c732a04/

[heady hearth]

Never before has a generative music ai so faithfully sung my metatag prompts nested in the lyrics field so faithfully😁

V3 will be fun to play with, figuring out the fresh quirks of the alpha is much easier with the bonus 300 creds.

https://app.suno.ai/song/30f76118-1d04-476e-9238-9933c7ab370c

[jade arrow]

For reference, this is what I see in my create tab.

[half dragon]

how about try in Incognito mode?

to prevent browser funny caching business

[fallow gull]

Small improvement in Web UI, when you report a bug, you can enter a reason.

[jade arrow]

Gave it a go, but it didn't do anything.

[daring lava]

Please begin a support ticket at [e-mail] and send me a dm with your account info (email) and I’ll try to expedite it

[main breach]

I guess it was an update after all! 😂

[warm oak]

I was liking what I heard but now everything is kinda frozen. I've been waiting for about 10 minutes for generations to get to the point where I can continue. To quote Steely Dan..."Wheel's turnin' round and round..."

[main breach]

Probably just enthusiastic ppl hammering it at the same time, I've generated some stuff earlier and it was great 👍 only an issue with a bug when you are on mobile browser, you continue a song but it doesn't indicate that it has

[fallow gull]

Text in square brackets gets spoken in v3:
[VERSE 1]
[First rapper steps forward, rhythm picks up]
...

[graceful cliff]

yeah some jobs are failling

just retry

hopefully we refund you credits...

[warm oak]

I'll think I'll let everyone get their ya ya's out and come back later.

[rare rapids]

https://app.suno.ai/song/889b70e3-ac41-4230-9a54-3bf87b0bdab5 https://app.suno.ai/song/b83bf83f-7366-402b-aecf-010b05b7f1c8 https://app.suno.ai/song/caf05024-2e0c-42af-9f10-a4c88e009429

it knows hyperpop now

[buoyant osprey]

Can you guys help me with trying to get this type of sound? I have used this Slow dark scary trap/rap hard-hitting 808 for v2 and got some songs I liked but I always feel like I missed the mark. Should I try to use something different for v3? I really want to get something like the background vocals.https://www.youtube.com/watch?v=0ekIKWmFXQM&ab_channel=2Scratch

[silver oracle]

how do you make a song without lyrics, just instrumental?

[buoyant osprey]

if you have v3 you can switch to instrumental

[limpid spade]

Has anyone tried 'Continue From xx:xx' ? I only just jumped on the PC and saw that option.

[full kettle]

I'm guessing it's down right now?

[fallow gull]

yep

[crude orbit]

yep it work

[half dragon]

my generation is still spinning... maybe server is overwhelm

is it slow for everyone?

[eager topaz]

Yep. Got a masterpiece that won't finish 😢

[half dragon]

ok it's gone... no more spinning but the entries are gone

wasted my credits!!!

good that they gave us some free ones to try out 🙂

[eager topaz]

Wow! Managed to finish!

[half dragon]

I guess I'll wait it out... too slow to test it out

[daring lava]

10 credits = 2 generated songs

[molten portal]

Is it possible to change the voice for each part within one prompt？？

[eager topaz]

V3 is saving money. I've done 2 full songs each taking 3 parts to complete. Previously it's been taking me 5 or 6 , plus I'm rejecting far fewer parts due to improved quality. Well done Suno team, ace people!😍

[half dragon]

Haven't tried, but I've seen suggestions where you give them name in bracket [John] ... [Mary] ... within lyric

[molten portal]

thank you! But it didn't work

[half dragon]

they also mentioned .. it's kinda hit and miss...

[half dragon]

You've seen this doc? https://www.notion.so/Voice-Metatags-4a84f912c5f04ac9aa75f3382e683772?pvs=25

[versed valley]

Welcome! For help, check out Suno Help

[gritty tinsel]

#🐣┃how-to-chirp

[molten portal]

thank you! You could change the voice for each part without inputting the music style!

[half dragon]

it's working?

[full kettle]

O

It's working for me

[molten portal]

The probability is low...

[cold flax]

I'm not able to download one of my full songs, audio nor video download is working

[leaden plaza]

V3 is out?? i tried generating some new songs and it shows V2. or i need a paid membership?

[teal raft]

Yeah you need to be paid

[leaden plaza]

ok Suno has my money now

[willow wren]

Can i continue V2 Songs in V3?

[tired schooner]

how to access V3 please tell me anyone?

[fiery sinew]

V3 is out for paying customers. A bit longer first part, a much better quality, alternatives closer to each other than V2. a little bit hallucinating.

[tight frost]

Help them by publishing a paper !

[jade mist]

V3 is good! Ai leak account, here’s the money😂😂😂

[broken lotus]

yes, with some exceptions (a very early v2 beta that most never used)

[runic dock]

Is the Intrusmental really working?. Never been able to have it running for me yet. I get hallucinations or noise instead.

[exotic fiber]

try with slow grime, grime, grime trap

[buoyant osprey]

Ok, thanks!

[exotic fiber]

https://app.suno.ai/song/aa835d8e-39fd-4176-8b88-5d24cceb319d

[buoyant osprey]

Oh is this with grime?

[exotic fiber]

slow grime, grime trap, Dark

[buoyant osprey]

Ok thanks

[exotic fiber]

https://app.suno.ai/song/396efa8b-fba3-48df-a900-cdcbdc0295b0

Grime, Slow Grime, 70 BPM

[cold zenith]

are you sure ?

[buoyant osprey]

Nice

[exotic fiber]

https://app.suno.ai/song/831dac6c-334a-41e6-8e09-4b07625f700d

vocal, male vocal, dark vocal, Grime, Slow Grime, slow dubstep,

[cold zenith]

is it true that V3 is out ?

[exotic fiber]

just for paid users

[exotic fiber]

https://app.suno.ai/song/47afc4cd-a489-449c-925f-1735a146cc10

vocal, male vocal, dark vocal, Grime, Slow Grime, slow dubstep,

[cold zenith]

For pro and premier then ?

[buoyant osprey]

Ok thanks I’ll try adding these into the prompt.

[cold zenith]

the generation is way faster

what would be great, it's to get the audio in 192kbs at minimum not 128kbs. The best would be WAV

[exotic fiber]

more than welcome, also.. as Grime is like a fusion of garage, jungle, dancehall and hip hop... try also using: garage, jungle, dancehall, hip hop

[buoyant osprey]

Ok thanks!

[exotic fiber]

https://app.suno.ai/song/e73c2b88-a594-4896-b4c9-ca2a1268910f

here an example with: garage, jungle, dancehall, hip hop, dubstep

[cold zenith]

what's the encod of the musique in V3 alpha ?

still 128 ?

[shy schooner]

I would like to subscribe to Suno, but unfortunately you only offer credit card payment.

[neat nexus]

I have heard that people have been successful with prepaid cards. Might be worth a shot?

[shy schooner]

I have no idea how that works. Credit cards are not that common in Germany, at least not the American providers. They should simply offer Google Pay or PayPal as well. But surely the fees are too high. I'll have to see about getting a prepaid card...

[neat nexus]

I mean if it works? 🤷‍♂️

As someone who only recently got a credit card themselves, I know this pain. It's not fun.

[shy schooner]

Yes, I don't think it's as easy as a Paysafecard at the gas station. You first have to enter all your details and then apply. Annoying.

[neat nexus]

Nah I mean my credit union insisted that I build my credit. Was a good thing I did. I needed a car and without having the credit card I don't know if I could have built the credit enough for a loan.

As for pre-paid credit cards, I don't know how it is in your country but here you can get them at Walmart, even Aldis. Backed by Visa.

You just load the cash on them when you buy it and use as electronic currency online, like a credit card or debit card.

[shy schooner]

Okay. I'm sure it will be similar. As I said, I've never needed it so far. But all the AI companies usually only offer credit card payment. OpenAI now also offers Google Pay.

[final hazel]

Hey @shy schooner, https://n26.com/ ist ein deutsches Unternehmen für Prepaid-Karten.

Ich benutze es auch, obwohl ich Niederländer bin. Tolle Firma, alles bestens.

[shy schooner]

Danke muss ich mir mal anschauen. 👍 Die Bewertungen sind jetzt nicht so toll.

[final hazel]

Gerne!

[daring bay]

v3 problem I'm having, my songs end up getting cut off right in the middle of the last word or earlier

[teal raft]

You can continue from any time now

As in you can contrinue from before the line was cut

[latent swift]

Hi. How do I change the style during a song?

[daring bay]

kinda sucks to waste credits for just a few additional seconds of song

[teal raft]

Sure but it is a solution

Didn't have one before

[daring bay]

I did have a solution: post it here so that the suno team knows its a problem that needs fixing

[teal raft]

If you're that patient then by all means

[untold briar]

use short continue times, example https://app.suno.ai/song/a2c3839a-1e38-4e72-abf8-7057ef5fd0f9/
you can think of the size of the clip (time 0 of that clip -> your continue time start) as a slider, the smaller it is, the smaller the 'style prompt' power inside the audio of the song itself, compared to the strength of your actual new text style prompt.

conversely, to avoid this, be sure to "get whole clip" and create larger chunk sizes

[crystal prairie]

Trending looks over all public whole songs created over the past several days, calculates a score based on the song's playcount, likes, some quality checks, the song's age, and serves up the highest scoring songs. Sometimes a song makes its way onto trending just by being very new, which is probably what happened here.

If your song is on Trending but you don't want it to be: You can set your song's Sharing status to Link Only, and it will be gone in the next update (You might have to wait up to 10 minutes).

[limpid spade]

Question about folders. Someone posted a screenshot earlier and showed 'clips' and 'playlist' options. I don't see those on the web version at all.

[cold zenith]

Someone knows if the music is still in 128kbs ?

[limpid spade]

All of my V3 downloads show as 192kbps in my windows folder

[tardy cape]

Hi! There is a specific style of music i want to try to get close to making. Im not talking about copying the style of an artist, but rather a genre that doesnt have a name. Here is an example of such songs: https://youtu.be/1iNdRGZKQyU?si=SvGLanLywglaQvJR

https://youtu.be/G5QEBPFDm9c?si=vSIgwJ0DGj9Sw-tv

[jade mist]

https://app.suno.ai/song/dcd2fb75-689c-4590-bd2e-1aaac3a4b1f8

[void charm]

Damn it kinda did to the chorus what would be most likely for a jazz/swing song with that solo and repetition and also with that double repeat of the last line for the ending

[clever current]

what is chirp and how is it different from the suno.ai app ?

[vernal meteor]

Chirp here on Discord is an earlier version of the AI. Lower quality.

On the website you get v2 and if you pay you can access v3

[azure saffron]

@shy schoonerCheck https://wise.com/ I previously used Deutsche Bank and Commerzbank, but have been with Wise for almost 5 years now. It's the best for me, offering both USD and EUR accounts. It's truly the best option.

[ivory terrace]

Thank you so much!😊

🤔 🤔 🤔 🤔 Does anyone know how to edit lyrics after the song is completed?

[proper shell]

So quick question, my friend and I were playing with suno last night, and he put in a prompt that the ai blocked, and it said it was flagged for moderation. Is there any chance my account will be banned from the bot because of this? I have a paid account if that means anything.

The prompt contained discriminatory language

[proper shell]

Don’t think that’s possible

I wish

I have so many good songs that I accidentally misspelled lyrics on

[final hazel]

I wouldn't worry about that 😉
It only stopped you from using that lyric..

[proper shell]

Okay thanks

[ivory terrace]

I'm also a programmer and I think that feature wouldn't be difficult to add, it's just a little more complicated than the issue of modifying names because it involves the issue of creating videos. Hopefully that feature will appear soon!

[crystal prairie]

we are absolutely not banning users on the basis of moderation issues.

[cold zenith]

Nice that's an improvement then 🙂

[tardy cape]

anyone have an idea on this message i replied to here?

[tight zephyr]

v3 makes kick ass intros. https://app.suno.ai/song/27393f62-0e6b-4a8c-aa5c-12ae72a18e97/

[fallow gull]

I have seen examples of getting different voices in verses by using "John sings", but what about getting multiple voices in the chorus? Like for the chorus you want the whole band to join in.

[thin flint]

Hmm. What's the best way to end songs in V3? I found a good place to continue to try for an ending and simply prompted [End] in the lyrics. Is it a matter of luck still?

[final hazel]

here is what I do. When I continue my song, I clear the "style of music", and replace it with "end". Works for me 3 out of the 4 times

[thin flint]

This chiptune is a bop.

[thin flint]

Making duets is still hard in V3, I see.

[torpid elbow]

Any idea what genre is the song ‘like i do’ by j tajor. I want similar feeling music result on suno

[heady hearth]

I've burned through all my V3 funbux, so I can't experiment: has anyone tried to smoothly transition the singer from one vocalist to another? That could be a very interesting effect, if it's possible to do.

[bitter fern]

is there a way that if we give the audio the ai adds the vocals to it?

[torpid elbow]

Does suno work on this type tag [instruments:guitar piano drumset] ?

[half dragon]

Not possible

[deft girder]

can you not download v3?

[deft girder]

lol chiptune was the first thing i tried on v3 too

also got a similar banger

[deft girder]

i managed to get some good results by prompting male and female

[broken plank]

Anyway to only output instrumentals?

Oh, there's a switch on the pro version lol

nvm

[deft girder]

lol i made a song about someone dying and i swear v3 is about to bust into tears midsong

so impressive

[eager topaz]

I spend a lot of time crying when I'm producing songs on Suno!

[deft girder]

haha is that because youre emotional or a perfectionist?

[eager topaz]

A bit of both, but, yeah, Suno does produce a lot of emotion. I'm always surprised when the tears start forming in my eyes 😁

[deft girder]

yep but also get up and start dancing or headbanging too

letting the imperfections rock on a otherwise gold take is painful tho

[versed valley]

Welcome! For help, check out Suno Help

[gentle musk]

Does anyone know is there a way to edit the lyrics on Part 1 of a song? I want to keep the verse and chorus but just change a couple of words

[jade mist]

Why are download options missing on Windows?

[gentle mesa]

I don't know if anyone on the training team at Suno reads these messages, but adding an invisible [end] token and training on them would help make songs have a better, less abrupt ending. 🙂

[jade mist]

Will continues get better once V3 is out of alpha?

[iron thicket]

Use « continue with this song » and do the update. You will need several attempts though as often ai may change the melody. In general you have good results if the rythm of the words you are choosing match the melody.

First feedbacks on V3 Alpha, quality is way better some results are much closer to release ready. Results on some genres are still muddy but overall an important step has been achieved. Congrats Suno dev team! What I like the most : it seems to me that variation of vocal melody and hook are better, you can get very unique results while before it was more stereotyped. Suno v3 Alpha you rock !

By the way I’ve tried to use V3 to glow up some of the songs produced with V2 but at this stage I was not really successful. AI changing melody and I did not see improvement on the quality with this use case. Let me know if someone was successful in improving a song made with V2 using V3 Alpha

[iron thicket]

Yes could help indeed, that’s being said I have pretty decent results with [outro]. But as usual with ai tool you need to test several times before getting exactly what you want

[cloud wraith]

https://youtu.be/JGrvlVfNhgE

[runic dock]

Still not doable. This feature should be a "Must have"

[iron thicket]

According to my experience, and as discussed previously, it is doable using continue with this song and using the new lyrics. The only issue is that with number of repetition the sound quality goes down. Anyway with patience and some post processing you can do it, I used this technique personally to have the same song delivered in multiple languages but with the exact same melody . If interested I can create a tutorial and you will be able to see the results.

[runic dock]

If you do continuation and you change the lyrics than how do you get the whole song without the first part that contains the original lyrics?

[iron thicket]

Once you get the song simply edit it in a daw and cut the part you do not like

[runic dock]

Yes of course. When i was saying "not doable", i was refering to a solution entirely made in suno with no external tool.

[slate matrix]

hey guys, is tere any way to remaster v2 songs with v3?

[final hazel]

Not really.. you can continue a song, but not start again. If you do, it will sound totally different (when you try to continue from 0.01 sec)

[slate matrix]

thanks, yeah already tried it, its close but the start gets really jumbled

[runic dock]

With V3 do we have new documented metatags?

[final hazel]

No, because to learn to talk to the new model takes time to understand.. it's like to getting to know a new person who speaks a different dialect. Some old tags work, some don't.. and it is still alpha so things will change too

[ivory terrace]

How is top trending calculated? My song "Playing Games" was removed from the top trending while still gaining 10,000 views a day.

[final hazel]

wasn't that answer yet? @ivory terrace #📚┃suno-school message

[ivory terrace]

oh, maybe Suno's calculation method is really much different from youtube or other platforms! I can only understand that, about 10,000 views a day is not a small number, I just find it a bit strange!

[final hazel]

🤷‍♀️ idk.. I have 2 trending there... posted yesterday. One has 5k views already

[tight frost]

How many views did you get on youtube for the same song ?

[ivory terrace]

I'm just planning on posting it on other platforms, starting with another video remake, not yet.

[final hazel]

But comparing Youtube with Suno is a bit far reached tbh.. here people listen because we are all doing the same, and have the same mind set.. Youtube is so much bigger and all kind of audience.. I wouldn't expect the same behavior there without having a big follower base before posting

[ivory terrace]

wow, can you share more of your experience to get to the top trending, I also have some new songs that are pretty good but can't appear, maybe it doesn't pass some quality test that I don't know!

[final hazel]

I have no clue tbh.. just shared it here in discord

[ivory terrace]

ok, maybe it takes a little more time to understand Suno's algorithm! Thank you for your reply, I see you are truly an enthusiastic person! Respect!

[tight frost]

Are you talking about the Explore tab on the website where only ≈30 songs are listed ?

[ivory terrace]

yes!

[ivory terrace]

oh, it has been renamed to Explore and not trending anymore!

[tired schooner]

guys how is V3? is voice clarity and quality good?

[ivory terrace]

All the issues you care about are good, the quality is completely different from V2, however if I create a song that combines V2 and V3, there is a part where V2 will produce a better melody than V3. Maybe it's just luck...

[tired schooner]

thanks bro for your reply

[ivory terrace]

Well, V3 has a situation where the lyrics stop before singing all the lyrics. This problem happens a lot, making it difficult to create the next part. I see this problem happening less often in V2. If it does happen, the singing rate is The lyrics in the next part are also higher!

You can listen to this song and evaluate the quality of V2 vs V3, from the pre chorus of the 2 songs created by V2, the sound compared to the previous V3 part I find smaller and the quality is not the same!
https://app.suno.ai/song/a47e5b52-f446-46dc-b796-b0f7ab34c820

[austere violet]

anyone else finding a lot of gens are timing out? 💔 Must be a busy time!

[final hazel]

no problem with me

[austere violet]

that's good then ^-^

[final hazel]

it comes in waves

if they keep spinning after a few minutes they are usually timed out

but won't take your credit

just remove and retry

[austere violet]

tyvm! ^-^

[meager inlet]

Why can I not use V3 anymore??

[final hazel]

I suggest you mail [e-mail]

See: #💬┃general-chat message

[meager inlet]

Okay yeah it was a billing issue. Strange I thought it went through since I had already received the credits for the month. But good now, thanks

[buoyant osprey]

does anyone know what prompt was used to get this? it's not mine.

[final hazel]

that's everybody's guess.. only the creator knows

[buoyant osprey]

well I guess I should ask what do you think it is. It sounds like rock or alternative but I'm not sure

[final hazel]

You might have more luck asking in the #💬┃general-chat where more people hang

[warm oak]

V3 doesn't seem to respect the [verse] [chorus] boundaries as well as before. I'm also getting a lot of gibberish. Been trying for an hour and I have nothing to show for it. Frustrating. Last night it was great, but not today. Bah humbug.

[tight drum]

I tried typing in lyrics for a second series of songs and Suno AI Chirp Bot said it was too long. I put of EXACTLY 750 letters as it was offering.

[fallen radish]

It was mentioned previously that when they added the "continue from anywhere" feature that this would essentially allow you to wipe out the bad part one lyrics... i.e. something like saying "continue from 0.1s"... so you'd get the same tags that built your song, but you'd be able to replace the lyrics. I'm trying to find the docs on how the new v3 "Continue from anywhere" actually works... AH.. I just noticed, it's hard to see.. when you hit continue there's a mm:ss box beside the word continue from

[exotic fiber]

I always use MAX -1 to avoid this type of errors using apps...

in ur case, 750 -1

[exotic fiber]

great song.. sounds like a sadly emo rock for me, u can try with that

[thin flint]

In the style prompt or where?

[exotic fiber]

usually in the style, male and female, duet

and, you can try adding
(Jhon) bla bla bla
(Mary) bla bla bla

to the lyrics

However, take into consideration that duets are kind a 50 50 or maybe less chance to tet them.

what i also do to get different voices results is adding boys band, girls band to the style, and then, adding different names in the lyrics.

[exotic fiber]

https://app.suno.ai/song/4100ced5-5d4a-4d1a-9511-4daf9bf4afd6

got this using emo rock, rock, melancholic, male voice, slow

not the same but is on the way... try using (whisper) or (whispering) in the lyrics to see if u can get that whisper kinda voice...

[thin flint]

@final hazel 3Daizy created a technical achievement with this. Multiple voices, genres, etc. Do you change the style prompt with each clip? https://app.suno.ai/song/8db92e67-6b31-4726-a0ea-08fe35b571f6/

It didn't do blues though.

[final hazel]

100%

I am not an expert in styles.. it sounded like blues to me.. I did a second one in more electronic music and now working on a third one.

[thin flint]

Well, I see you have piano written in style but I heard no piano there, just mostly guitar and sax for the jazz parts.

Still it's pretty awesome all the same and now I want to do one.

[final hazel]

I also change the style with every clip.. so what you see in the beginning is only for the first clip

[exotic fiber]

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

[final hazel]

want to get a preview of the one I am working on?

it has blues 😛

[thin flint]

Sure. 🙂 @final hazel

[final hazel]

wip https://app.suno.ai/song/e6c94132-3301-49d6-8fd7-299648153fb1

[eager topaz]

That is amazing! Great prompting 💖

[buoyant osprey]

ok thanks

[final hazel]

https://tenor.com/view/elvis-thankyou-gif-6754675021318788732

[exotic fiber]

@final hazel is on fire...

great testing, im learning a lot with ur prompting for v3..

[shut gyro]

https://app.suno.ai/song/6cff6b02-cc9a-4e1e-a58a-34126a2ab90c

[final hazel]

Pleased to hear that.. that was my intention... I burned more than 9000 credits to get to know V3 so far 😄

[eager topaz]

Wow! I thought I was bad at 4000!

[final hazel]

I wouldn't worry about it.. it will pay off in the end.. YOLO

It's done.. else it became too long lol https://app.suno.ai/song/24410aea-61b6-4107-a846-3eee367b466f

[broken heart]

90% of my continuations are only doing 10 seconds, maybe half a verse and then just stop. V3. Using my credits.

[final hazel]

Are they continuations?

[broken heart]

Yes. Most of them are timing out as well.

[final hazel]

Okay.. do you change the time of where you want to go next?

[broken heart]

I did. Had to really because the original generation did 55 seconds of song and then another 55 seconds of silence. So I continue from 55 Second Mark.

[final hazel]

Okay.. please check if the seconds are still correct. I noticed that sometimes when you jump from one to another it is somehow stuck and therefor wrong.

try to clear the continuation and set it again

and when it get stuck (i had it too) it will not give you new clips or very short ones.. because it's confused

[broken heart]

I will give that a shot! Thank you so much!

[final hazel]

please let me know if it helps.. would be great if we could figure out 😉

[lucid anvil]

Is chirp v2 or v3?

[pliant cave]

You have a lot of creativity, well done.

[pliant cave]

Try barbershop quartets/ harmony. I've gotten some amazing stuff.

[broken heart]

V3 is like super broken. I would like to get my credits back please. I spend more time reporting Bug after bug than anything else. I guess thats ok but it is taking my credits. I think V3 sounds amazing, you guys are amazing, I would just like to get the credits back that I spent on testing it out. Im poor or I wouldnt care to much. I also wouldnt ask if I had gotten anything remotely usable.

[rose ore]

so this is interesting
I just figured out that using "high quality" actually improves the quality of renders on v2

[teal raft]

yeah mc months ago suggested it might work

I assume they were talking with privileged knowledge
I think it was mc

[broken lotus]

Are you suggesting my prompt Justin Timberlake with a sock in it at the bottom of the well is causing some sort of effect on the music quality?

🤔

there goes my whole concept album. 😫

[teal raft]

'why my song sound bad my prompt is 'garbage butt garbage 2: garbager also really bad and ugly can music be ugly?'

[rose ore]

God I hope not, that sounds hot

[untold briar]

It's pretty hard to make a reliably garbage prompt in V3 actually. Try it. https://app.suno.ai/song/7f80e40f-c864-40d4-b821-ecc5bcceaacf/ https://app.suno.ai/song/d07f3daa-46d5-4492-bf4d-e119d03a6b30/

[limpid spade]

Mixing question - The file on the left is a song from a commercial artist, the right is a V3 song. Are all generations Limited to a maximum output level?

Follow up question, for commercial release would it be worthwhile mastering in a DAW to at least level match? (Actually, I think I will set up a gentle limiting template).

[broken heart]

How do you do High Quality? I cabt seem to find that button.

[rose ore]

add it to the genre prompt

[eager topaz]

Yes, master in a DAW. I normalize and render to Peak -1.
The file on the left looks like it's on the verge of overload/distortion. Spotify and others set the maximum, but if you render to -1 db that is usually ok, although there are lots of differing opinions!

[broken heart]

But how? What is the prompt to add? Sorry to sound stupid.

[limpid spade]

'High Quality'

[broken heart]

Boom! Thank you!

[rose ore]

so if you're adding, for example, "rock, 90s, pop, guitar" normally. Just add "high quality" or "high quality stems" or "high quality vocals" and it changes the output. Makes things clearer

[broken heart]

Becaise I like V3 but I cant get anything out of it other than jubberish.

[rose ore]

V3's been odd. Seems to be mostly working now

[broken heart]

Whaaaat! I had no idea. You rock! ♫♪♫♪ Thank you.

[rose ore]

🙂

seems to work reliably on v2

[limpid spade]

We really could have a Prompt channel here for just these discussions 😄

[rose ore]

nothing works reliably on v3 yet

[rose ore]

I agree

[ivory terrace]

🤔 V3 often feels longer than the lyrics I create, how can I solve that problem?🤔

[limpid spade]

"High Quality, rock perfection, delay, guitar virtuoso, over-produced" Not sure if over-produced did anything but i like the quality.

[cold zenith]

WOW

I just discovered something guyz....Your previous Songz in V2, put an continue on those put [instrumental] only, and choose V3. And get the whole song. The entire of the song will be V3 sound ^^

I didn't think it would works, but i tired and it works

i'll try one with hight quality include

I said [instrumental] for those which the songz is finished

[fading heart]

I have to say I love V3, the quality, the coherence and creativity specially for instrumental songs is crazy!
One thing missing for me is to upload a song and continue from this one. Is this feature planned?
The tip of the iceberg would be to get the individual stems.(is there a good solution right now?)

[cold zenith]

that would be grea

The rework V2 to V3 from a Previous songz is incredible...

( I know some people here don't like to gives tips hahaha, because they won't be able to give a formation on youtube and their website for 100$ ) but i don't care.

I'll share what i know and discovered for free as people had helped me for free too.

[alpine venture]

OK so my primary criticism is that you cannot edit lyrics when using the continue feature. This creates issues in terms of prompt adherence since it will one-shot learn to ignore portions of the prompt.

For example, if you have some lyrics at the end of your song that don't "make it" in the time limit, it is possible it will begin to ignore any music tagged similarly.

I manipulated this in v2 to produce controlled instrumentals by waiting until I rolled one where it read parenthetical as instrumental instruction, and then running continuations from there.

In v3, being able to manipulate the FULL sheet, as well as the start position of the current "clip" would be MUCH MUCH better, and create a huge improvement in prompt adherence. It would also allow for full lyrics inclusion from the start, and behind the scenes you could just slice it however you want and place some sort of a marker where the model "believes" it read until. Then, we could modify this position for accuracy, or even in advanced cases, test manipulation of this to create more unique sound and instructions.

if you are trying to produce actual unique or interesting sounds, lyrical adherence is key since it has a massive influence on the model, although v3 is def more close in terms of how it follows style

Here is a great example:

https://app.suno.ai/song/c4558140-8e28-4056-b55d-3e42de14df0c

you can see how over generous I need to be, and towards the end its basicslly a crapshoot whether or not I will actually get adherence

because it's so "untethered" from the written version

[untold briar]

Try backup up a few second susing the custom continue time, to a time when there is a cleaner break between lyrics (between lines, at least)

[alpine venture]

does it actually affect the models perception of "location" on the sheet lyrics?

if not, I think the continue feature will at times be frustrating, as you could end up with a particular stanza it just refuses to read (I've had this) because it believes it is supposed to skip it

[untold briar]

Well you can change the lyrics in the continue to be whatever you like, exactly. That is all it will see, it won't see the prev clip prompt at all (except in the audio itself, if it's embedded in that already, which can happen. But if you back up enough it often goes away.) This should let you 'clip off' the part of the audio that is making it think it 'already sang those words'

[tight zephyr]

I agree storing the full lyrics and using that for continuation would be a life saver here..

[untold briar]

There may be some bugs around continues in some complicated cases, especially with older V2 clips if you try to continue them with V3, and especially with clips that are 'get whole song' of a 'get whole song' (complicated situations) but for V3 - > V3 it should work.

[tight zephyr]

It's just sometimes hard to find out where it stopped since it may add some lyrics too

v3

[untold briar]

FWIW I have a personal feature quest in to be able to fully customize both the START AND the STOP position of continues, in fact that's like 1 of like 8 custom continue features I requested, lol. So I hear you. (And some of my features are about chanigng the "location" in a deeper sense, like you mentioned.) But I think that particular problem can be solved with the current continue time feature.

[tight zephyr]

It's called Inpainting when doing image generation, that would be awesome

[untold briar]

Inpainting is in the conversation, if you search up dev posts. Timeline who knows, but devs are thinking about it for sure.

[tight zephyr]

But to just paste all of your lyrics, and hitting Continue without listening or selecting the start position would help save a lot of the manual time

Im now at 2500 credits on v3 and will post a list of comments later, good and bad

[tight zephyr]

Great

[jagged loom]

What should I do to give music commands in the style I want? Sometimes it is meaningless, and it does not read the lyrics I write.

[runic forge]

Any good way to end a song in V3? I can't end anything with [Outro] or [End]...

Nvm. I continued from 2 seconds before the end and it ended with [Outro] 😉

[cold zenith]

[fading end]

[runic forge]

I hate fading ends but if other ones won't work, I'll take it. Good to know, thank you.

[tight zephyr]

[Cold end]

[digital talon]

Has anyone been trying to use specific musical terms in prompts in v3, to exercise more control over the music? I've been playing around, and I think it seems to work quite well, although I'm not a good enough musician to make a full assessment of its abilities. Here's some examples:

Prompt: fast looping analog synth arpeggios
https://app.suno.ai/song/b118e010-f5ce-4e64-bf9c-d28674603506
https://app.suno.ai/song/ff2263ad-c021-47f7-b3e4-344b87bf274e
Prompt: fast looping marimba arpeggios
https://app.suno.ai/song/2b37e8b6-4571-4ef0-9c3c-357e87e866cb
https://app.suno.ai/song/cd0e9978-03d8-4056-861c-6bcf19fdea9f
Prompt:fast piano Broken Chords
https://app.suno.ai/song/4ce63f3b-2a5f-4b20-9843-9988f65e7f18
Prompt:piano c-major broken chord
https://app.suno.ai/song/9ae0bd16-d0de-4dae-9278-6bdb35c5bdb7
Prompt:piano d-minor broken chord
https://app.suno.ai/song/88d0bea0-5056-4d23-bf83-19629cf5f4d3

It seems to offer a lot more control than v2 and accept more complicated prompt descriptions, but what has everyone else's experiences been like with this?

[limpid spade]

In doing orchestral tracks I've found prompts like staccato violins, legato cello etc seem to work more often than not but it doesn't know ostinato.

[digital talon]

Yeah, playing around it seems to handle more common terms well, and in combination with other instructions like fast or slow

[limpid spade]

"high quality, orchestral, staccato viola, legato french horn, spiccato violins" I like things like this but also adding moods like romantic, emotional, heroic, victorious etc help too

[cold zenith]

it's beautiful

[thin flint]

Interesting. So you have to be very freeform with this style.

[thin flint]

I like the actual bird sound opening. Can't do that reliably if you actually want more of it though, I've tried for rain sounds and such.

[alpine venture]

In general I find that free-form, avoiding unnecessary line break, allows the model to appropriately apply its "creativity" to how it produces the content, which works best for non rhyming or non-metered

it will make its own

and it often works

[thin flint]

I don't get this. Do a continue on the complete song, prompt with [instrumental], and then get whole song from the generated continue clip?

Seriously wish ther was a way to quickly find stuff in library. I want to go back and continue/remaster old V2 hits but I have to keep scroling through page after page after page.

[manic cradle]

thats cool. Testing that out. What is high quality include?

[digital talon]

You need to go to the full song from v2, select "Continue from this song", select v3 Alpha, then set it as an instrumental. I found before I could continue with making the song I also had to put the timestamp (which I put at the end of the song) and copy and paste the style into the style field. Then when the new clips generate, made the Whole Song and you'll have the original song in a v3 version, which cleans up some of the rough edges more than anything.

So here's one I just did:
v2 - https://app.suno.ai/song/91c4baee-829e-473a-b3cf-6a0d52070a31/
v3 - https://app.suno.ai/song/fb610d4c-8873-4062-bdbc-b3c428ac60f1
The best section to compare starts at 4:00

[thin flint]

Does all the instrumental checkbox do on the create page just litterally add an [instrumental] prompt to lyrics? When you create a new one that's all it seems to do, unless there's some enforcement prompts on the backend that goes along with it.

[digital talon]

This is probably a better v2-v3 comparison. You can hear right away where trilling sounds and blown out bass sounds have been dampened.
v2 - https://app.suno.ai/song/5185676a-f3e2-4861-ac00-cff7af99e197
v3 - https://app.suno.ai/song/b1deb5c4-d08e-49e1-b261-96b26404d142

[thin flint]

One thing that V3 is better at? You can put more than three descriptors into the prompt and it won't be overloaded it seems. Curious what the limit is now.

[digital talon]

I've found you can make them more complicated as well

[broken lotus]

🙏 Sensei, what is the word for sustained notes with a bowed viola (or violin, or cello)?

🎻

[broken lotus]

ugh got it "legato"

https://app.suno.ai/song/a46ed231-d081-4f52-99ad-2265ab65a893

[versed valley]

Welcome! For help, check out Suno Help

[wild lark]

https://app.suno.ai/song/99c22da1-4493-4dd0-b968-76820d78b4ce

holy cow

https://app.suno.ai/song/34cd2252-bee5-402b-a17a-ce806da39892 ok im insanely impressed with v3

[exotic fiber]

-6 is the magic number for streaming services + normalizing EQ levels... however, what i recommend is to RE MAKE the track on a DAW using ur knowledge of music production using the Suno Track as a guide, also processing the Vocals and making the track all over. Why? Sound quality...

As I explained i few days ago, the quality v2 offers is 128kbps and v3 192kbps. a LOT of info is missing, and mastering, leveling and noise cleaning are not going to bring back missing information.

Take Sunos tracks as a guide and inspiration for what u want to make. or, wait until Suno improves the output at least to 320 kbps and then, with some tweaks, u will be able to use the tracks directly for commercial use.

right now, the quality is not good for that purpose.

[fallow gull]

I think v3 captured the spirit of Kwela music quite well in this. A bit to midi-like, but overall well done.
https://app.suno.ai/song/25a91592-c610-4092-b531-cd5684756209

[wild lark]

https://app.suno.ai/song/a982978c-6543-4f4f-9d83-7fa2af9c36cb

so beautiful....

[gentle hill]

https://app.suno.ai/song/fa1edd77-648e-4c80-a1ea-c07ed71b2e97/

[pliant cave]

Really? It's hard to believe this works. I know MJ had this little trick in version 5, but I didn't think Suno's models would pay attention to it, especially version 2, given how picky it can be with style tags and whether it goes wrong or not.

[rose ore]

I know. It's weird. Clearer neurons maybe?

[pliant cave]

Are you a musician? I've heard a few people suggest running side businesses of using Suno to quickly mock up songs, then actually produce them with real instrumentation augmented with the vocals, etc., or doing remastering tracks for people that way. I'm sure the demand is there for that, but it could become redundant if/when they get the quality to that level. As an untrained ear, it's almost there already to me in v3, but I'm sure musicians can tell.

[pliant cave]

I'm a software guy, but I'm not too deep into ML. I'd assume the material trained on would have to have some tags in there somewhere, like "high quality," for the model to pick up on. Who knows, maybe, but it seems unlikely. Well, obviously the developers know, but in terms of our speculation.

[runic dock]

I found this "High Quality" mention a bit confusing as it implies V3 default quality is low.

[broken lotus]

it's exactly like typing "4k Octane Render" into an image generator...
(you're not getting 4k)

[rose ore]

oh yeah, good point, yeah, same idea

[broken lotus]

I use 'Studio Recording' but it has too many letters, lol

[rose ore]

I know, I have the same problem all over the place

[broken lotus]

Will my prompt fit in the overhead luggage compartment...?
Do I need to pay extra...?

[rose ore]

hehehe

[shadow timber]

When you type “4k Octane Render” into an image generator, you’re not getting 4k, but it is sourcing images tagged with 4k in the training data, so it does make a difference. Suno could be doing something similar.

[rose yacht]

Oh, maybe I should stop including "wax cylinder recovered media" then

[shadow timber]

You could probably get some interesting results with that, I have gotten interesting results from “garage band” and “amateur”

[rose yacht]

I haven't tried too many effects like that but I did try wax cylinder once, it's appropriately low quality. I assume they have lots of old recordings since a lot of public domain recordings are archived online one place or another

[exotic fiber]

im not a pro, but i know a little bit.. i have self learned... dj producer as a hobby since i was 16... now im 40 xD and.. i see suno, even with great quality output, more as a inspiration tool. u can get new ideas for your productions. and when u tweak this new ideas into a new original track... that's where the reward enters the formula...

(My point of view)

[exotic fiber]

i dnt know if is just my brain playing tricks with me, but, if i use FLAC on the style, i think i get i much cleaner sound.. but, as i said, maybe is just my brain playing tricks :p

[exotic fiber]

but u get something similar in quality to it (cleaner image, less aberrations). so maybe, uf this suno model works and was trined something like image generations, taking the titles of the songs and/or tagging. If we cal the "High Quality" pr "FLAC" we will tell the tool to just look for that material, avoiding maybe low quality sounds and just focusing on good quality trained sounds...

(if this works as the image generation, can work, otherwise dismiss everything i said xD)

[broken lotus]

'high fidelity'...

I stopped using '1960' because it made the audio worse..., I could get the same sound generally with other words... but 1960 seemed to make everything sound like a 45...

[exotic fiber]

umju... thats a pro/con because if we want to recreate something like a 70s hit song, it would come usually with the 70s audio quality xD

the song will be good, but is recreating everything xD

[vernal meteor]

I've tried years, I've tried "1989 hiphop" (example) and it seems to do nothing

[midnight shadow]

how do I see further into my recent history? It seems that it's only able to load the most recent songs it's created in browser and wonder if there's a way to find all the tracks I've created.

[final hazel]

check Library

[midnight shadow]

oh got it - thanks!

[broken lotus]

next time you see Tony ask him about years...
I remember a comment a long time ago that they only trained decades
1989 same as 1981 🤕

might have been v1

[versed valley]

Welcome! For help, check out Suno Help

[tight drum]

Is there anyone trying to get the same chorus to get repeated but it never sounds the same as the last chorus? That’s what’s been happening to me.

[naive glacier]

i have question if you pay for the pro memeber ship. Can you then download stems in 24 bit wave?

[eager topaz]

No, it's still MP3 192

[naive glacier]

like if you only want some parts and then produce the rest yourself? so i can download every part? rip XD

[untold briar]

Do you have a clip you can share with that problem? The clip with that you are continuing from

[eager topaz]

Yes, same here. Sometimes it's the same, but very hit and miss. I don't have an answer to force the chorus to be the same.

[tight drum]

https://app.suno.ai/song/dde5c68d-4796-4780-83f1-3033fa9cfe4c/

[eager topaz]

You can download the audio for every part and edit in a DAW and merge them, if that's what you mean

[naive glacier]

yeah but i wanted it as wave file like "Higher Quality" you know 😄

[eager topaz]

No, can't do. Still only MP3 and even if you render it to WAV the quality stays the same.

[tight drum]

I sent the link to the song.

[naive glacier]

maybe it will come someday 😃

[eager topaz]

Perhaps it will! The instrumentals are mostly ok quality, but vocals are still hit and miss for quality.

[naive glacier]

yeah and i think it could be fire just to make vocals from AI and put it back right in FL Studio haha 😄

[full kettle]

I am generally fine as long as the chorus is in the clip I'm continuing from.

[shadow timber]

I’ve been finding that it makes SOMETHING into a similar sounding chorus, but not what I say in the lyrics field is the chorus. It ignores that. And often it makes the chorus completely different anyway. V3 is much flakier than v2 in regards to choruses.

[eager topaz]

This is Suno COOL!

[civic yoke]

does the AI recognize verse, chorus, middle 8's, pre-chorus and all other song structure?

[half dragon]

it understands [Verse], [Chorus] and [Pre-Chorus].. not sure about middle 8's though... try it!

[eager topaz]

[Bridge]

[limpid spade]

[bridge] and [build up] worked well in V2

[eager topaz]

[Bridge] is fine for me in V3. I've never tried [Build up], but I will now 🙂

[civic yoke]

and if what I've read correctly, V3 does not recognise Key and BPM choice yet?

[eager topaz]

Not for me yet, no. I was hoping!

[civic yoke]

Can't wait for that to be added!

Another question, does it matter if Verse is written with () or []

[teal raft]

() are for stuff like this:
"these are song lyrics oh baby (oh baby)"

Your singer would repeat the stuff in ()

Or a different singer

Sometimes it's silent because AI but that's the intent

[half dragon]

my understanding () is used for backup singers? works well in my lyrics

[teal raft]

yeah exactly

[wary python]

When giving a song description when I want it to generate full lyrics for me, is there a way to give it a word to use for the style of the song like "glitchcore" but not also have it use that as a lyric? Seems to happen most of the time I generate a song

[half dragon]

I've seen prompt that has [Chorus in minor key] but can't confirm if it's actually doing it

[alpine venture]

https://app.suno.ai/song/eae6f0de-0699-4eef-892e-44429e7198d2

[eager topaz]

I've seen that too, also I've seen chord names put in as prompts, but I haven't seen any actually working.

[alpine venture]

I keep coming back to this one

bc from a technical perspective its impressive

what I mean is, the way it kills and revives the beat

in the end follows multiple clean patterns

like, it slowly condenses the themes and tunes of the song, while maintaining the "meta pattern" of false stops and then bombastic, explosive starts

[half dragon]

I was going to try but then I'll be wasting my credits 🙂

[alpine venture]

any prompts like that, you need to basicslly get it working somewhere in the existing prompt, then make sure it is adherent before proceeding

it will one shot learn better adherence that way

but this greatly limits your flexibility

[noble summit]

Is there a paper available on how this model works ? I’m very keen to learn that if possible.

[cold zenith]

take the last part of the song before the complete one, continue it insteasd of get the whole song with a instrumental in V3 and then get the whole song

[digital talon]

Anyone know if the recent maintainence fixed the repeating lyrics issue?

[civic yoke]

So how does the continuation work? I find the documentation a lil' confusing. Do I keep the original verses I have in the box or do I clear it and put the new verses in?

[untold briar]

The clip in the continue box is the clip you want to add on to. Also, if you select V3, you can choose a specific time instead of the end of that clip.

This custom time feature is more powerful than it looks, it allows some more unusual workflows and songs

[civic yoke]

So I get these which seem to be separate from the original clip?

[untold briar]

Right. So that is part 2. If you look in the ... Menu now there will be a new option "Get Whole Song" that will put all song parts together, for no charge

[civic yoke]

ohhhhhhhhhhh

I see now

wonderful

[untold briar]

All previous clip, just backwards in time

[rustic gale]

Is there GPT's with all rules for Chirp? Maybe some one already created one.
If no. i have one:
https://chat.openai.com/g/g-l4WpIlAxX-songwriter-s-muse

But it has only basic things, like what Tags to use and how to generate Style for a user.

[sleek jay]

What is the preferred word for the noise that is in the old music source, if it is a word?

[exotic fiber]

static, old record, try using an old decade like 1960s

[neat lake]

I wonder if the amount of people generating at the same time affects the results, because first couple days I had amazing results and now they are terible and muddy, probably worse than on V2 sometimes 😄

https://app.suno.ai/song/e7dd2418-8f6d-49d9-9801-b8485b4260a9

[broken lotus]

you mean..., in exactly the same way that youtube videos are more boring when too many people watch them...?

eyes use up the mojo.

[wet minnow]

for me its the genre. it decides if the output gets "muddy". if i choose common genres it works brilliant... if i mix weird genres it gets less accurate. (on v2)

[neat lake]

Nah, just remember using another AI for something else and when a lot of people generated at the same time, it either took ages or the results were poorer, I've no idea if that's a thing actually 😄

[broken lotus]

I'm pretty sure it's just put on a jobs list, and you'd have to wait in a queue....

They don't 'remove instruments' or anything, because the musicians are too busy... LOL

or you get the less-good singer because the better singer was already booked for a Vegas show... 😹

[grizzled lance]

The continue function for v3 is unusable, can't keep the same style going, changes the singer for no reason, repeats words over and over or doesn't include them at all. V3 is amazing don't get me wrong, sorry if this has already been brought up I'm just frustrated after using hundreds of points to try and continue a generation I really like and it just won't do it. 😤

[broken lotus]

If you want to vent... 🤕 noted!

if you want help getting a full song that lasts several minutes..., ask 😉

https://app.suno.ai/song/61d49eba-a1df-4697-8849-12cb11950062

👆 this song went off the rails... but I was able to stitch it up, and find a continue from time, that let me get to the ending...

(for example)

[grizzled lance]

Ahh cheers I'm just venting 😅 I don't know if I'm doing something wrong or if it's just the "alphaness" of v3

[thin flint]

The "my engine's alive
when I drive
you-ooh-oooh" section that starts with sultry male vocal kinda sound female to me.

[broken lotus]

Venting is GOOD! 😋

[broken lotus]

it decided to give me a duet...
I went with it.

And my songs don't need your heteronormative gender policing, thank you very much.

LMAO!

[thin flint]

Heh. Right.

[thin flint]

@final hazel Hey, trying this style mixing stuff. I got a good intro, but on the continues, how do you format the style? I'm getting the same style as the first clip; is it a matter of regenerating until it switches? I've got tags in the lyrics to try and enforce the new style.

@final hazel Unless I'm just starting with a style where Suno thinks it'd be impossible to logically switch, I don't know.

[neat nexus]

Are you doing this with V2 or v3?

[thin flint]

V3

[neat nexus]

V2 takes a while to "swing the boat" in another direction, v3 is better at making complete shifts.

That's why I asked.

Huh...

It could be an issue of the styles not being known or recognized? Or just bad luck?

[thin flint]

I'm trying to go from future pop to indie rock... When @final hazel did it, they went from ghospel to jazz to blues, though I'm trying to completely switch the genre from Electronic genre to non-electronic and I wonder if that's messing with it.

We shouldn't allow self-promotion here 😦

Wel, anything not Suno that is

[broken lotus]

they spammed every beta channel, so 🙄 it was just spam...

[final hazel]

Here is what I do, when I switch to the next part, I completely change the "style of music" box, and ussually when I mix to the next, I set in on instrumental first. I will then create a long instrumental there, but I just cut it off where I want and go back to custom mode with lyrics.

[broken lotus]

👀 this is a good suggestion... I have been missing my instrumental interludes, and forcing a lead instrument to take over the voice is sometimes too 'in my face' with that instrument...

I am not using the Continue from Time feature enough... 😓

[final hazel]

https://tenor.com/view/share-the-love-benjamin-engvid-sharing-is-caring-spread-the-love-gif-7082848330196318409

I have the feeling it also helps if you put 'intro", or "verse", or "solo" etc in the "style of music"

[tight frost]

Funny, I had this impression before, like the model is trying to mimic what it generated recently.

[half epoch]

better with bracket ex = [intro] [Key: D Minor] [Tempo: 110 BPM] [Verse ] [Chorus] [Bridge] [Outro] [end]

[final hazel]

Not in music style

Not for me at least

Then it starts repeating the words

[half epoch]

is in custom mode " lyrics "

[final hazel]

No I post it in music style with V3

With every part I change it

I’ve already burned about 9500 credits with v3 and this works for me

[half epoch]

ok i will try also , thx

[final hazel]

@half epoch this is a good example https://app.suno.ai/song/8db92e67-6b31-4726-a0ea-08fe35b571f6

[half epoch]

ok a can see the structur ! thx @final hazel

no trick to inccrease the quality ? like more clear ? no saturated ?

[thin flint]

Hmm. yeah, that's the thing, I switched style to indie rock and a few other keywords too and style didn't switch, so just wondering if I was doing something worng.

[half epoch]

https://app.suno.ai/song/0333db59-e944-46b5-924b-346bba108274

[final hazel]

Did you do instrumental like I said?

[thin flint]

No, I did it before I asked about this, I'll try that, but what about those vocals you have in the beginning? Will it still produce vocals despite instrumental being checked?

Or do you just use continue time feature to add vocals over the instrumental?

[final hazel]

I switch back and forth with instrumental option and custom

[thin flint]

Oh, so the instrumental option ensures the style actually switches with the proper instruments?

[final hazel]

For me it does

[spare cargo]

If there was a way to select a point in a chirp where you want to continue from, that would be super powerful.

V3 is really amazing, but after playing it I found myself wishing the limit was shorter.. there's more control if you generate a clip you love, and then grow it out from there; running as many 'continue froms' as you want to find the next perfect section...

[teal raft]

That is possible with v3

[spare cargo]

OMG

I LOVE YOU

[teal raft]

It's hard to see I know

[spare cargo]

haha

[teal raft]

[spare cargo]

That is so great. Thank you.

[thin flint]

Even as an instrumental, the style won't change. I guess it just has trouble switching from electronic synth to indie rock or pop rock etc. @final hazel https://app.suno.ai/song/c5ca114e-c78a-43da-bd3c-dc908af0f07a

[final hazel]

maybe not combability? I did it with several styles 🤷‍♀️

[sinful abyss]

Anyone else having problems continuing on v3? It seems none of my songs go on right

Every time model hallucinating

[neat nexus]

It can happen and it can be annoying. Depending on where the hallucinations are and what kind they are, resume from time continuations can help a lot.

[sinful abyss]

And still impossible to get 100% proper intro 😦

I see no progress in stability and predictability

[neat nexus]

I seem to do OK...

[link removed because song is not released yet.]

[neat lake]

Would like to ask you guys for observation which I can test. What do you find V3 or Chirp in general to be good at and not so good at? A broad question ik, but sure there are things I missed, which will at least make me feel like I'm not completely wasting my creds 🙂

[neat nexus]

V3 Alpha gets emotional. I mean Chirp has always been able to emote. But I think it gets even better in v3 Alpha.

[hidden seal]

Hello Everyone
Suddenly the drop down menu with V3 Alpha has disappeared from my acocunt.
I still can see it in my friend's account
Please Help

[final hazel]

check if you have open payments

[hidden seal]

What do you mean?

[final hazel]

Is your last bill paid?

[hidden seal]

I have credits, i guess so

[final hazel]

I would double check. Suno said that it's the main reason

[versed valley]

Welcome! For help, check out Suno Help

[hidden seal]

i have 2500 credits that was renewed today
do you think there was a problem with my payment?

[final hazel]

I think so

Click on subscriptions, then billing details.. you can see what is paid @hidden seal

[silver plank]

Hello , you know how did a specifical longuer for song ? Like 1min30 or 1min?

[obtuse heart]

I forgot where to type /chirp, because when i use this command in chirp-beta it doesnt work...

[broken lotus]

Have you tried the website? the v2 bot there can make 80sec the 1st clip, 60sec continues.

[limpid spade]

https://cdn1.suno.ai/fd051be2-4441-4400-ac7c-39a525f37068.mp4 some style prompts can help. Here i used "News broadcast, news reader, identity, short" to get 30 seconds or so. I used ident "a short sequence shown on television between programmes to identify the channel." but I think Suno made it identity.

[broken lotus]

that is perfect

[balmy wing]

hi everybody!
i need some help please
when i ask suno make some trap music, it makes rap music with voice, how to set promt to make the same without voice

[drifting scroll]

Really loving the v3 update! Is there any technical information available about how Suno is accomplishing this? I'm a comp sci type and familiar with the basics of LLMs but not sure how much applies to music...

[final hazel]

I doubt they will tell how the secret sauce works... but the have a previous model Bark for voice.. that is open source

https://github.com/suno-ai/bark

[halcyon flower]

Is there any way to use V3 Alpha in Chirp? Thanks for any answer!

[final hazel]

No.. Discord technical limitations.. so it will stay V1

[broken lotus]

forever?

[final hazel]

I don't know that.. But I doubt discord will change for Suno

[tight frost]

Hey I never tried duo singers, how to do it with v3 ?

[broken lotus]

mmm, I had one just happen...
I stopped trying for them, so not sure if v3 is 'better' at it...

[tight frost]

Is it equally good as v2 ? if not I'll use v2. What are the tags ?

[final hazel]

V3 is like your lover... if you say the right words, it will respond way better than v2 @tight frost

😂

[tight frost]

Problem : I already told v2 it was my lover

[final hazel]

then you are set... stop looking if the grass is greener with V3 😛

[halcyon flower]

Wow, okay. So, Chirp can't even do V2? Only V1?

[final hazel]

yes

[eager topaz]

Just when V3 starts to infuriate me it throws me something that is sooo creative and something I would never have thought of doing, so I'm happy to call V3 MY lover 🤣

[tight frost]

So I did a duet with v3 but got only 1min long song, should I add the verse and chorus tags in addition ?

[broken lotus]

the tags seem a bit iffy in v3...
When I got a duet as one of the gens, I just let it flow and it kept switching between the voices...
(I didn't interrupt it)

https://app.suno.ai/song/61d49eba-a1df-4697-8849-12cb11950062

(but I didn't prompt for it either)

[obsidian lintel]

hey can anyone here help with an issue i have with the website version?

im trying to make a song about attacking some aliens in game called elite dangerous and it says that the prompt was flagged and it cant generate it...i have the lryics here if you want to see them there is nothing in them...

unless maybe kick thier ass is what it doesnt like....

[teal raft]

Do you have a lyric that says "kick their ass" literally?
You can try changing "their" to "there"

That's incorrect but will be pronounced the same and that might be the trigger

A while ago I gen'd a little thing about beating the shit out of a vending machine

[obsidian lintel]

i did have that yes i canged it and still the same...

[teal raft]

No problem, but voilence on someone else was a no go

[obsidian lintel]

oh ok i mean its not real things just aliens in a game but i see what you mean.....

but yeh that didnt fix it

[teal raft]

Yeah the system doesn't like violence against people or 'people' or names or whatever

[obsidian lintel]

ok will try changing that see how it goes thanks

[fallow gull]

You could try "kik", but sometimes the AI tries to correct misspellings even if that is what you want.