📚┃suno-school

[fleet edge]

there are a bunch of apis accessing suno around

[broken lotus]

The are all illegal

[fleet edge]

ups

[elfin storm]

How come there's no answer?

[fleet edge]

im wrapping up very late here , working early tomorrow 🥴

[hollow furnace]

Curly brackets just for End, or for everything?

[fleet edge]

but id love to catch up with somebody in the company

[final hazel]

end

[broken lotus]

Not for everything, use square brackets for metatags communicating with the ai. The curly brackets is not talking to the ai, it is talking to a process that stops the AI.

[final hazel]

If you check this video, you can see how the staff made their lyrics, and they you see all use {end}

https://youtu.be/kB3W5xLza_k?si=lF2sPwlQGIHYWanc

[languid socket]

Hey guys, yall know how to add people talking in the background, not singing but like background noise?

[hollow hare]

Unfortunately, there ia no simple solution from what I gathered (though I did not search much).

I do not know any solutions aside from basically regenerating it if it was in the base generation.

If it was in an extension, going to an earlier part and extending (as usually additional parts kind of reinforces the reverbing and it compounds when the AI detects and emphasizes it). This is a bit more manageable as you can usually stich the parts together in a DAW.

From what I had experienced, it seems to be an issue that popped up when I use a diversive genre or ones that have a lot of vocals during a chorus.

Likely narrowing down to a specific subgenre and redoing an extension once a reverb is detected would help. Or looking into how to edit it down in a DAW

I am sure that looking in this channel will provide you with a better solution than just that

[hollow hare]

Honestly, it would be easier to use a DAW and pluck the crowd talk into the track itself (gives more control over the volume, when if occurs, etc.).

You can likely generate the crowd talk in Suno but better off looking for free samples available online or maybe using Bark, Udio, or other audio generating services to make a specific crowd conversation

[languid socket]

Ah, I see, thanks.

[hollow hare]

I forgot to mention that you can provide (parenthesis) at the end of a line to have a back vocal echo the line which might help you too (assuming if the AI bothers using it, sometimes it gets ignored)

E.g.

"Don't have a cow man" (eat my shorts)

Man, calling Principle Skinner a weiner (instead of wiener) is such an old reference... maybe back in season 1? Ugh.

Edit: It was specially a reference to a spray painting Bart did back in S1E2... yeah...

[languid socket]

Thanks, also what a weird example.

[hollow hare]

I was just rewatching the classic simpsons lol

[languid socket]

Oh yeah that makes alot more sense

Can't believe someone remebers that.

[elfin storm]

Okay

[slow edge]

(Sadly still getting a 'system error' notice on this one, when trying to extend, not sure why, a pain as I like the generation too, hoped it might be just a glitch that would sort itself :((. But seems not, sadly.) Anyone know if it can be fixed or where to go in particular to report this kind of (won't/can't extend) error?

https://suno.com/song/be3de8a6-befb-4fa3-8459-74c9370d7486 I wanted to extend this, but just keep getting a system error, I reloaded page and, tried from a different point in the music, still getting system error 😭 (and Yes I do have credits left).

[elfin storm]

you experienced the same as me

[hollow hare]

What did you put as your extend from time? What did the error specifically state? (Providing a screenshot would be very helpful)

[slow edge]

Really annoying when it's on one you really like and would like to extend. The other issue i am noticing on quite a few songs further extensions produce, poorer sound quality, they get scratchier, distorted or tinnier with further extensions.

[unique stone]

Post it in #🪲┃bugs-web if you haven't done so already. I was going to ask if you'd deleted any parts but I see this is just the 1st part so it does seem like something is broken there.

[unique stone]

Sometimes doing "get whole song" before extending further can help with this. But I get this sometimes as well - it seems to collect a bunch of frequencies and then just intensify them to the point where it rings a lot

[slow edge]

difficult to get screenshot as the error window just flashes up for a couple of seconds, but its a small box with

! An Error occured
A system error occured.

[elfin storm]

My song has an echo and there are the same tone of voice in several verses of the same song and the instrument in the song has a sound tone that is not classy

it's really annoying when editing and it doesn't sound clear but when it's put together there's a double sound and the singer's voice doesn't rhyme with the notes written

[slow edge]

I've tried from several extend from times but same error.

[elfin storm]

Suno Ai didn't fix that

[unique stone]

It sounds like there's something very wrong on the server end with that song if you can't extend it from any point. If you post it in the bugs channel one of the devs might be able to get it working.

[open imp]

Wh-?! Furrybeans?!
Are we related?

"Eyyy, it's me, your cousin from space!"

[fair shard]

Hey, I'm kinda sleepy and by a mistake I bought the yearly sub, instead of the monthly one, already sent an email to support but was kinda wondering what are my chances of getting the money back and switching the sub to just regular monthly

[hollow hare]

Should be a pretty good chance as it is a simple process.

My assumption is that they will cancel the billing and take away the paid credits given and ask you to subscribe to the monthly one.

[lofty merlin]

Wait, that's a thing? There are actual tags that have metameaning outside what the AI model interprets? Why is this not better documented?! 😅 The number of times I could have used a forced end to stop generations from going on too long and repeating lyrics.... Thank you (and @final hazel ) for this tip! ❤️

[broken lotus]

😆 it was a Dev-leak during v3alpha (the song would end but the clip would keep going)
It's apparently the process that turns off the audio clip... a 'forceful quit'... not sure if it can be ignored?

I haven't seen any other {} commands. I know some people were trying random things to try to break the bot, lol. I don't think they found anything.

[lofty merlin]

It doesn't seem to work 😦 I put {end} after about 30 seconds of lyrics, and it still generated an extra 1:05 of mostly instrumental, plus repeated lyrics.

[broken lotus]

🤷‍♀️
It may not be the same function now. maybe it is just [end] but with fancier design...?

[dark pike]

How do I get a song to generate longer instead of cutting off early? For example, some of the song chunks I have made are 2:00 long while others are 1:00 or less. The biggest problem is it usually cuts off the last line or two of my Chorus. Once that happens the AI can't seem to generate something that matches the previous Chorus lines.

[signal quartz]

有台灣人教學suno 怎麼做預告片配樂嗎？

[hollow hare]

Having more lyrics helps a lot. I find most things I generate is too fast for just 2 verses + Chorus. I usually add in another chorus or two and a bridge and an outro

You can also try putting in more Instrumental sections too to help continue (several tag examples are listed at https://suno.com)

[cerulean swift]

are there any plans for self hosting & running suno song model locally?

[solid widget]

Hi all, I tried to extend a song with lyrics but the results are all fade out outro even i did not type any [Fade Out] or [Outro] prompt in it, please help me and tell me how to fix it, thanks

[boreal rose]

Does it matter what symbols I use to prompt AI to stay focused? ( ), { } or [ ]

[dark pike]

The problem is there is no control to clip earlier portions of a song chunk; instead it only adds to a splice on the right end. I hope the devs add a feature that allows adding to the front end of a song chunk or provide more control with a seed. I've had the issue of a crappy Verse followed by a beautiful Chorus only to realize there's nothing I can do to save the Chorus. But I'll try your suggestions and see if I can get more options with more content. I'm just worried I'll still run into the same issue as I edit the end of my songs.

[azure crane]

Any ideas on how to just have the background music without any melody in an instrumental track. Can’t understand why it always create a melody in the intro of a song. Trying with several tags but nothing works.

Wonder also why they removed the Extended link from extended parts?

[echo plinth]

Good afternoon, Suno's subscription ended yesterday. There were 9,500 credits on the account, today I paid for the subscription and there were 10,500 on the account. Why did I get 1000 credits per account instead of the 10,000 I was supposed to get?

[normal grove]

Credits don't carry over.

[low coyote]

Heyyy I was walking in the city touching my screen with one hand and accidentally flagged a song 😔 is there a way to cancel the action?

[broken lotus]

subscription credits do not roll-over
'extra' credits do
https://suno.com/account

[broken lotus]

when your month refreshed, you lost the unspent 9500, received your monthly 10000, and you have 500 extra.

[echo plinth]

It's a pity that I only found out now. I specifically bought a subscription a month ago and didn't spend any credits to renew my subscription this month and make music on vacation.

In fact, I just clicked on the buy subscription button and had no idea that I needed to scroll down the page, where this warning was in small print.

I wrote to support, maybe it will be possible to return at least some of the credits.

[unique stone]

(This was discussed earlier in #💬┃general-chat where the question was also posted.)

[steel night]

Is there any list of keywords available I could use in the Style of Music field?

[broken lotus]

here's a big ol' list
https://www.suno.wiki/faq/style-and-lyrics/styles-and-genres/

and if you like dancemusic, here's another big ol' list by Mista.Lewis
https://www.suno.wiki/faq/style-and-lyrics/prompts-for-ai-dance-music/

[final ice]

question how do I tell suno ai add a bridge to a chorus ? I am using the [bridge] tag and its not working

[broken lotus]

you might try [pre-chorus] instead

but (without seeing your lyrics) the word-pattern might dictate the song more strongly than our metatags do.

If the meter and word count of the bridge is too similar to the verse or chorus pattern, Suno will often blur them together....

[placid spade]

Thanks for having me! Lots of topics above - just looking for a short answer on if I want to make the song longer - how do I do that? I tried extend but it gave me a different version of the track. Thanks all!

[broken lotus]

switch on Custom Mode.
Change the lyrics for each Extend...
Extend from part 1 to make Part 2
Extend from Part 2 to make Part 3...

[final ice]

thanks I had to regenerate twice before it finally added the bridge I wanted into the song I forget every tag don't work immediately

[thick stirrup]

what would be the motivation for Suno to do this? if they let you download their model, and use locally, you don't need them anymore. And then, how would they make money?

[unique stone]

Some apps can run locally but require a subscription to use, so they could to that. The advantage would be you could use your own GPU so have exclusive use of the hardware, and potentially unlimited generation / no credits involved.

It's probably easier to develop/support, being a "web app" though.

[thick stirrup]

but they would still have to let you download their model weights, which is probably something they don't want to do, as that's their "secret sauce".

no matter how encrypted they are, they would have to be decrypted to send to your gpu...

[unique stone]

Valid point also.

[steel night]

Is it possible to get back Credits for
a nice song but unfortunately awful audio?

[steel night]

Thank you very much!
This is what I’ve been looking for.

Could it be that the audio is getting worse the more I try? After 5 tries more and more it gets tinny and rattles.

[stuck pollen]

been noticing the exact same issue today, 200 credits gone now and the vocal quality in particular is sounding like an old damaged cassette tape

[broken lotus]

as we extend and extend and extend...
it is like a Xerox of a Xerox of a Xerox...
the quality degrades each time... a very 'clean' song in the first clip may stay relatively clean, but a song with lots of distortion guitars and reverb (metal, shoegaze) will degrade faster because there is a lot of sound in the high frequencies...
(genres like DnB, Electro are more bass, and don't degrade as fast)
Unfortunately the Extend hears that 'ringing' reverb and starts to duplicate it as part of the song....

[opaque pasture]

No idea if it's a "technical thing" but can you see the likes of songs in your library when you are a free user? 😅 Is that a setting?

[broken lotus]

I think it will show on the song's page.

[stuck pollen]

so this suggests that the AI isnt storing any sort of meta info about a track. it just using the mp3 to extend from? could we send a message via the "style of music" to tell it to quit doing certain things?

[steel night]

Does mean? … Guys, we pay money for it.

I can expect consistently good quality. Otherwise you'll have to reduce the credits for each new attempt!

[eager owl]

I always use custom mode and use my own lyrics. Sometimes I could get a very good melody/music but couldn't use it because of hallucinations. It would be nice to be able to regenerate the same melody until all the hallucinations are corrected. Is there a way to do this?

[broken lotus]

it might be possible to 'interrupt' that noise with a
[break]
or by sending it in a new direction...

I have found some genres will switch between a quiet verse and a big distortion chorus that rocks out... when it gets to the next verse it will chill again and stop all the noise...
dreamcore and sinistercore seem to do this...

My method is to ruthlessly avoid any clip that starts that ringing sound because it's very hard to get rid of and will kill the vocals by the end

especially don't Extend from a section that has a lot of that noise...

[broken lotus]

Devs are calling this 'in-painting' (like from image-gen)
it's expected for v4 (currently in the works, but no eta)

[opaque pasture]

Found it! Had to go to my profile

[eager owl]

Thanks. I hope v4 comes out soon and with better sound quality. It'll be great if you can also upgrade your v3 generations to v4.

[steel night]

Suno is not willing to extend my track. Tried 10+ times now. Credits Refunded. But not working either.

[steel night]

So many failures today and over 400 credits used. I'm really considering not renewing my subscription!

That's a shame, because I've actually had a lot of fun up until now.

Actually, suno is like gambling. I already feel addicted.

[unique stone]

Apologies if you have already tried this, but does changing the "extend from" time help? Sometimes Suno wants to end a song and changing this helps it to keep going.

[steel night]

I’ll give it a try

[unique stone]

Even just a few seconds difference can make it forget that it was going to end. (V2 used to not have this option and it was VERY frustrating when a good song stopped)

[steel night]

Yes it is frustrating.
I tried but not working. Anyway I will let it go for now. Thx 4 ur help.

[placid spade]

I tried that but the beat of the song for part 2 and 3 are always different from the original

[broken lotus]

Then you are doing something wrong.
I suggest looking at other people's songs where you can see the structure, and hear the extends following the original.

[stuck pollen]

i often find that getting the whole song and then extending from this fixes a lot of the issues that i hit when the extend just wont work for whatever reason. A "remaster" option would be amazing though. It could fix a lot of the annoying issues that crop up over extensions such as degradation of vocal quality and would in turn potentially allow us to remaster a V3 song when V4 comes out to improve it.

[sly goblet]

Hello, I need help.
I’m trying to create several songs, but to continue with the clip it only plays 3 seconds. For several times and continue to generate part 2 for 3 seconds.

[broken lotus]

listen to the end of the clip you are extending FROM... it has ended within the clip.

Use Extend From Time, and set the timestamp a second or 2 before the ending stops (or the extended clip will continue the ending and end again)

(this is why you are refunded)

[keen cloak]

is there a meta tag to force a guitar solo?

[sly goblet]

Thank you for your feedback.
Yeah, now it worked. Have a good job.

[neat nexus]

Force? No. But you can try using [guitar solo] and seeing if Suno feels like doing one today.

[broken lotus]

[virtuoso guitar]
[acoustic guitar interlude]
...?

[keen cloak]

[guitar solo] does not work, will try the others

[teal raft]

virtuoso is king for me

[neat nexus]

I tend to have better luck by pairing more than one tag together.

Try something like this:

[emotive blues riff] [guitar solo] [melodic instrumental]

[stuck pollen]

it does work, you just have to try it a couple of times. Burns through credits and is annoying but it does get it done. ive also found putting "guitar solo" into the style of music encourages it a bit more

[neat nexus]

Oh yeah!

[teal raft]

Most problems in Suno are solved by throwing credits at it

[neat nexus]

I generally include extended solos in the style for certain types of songs.

[teal raft]

Also if you are having a good time managing the breakdown with multiple generations I'll get real wild with cutting generations down to seconds

and slapping them together

breakdown = the noise that tends to build as you pile on generations

[glacial scarab]

Yea.. I really wish there was an "unlimited" plan for like idk 90$ or something.. maybe reduced priority on peak times or whatever to make it still profitable but i "need" it.

[teal raft]

pat spoke directly about a tier above, said "maybe later, not a priority right now" so

[glacial scarab]

Well I am hyped.

Think this month I might do a second sub. Test a lot of stuff and then take a break. Postprocess what I have write a few "real" songs (Most of mine are comlete/partial memes) and then resub later. Might burn myself out otherwise I fear. And the stuff I really want to test would require too much credits currently 😄

[steel night]

Same experience here today.

[unique stone]

This is probably what I'll do if I find myself busier editing stuff than generating. I know yearly sub is tempting with it's discount but being able to take a break is kinda nice. No commitment. Wooo

[stuck pollen]

the recycle bin for your account may explode from the unlimited amount of junk that comes from unlimited creations 😂

[glacial scarab]

Ha my library already a junkyard with some finished pieces throughout... could use the time off to clean that out as well I suppose

[jagged narwhal]

I have a question about bark but is it gone now?

[hollow hare]

It is available through huggingface though I do not know how many people available have experience with bark itself: https://huggingface.co/spaces/suno/bark

Source code is also available on github: https://github.com/suno-ai/bark

[unreal trench]

true indeed

[warm palm]

Is suno available to clone voice

[broken lotus]

Suno uses a generative voice (based on Bark). they do not have voice cloning.

[spring spire]

Hello, i still spend many time in Songs, but to many times i get an Great Rhytm but the Sound of Voice and Synthesizer are to high ( overrides) any Times also the Bass, Sorry but louder is not better. And nost of the Voices, also Male going to high. Why can i set not a Prompt for example Max 95 db ore another in this art, ore for the Voices a Setting for reducing. One of my Great wish, if i have a Sound what i like, but i have a Part where the Text sounds not goog, that i can Parking for this Projekt this sound and ich can make a Finer Tunning with the Text. So a little bit more me ore not Sound made ny AI, only Supported by AI. To paid Credits also for Fine Tunning is no Problem. So i have many more Ideas to make SUNO better ore to one oft the best. But please Work on the Overrides, to make a song/Voice Louder is not so hard than a delete of a Override. And sorry for my funny English, but its better than the German of any Persons.

[winged fractal]

Is there any way to change genres mid-song without extending? Like with tags so you could do something like [Rock] or [Pop] to influence the feel of a section or singer?

[wheat quail]

Just wanted to know, considering my goal is to create a game, how and how should i credits Suno?
(Anyway tbf with players they need to know it's AI generated)

[wheat quail]

gonna give it a fair try

[glacial scarab]

yes and no. If you put multiple genre in the style it can work. But you can not control it 100%. It will switch sometimes between them but also not really be the genre at the same time. But it is not predictable just luck. For example this is a single gen with this style:
scary|evil, melodic female singer-songwriter soloist violin reggae, death metal virtuoso, polyphonic violin virtuoso

[wheat quail]

https://suno.com/song/fb3890ff-2696-41e5-9d2f-ae800f66d490
Another try so you can fairly decide if it's ok or not lol

[lucid geode]

Is there any way to make the lyric hmm with song for a short bit?

[stuck pollen]

Is Suno having some technical issues today? It feels like it is a total mess. Im 700 credits into a song now and it just will not repeat the chorus a second time. It just keeps changing the melody. Sometimes it will do it but there will be glitches in the sound. Like i understand this is a relatively new AI but it has never been this bad for me before and i am wasting my money trying to get it to do something it previously did and wont do today. I normally get the whole song and extend from this as it will help it get back on track for the pattern of the song but this just isnt working for me today. just non stop glitches and issues all day

[unreal trench]

i experienced the same a few days ago and it can be frustrating. i heard that if you reduce the time of the extension by a few seconds, you may increase the chance of having a better hit in part 2 or 3 (or whatever part it is)... give that a try, i hope it helps.

[winged fractal]

Anyone have any tips for reducing the frequency of the lyrics 'melting' into the instruments? I've been getting several compressed songs that just start to sound like noise.

[stuck pollen]

saw someone mention this and have tried it. its hard to get the lyrics right. i dont normally have issues with this either but it will spit out words all messed up and then get back into the flow but sure the song is ruined after this.

[winged fractal]

The lyrics aren't much of the problem, it's more the actual composition where the sound metls together and you can't differentiate between a voice and an instrument

[nimble smelt]

Help my web is glitch

[nimble smelt]

Sensor warning for flashing or shake

[true jungle]

I have a problem can I md an suno admin

[unique stone]

Someone here might be able to help if you say what problem you are having?

[true jungle]

is about premium money I paid one month premium but now suno automaticlly pay another month and I want to cancel it

and I called premium but I want my money back

[unique stone]

Ah, so it's not a technical issue. You'd need to e-mail them at [e-mail]

[runic vale]

hello i have a question. where can i find explore random creation feature?

[true jungle]

I e-mail them but they ignore me

[final hazel]

https://youtu.be/kB3W5xLza_k?si=lF2sPwlQGIHYWanc this?

[runic vale]

no when i opened this site some minutes ago .there was a list of random music of other's creation

is that new feature?

[coral coyote]

https://suno.com/song/1d86ad97-5a60-4fa4-9900-33e85203a6d7

https://suno.com/song/19480c80-da4c-4674-aaff-758af7d7fb18

[hollow hare]

Both the Explore and Home pages are new (well, Home was updated and used to be called explore in the past iirc). The Explore page has a carousel of different genres displayed you can, well, explore and listen to. The Home Page has been updated so that it lists the top, new and trending songs, different genre playlistsz and music you liked

[coral coyote]

When i open create my song list is shuffling...Anyone else with same problem?

https://suno.com/song/e3d79a67-6964-4d24-acd7-c87fec182e24

[runic vale]

it's not categorized and completely random ,listed just like my creations, not the carousel one btw

[hollow hare]

Can you show a screenshot please

[runic vale]

sorry i didn't have one . it dissipated on reloading

I found the answer in this thread before, it seems a glitch......

i thought it were new feature

[nimble smelt]

Anyone else having this shake?

[unique stone]

Yeah someone posted in #🪲┃bugs-web - what browser and device is this on?

[nimble smelt]

Chrome iPad since it's the only one I can download on

[unique stone]

Ok I've mentioned this in the bugs channel in reply to myself linking to your message here so hopefully when the devs see it and the other person's issue it'll be easier for them to follow 😅 I might try Chrome on PC later and see if it's the same

[neat field]

how can i see someone else's Trending song video, showing all the prompt animations? so like if i go to the below song, how do i see the video that would naturally show up in discord if i directly linked it here?

https://suno.com/song/1a9b5713-0f0f-4c04-97e2-dad2597e0329

[glacial scarab]

Make own Discord Server
Post Video
Profit

There might be other ways but if on is strict those would be TOS 👀

[fervent siren]

It’s happening on my safari too. Plus trying to type and delete anything is a stuttery slog.

[wintry saddle]

This is a mobile chrome issue - you can hit settings and set it to desktop mode and it stops.

[nimble smelt]

Mine is switched to desktop

Has always been

[wintry saddle]

rotate screen / turn rotate off

[solar snow]

It used to generate 2 minute songs, but now it only generated 1 minute, anyone know how to fix

[hollow hare]

Add more lyrics, instrumental tags, etc. https://www.suno.wiki/faq/metatags/

Basically want to pad it up, esp. If you are using the default 2 versus + chorus

[solar snow]

I'm using my own lyrics and correct tags. the problem is it will litterally stop mid word or verse once it reaches a minute. it used to do 2 minutes

[hollow hare]

Is that during the first/base generation or Extend? Extend is limited to 1 min as the other min is used to analyze to song style before the continuation point

[solar snow]

the initial generation

[hollow hare]

Hm, strange, don't think I encountered it stop talking in the middle of a sentence at 1 min during initial generation.

Verify it is lacking the Part 2 tag (in case there was an accidental extension, I did that once before). If so, you should report the song as a bug

One thing that might help is having an ending metatag (e.g. Fade out, big finish, end, etc.) Which will tell the AI it is not done yet

[solar snow]

it's been happening for about a week, it just randomly stopped doing 2 minute videos, it's not just this session. it's almost like I have an account limitation because I made too many songs

[hollow hare]

I noticed mine were around a minute as well these days but that is because the styles I have been experimenting with were on the faster side

[broken lotus]

try adding
[interlude]
or
[Break]
after the last line in your lyrics.
It may still be short, but it should play some instrumental before ending

[solar snow]

This is what it looks like, I have [verse 1] [chorus] [verse 2] [chorus] [verse 3]

and it is only generating 60 seconds.

[solar snow]

the problem is that it is only generating 1 minute, no longer doing 2 minute generations. I understand it for extend, but this is for the initial generation

disregard, it just worked.. it says 1 minute but plays for the full 2 mins

[teal raft]

If you play it while it's streaming (before it's done) the bar shows 1 minute but like you see, it plays past

Any song you thought was 1 minute before is more than likely past that if you go back and look

[solar snow]

Negative, the previous ones would litterally cut off mid word

[teal raft]

Well that happens sometimes too yeah

[merry storm]

this might not be technical, but i couldnt find an answer when browsing the website. Who owns the music created by Suno? Is it the creator' or does suno own it, or is there a specific premium subscription that then gives ownership?

[glacial scarab]

it's in the FAQ and in the TOS.
If you pay the songs are yours if you are free they are sunos

[merry storm]

thank you! im a bit tired so thats an oversight on my part

[teal raft]

Also if you have any paid sub, they are yours

Fully

The only difference between the sub tiers is credits you get every month

[merry storm]

if i have created a song in free, i like it, and then buy a sub, is that first song also mine?

[teal raft]

No

The generations are tied to when you make them and as of right now there is no way to separate them

Afaik there is a watermark

[merry storm]

thanks for the info!

[sudden spear]

I have a generated song with voice syncopation I like but it has no intro. Plus I want to use the voice and same singing style in a custom song with custom lyrics. Is it possible? I tried extending at very beginning and adding an intro but doesn't work.

[covert reef]

im trying the chirp command in a private message but it doesn't work

[final hazel]

Because the bot is gone. You can now only make music via https://suno.com

[covert reef]

thats a shame, but ty!! is the software free?

[final hazel]

You get 50 free credits per day

[covert reef]

sorry i don't know what that is? is that good, is it to limiting?

[final hazel]

You need credits to create music. 50 credits means you can hit the create button 5 times.
And if you want more, you can take a plan.

[limpid nova]

also [outro] works well for me 🙂

[covert reef]

how cruel 🤣 ... do you know any free alternatives?

[hollow hare]

Most, if not all, AI generators are locked behind some sort of credit system because of the resources the AI takes.

Being able to generate 5 songs (total 10 tracks) per day is a pretty good deal as some services only gives a limited trial credit that cannot renew.

Another music AI available is Udio. They also have a limit but I cannot recall what it was. I found that Udio generates better vocals but I do not love the styles as much as I love Suno's

[covert reef]

ty for replying! i think ill sign up with suno then, do the 50 credits perday stack up or reset so you only have 50 per day?

[hollow hare]

Just replenishes to 50, no stacking (AIs are too intensive to stack).

The daily credit refreshes at 0300 UTC+0 (<t:1713236400:t> in your local timezone)

[covert reef]

thats not even enough -_- .... can't even experiment with that..

[hollow furnace]

I find even a paid account is not enough as the fever grows ...

[glacial scarab]

then buy a sub 😏

[unique stone]

I was just thinking, some of the paid subscribers experiment to the point that they run out of credits too xD

[glacial scarab]

wonder who those people are 😏

[hollow furnace]

what is the best way to ask a song to steadily build in intensity? genres, tags inside, other?

[hollow hare]

There is the [Build] tag IIRC
But that could have been more for a beat build for the [Drop]... https://www.suno.wiki/faq/metatags/instrumental-tags/

I wonder if the AI would recognize [intensify], no credits to test it out though

[hollow furnace]

I see the end of the month will be rough going 😉

[lyric anvil]

Has anyone consistently used two voices in a song? One for verses one for chrous? Having some trouble with it.

[broken lotus]

[rapped verse] and [pop chorus] sometimes works...
some tips here:
https://www.suno.wiki/faq/style-and-lyrics/duets/

[lyric anvil]

Does anyone know the correct structure for multiple tags?
EG: [Verse, Male, ] or [Verse And Male And ???] ?

[dark pike]

@lyric anvil Here's a video that may help with the tags. I'm still working on mastering Suno too as it doesn't appear to have in depth control over what is generated.
https://www.youtube.com/watch?v=lj8utBd_IUM&list=LL&index=2

[lyric anvil]

thanks for the responses everyone 🙂

[lyric sonnet]

I started a stupid song about socks but it evolved into an impressive groove towards the end and now i want to make it into a song with different lyrics by extending, but somehow suno keeps ending the song immediately, or falls back into old lyrics, or quality degrades massively; any way around that? Am I leading the train off the tracks and it crashes?

[dark pike]

Dang... I've been in a similar situation but don't have a good solution that works either. I'm curious as I'd love to fix some of the previous songs I've written. I'd REALLY love a way to splice song chunks at the beginning and not just extend at the end too.

[lyric sonnet]

hoping for the next version..

[plain orbit]

Moderation already couple hours doesn't allow to generate music for subscriber. This is normal?

[hollow hare]

Sorry, I did not understand that. Are you saying you are getting a moderation error when trying to generate your lyrics or something else?

[plain orbit]

When trying to generate music yeah with lyrics. It doesn't allow to generate at all always a popup from bottom about flagged for moderation. Couple hours later same message. How long to wait for that moderation?

[hollow hare]

Ah, it means it got censored from ChatGPT not liking something as it is the moderator. Ask OpenAI's ChatGPT itself about "what would censor the following: [content]" and it should tell you.

Usual causes are using swear words or interpreting harmful situations. Even if it worked before, it could be a small change caused its interpretation to become different.

[plain orbit]

I'm not about chatgpt. I'm about Suno.ai site.

[hollow hare]

ChatGPT is the one that generates the lyrics for Suno.

[plain orbit]

No lyrics is mine in my case

[hollow hare]

???

[plain orbit]

I use my own lyrics. Still suno.ai not generating music with my own lyrics.

Moderation

[hollow hare]

So you have Instrumental and it is saying your lyrics is moderated?

[plain orbit]

Not lyrics moderated

[hollow hare]

Well, not sure what else I can tell ya. I know Suno uses ChatGPT to check the lyrics for moderation.

If you want to find out the cause, like I said before, you have to go to it and ask what is causing censorship in the lyrics.

Otherwise you will likely have to delete line by line to find the issue if you do not want to do that

[plain orbit]

Here is exact error message: Couldn't generate that. Prompt flagged for moderation

I sent an email to suno support

[broken lotus]

You don't need to wait for human moderation (there isn't any)
it is telling you there is something in your lyrics it doesn't like (probably violence, self-harm, or explicit sex).

Nothing else will happen. You are suppose to change your lyrics and try again.

In reality, this is vague, and we don't always know what the AI objects to. That's why Paimon suggested that you paste your lyrics into free ChatGPT and ask "why are my lyrics being moderated?" GPT will give you a list of reasons.

[rapid bluff]

or paste it here if you want and ill tell you lol

[plain orbit]

That lyrics was fine to generate music, after couple times i tried to continue the same lyrics moderation started. So same lyrics was fine and no moderation. And then suddenly moderation appeared for same lyrics?

[rapid bluff]

unless it in a different language than I wont tell you

[broken lotus]

Yes, it is AI and often quite stupid...

we have to work around the limits of the technology

that doesn't mean 'give up' that means understanding what the AI is thinking, so we can get it to think something else

[plain orbit]

So ai moderates my own lyrics too?

[broken lotus]

for violence, for slurs, for explicit sex, for self-harm...
all the things chatGPT censors, yes.
the censor is chatGPT....
... then the lyrics go to Sono to make music.

[thick stirrup]

yes

[plain orbit]

I found the word it doesn't like - it is "skeleton"

Very harm word

[rapid bluff]

its not words it looks for, its contexts of said words

[plain orbit]

My appologies

[broken lotus]

it is AI... probably not a single word. It is the meaning.... or, what the AI thinks is the meaning.

[plain orbit]

I changed that word for similar but different and all good

[broken lotus]

that's how we do it. 😉

when it MUST be that word, we misspell it so GPT doesn't censor, but Suno sings it.

[quasi geyser]

Всё не по-русски я не понимаю

[hollow hare]

https://discord.com/channels/1069381916492562582/1154779864851812422

[regal river]

hi, if i subscribe, do i still get the 50 daily credits ?

[light harness]

I would like to know when it will be possible to improve Electric Guitar sound effects? Now it sounds too keyboard, especially for solo. @hollow hare

[hollow hare]

From what I read what others had said, no, not while you have any of the subscription credits available (as it gives you all the daily credits for a month at once plus an additional bonus amount). However, if your credits bottoms out to <= 50 credits, then it will go back to the daily refresh until you top off or ithe next month credits refreshes (the credits do not stack on the next month).

Hopefully someone verifies what I had mentioned soon

[hollow hare]

Unfortunately I am just a regular member, so I do not know.

However, maybe some onomatopoeia may help?
E.g.[shredding electric guitar] Bwiiing-BRANNNG CHUGA-CHUGA-CHUGA(Don't judge my poor sound effect skills ;_;)

Based on the last section in
https://www.suno.wiki/faq/metatags/instrumental-tags/

[light harness]

Thank you, buddy.

[wispy nebula]

What is going on? Genres aren't recognized at all right now. Wasted a lot of credits.

[thin nebula]

Keep in mind it could be copywrite from another song. I've had that happen....

[thin nebula]

Mines working?

Refresh?

[graceful condor]

Hello, good morning, I am currently trying to delete my Suno.ai account and unlink my bank card as well as cancel the pro plan, however, the website does not allow me to do so, I cannot delete my account or cancel my pro plan, I have recorded a video where I explain this problem. The email with which I created this account is [e-mail]. Given this problem, I would like you to delete my account from your servers

[orchid ember]

It seems that suno has been getting worse in recent days. Garbage is generated instead of vocals, a week ago the vocals were much cleaner. There are a lot more artifacts in the instrumental than there were. Distorts the structure of the text. Developers, return everything as it was!!

[autumn jolt]

Interesting article, especially the thoughts around what to call people who use Suno and the lack of control. https://theconversation.com/ai-can-now-generate-entire-songs-on-demand-what-does-this-mean-for-music-as-we-know-it-228937

They clearly haven't realized you do have control, if you count burning through 20-30 clips to get what you wanted. 😅

[cobalt spear]

for 7 day ore more. for one song i just need hundrets of credits and hours of work-
4 weeks ago, i made 10 songs in one day.

whats goin on Suno??

[autumn jolt]

Yeah, I've noticed this. I'm kind of taking a break at the moment, wondering if I should wait for a new version. Seems a lot harder to generate clips I like and extend songs that continue the vibe, e.g. remember the melody, chorus, etc without tons of static/artefacts

[cobalt spear]

if i create the song after 2 parts, its horrible as well 😦

[light harness]

Just wondering why vocalist is still active in the EXTENDED part when the INSTRUMENTAL button is turned on? and prompts were given for [Guitar Solo], [Acoustic Guitar Fitting]

@broken lotus

[obtuse fulcrum]

Trying to create a song that has an robotic like voice that I can't quiet get it right, has anyone tried and succeeded ?

[broken lotus]

I have used:
monotone
vocoder
robotic
most success is with genre that have robotic voices (synthpop)

[wet minnow]

https://suno.com/song/c4f8aee3-4346-4562-8bd7-09aebb9bd30c 😄 like wet mentioned ... robotic is quite good, also vocaloid

[broken lotus]

Instrumental button is more about the UI, I think... (just closes the lyrics box)
I sometimes add
[instrumental]
to the lyric promptif it was singing before and I need it to stop.

[obtuse fulcrum]

I'll give it a crack, thank you

[grizzled slate]

Good Morning All! Two questions:
1, Can you add BPM to Style of Music, or is this dependent of the SM I'm using?
2. How do I create a gap in the lyrics? For example, I want to sing, I [Pause] Want [Pause] That [Pause]. Suno seems always to put words together, if you know what I mean. I want to break up the verse or chorus to sing as I want it to be sung.

So, does anyone have any working solutions, please?

One of the significant things I've found is that the lyrics and the lyrics tags you use are very much dependent on the style of music you are using. It is frustrating that you try a Lyric Tag, and it works, and you think, Great! You then come to try it again, and it doesn't work!

Having said that, even if you get a Lyric Tag to work, there is no guarantee it will work in the same SM it worked in before, either!!

Constant Lyric Tag would be a great boon to Suno! This is my opinion, so if you disagree, that's fine!

[autumn jolt]

Hiya, re 2 I think all you can do is include tags and line breaks and keep generating clips until you get the result you want - or something close to it - see below. It's hit and miss, though. You can structure something like this and Suno will often launch into what sounds like a chorus halfway through the bridge. Just gotta keep smashing the create button and play the numbers. 🙂

[Verse]
Line 1
Line 2
Line 3
Line 4

[Bridge]
Line 1
Line 2
Line 3
Line 4

[Chorus]
Line 1
Line 2
Line 3
Line 4

[broken lotus]

I...
try...
to break up the lines....
like this...
sometimes...

[autumn jolt]

Yeah, just try different formatting. You might be doing the 'request' right but only get things to flow as you want 1 time in 10/20

[broken lotus]

i easily spend the most credits whittling down my lyrics to fit the rhythm and cadence I'm hearing in the clips.

by the time I am continuing the song, I have that rhythm down and am getting (semi-) consistent results...

Lyric pattern is definitely stronger than metatags...

(admittedly, I am the type that likes to 'collaborate' to find the song, I'm usually not trying to force a specific sound)

[weary wharf]

"Expressing emotion without feeling it"
Hmm... I think you can teach AI about what describes which feeling when singing. That depends on the training data and your description quality.
Still won't feel the feeling, but will understand how to describe it via music.

[wet cargo]

I think that to make better music with AI, we should create a studio, where you can cut clips, modify the lyrics in pieces, change the music and, very importantly, be able to add instruments, rhythms, literally like a studio. like this:

[autumn jolt]

Yeah I do edit lyrics as I go sometimes depending on what I'm hearing in the song. One of my favourite things with Suno is when you get a certain line or several lines sung in a really catchy way, and I wasn't planning to use that again, but I build the later parts of the song around it to repeat it. That feels weirdly like collaborating in a way.

[autumn jolt]

I mean, a lot of singers can phone it in and sing something that sounds expressive without feeling it! I don't really attach a label to what I'm doing with Suno. I'm making music/songs, that's factually indisputable. Am I a creator? Consumer? Curator? No idea.

[autumn jolt]

I'm not sure how well you can import a Suno track into a DAW and separate anything. I signed up for Soundtrap and got immediately overwhelmed! I love this idea, at the very least being able to remove the vocals from a track. But I would imagine there would be an issue that the different sounds in the track aren't actually recorded individually? Suno could remove what sounds like a guitar or drums but it wouldn't necessarily be clean separation?

[broken lotus]

Agree.

the image-gen equivalent would be to render the full image, segment the image into layers, and then regen each segmented layer to create what wasn't there in the original.
Then composite everything in the right depth layers and deliver a photoshop/aftereffects file.
😏

these do exist, as services for fx. they are expensive

[weary wharf]

I use RipX to get the stems, clean + mix and sometimes re-create the seperate piano / bass / drums parts. Keep the vocals.
I correct for some mistakes, move a few things around.
Also for a lot of times, come back and create some additional vocals re-extending from some parts, like add 2 more chorus right after the chorus to get some different vocalizations out.
It's still a cretive process, and only thing that is made simple is finding the best melody for the lyrics. And still I need to spend a lot time to find that.

[karmic rampart]

Is there a list of commands like "[Verse 1]," "[Chorus]," etc? I'm wondering how to do male voice, female voice, male narrator, robotic narrator, intro, outro, piano solo, etc. Thanks!

[broken lotus]

you have it intuitively, you just need to put those in square brackets at the right time in the lyrics...
[male narrator]
[piano solo]

https://www.suno.wiki/faq/metatags/

[brave scaffold]

Hello, I have a paid subscription on Suno and I make songs on Suno with my own lyrics. Can I publish them on digital platforms such as spotify?

[broken lotus]

yes, songs made with a paid subscription have a commercial license, you own them and can sell, upload to streaming services, use in professional/commercial projects.

the songs keep the license, even after you un-subscribe

[brave scaffold]

Thanks

[brave scaffold]

There are songs that I created before getting a paid subscription that I really like, is there anything I can do to use them?

[dusty marlin]

Hi, just askingg its not allowing me to generat some songs??

[karmic rampart]

Does Suno care about capitalization of the commands in brackets?

[broken lotus]

We all had those early songs that convinced us to subscribe... tbh, I feel I am a lot better at prompting the AI and adjusting my lyrics... I had a strong emotional attachment at the time, but in hindsight, they were just songs... and I have recreated them since.

I think you should remake them with subscription..., there is a digital watermark in the audio... for a commercial project it just makes sense to own outright.

(I have been here since September, never heard anyone say that Suno granted commercial rights on their early songs)

[broken lotus]

there is a content censor, which is ChatGPT.
You might try pasting your lyrics into free chatgpt and asking why they are being censored. Chatgpt will write you a list of reasons....

[broken lotus]

or you may be out of free credits for the day?

[broken lotus]

I don't think capitalization matters.... our words get broken into tokens like image gen or LLM..., long phrases can get lost in translation...., and we can make up new words (portmanteau) that only an AI would understand... based on how our text gets chopped into tokens for the AI to interpret....

[dusty marlin]

ohh i think thats why!!!!!

thanks!!!

[weary wharf]

Not from my knowledge. Lowercases tend to bleed into lyrics though. I prefer [CamelCase]

[gaunt radish]

hi why keep error generating lyric?

[hollow hare]

Likely it got censored from ChatGPT not liking something as it is the moderator. Ask OpenAI's ChatGPT itself about "what would censor the following: [content]" and it should tell you.

Usual causes are using swear words or interpreting harmful situations. Even if it worked before, it could be a small change caused its interpretation to become different.

[little cypress]

I think I lost one of my songs, is it possible it was automatically deleted? Can I restore it in someway?
It gives me error 404
https://suno.com/song/30ec1305-5693-40bc-8be4-bbead648fdc3

[true bridge]

Suno is broken. 20% of my instrumentals are throwing in what sounds like scrambled copy-written songs. https://suno.com/song/af4aa271-983e-438f-b75f-b72b62fc0d8a

[final hazel]

This is called hallucinations. It happens if you have chosen a genre that normally has lyrics in it. Then the AI wants to sing or rap. It's the equivalent of chatGPT making up things. So it's not broken... it's just something the AI doesn't get yet.

[little cypress]

yeah but instrumental should be enforced more strongly by suno to avoid that

[final hazel]

it takes more development to understand how to do that.

[onyx fulcrum]

Just came back after a couple of months, I am not seeing any chirp channels?

[final hazel]

Because chirp is retired. You can now only make music via https://suno.com

[onyx fulcrum]

ty!

[oak trellis]

/home/bark-tts/venv/lib/python3.11/site-packages/torch/nn/modules/conv.py:306: UserWarning: Plan failed with a cudnnException: CUDNN_BACKEND_EXECUTION_PLAN_DESCRIPTOR: cudnnFinalize Descriptor Failed cudnn_status: CUDNN_STATUS_NOT_SUPPORTED (Triggered internally at ../aten/src/ATen/native/cudnn/Conv_v8.cpp:919.)
return F.conv1d(input, weight, bias, self.stride,

what can trigger this?
i can mostly generate an audio file and sometimes it just crashes and the whole system reboots.

import torch
from transformers import AutoProcessor, BarkModel
from scipy.io.wavfile import write as write_wav
from IPython.display import Audio

def main():
device = "cuda" if torch.cuda.is_available() else "cpu"
print(f"Using device: {device}")

processor = AutoProcessor.from_pretrained("suno/bark")
model = BarkModel.from_pretrained("suno/bark").to(device)

while True:
    text_input = input("Enter the text to convert to audio (type 'exit' to quit): ")
    if text_input.lower() == 'exit':
        print("Exiting program.")
        break

    voice_preset = "v2/de_speaker_8"
    inputs = processor(text_input, voice_preset=voice_preset, return_tensors="pt")
    inputs = {k: v.to(device) for k, v in inputs.items()}

    with torch.no_grad():
        audio_array = model.generate(**inputs)

    audio_array = audio_array.cpu().numpy().squeeze()
    sample_rate = model.generation_config.sample_rate

    # Save the audio to a file
    output_path = "/var/www/html/bark_out.wav"
    write_wav(output_path, sample_rate, audio_array)
    print(f"Audio saved to {output_path}")

    # Optionally play the audio in a notebook
    # Uncomment the next line if running this script in an IPython notebook
    # display(Audio(audio_array, rate=sample_rate))

if name == "main":
main()

this is the code, dual 3090 setup

[grizzled slate]

Hi and thanks for the reply... I'll give it a try!

[grizzled slate]

Hi and thanks for the reply. This is an interesting one.... if you take out those lyrics, what do you think will happen... Even how you write the lyric tags is strange.... I've tried [ Verse ] and Verse... and both work, sometimes! It is hit and miss, but this is where I think Suno needs to develop. 🤞

[hardy lodge]

Hey i'm new here and new to AI all together got a couple questions. 1 when making metal music such as alt and nu metal. is there a way to make the voice clean instead of screaming and vice versa? Having a prompt for that would be good i think. Also I've noticed with the more comfortable i get and the longer mysongs get, the more robotic and tinny the vocals become is there a way to avoid this?

[graceful condor]

Hi! I'm no expert but I have this happen to me as well. Robotic voices I think are kinda like a lottery honestly, sometimes it happens and sometimes not. For longer songs, I try to slow down the tempo during the bridge, which sometimes fixes the voices

[hardy lodge]

You can slow down the voices? I've mainly been using [Instrumental interlude] even for solos as [guitar solo] doesnt really work lol

[graceful condor]

yeah, sometimes if you put a [Bridge] the AI slows it down, but if not you can try to reinforce it (tho it doesn't always work)

I did this in one of my songs because the voice was getting out of hand lol, it did help a bit

[hardy lodge]

oh i'll try that ty

[unreal trench]

😖 .. just burned tons of credits again because Suno just wont take the entire text of my extension....

[hardy lodge]

It's ok I burned through over 2k cause i couldnt find a first 2 minutes that i liked. x-x

[broken lotus]

Just extend on the extend?
No need to keep burning credits.

[hardy lodge]

Are you allowed to use lyrics from songs you've heard before and mix it with a different genre of music?

[grizzled slate]

I've reached that, Head Banging up the Wall! Moment!

Using an Instrumental as an Intro to the Song. Extend it, set the time to extend the instrumental, add some lyrics and click Extend!

No Instrumental Intro, just straight into the first Verse!!

Today has been a tough one. I'm calling it a day, but I could really use some understanding and empathy from you all. 😩 If you know what I'm going through!

[unreal trench]

trying to make part 3 with another verse and a chorus but it only give me the chorus and totally ignores the verse

[broken lotus]

There is no reason to tag me just to complain.
I can't give you advice without seeing what your prompt is, or hearing the music.

"It works for me" isn't very helpful..., but it works for me.

[hollow hare]

Taking a direct copy of someone else's lyric is a copyright breach if you do not have permission (people who perform covers needs to get a mechanical or synchronization licenses to use the said lyrics).

That being said, if the lyrics you use fall under fair use (https://support.google.com/youtube/answer/9783148) or other copyright exception (https://www.youtube.com/howyoutubeworks/policies/copyright/#copyright-exceptions), such as parodying the lyrics, there should be no issue

[broken lotus]

Generic advice:

try adjusting the timestamp on your Extend From Time. I have gotten very different results just setting it 1sec earlier.

[quaint kelp]

Until mid April or so, Suno was mind-blowing—I've got several straight-up hits in my playlists, and I'm not even joking..... Now it is uninspired and boring at best, usually inconsistent, and the sound quality (with few exceptions) went to 💩

[unreal trench]

sorry if there was a misunderstanding; i wasn't trying to tag you just to complain. it wasn't a complaint, rather than a general observation and a sigh... and thanks for your generic advice: i'm aware of this and have tried it, yet Suno ignored the verse. i'll give it another go sometime later - but thanks!

[hardy lodge]

The song i have in mind would be non-profit and no where but suno. The song is also not on cpdl or pdinfo so i should be good right? If not I'll make it private on suno.

[hollow hare]

Copyright violation is still copyright violation, whether you are monetizing from it or not, and you will be held liable to any consequences that results from it.

E.g. your first agreement with Suno was to use original lyrics or those you have the rights for. Violating this (which is also part of the ToS) can result in the termination of your account

https://media.discordapp.net/attachments/1069381916492562585/1236654640893001758/TermsForLyrics.png

[hardy lodge]

mechnical lisence it is~ ty sorry im new to all of this and still dont fully understand all the details

[hollow hare]

Why not just make a parody of the song? Way less headache and free. (Better yet would be to make a song inspired by the other lyrics, just try not to use the exact or very similar lyrics that can get you in trouble)

[hardy lodge]

I could try that, i honestly thought it would have been ok cause the genre is completely switch but a parody might be fun as its already a silly song

[hollow hare]

The change of the genre is one thing but the lyrics is a separate and distinct matter. It is why the free/paid plans on Suno denies/allows monetization of the music itself but clearly states in the FAQ, despite what plan you have, you still retain copyright over the lyrics (https://suno.com/docs/faqs)

It is just to help CYA. After all, "I thought it would have been ok" is not a good excuse if you had to repel a decision if the account ever got terminated or if it ever went to a courtroom (very unlikely but there are certain artists or companies that would definitely do that, especially those company trolls)

[hardy lodge]

true, ok ty. I'll do better in future.

[cold zenith]

i'm very tired of those double voice after the third continue. You can't make a music of 4min and more without them and the decrease of vocals

you got a good results and it ruined it

😫

[hardy lodge]

yeah i posted about that earlier its quite annoying

[cold zenith]

For the moment to keep it maximum is to have a music of 3min, but it's very annoying

Listen to that...

[wanton mulch]

Anyone successfully get a lyric whispered? If so, how did you do it?

[hardy lodge]

try adding whisper in style of music

or [whisper]

[wanton mulch]

i did [Whisper] and it didn't work, does the cap matter?

[hardy lodge]

In my experience yes. I get better results if i make the cap.

[glacial scarab]

asmr style tag might help

[torpid latch]

hi, is there any way of changing a song to instrumental after its already been created with lyrics? if not, I think it would be pretty nice if suno created 2 files, voice and music...

[hardy lodge]

could try post production in a seperate software

[torpid latch]

Im too stupid to understand what you mean, can you be more specific? I have experience with sony vegas for example, you mean I use an app/program to remove the voice manually? I think that would make the quality of the music very poor...

[hardy lodge]

Like a software the deals soley on music could look around to see whats out there that exels in isolation

if you can isolate you can remove

[chrome topaz]

How much songs i can make? ,,,, bcs i cannot use it now 🙂

[orchid ember]

After 1-st continue...

[hardy lodge]

1 song is 5 credits, it generates 2 songs at a time. if you dont have credits you can't make songs

[azure saffron]

What's up, mates? I've been here from the early days, and regarding the dropping quality, I can confirm that something is definitely going on. I've been creating music with Suno almost every day for about six months straight and have noticed these fluctuations. We saw some changes in generation speed shortly after V2 was launched. It was quite interesting because, after users started reporting a drop in quality, the developers rolled back to the "normal" generation durations, and the quality improved, but not back to 100%; it was more like 80% of the initial quality. I had solid V2 outputs, and you wouldn't believe they're from V2. The same thing is happening with V3. The times are shorter, and the quality has dropped. You can say what you want, but I make songs like a factory and notice any changes immediately. However, some tweaks may help you. Always add "single voice" at the end of your prompt, and if you make heavily loaded mixes, add the word "minimal." Last week, I had to add these two words to nearly every single prompt. And yeah, now I need maybe 20-30 runs to catch a good output, whereas initially, a good song appeared after just 5-6 runs.

[vestal rapids]

are "extend" generations always limited to 1 minute?

[hardy lodge]

yes

[vestal rapids]

thank you

[hardy lodge]

Where exactly would you use single voice in the style of music? or where you'd put the lyrics?

[slim eagle]

Does anyone here know if there are some kind of language tags? I’m making a lot of danish songs, but my first two lines always sounds like a mix of English, Swedish and Norwegian!

[hardy lodge]

you need to use your own alphabet

[broken lotus]

stop trolling.

[dawn mauve]

Put your lyrics into a translater then copy that and paste it onto ur lyric bar

That’s what I do with my French & Japanese songs!

[slim eagle]

And you don’t experience problems in the beginning of your songs?

[dawn mauve]

So like English to Danish is what ur after… maybe

[dawn mauve]

Nope, the French and Japanese are always smooth and fluent!

Would you like an example song?

I have one of my French songs somewhere…

[slim eagle]

Maybe some random danish text be for an [intro] tag could resolve my problem

[broken lotus]

This happens with portugal/brazil... and mandarin/cantonese...
I don't know a specific fix for you, but maybe try something in square brackets like a metatag, to instruct it to be Danish.

[slim eagle]

I wouldn’t know if your French song sounds French, lol

[danish] [dansk] doesn’t work.

Maybe [danish voice] or [danish lyrics]…

[dawn mauve]

I’m sorry that i couldn’t help 🥺

[slim eagle]

Yeah! Still love Suno

[broken lotus]

but in Danish...

[lyric sonnet]

if i tell it to write a song in Japanese i get a song with English lyrics and a Japanese topic. Writing the prompt in Japanese helps

[slim eagle]

Clever. I’ll give it a try

[vestal rapids]

I am using Suno to create classical music, it would be nice if Suno knew how to finish a track, with a coda or some other device or technique

[broken lotus]

actually
[coda]
might work...
(not sure what it would actually go back to... depends how far back in the song it was last heard...)

here are some other metatag tips for instrumentals...
https://www.suno.wiki/faq/style-and-lyrics/how-do-i-make-an-instrumental/

[vestal rapids]

thank you, I'll read that

[winged fractal]

Anyone find a way to stop the voice from being heavily autotuned sounding?

[final hazel]

change the genre to music that has less autotune in real life too

[winged fractal]

It's been consistent in all genres, especially after extending a track

[broken lotus]

'autotune' is a specific vocal effect.
If that's not what you mean, it's going to be hard to help.

[winged fractal]

It is a specific sound that occurs -- the voice 'melts' into the track and often has a robotic twinge. You can often check it with audio extraction tools and it can't distinguish the voice from the track

[cobalt spear]

is there a statement from suno?

[static girder]

sometimes it just generates absolute junk and the 2nd creation is the absolute opposite and fits perfect to the prompt. maybe the rng knob was turned up to 11?

[cobalt spear]

and the quality to 0.3? 😦

[static girder]

thats part of the rng 😄

[cobalt spear]

its a feature, not a bug

just believers get some bangerz

iam a fkn believer

[static girder]

#me2

[quick ermine]

hey everyone :)
i just found the website and already love it! i have one question tho, how do you get it to generate more text, longer lyrics? now its only 2 verses on every try i do (plus a little something in the middle). So when its done generating the song, half the song is with lyrics and the other half is just instrumental. I tried puting in "a 3 verse song" in the box but that didnt help 😅

[wet minnow]

you can use chat gpt to make lyrics for you (its the same as sunos internal one but with a little more control) i know sunos textbox limit is around 1200 characters ... so i ask for "write lyrics about bla bla. 3 verse, 1 chorus. 1200 characters"

[quick ermine]

oh ok, ill try that tomorrow when i have more credits again haha. thanks!!

oh one more thing, if you buy credits or become a member or so, can your "Song description" take more text so you can give it a more detailed instruction?

[gleaming fable]

https://suno.com/song/a0a5bbd5-fa3d-47fd-b1aa-d62599a1c099 i know this maybe confusing it a bit but its happening on a lot of tracks of mine....but as a V3 its crazy poor

[quaint kelp]

Nope (I follow Suno.ai and all the founders on Twitter). They probably spent a whole lotta money on GPUs when they launched v3, and they switched off when they got the subscriptions, and now they rollin and I'm hatin.

[unique stone]

As official as you can probably get is from one of the developers confirming that they don't change the models once they're released - #🪲┃bugs-web message

Audio comes from the model, so that determines the overall quality. V3 release was late March. Any quality degradation since then will vary but won't be the result of changes being made (since the models are "frozen".)

And here's one of the other developers confirming that they are working on v4, several days after v3 was released - #💬┃general-chat message

[lyric sonnet]

could be perceptual bias (for example if Suno produces better results with simpler prompts, and the complexity of your prompts has increased). That said I wonder if the 'model' has parameters which can be tuned for speed/quality without model change?

[unique stone]

Sometimes I get a garbage output at the same time as a good one so I don't think they do that (and I don't see why they'd do it randomly.) Closest comment I've seen from a developer is basically confirming that they do not adjust the quality based on how busy the service is.

[quick ermine]

where do i put custom lyrics then that chatgpt makes? 🤔

[wet minnow]

in suno activate the custom mode above the prompt box. the lyrics field should appear together with all other options

[quick ermine]

ahh ok, ofc haha. thanks again!

[verbal creek]

how do i delete my account?

[bitter field]

https://suno.com/song/57204c87-3de5-44cf-8db6-aba635754bf0 I used to get pretty decent songs.. but lately they have been ending up devolving into a tinny mess... skip to the last 20 seconds of this song... anyone else getting this?? Or know what I should not be prompting?

[teal raft]

It's just a timeout

And 3Daizy is quite lenient imho

[final hazel]

Still breaking the rules.. and you are just a sock puppet

[teal raft]

nooo that was a real other person

That always happens

So real

[final hazel]

yes.. with same name...

[teal raft]

[final hazel]

lol

[eager topaz]

Yes, it's been happening to me, too. I've been using the same prompts that I've always used, but now I just get poor quality.

[hollow hare]

@old cedar You will want to use the [Fade] tag or its variant. I am partial to [Fade Out, End]

Here are several tags to End a song (be sure to check out the rest of the wiki, very useful):
https://www.suno.wiki/faq/making-music/how-do-i-end-the-song/

[old cedar]

Thank you so much! I really appreciate it!

[graceful condor]

Why does the chirp channel not load for me?

[final hazel]

Chirp is retired. Please use https://www.suno.com/ to make music.
Also have a look at the community Wiki for a guide on how to use SUNO. https://www.suno.wiki/

[graceful condor]

ok ty

[hollow hare]

It depends on the genre, as some are more skewed to a certain gender, and simply RNG on what the AI selected. Usually to get a certain gender (esp. males in most genre), having more reference to the gender vocal in the style keywords gives more weighting.

If you had generated a great track but it was using the wrong gendered voice, you can use other AI tools to swap the vocals, tools like kits.ai (see #🗣️┃chill-chat message and #💡┃suggestions-web message )

[old cedar]

@native rock I usually choose create twice because then I get four songs. Out of the four songs usually only one is what I really want and will have the right gender singing it!

[high ore]

if i upgrade my pro will it refund or just add $16?

[broken lotus]

I think there's an option to just start the new billing cycle
(I haven't actually done it)

[high ore]

ok ty also how come the first render isn't always 2 min?

[broken lotus]

I don't know.
try adding [interlude] to the end of your lyrics,
try more words,
try a shorter (simplified) style prompt...?

[high ore]

i'll try the interlude

[sick pendant]

new half song method?

[hollow hare]

Usually done when a track failed to generate (below like 7 sec) and you got refunded for it. You still need a min. of 10 credits to generate though

[high ore]

i ended up spending 400+ credits and only had 1 good seed hahaha

[sacred arch]

Recently, I've noticed that about half of the songs generated are always less than 2 minutes, roughly around 1 to 1:30 minutes. It feels like my credits are being wasted since the songs often seem too short and incomplete. I'm creating instrumental music in custom mode. Any advice?

[hollow hare]

I recommend not using the instrumental switch but placing in various instrumental tags. Not only will this give you more control over certain portions (e.g. build and drops in dubsteps/EDMs/etc.) but will be able to pad the time to 2 min.

https://www.suno.wiki/faq/metatags/instrumental-tags/

An example of the tags I used (the switch at 40s happens because of the drop):
https://cdn1.suno.ai/46531a71-c9de-45c7-bf6f-52e72f984b34.mp4

[light harness]

I agree. And yes, I did turn off the costume mode and there was no lyrics. Still.... weird.

[sacred arch]

Thank you for the recommendation. Yes, I did use custom mode and didn't activate the instrumental switch, but I haven't tried using tags yet. I'll give that a go and hopefully see better results. Thanks!

[hollow hare]

You're welcome, I wish you the best and hope you make some great experimental progress 🙂

[wispy nebula]

There was a whole month where I got virtually only 2 minute songs. Some days, now, it's different. All random.

[weak kiln]

I just made the following prompt for chatgpt to add to the end of building a song:

I need the first section of the song to be about 1 minute 30 seconds and fit in 1000 characters. I would like this song to have 4 parts. the first part being two minutes and the following parts being 45 seconds minute. each section can only be 1000 characters. Give me some genre ideas for this song before continuing

so I would say:

Write me a song about Honey Badgers having Tea. I need the first section of the song to be about 1 minute 30 seconds and fit in 1000 characters. I would like this song to have 4 parts. the first part being two minutes and the following parts being 45 seconds minute. each section can only be 1000 characters. Give me some genre ideas for this song before continuing.

I get 4 code blocks i can copy and paste into suno. I also told chatgpt to ingest the suno wiki and to put all tags in [ ] Been getting better results

[weak kiln]

had to make the song and this is what i got

https://suno.com/song/880af4b9-83a6-4167-9c90-bc63e846f243

[zealous stratus]

Guys i wanna make a funny sad song in urdu for my friends what should i put in the prompt so it's like a rap but clear

[ruby owl]

Hello, I noticed a very big problem with the vocaloid's voice in AI. Problem: when I use a song with a vocaloid without j-pop. Then the voices turn out to be ordinary, or 1 verse is normal and then everything sings in a human voice. Because of this, it is not possible to create a song of a certain genre. After all, it is created as a j-pop basically. And often because of this, there is no desired result. Because of this, I am very limited in using genres with vocals and voice. Here are some examples where there is no necessary vocal voice or it is only at the beginning and so with many of my created songs. https://suno.com/song/0c9ae675-4e33-4d90-bba1-2a6c295bef82 https://suno.com/song/c8c7c1a5-ba55-4e20-bffd-9ccf6d7914b6

[storm bane]

Question for someone who is more well-versed with the engine generating the music: Would it be possible (in the future or right now through the right set of commands) to create an acapella version of your own Suno track? Or is it that the vocals and music is generated together and cannot be split into stems?

Using Suno a lot to test ideas for lyrics/vocal styles for music genres I produce, and would love to be able to make vocal demos for my own tracks to show prospective vocalists. Save me from having to record my own voice, AI!

[unique stone]

At the moment everything is generated together.
There are other tools you can use to do a stem split (vocalremover.org or Ultimate Vocal Remover for example) but these remove some of the detail from your song - sometimes quite a lot, depending on style etc. It's something that is often asked for though so eventually it'll likely be possible, but at the moment it's probably very difficult to do (i.e. the developing and training of the relevant models, along with a higher amount of computing power needed.)

[hallow sleet]

Anyone here who can give me some tips on how can I make a song as much similar as possible to some popular song like Money from Boney M for example? or some Abba songs etc. Since Suno is not understanding the artists

[dense veldt]

is there a way to get quality exports instead of 192 bitrate

[unique stone]

Not yet (there's some recent discussion about this in #💡┃suggestions-web today)

[dense veldt]

oh thank you

[graceful condor]

Are anyone else's songs not showing up in New? It used to be pretty consistent for me but lately my songs don't appear there

I don't know if there's a change I'm not aware of

[jolly tundra]

There is a way, like some AIs, to directly give us seeds, so that next time we can get the same melody by using the same genre and adding seeds.

[mystic salmon]

Hey, is it possible to somehow get more credits without the subscription? It seems there is some info out there that you can get 10 credits by doing prompts but im not sure if thats real, or if I am doing anything right? Just curious if I am missing some big feature.

[wet minnow]

do you know this little write up? https://www.suno.wiki/faq/polish-and-production/how-do-i-get-my-song-on-trending/

[mystic salmon]

okay thanks!

[broken lotus]

no because the melody is generated (in part) by the lyrics. Change the lyrics and you have a different melody

[mystic ember]

Hey guys do you know if I made a song in free version of suno, can I import it to a pro version to make it commercially used or edit some lyrics on the same song?

[broken lotus]

There is no way to convert it to commercial... (I have never heard anyone say they did)

I suggest remake the the song with REUSE PROMPTS

It won't be exactly the same (it might be better) but you will own it, no compromise.

[mystic ember]

thank you so much!

[serene viper]

I have a question regarding my subscription. My account is scheduled to renew today. Should I expect to receive the credits today, or will they be allocated the day after the renewal? Not sure where else ask this.

[broken lotus]

the theory is that monthly credits refresh around the same time of day you subscribed...

(I haven't really figured out what time mine refresh, I always seem to assume the wrong time of day)

[serene viper]

Interesting

Well thanks for that.

[unique stone]

Mine renewed at 3am, which is when I subscribed xD

[serene viper]

I don't really remember when it was when I subscribed tbh lol

I guess it renes by exact date and hour, not by days lol

I wonder what happens to the folks that subscribed on a month that has 31 days

[unique stone]

I looked at my payment confirmation email to see the time it arrived 🙂

[broken lotus]

or Feb 29th...

[serene viper]

Lol true

[final hazel]

You can see in your account. Just click on subscription. You will see the renew date @serene viper

The time is depended on why you started one.. so if you started on 1 pm at the 9th, you will be getting it at 1 pm at the 9th next month

[broken lotus]

we need to know the exact minute, so we can use the last credit a few seconds before it flips...

[serene viper]

Haha yeah

I need to prepare myself better for next time then

But appreciate ya'lls help

[final hazel]

You can only check the time of the first email of your plan.. then you know the time

[serene viper]

Do you know the email recipient that is sent with the subscription?

Easier to search for it that is why I ask

[final hazel]

look for suno?

[serene viper]

I guess wanted a unique ID better but I can search that way too, thanks. lol

I got it no worries

Thanks

Damn gotta wait until 11:11 PM

🤣

[final hazel]

😹

[graceful condor]

Ooh, thanks! I guess my latest songs might have hidden meanings lol

[serene viper]

Also, if I buy I don't lose the credits, I only lose the monthly credits if I don't use them by my renewal date but not the extra credits?

[wet minnow]

bought extra credits stay. only the renewing ones go away

[serene viper]

Awesome, well appreciate the help.

[obtuse idol]

Good morning, I'd like to know if there is any syntax or method to be able to custom describe a piece as instrumental but be able to describe it as well as i can describe a song with vocals?

[wet minnow]

you can try to guide the song with metatags in the lyrics field. like [instrumental bridge] [intro] [outro] ... there is no official list, because a ton of things work. but here is a general guide to metatags: https://www.suno.wiki/faq/metatags/song-structure/

[obtuse idol]

great place to start! Thanks!

[mortal drum]

Why chirp bot does not work?

[jolly tundra]

OK! Thanks!

[hollow hare]

Chirp bot is retired, use the website https://suno.com

Some helpful starting links: https://suno.wiki
https://youtu.be/lj8utBd_IUM

[hollow furnace]

Does anyone have an example of a song they made on Suno (public), remastered with Vocal Remover/audacity or some simple tools, that came out improved? It seems that some people are doing this. I'd love to see an example.

[balmy spade]

I think I'll stop paying for this service until the sound corruption after blocks 2 and 3 is fixed. It's getting unusable in most of the songs.

[hollow osprey]

Is there an API for embedding outside of suno?

[woven jungle]

every time i try and use the website it says "an error occured" "a system error occured"

[hollow osprey]

Thank you

[ruby owl]

Guys, did I correctly address this topic here? Or should I write it in bugs?

[potent tendon]

I want to know too

[broken lotus]

I see this error when I use REUSE PROMPTS on a full song, and the lyrics are too long for 1 clip....
I have to remove text before I can even edit it...

[broken lotus]

Not that came out improved LOL

but btw...

RVC Replay just got a built-in voice trainer.
it will auto strip the Suno voice and replace it with your own cloned voice.

[woven jungle]

I haven't even made it that far, all I did was try and navigate to https://suno.com/create after creating an account. it just spins forever and if i try and create anything I get the aforementioned error

[hollow furnace]

ooh have you tried it

I wonder if the voice might be higher quality this way, less vocaloid

[hollow hare]

For the error, simply use less lyrics. You can extend the song by 1 minute each time if needed https://www.suno.wiki/faq/making-music/how-do-i-make-the-song-longer/

If it did not pop up the error but kept spinning, try checking your library (and refresh) to see if it had finished generating, can take a few minutes

[broken lotus]

I tried it..., but was my first time training and not sure what I'm doing...
I have some Elevenlabs recordings I made last year... training on that because it's very 'lively'
I need to try some more, but their current controls are sparse... hard to get it wrong...

[hollow furnace]

it's really a cool idea

[unique stone]

Using an AI stem split to separate vocals from instrumental and then putting the vocals through a service with AI voices seems to work quite well. I briefly tried it and had to record myself as well to patch over some of the pronunciation. With enough effort it'd probably give very good results.

(I just tested it a month or so ago to see if it'd be feasible)

[hollow furnace]

did you use your own voice there?

[blissful cedar]

Hi Everyone. I'm struggling with some of the songs' output quality because of background noises, artifacts. I have tried to clean it up with Audition, but it's not much better. Anyone found a solution around this? suggestion? is RipX the only solution?

[woven jungle]

I didnt put any lyrics in - i just tried to use a stock out of the box thing

[hollow hare]

To reiterate, you used a random prompt and it gave the system error response? Weird.

Assuming you have no add-ons that could potentially interfere with the process (e.g. NoScript or uMatrix blocking the script), that is very strange.

Did you try again? If not, try using the custom mode as you get more control over the input

[woven jungle]

I tried both ways, I just cant get it to do anything. I'll try and create a new account i guess

[frank bison]

Hello everyone,
I`m new to suno and added @versed valley to my own server but the commands wont start. Is that normal and it repairs itself or not?

[final hazel]

There is no discord service anymore. Only the website.

https://suno.com

[frank bison]

thanks for the help anyways

[worthy arrow]

has anyone be able to get Djent style guitar, 7 string, 8 string output from Suno? I've tried many different prompts but the closest i can get is Octanecore which isn't even close to Djent. Also wonder if anybody has been able to get actual Screamo Vocals for their Output? Sometimes i wonder if downloading the source on github and training it somehow on those styles of music would yield these types of results.

[broken lotus]

RipXPro can erase a lot of that AI ringing noise...
(you need the Pro version to do it)

A big EQ or a frequency-targeted filter might help suppress it.

I'm pretty ruthless about rejecting any clips where that ringing is already audible because it will only get worse...

(It is also much worse with some genres I don't seem to use, metal guitars and high-pitched singing... The problem is slight to nonexistent on other genres without 'wall of guitars' sound, electro and RnB stay very clean)

[broken lotus]

Suno source is not on github. it is not public.

I remember doing some djent experiments after someone mentioned it here....
I think I just used 'djent'... not my genre so I couldn't rate it.

[blissful cedar]

Hey Thnx. You're spot-on about the genres. I made lot of Alt Rock so I'm always facing that problem. Though, I just created a song with prompt "Dark electropop, emotional, epic, backup singer". the song is super cool but the ringing noise is killing it!

[broken lotus]

try one with the 'simple' prompt... it uses a 2 or 3-word style and they are always very clean sounding...

based on that I have been stripping down my style prompt, trying to keep it just a few words and not describe every instrument, I get much cleaner voices...

of course that's a compromise from having more control over the prompt...

[gleaming inlet]

Um, hello. How many tracks u can generate at one session? This thing have a timing, or it's about number of tracks per 1 account?

[wet minnow]

on a free account just 1 job / 2 parts ... on a paid plan you can click 10 times i think to make 20 parts at once

[wet minnow]

or do you want to know if your credits are empty? you get 50 credits a day wich is 5 generations (as a free user)

[blissful cedar]

I have actually tried twice to keep the prompt to "Dark Electropop". the results were not as good as the longer prompt. but I get your point. thnx again

[gleaming inlet]

Is there an opportunity to get more credits? Cause i dont see the button "subscribe" or smth on site.

Oh duck, i seen it

[final hazel]

https://suno.com/account

[wet minnow]

if you click on your credits it should take you to the subscription page. there you can see the monthly or annual prices.

[gleaming inlet]

Sorry guys my bad

I just use a vertical monitor, and the button is not visible at all on ma watch :)

[wet minnow]

no problem. we are here to help with exactly these kinds of things 😄

[wanton holly]

"Hello, I cancelled the subscription a month ago, yet it still deducted the amount from me! I would like to request a refund

[hollow hare]

You will need to contact [e-mail]

[wet minnow]

all billing is done suno internally . please contact [e-mail] to get that fixed. we here have no access to billing

[worthy arrow]

https://github.com/suno-ai here is the link to it. when you said its not public, were you just referring to the nice frontend that suno has setup to use bark?

I tried the "Djent" tag but that doesn't produce anything close. i appreciate the feedback through. maybe in a future version it'll support

[rapid bluff]

thats bark, their text to speech model, which is how they started

[broken lotus]

When you find the Suno /Chirp source code on that github, let us know.

😉 👍

[worthy arrow]

lol sure

[nimble stratus]

Is there a way to more efficiently get rid of songs that are in the trash permanently

[wet minnow]

you can multi select with ctrl and shift

[golden cove]

Hello guys , my question is that in the free plan when you create a track most of the time the end of the track the vocal get "tired" and be really synthetic and distorted, this problem can be avoided with the paid plans ? i mean, when you generate a song with the paid plan, the vocal can be stable most of the time ?

[nimble stratus]

Thank you for your response but I'm using an iPad my computer is ancient technology and not functional in today's market. Is there a way to make it work with the what are the little trash cans I see trash

[wet minnow]

free and paid work the same. you have to avoid that by extending the song before it degrades and select a part that is clean. takes some tries but is doable.

[golden cove]

thank you

[wet minnow]

let me test something on my iphone... i will report back shortly

[wet minnow]

tap and hold tracks to multi select... but wait i can select them by just tapping on them once

[graceful crow]

Hi please do you know the style can I use for Haïtian kompa Music. I put kompa and the Music was not kompa Music

[zenith depot]

11Labs got Spitfire rap working.
Have Suno do this.

https://x.com/flavioschneide/status/1788654503790628935?s=46

Flavio Schneider (@flavioschneide) on X

ElevenLabs Music model rap.

Twitter · May 9, 2024, 7:37 PM

[spice root]

Is there any method to getting Suno to pronounce a word a certain way?

[wet minnow]

you can use phonetic writing... just write it how you would speak it

[ionic meteor]

when I write in French, it's a bit different, sometime, there is no other way to write the word but the AI pronounce it weirdly

[lost hollow]

Is there any library or source that contains all command like (interlude) or (chorus)?

[golden cove]

https://www.suno.wiki/faq/metatags/

[lost hollow]

thanks a lot!

[dusk kettle]

Out of curiosity but does anyone know if when you report a song for a bug then delete the song from your account's library and trash, do devs still get to see the bug and what happened to still improve the AI after deletion from your account or should I leave the bugged stuff in the trash for a while to help them develop the AI?

I got thinking about this after deleting a lot of songs and reporting them under the bug report for not following lyrics correctly. I mean if someone deletes something from a drive shared between people then its gone forever and you can't see it anymore to check out what was bugged so I would only assume this would be the same here with the files on the Suno server under your account. I mean I might be wrong but wanted to see how that worked.

[tribal beacon]

Can you let us edit the lyrics that appear in the exported video? Sometimes the creation process using extends gets messy and the lyrics are completely wrong. Editing the photo and the lyrics that appears in the exported video would be neat!
I take the opportunity to say that your generative AI model have been giving me the most joy I have felt for awhile. Thanks!

[somber sequoia]

I have a question. I subscribed to the Pro Plan, but I can only find versions v2 and v3. Is the v3 alpha no longer available?

[dusk kettle]

Nope when V3 came out they got rid of the V3 alpha.

[inland juniper]

Hi everybody! I am trying to generate a song, but the prompt gets flagged for moderation. But I have no swear words or anything in it..
Can I paste the prompt here so you can help me figure it out?

Or should I ask in other channel?

[hollow hare]

It got censored from ChatGPT (which moderates the lyrics in Suno) not liking something. Ask OpenAI's ChatGPT itself about "what would censor the following: [content]" and it should tell you.

Usual causes are using swear words or interpreting harmful situations. Even if it worked before, it could be a small change caused its interpretation to become different.

I remember people getting censored for using bones, ribcage, and skeletons because it made ChatGPT assuming some sort of horrific tale.

[inland juniper]

Mmmm, I'm not using ChatGPT or anything. These are lyrics of my own, so I'm having a hard time finding out

[hollow hare]

It is not you using ChatGPT, Suno uses it for the lyric moderation

[inland juniper]

But I do use the word "slicing" in the Chorus. Could that may be?

[inland juniper]

My bad, sorry for the misunderstanding

[hollow hare]

No worries, you are not the first and likely not the last

[inland juniper]

This worked like a charm. Thanks a lot!!!

[hollow hare]

Great to hear and you're welcome 🙂

[still peak]

does anyone else have this issue? i'll be making a song, but the longer it goes, the less it sounds like the song started, IE, sound degredation.

[hollow hare]

Yep, it is understandable when you learn that the AI does not remember your initial generation so extensions spend ~1 min before the continue point to relearn/mimic the styling.

If you have reverb within this duration, the AI will pick it up and include it as it thinks it is part of the style.

As you extend more, it gets more prominent as the AI "hears" it increase.

So, currently, you must treat the problem when it happens by generating a track with little/no reverb before extending (though there are some songs that goes well with it... like songs about digital/memory corruption)

[still peak]

hm. thanks

tyhis'll be interesting...

ok it's working. sad though, some of this is good stuff

[hollow hare]

One option is going to an earlier generation and extending from there. Of course, you cannot combine this readily with your later parts in Suno but you can use programs like Audacity to stitch it together (made easier if have some sort of instrumental bridge)

[still peak]

i have reaper. been using it for fifteen years.\

[mystic salmon]

Are there any prompts I can use to make the singer use a calmer voice during certain parts?

im doing like a screamo thing, but i kinda sorta want some moments to be relaxed

[hollow hare]

Maybe [spoken word]? Though that will likely give a story telling portion rather than something lyrical, hmm..

Maybe one of these tags may speak to you? https://www.suno.wiki/faq/metatags/voice-tags/

[mystic salmon]

ill give it a go?

ill tell you how it goes

[hollow hare]

Let me check something quick, I think I seen a clown music with what you want... though if I cam find it again is another question

Nope, cannot find it 😦

[mystic salmon]

aw :(

[lyric hare]

why does every song i try to make refuse with "content flagged for moderation"?

[hollow hare]

I did find out [soft spoken] and it's variants but that is more like ASMR style

[lyric hare]

im about ready to pull a chargeback on my CC over this.

[hollow hare]

It got censored from ChatGPT (which moderates the lyrics in Suno) not liking something. Ask OpenAI's ChatGPT itself about "what would censor the following: [content]" and it should tell you.

Usual causes are using swear words or interpreting harmful situations. Even if it worked before, it could be a small change caused its interpretation to become different.

I remember people getting censored for using bones, ribcage, and skeletons because it made ChatGPT assuming some sort of horrific tale.

[lyric hare]

its horsecrap.. Im trying to write songs about the autistic experience, its messy, its painful. Sorry i cant write songs about unicorn farts

literaly ablism in an app

and there no way to reach any support over this?!?1

[hollow hare]

You would need something good if what you want goes against the ToS

https://media.discordapp.net/attachments/1102676217687195679/1236146831902244874/image0.jpg?ex=663edbf3&is=663d8a73&hm=a62c0999248f23e0e16dc953859616ca448b62c232774f7ea210d7526fe34632&

[lyric hare]

like i get censored by the allistics evey moment of every day, i finally find a way of expressing myself, and boom censored again

none of that.

[hollow hare]

Then ask ChatGPT what it has an issue with and fix it

It's an AI, you can be innocent and it can think of crap

[lyric hare]

i writing lyrics about masking, being forced to mask. Now i got to mask even more through crapgpt!

[hollow hare]

*shrug* Get creative about it. Phrase it another way or change the spelling to be phonetic (though remember your account is liable for suspension, if not more, if it goes against the ToS)

[lyric hare]

i would love them to try, boom human rights complaint

[hollow hare]

As I said before, it is an AI checking, so it is dumb like censoring "Puerto Rico" because of Rico being an artist.

Until an exception is made, you could say something like Puerto Ríco, Reeko, etc.

[lyric hare]

yea ok, so just dont talk about the autisic expeience.. it scares the allistics. FML. same crap as always

I'll just write songs about how aweome it is to be a neurotypical.

[hollow hare]

More likely it is how you went about phrasing it that ChatGPT has something against. Like what this meme implies: #😜┃suno-memes message

[lyric hare]

cool.. tone policing, by AI.. love it.

[hollow hare]

It is what it is. Companies want to limit liabilities and one of the efficient resources without getting rid of custom lyrics is to use AI for censorship ).

If you need some suggestion on what to change to make it past the censorship, ask ChatGPT for suggestions

[lyric hare]

sounds like a good point to work on an opensource model. Like every company, take a great idea, and gurardrail it to death.

[hollow hare]

E.g.

(I was asking it what would censor the following and it didn't like my example)

[lyric hare]

so if im singing about how masking forces some autistic people to become suicidal, how do i get past that?

[hollow hare]

What did ChatGPT say specifically about that?

[lyric hare]

i dont know, i dont use crapgpt I use a unguardrailed models.

i dont even have an account

[hollow hare]

Well, there you go. It may not even be about what you are complaining about

[lyric hare]

well fortunatly for me ive got 70,000 folowers i can put suno on blast with

they could have had a support email, but if they wanna make this public and ugly i can serve it up.

[hollow hare]

Cool, less people using the site (note that I am a regular member). As I mentioned earlier, ask it (https://chatgpt.com) something along the lines of:

What moderates the following:
[Content]

If you are not willing to do that and wait for a response from dev member, you will be waiting for quite a while

The only email I know for support is for billing. Not sure other ways you can get in touch with them besides one of the feedback channels

--

As I mentioned earlier, you just gotta find the problematic line(s) and either rephrase it or substitute some phonetically spellings so it bypasses censorship (though you are still held liable for the content generated if you were avoiding swears, etc. And someone flags it)

[graceful condor]

"hypasses" is best bypasses

[lyric hare]

literally talking about how making makes autistic people wish they were not alive. true experinces, authentic experiences. If they want to ban me for that, i'll go abosolutly ballistic to the media about it.

[hollow hare]

Well, there is the issue then, sensitive content material is likely the basis that ChatGPT is censoring it then

Should be easy to bypass by changing 'alive' to 'álive' or the like.

Could rephrase to be like "better off not being there/existing/etc." Which has less objectionable phrasing than "not being alive" (AI be dumb like that)

[bold current]

Aside from my Expanded starting point, which I starting to believe has less and less effect on the sound, is there a way to get rid of that constant Metal Grinding sound in the background that seems to show up now in all my song by the 3rd + Expand? No matter what genre I'm doing it in, I can instantly pick it up and notice it.

Is this a ploy to get by the Copyright AI checkers? That would make more sense why it's contantly there

[graceful condor]

Welcome to algospeak.
https://onezero.medium.com/why-were-creating-language-to-hide-from-tech-s-censorship-systems-ad024f083c89

[eager topaz]

Happens to me, too. I've stopped using Suno until the problem is fixed.

[lyric hare]

any suggestions for alternatives? Im mainly prototyping songs, getting them professionally recorded later.

[eager topaz]

You could try ripping tracks with RipX Daw Pro and trying to clean the tracks, but it's a huge job and I find it's not worth the effort. The Noise remains to a degree, making the songs unusable for any professional application.

[lyric hare]

im giving it to an artist on fiver to totally re record, so the noise isnt as much as a problem as the censorship

[bold current]

What's horrible about this is it makes all my songs sound similar regardless the genre and I have stranger out in the wild able to recognize I used SUNO without me even telling them

[lyric hare]

there is still a place for musicians, its just making the prototyping of a song easier

[bold current]

It's like a SUNO AI Soundprint

[eager topaz]

Yes, it's very easy to recognize!

[lyric hare]

yea i noticed this too in the last day or so.

[graceful condor]

https://suno.com/song/0814fb3c-9d87-4313-a70f-fca109f1d2f0

[lyric hare]

for my purposes it works but i can see how it would be unbelievably frustrating if you wanted to have it in one go.

i noticed that as soon as i got a paid account, boom, the quality was crap

[bold current]

Yeah just had an entire [Bridge] skipped and didn't realize what was going on until I paid for a few Creates that it was being censored.

Paid account here too and I purchase credits on top of that as I make a LOT of music

[lyric hare]

i appreciate the attempt at the rewite, its missing the impact though. thanks for trying though, ive lost so many friends this way, as soon as i had the ability, i wanted to make a song that really gets the impact.

but i get what you are getting at.

i think if i have to chatgpt mangle 30 hours of work, im going to twist it into a song that calls out suno and mocks them.

i like the twist on the instrumentals though, interesting direction. I was going with indy grunge metal

[lyric hare]

for me this is a whole new theraputic way of expressing myself. Ive always had lyrics flowing in my head, and the music, but i cant sing or play a note. Im mostly nonverbal. (selective mute)

this song was about where i was a year ago, and singing about it (through AI) was a way of shedding skin

[bold current]

Oh I fill ya.. I'm a musician and have written music way before AI or Suno.. but there is something theraputic about making songs on the fly with mixed genres.. it's an amaing piece of software, but needs up now up it's game

You can't sit around with anything on the market in AI.. after 6 months you're considered old

[lyric hare]

the lyrics have been running around in my head for over a year. when i finally heard it in verse. i broke down and weapt, let go of soo much weight.

[eager topaz]

There is V4 on the way.

[bold current]

That's what is keeping me here. Can't wait and hoping it addresses these major issues

[lyric hare]

i wish we had more control over the melody, seperate tracks? wouldnt that be awesome to be able to control the mix. (that is my instrument. I can melt your face off with a mixing board)😂

the music is in my head, i just cant get it out.

[bold current]

I use DAW composing software. I can recreate most of what I hear already in Suno, but instruments are hell of expensive.. into the $10k range easy. That's where Suno is a huge asset, is you don't have to buy expensive instrument/beat packages to make a song. The vocals are great most the time as well. Amazing in fact

[lyric hare]

at least i have one version of it. I dont know why it didnt get censored the first time, but i wanted to make a cleaner version. I didn't know the muddy sound was endemic, i thought it was something i did. but i got one version of the song, and its pretty good.. enough to get somone to re-record it

[lyric hare]

im more of an analog guy, give me a mackie onyx 4880 and im set

i guess i need to learn the new digital stuff

I am more experienced at live

[lyric hare]

suno is amazing (Up till this point) at prototyping songs.

[bold current]

No doubt there. I'm nervous to submit anything created entirely by SUNO to my distributor as I also create music myself, so there is a risk there if I go AI.

[lyric hare]

Oh the AI paranoia is real, but its because people dont know how AI works.

I develop AI models for disability aids, and even as a disability aid people lose their minds about it.

i had to give a talk about it at a university because they were refusing to let people with communications disabilities use it to communicate in school. people think its a big brain in the sky, but its not.

from my talk

"Being in awe of AI's capabilities makes total sense. This technology unlocks amazing new forms of human expression and collaboration, especially for those facing disabilities or challenges.

At its core, AI detects patterns across the vast pool of our collective intelligence and creativity over time. Everything it generates emerges from recombining that human knowledge in novel ways we couldn't easily do alone before.

This is where AI's power lies - giving those who have struggled with conditions like autism, dyslexia or speech issues a new co-pilot to bring their inner brilliance to life. It expands the boundaries for joint idea exploration and creative expression among all humans.

AI facilitates our collaborative intelligence, amplifying our abilities rather than replacing them. The amazing creative potential still flows from people. AI just helps us combine it in incredible new ways we couldn't previously. It's not an independent conscious entity, but a tool that unites our skills and knowledge like never before.

So while the capabilities seem almost unbelievable, we should embrace AI's role in empowering the full scope of human ingenuity and helping it find new forms of expression."

in 10 years we are going to look at this, like meh.. its a calculator

but the people who are in jobs that will have to adapt because of the tech, just like the trade of "computer" back in the 1900's was replaced by mechanical and then digital systems, are threatened. no one likes change when they are stairing it down the barrel

[umbral dragon]

will we ever get an API 👀 there are many possible interactive experiences can revolve around Suno or have Suno a part of it

[fiery pier]

A.I. generation at this current stage is just a 'supplement', and not a 'main meal'. It can generate awesome stuff. But it always lacks a certain element. Not the 'real deal' yet... But a great assistent to spark new ideas or get you going in a new project.

[umbral dragon]

so what

[void marten]

Hello im new to bot, i already add chirp to my server, but when i type /, the chirp command doesnt appear, so what to do?

[wet minnow]

Chirp is retired. Please use https://www.suno.com to make music.
Also have a look at the community Wiki for a guide on how to use SUNO. https://www.suno.wiki/ ✌️

[void marten]

Ah i see

Can i use custom lyrics in suno?

[gusty wren]

yeah, go into custom mode. you can have it generate random lyrics, lyrics based off of a prompt or stanza, or insert your own lyrics

[hardy lodge]

Is it possible to swap and change vocal styles in a song for example

[intro]
[Fry Scream]
Lets go
[Clean vocals]
(verse)

[wet minnow]

depending on the style and sunos mood this might work. be prepared to not fully get it working and fix messy parts with extensions. suno tires to figure out what you want and its quite good at that. but its still more or less luck.

[hardy lodge]

alright no prob. I'll try a few things see whats what next time i have credits.

Until then if anyone does figure it out even if its just luck lemme know pls a clean verse and scream chorus would be cool.

Just so i know if it is actually possible.

[azure crane]

Have been asking this before but never got a solution that works. Is there any chance to silent the melody played but let’s say a piano and just have the background music sounding in an instrumental track.

[wet minnow]

i had luck with [melody break] . but to be fair i just tried it one time. i have no idea how reliable it works.

[azure crane]

Tried it now and it just added another melody 😅 . It seems that "removing" is not a thing that AI understand.

[raven ridge]

Hello everyone! I have a question about the quality of the Sound. In which Quality can you download if you have "Premier"?

[broken lotus]

all tiers (free, pro, premier) have the same AI features, and create the same quality. (and downloads)

(at least right now)

[neat field]

i've found that using two genres with the adjectives "female" on one and "woman" on the other produces more consistent deep female singing voices (vs soprano and mezzo-soprano)

i have not tested prompt Contralto tho... and maybe i should

female pop, woman alternative folk something like this seems to produce more lower register woman singing but not always

woman post-hardcore, emotional opera

this produced some very interesting voices in generations on rare occasions, there is a "growling" in opera that hits right with the screaming of post-hardcore genre

[somber sequoia]

Good day guy! I have a question, is it conceivable that new genres could emerge in the future, such as "New Jazz" or "New Jersey Club Type beat"?

[fast void]

any ideas why all suno song are very noisy? even with some simple prompt like "melodic happy hardcore, uplifting and energetic, 180 bpm"

[broken lotus]

that's not a simple prompt...
simple prompt is like 3-words...

[broken lotus]

try consolidating
"happy, uplifting, energetic" (all mean the same)
to 1 word, like "jubilant"

[fast void]

well, happy hardcore is a genre so "happy" must stay there. so its "melodic genre, 1 emotion, tempo". pretty simple for my taste

[keen cloak]

is there a filter for our own library?

[still peak]

whaqt about time signatures? can i input 6/8 time signature? will the tempo be affeted if i input a different one in an extension?

[winged fractal]

I'm not sure if something changed, but for this month most of my tracks have had issues with audio compression and lyrics being ignored on extensions -- it'd skip ahead several lines.

[broken lotus]

"Why my songs have noise?"
gives advice
"No."

🤷‍♀️

[winged fractal]

It's done the same over half a dozen genres and is exacerbated the longer a track goes

[broken lotus]

nothing has changed. Compression has not changed. Lyrics have not changed.

It is possible your prompting has changed, or RNG, or you are exploring genre control a bit more with mixed results

or honeymoon period is over and you are starting to hear flaws you weren't listening for before.

[winged fractal]

Something has definitely changed, as even a simple instrumental prompt unchanged and extended has issues not found from one month prior

Here's a simple one-instrument example:
https://suno.com/song/e8abcc9e-fbd8-4bc0-94a9-b39806e53d44

[broken lotus]

Nothing has changed. You are welcome to your opinion, and I'm not denying you hear what you hear in your own music
But I can't allow you to keep posting false information.
Nothing has changed.

[gentle rose]

I have 3000 unused credits this month cause every Suno generation comes degraded and is giving me ulsers. I'm using Udio now with no issues.

[still peak]

you guys are usin AI to create music. it's not going to be a flac file for a long time.
use the extend option from a few seconds before it goes wonky.

@winged fractal you should've extended it again from like 2:19

[neat field]

what bands are in this genre? just curious

[winged fractal]

Thanks, I'll check it out.

I won't be continuing my subscription and have requested a refund after the moderator exchange, no way I can use a tool if that's what I can expect from Discord interactions.

[winged fractal]

I experimented quite a bit with it so far, but it seems to be a natural byproduct

This was the other starting track: https://suno.com/song/12b6a783-d2dc-4a4a-915f-a0c63f9ad862

you can notice it has the same issue appear towards the end of the song. The prompt is only "Medieval acoustic" with an auto-generated title and no extension, but the same problem occurs naturally in the generation

Previously I had to re-extend or re-create to get different stylistic choices

[unique stone]

Here's part of a V3 alpha generation from 23 February with a pretty awful cello, exhibiting similar quality degradation.

[winged fractal]

Do you think its related to certain instruments/tone that confuse the model?

The change from 0:28 to 0:38 there is similar to what I found in this one:

https://suno.com/song/490fe10a-ba0f-46cf-9bd5-8cd630bba3e8

I made a long track very early on as a test and didn't find any quality issues occur for quite some time (around 6ish minutes here):
https://suno.com/song/1ce09bc3-f42f-488d-9040-ec8f4febedf8

[glad current]

Is there any way to adjust the AI's pronunncation? I'm trying to make a song for a world of warcraft video and my druid's name is Bearijuana, but the AI pronounces it like "bear wiener" lmfao..

https://suno.com/song/26cd900e-e2f6-472f-90e3-f4d3fd91bc3e

[winged fractal]

break it up into phonetic parts. e.g., "Guy Den" for Gaiden

[glad current]

Hmm so just spell it wrong. Not bad.

ty

[broken lotus]

bear-ee-wanna?

[unique stone]

I've been trying to work out what sets it off. I have a music box song that is troublesome at the moment but the bit I want to work with already started going bad so it's tricky.

It just sometimes seems to gather more and more noise and then further notes/chords amplify that noise etc. I do hope it gets addressed, only option is to cut at the point where it starts to develop, sadly. But it does seem to be a long-standing thing rather than a recent change.

[broken lotus]

the steady degradation in certain genres has been present since v1.

(it was much worse)

[winged fractal]

It seems significantly more pronounced than in April for sure, I have like 66 pages of generations at this point and the hit/miss ratio has drastically altered for me in all genres

[lyric hare]

suddenly "content modeation" LMAO

[winged fractal]

In the V1 I didn't use for The Drunken Bard track you can hear the degradation begin at around 11 seconds, even with only one instrument

I'm looking through past tracks trying to see if I can find one I missed

or if it's related to certain tone/pitch/instruments

but so far its "if it was created in April it is fine"

[lyric hare]

yea ive done a few experiements and it seems to be degrading around the 1m mark

was going to play with keeping lyx short so i can extend sub 1min and see if it makes the garbage go away

[winged fractal]

This maintains quality the entire way through: https://suno.com/song/6eb1161a-15dd-4080-969a-0db64fe45b8c

[unique stone]

I asked one of the devs before if they change the models after release because there's often complaints about noisy generations - they confirmed that they don't change the models. The audio comes from the model so I'm not sure what else could be happening if the model has remained the same.

[lyric hare]

when was this made?

[winged fractal]

Faster was made Apr 4th

I made a big band jazzy song on March 29th: https://suno.com/song/8ca958df-22f9-4984-a879-ac8c23190de5

and those naturally have that old staticy crackle in them

so you'd expect the quality degredation to be more prominent in them

but it doesn't seem to have the issue

[lyric hare]

i kinda doubt that statment, i work in vocal ai, and we are always RLHF's it. i would expect that they are fine tuning the model based on thumbsup/down otherwise why even have that there?

[lyric hare]

yea i feel like the bad parameters crept in around the beginning of the month, because thats my experience too. everything in april was good, then it went to crap

[winged fractal]

Yeah I can't find a single March/April track with the same issues

wait I found one

https://suno.com/song/1b9c32ba-ff99-482b-aa38-cae90051a622

April 23rd, quality drop at aroudn 0:53

that's the only one so far

[unique stone]

Thumbs up/down was used during V3 alpha to help them train the final/released V3. I remember a dev suggesting this why we have choirs all the time in V3 (as people must've thumbs-up'd that kind of vocal a lot)

Honestly if a dev tells me how something works for the service they are developing, I'm inclined to believe them. 🤷‍♂️

[winged fractal]

I'd believe them if there wasn't reproducible examples of something being different

Though I'd imagine it doesn't even have to be something that changed within their model

[glacial scarab]

"reproducible examples of something being different"
Give me 10k gens with same prompt from alpha and now

[winged fractal]

They don't give you enough credits to do that or I would

[lyric hare]

I think the whole model is doomed to fail honestly, its obvious they are training on mixed down music, so the model has no way of picking parameters for guitar, singer, drums etc.

if they trained as a MOE on the seperate tracks, this would be 1000x better

[winged fractal]

It clearly understands the difference between the instruments though, you can ask it for guitar solo or drum solo and it provides it

[lyric hare]

its what we do for AI voice reconstruction

[still peak]

omfg

[lyric hare]

it picks the closest parameter to a "guitar solo" that it has in its model, but it doesnt know what a guitar is.

[still peak]

watching people be more pedantic than me is really nuts.

hilarious but painful

that said, i'm trying very hard to make a macgyver-style 80's tv show theme and it's incredibly difficult haha

but i got it.

[winged fractal]

What genre do you even use for that

[still peak]

ha.

"macgyver tv theme"

i tried airwolf too but it didn't like that

i'm doing a girl punk song about getting your period at the mall

[grizzled slate]

Anyone got BPM to work?

[frigid orbit]

So does the Explore page follow the styles that were set in the first generation?

[tender helm]

i am not able to open the chirp page.

[hollow hare]

Chirp bot is retired, use the website https://suno.com

Some helpful starting links: https://suno.wiki
https://youtu.be/lj8utBd_IUM

[supple mason]

Heyy does anyone on here know anything about the python system for using bark? I am trying to work on a llm project with it, yet the speeds are unusually slow even on newer hardware, and I was wondering if anyone knew of any good ways to speed it up, as I am stumped.

[lyric marsh]

Did they change the charactor limit on custom mode? I can barley paste in a single verse?

disregard. I logged out and back in and it is working now.

[wet minnow]

it should be 1250 characters. it alwasy was around 1200. but yes.. it could be more for the first part.

[open oriole]

Is there any way to control the amount of distortion or eradicate it during build up? Sometimes vocals can distort. When you get a track that sounds awesome only to find the build up distorts.

[plain prism]

Is it possible to only change the lyrics and recreate the song in the exact same style?

[vague night]

I don't know why my prompt is being flagged for moderation. (lyrics) really don't get it. I can't figure out what exsactly is the issue.

[untold briar]

The easiest way to figure it out is often to ask regular ChatGPT to make a guess about why the prompt is being flagged.

[untold briar]

Maybe check Huggingface Discord as that is probably the most popular Bark implementation right now. You can DM me as well, as long you don't mind a delayed response.

[vague night]

so violent and hateful language? well what the shit. Thats like every death metal song out of the window 😄

[rapid sigil]

How can I extend the song 2 chorus sections with same style than before?

[untold briar]

I'm not sure I understand exactly, but in general when you want to have as much musical consistency as you can, you should 'Get Whole Clip' on any clip (except the first clip of course) before Extending it.

This would fall under post processing, yeah. Either just general music 'mastering' or 'stem separation' which is a way of processing audio and turning it into separate tracks for the voices, instruments, etc, which can then be adjusted in an audio editor.

[rapid sigil]

Thanks, I will try it.

[untold briar]

But if the two [chorus] verses are too far apart, it sometimes isn't possible, at least for now.

[vague night]

well... so I changed Hatered to Hate and then prompt wasn't flagged. Geez. Thats abit... much

[open oriole]

This is what I mean when I refer to distortion:
https://audius.co/rnp1969/paper-areoplane

It's also sounding compressed in the main final choruses.

[low remnant]

Mr Fly will probably correct me if I'm wrong because he knows everything Suno. I believe the look back window for previous chorus is 60 seconds. Kids, don't mix 8 line verse and slow introspective delivery. And, of course that chorus has to be in the segment you are extending from, so if not, tap that get whole song button and extend that. The risk of get whole each time is that sometimes aspects of the song can merge together and confuse the AI. Voices can start to sound like instruments or two instruments can merge or other audio degradation. Give the AI enough context to do its job but not much more. It's an art in itself sometimes balancing the two.

[lone oak]

I heard that SUNO is improving the reliability of the language. Can anyone reveal the date of the next update?

[final hazel]

From who did you hear this?

[lone oak]

Email from suno support yesterday

[final hazel]

Okay... well then it's only known inside the team. It's not officially known knowledge yet.

[vestal rapids]

I hope generation length becomes longer sooner than later

[cunning notch]

I hope we can remaster old V1 generations and even extend them. With that feature we could even add extending our own audio (for example extending our own demos)

[hasty cove]

Will there be a way in the future to download the midi files of a song? Say if we want to use the different Virtual instrument quality?

[woeful egret]

Hi, is it possible make male and female vocals in 1 song? If it is, can someone explain how please?

[wet minnow]

yes. its possible but difficult. have a look at the wiki page for that : https://www.suno.wiki/faq/style-and-lyrics/duets/ And here is a example i made . But keep in mind: every cut you see is a time where i had to extend the song because it got confused and i had to fix this. its not as easy as it looks and costs some credits to get it really good. or you have just luck and it works. 😄 https://suno.com/song/1b4f8607-1a8c-464e-b57d-d0f6089b28e8

[pulsar blade]

oh wow thanks for this. I came here to ask the same question lmao. cheers!

[hollow hare]

One of the keys I saw people try with the more successful ones were using male and female names in tags, that might help, but it is still hard to achieve.

It may be easier to make something like an instrumental part 1 then Extend it twice (one extension where you specify for male vocals in the styles, the other for females) then edit them together in a DAW.

But it is much easier (and cheaper) to make a full track then swap the voices gender using vocal swapping AI tools like kits.ai (see #🗣️┃chill-chat message and #💡┃suggestions-web message )

[rain mist]

when i'm extending at the moment, it doesn't seem to copy the style of the last chorus? any tips?

[broken lotus]

how long ago was the last chorus...?
needs to have been heard with about 1 minute of the time you are continuing....

[rain mist]

aaaa there's an instrumental part in the middle

[broken lotus]

it may be possible to save, with Extend from time, and sacrificing some of the interlude...

or start a new branch just to get that final chorus, and splice together in a DAW...

[rain mist]

i am technological inept so i will just suffer and deal with it, maybe use a bridge tag instead of chorus to mix it up

[pulsar blade]

thank you!

[fast void]

this was not an advice. a few weeks ago i was able to create music without noise with very complicated prompts, now simpler one outputs garbage. your response is same as "go for a run" or smth xD

[broken lotus]

I post REAL EXAMPLES every day hon.

You can indulge in fiction. But it's just your fiction, not even an opinion.

Maybe have a look at the 20 odd songs posted in the last 6hrs in https://discord.com/channels/1069381916492562582/1069381916492562585 ?

I can tag you in each and every one if you'd like?

ALL HAVE EXCELLENT VOICES perhaps yoiu would like to learn how?

[fast void]

ok, sure, lets compare any song from showcase. someones prompt: "Castalia music, female voice, chill, city pop, dreamy cute funk". im not sure if it so simpler than "melodic happy hardcore, uplifting and energetic, 180 bpm"

[broken lotus]

😐

does that random song have better sound quality than yours? If it does, then take their example as advice. 🤷‍♀️

You don't have to listen to me. Listen to other songs. Steal from the ones that have the sound you want. No one is going to force you to try a different prompt style....

[fast void]

yup, that random song does not have issue with noise. but your advice is invalid since my prompt is not any more complex than the others. and im not taking about just that one example, but a lot variety of them that ends up with noisy output, which btw sounds like could go away with more model steps.

[unique stone]

I don't know if this helps or how it compares to what you are experiencing or how close it is to what you were aiming for but here's the first two generations I got for the prompt "happy-hardcore, melodic, energetic, uplifting, fast":

https://suno.com/song/2b68ab16-c266-4db3-9404-199d9c2453e9
https://suno.com/song/5ee37394-5505-4b2e-b167-b5cafa20176f

These are just instrumentals (not sure if you were looking to have lyrics.)

(it tends to ignore bpm so "180bpm" likely won't do anything, so I just went with "fast")

[fast void]

ok, so thats a real advice. and it seems that noise problem may be due to lyrics (not english). thanks.

[unique stone]

If it isn't noisy all the way through you can sometimes extend from an earlier time to avoid it

[unique stone]

I tried extending one of them with some words (describing the voice I wanted in the style). One of these skipped the verse but just to give an idea of the prompt and what it gives:

https://suno.com/song/b205aaea-5b6b-44f9-8a44-6ff98c4d67a6
https://suno.com/song/0f759537-d18c-4214-92d5-eb1206d8d88f

(English lyrics though as that's all I know sorry 😄 )

This is the first time I've managed to get happy hardcore I think (I didn't try "happy-hardcore" before)

[fast void]

those two imo start to sound kinda noisy. so maybe suno just isnt best suitable for combining happy hardcore (and other fast edm) with lyrics

[unique stone]

I don't work with this genre normally so I can't think what else to suggest. I notice a bit of a "fuzz" to the vocals. You might be able to use a separate program/service to split the vocals out and replace them but as it's quite a loud style of music, it might not give great results (harder to cleanly separate the vocal/instrument)

[fast void]

yea, i know, fl studio manages to split vocals and its ok-ish, but song cannot be very noisy by itself.

[wet minnow]

https://media.discordapp.net/attachments/539129467898167317/758017679134163005/explodyparrot.gif

[steel gale]

Hey All! For the paid subscription, do we have the ability to set the initial song length higher than 2 minutes and able to add more custom lyrics than the free version? I really dont want to make two files and try to splice them.

[wet minnow]

free and paid work the same. you have to extend them with the extend button... have a look here for a neat tutorial https://www.suno.wiki/faq/making-music/how-do-i-make-the-song-longer/

[steel gale]

Hope this helps. I have been failing at trying to extend them. 🙂

[wet minnow]

its super easy. basically just 2 steps. and suno makes one long mp3✌️

[steel gale]

ok. Looking good. One more. When I go to extend, I have the option to change the style of music but is not recognizing the new prompt. I would like to transition between male and female voices. Is that possible?

[steel gale]

Changing the lyrics worked so half way there. 🙂

[wet minnow]

have a look here for the duet thing... #📚┃suno-school message 😄

[steel gale]

much appreciated! You are busy. 🙂 Hope that makes it on the new feature list.

For now, appears changing the style of music for the extension track is not working.

still cool. Hope the duet feature can be added.

[wet minnow]

i have done a lot of experiments also including style changes. everything takes some tries and fiddling around with the continue time. style changes are also quite a challange... but doable. Its basically like with everything in life. if you want the better looking (sounding) stuff, be prepared to pay more for them 😄 . but even if a song is like 1000 credits. thats 3 bucks. basically nothing compared to a professionally produced super customised song. i ll give you the link to my suno profile for some style inspiration if you are interested 😄 https://suno.com/@mrtree

[steel gale]

You have been busy! Nice work.

[thorny cypress]

I got it to start with an instrumental chiptune and then put a voice in later. So it's possible, you just have to experiment.

[steel gale]

That extending does a really nice job!!!

also, not putting in lyrics for the last extension makes for an amazing end to it all.

very impressed with this tech.

[lime bane]

Hey yall! Is there an API for Suno?

[unique stone]

I find when trying to change styles or introduce a new instrument, taking it out of instrumental mode and putting e.g. [bridge] in the lyrics box can sometimes help give things a nudge. There's other things like [instant switch] that might work similarly.

[lime bane]

I looked around the webstie but couldn't find anything 😦

[unique stone]

Not at the moment.

[lime bane]

Maybe this isn't the place to ask, but are there any services similar to Suno that have a developer api?

[unique stone]

You'd need to find and ask those other services. I don't know of any.

[broken lotus]

doesn't google have a service with an api...?
afraid I don't remember the name...

[late trout]

any time i try to do the /chirp command it just comes back with "messages failed to load" is it not working right now or am i missing something?

[unique stone]

Chirp is gone, replaced by the website at https://suno.com

[late trout]

thank you

[merry storm]

if i get a monthly sub to the pro plan, can you unsub at any time?

[final hazel]

Yes, you pay a little more if you don't do the year plan, but you get the freedom to stop anytime you want.

[merry storm]

thank you, and credits, do they dissappear at the end of that sub duration?

[final hazel]

Yes, it's a "month" plan.. so refresh every month

[jade tapir]

Hey guys im new here, but ive been using Suno for months now, I want to know how you guys fight against having the song hallucinate

I write my own lyrics but everytime i extend a song, the lyrics start off well and then everything just derails

[thorny cypress]

Rule #1: make sure there's a chorus in every segment because if there's not, it'll forget what it sounds like and make up a new chorus.

Also I try not to do a whole segment of instrumental, unless the whole song is instrumental, otherwise it'll forget the entire song and meter and everything. You can put [Instrumental] in between some stanzas to get a short instrumental bit though.

[slender urchin]

what keyword do i need to write in the text box that suno understands that the chorus repeats and the volume fading out at the end of the song?

[thorny cypress]

Here's how I did it.

https://suno.com/song/e8e58cf3-2256-44b8-9401-821a4ce383e0

[slender urchin]

so one section with 2x the chorus + [outro]?

[thorny cypress]

Here's another example of how someone else did it.

https://suno.com/song/689b4118-d940-457f-ab6d-9e7de0115ce2

So there's probably multiple ways.

[grim skiff]

Does anyone know how long it takes for Suno to respond. I have a 1 year subscription and cant get my account back to the paid format? I emailed 4 days ago. No answer and I can't generate songs.

[final hazel]

What do you mean? What is wrong with your account?

[edgy roost]

When it says "A system error has occurred." When generating. What could be the problem?

[fiery bobcat]

Im having that issue to and the credits are showing either

[torn brook]

Do you mean in addition to our two stanza, or should always dedicate one of the two stanza’s to being a chorus?

[thorny cypress]

How many stanzas you can fit in a generation depends on the song tempo and stuff. Which you can't always control. But usually you can fit in at least 2 verses and a chorus in the first segment. I wouldn't dedicate an entire segment to a chorus unless you want a lot of instrumental stuff, or want it to repeat it when it runs out of lyrics to fill the time with.

I'm just saying, whatever else you put in a segment, it needs to have a chorus in there somewhere. If there's a segment with no chorus, by the time you get back around to it, it'll have forgotten what it should sound like since it can only remember the segment immediately before the one you're working on. If you're extending from part 2 to make part 3, it won't remember part 1.

[empty meteor]

how can i fix it

[thorny cypress]

Usually means it brokeded.

[wet minnow]

did you use the reuse prompt option? if yes, check if there are too many lyrics in the lyrics field. shouldnt happen...

[empty meteor]

nothing is written and there is a problem loading like this

[balmy crown]

suno just... ate some songs?

[thorny cypress]

A lot of people are reporting that.

[balmy crown]

like I closed the browser and when I came back, the songs it had generated were gone

alas

like it refunded my credits but i was actually okay with one of the songs (it was even two parts, but came out okay!)

has anyone reported the songs being.... uneaten?

i'm just like fuuuudge i would've saved if i had known lol

I'm trying to figure out if it's more or less likely that I'll get them back if I try to re-create them with the refunded credits >>

[midnight trout]

Please tell me I did not just lose the last hours worth of work

[balmy crown]

hahahahahaaaaaaa

I cannot tell you that

[errant osprey]

What does this mean..?

[balmy crown]

idk i feel bad for suno though

it's rough being in beta, and this is gonna make a lot of people really angry

[final hazel]

I just had it too. I've reported it to the devs.

[balmy crown]

dumb question, but how do you report it to the devs?

[smoky mesa]

I cant generate and delete my songs rn, is Suno down?

[final hazel]

I can because I am a moderator

[twilit veldt]

I had songs "eaten" too

[final hazel]

The devs are notified

[balmy crown]

oh, okay, that makes sense lol

the power of personal authority

I won't try and find a way to submit anything, then, because even if I did, I'm one voice in a zillion users right now

[final hazel]

https://tenor.com/view/40year-old-virgin-salute-wink-beautiful-pretty-gif-17334149

[balmy crown]

I guess I'll wait for the day to officially roll over and start trying to recreate the songs? I don't want to use the credits that have magically reappeared, because I'm hoping it won't.... overwrite something somewhere. This is more magical thinking than technical thinking, but I'm just gonna roll with it

This is just a good reminder to me to back things up offline

[hollow hare]

I always try to backup onto YouTube first for reasons like this (and "proof" of copyright)

[balmy crown]

I'm an idiot, but how would having something on youtube prove... copyright?

like I thought that in theory AI generated things were messy in terms of copyright right now

or is it just to have a record "I had this at x time"

[hollow hare]

It's usually to stop other people from taking the song from the home page and uploading it first

[balmy crown]

ahhhh

[hollow hare]

Albeit not much of a problem for me since my songs do not trend lol, my public ones are usually tests

[balmy crown]

I think my songs are pretty much only pleasing/interesting to me so I haven't told Suno to make any of them public lol

they are lovely for ME to listen to, though

[hollow hare]

Ikr, I can't stop listening to mine too

[balmy crown]

I have a little playlist of all my favorites (downloaded as backups to my harddrive) and I'm just like.... ahhh, magic custom songs

I even downloaded some where the lyrics went all wonky but it just sounds catchy -- but again, almost definitely just to me

[wet minnow]

yes. there are server problems. the devs are informed

ah i was back in time

[full hull]

mr tree how are u now

[wet minnow]

quite busy... like always when something is not working 😄

[balmy crown]

are you also working on this project?

[wet minnow]

i am not a official suno employee. i just look for the community in the discord.

[balmy crown]

ahhh okay

[lament shore]

my generated song disappeared

[wet minnow]

yes. there are server problems. the devs are informed

[balmy crown]

I wonder if they're the kind of problems where the songs can be recovered (like the data is still there, just not linked up) or if they're gone

I should've learned more about databases in school, I'm a disgrace

[lament shore]

Will the devs be able to restore the song? I am a paid subscriber.

[balmy crown]

I don't know if anyone in this forum actually works on Suno itself?

and they might also not know yet

assuming that they haven't pinpointed what's going wrong rn

[wet minnow]

we have to wait and see. but usually everything comes back that was already generated

[wet minnow]

devs are informed and also working on it

[indigo walrus]

but points is restored too (

[thorny cypress]

Was this what happened the last time when they gave us a bunch of credits to compensate for it?

[wet minnow]

everything will come back

[balmy crown]

i'm afraid to use the points that were restored in case I get in trouble for using points that I don't technically have (assuming that the songs still exist somewhere)

[balmy crown]

has this happened before, or something?

[wet minnow]

dont worry about that

[balmy crown]

...okay then i'm going to listen to the devil and try to recreate the song with the points that were restored

[wet minnow]

yes. during a server migration the servers were offline for some time and it was like the world ended for everyone

[final hazel]

the missing songs should be back

[thorny cypress]

I don't know. There was one day when it was down entirely and wouldn't even let you try to generate but I wasn't here then.

[balmy crown]

ahh I'm so relieved rn

thank you all for being reassuring when I came in like a chicken with its head cut off

[unique stone]

You're no chicken

You're a duck!

[balmy crown]

You are so right

for a second i weas like "how do you know that's my nickname" and then rememembered it's in my username

[delicate sedge]

can we please work together to expand the input profile to be more than 244 characters now. I was using wave former before this and I could fit so much more into the description and had file names past the windows 258 character limit. Those file names were descriptions combining genres and instruments as well as different music theory terms or directions or instructions a composer or a producer would make to get specific results. it is difficult to do this with the current 244 character limit.

[thin flint]

Can anyone DM me? Just started using Noisee and well, how does one make full videos for the whole video? Only get a minute by default.

[twin igloo]

cant extend songs tonight.

[glacial scarab]

uhh... maybe ask in the noisee discord?

[coarse pecan]

hey guys i have a question, i create a song with vocals, can i use the only vocal only for commercial if i have the pro subscribe?

[untold briar]

You can create a song using music styles like "A Cappella" which can make a song that only uses vocals, but you can't take a regular song and use Suno to make that same song with vocals only. For that, people often use "stem separation" software, for example, "ULTIMATE VOCAL REMOVER V5" is a good free one.

[coarse pecan]

i use moises.ai very simple and very good@untold briar

[gloomy stump]

why does suno always try to avoid creating rap beats? Its the first word I put in but it always avoids it in 8/10 times

[shell loom]

darn discord:)

[unique stone]

Could be worse eyeballs the memes channel

[carmine cobalt]

i think it works well if you do Hip Hop / fast rap

[oak leaf]

Why is Suno giving an error?

[hollow hare]

You have to be very specific about which error you are getting. Show a screenshot if possible

[hollow hare]

Chirp bot is retired, use the website https://suno.com

Some helpful starting links: https://suno.wiki
https://youtu.be/lj8utBd_IUM

THERE ARE NO OFFICIAL MOBILE APPS. Mobile apps promoting Suno merely redirects to the main site for a fee/subscription. Do not give your critical info (e.g. passwords, financial info) through these apps.

[oak leaf]

Oh ok

[unique stone]

Instead of replacing vocals with a different voice after using a stem splitter, has anyone found a way to "smooth" the original vocals out so they're less scratchy? I'm going to try a few things later on - the first thing I'm thinking of is deliberately encoding the vocal at a lower MP3 bitrate to lose some of the detail (that detail being what I expect makes up the "fuzziness".)

[broken lotus]

mostly I have used EQ and a bit of reverb to soften the vocals.... (stems or just isolating with EQ)

[unique stone]

EQ definitely will help in my case as I have a music box that's been leaving a trail of destruction throughout the song 😄

Reverb was something else I was going to try, though the vocals I have got from Suno are caked in it already - might try a very short-length reverb.

[lyric sonnet]

a narrow notch filter in the mids often helped me to improve harsh vocals, if you hit the right frequency or frequencies

it's easy to find where exactly if you invert the notch to a bandpass to find which band screams, then pull it down

[sacred arch]

hello everyone. I made an instrumental song that's about 5 minutes long, but when I uploaded it to my YouTube channel, 2 minutes of the song were flagged for copyright issues. Has anyone else faced this before?

[final hazel]

just dispute it. Because it is 99,9% a false flag.

It's an automatic process.

[sacred arch]

Thank you! I will do that.

[vapid peak]

hi i'm new, do you knwo why i can't access to #how-to-chirp and beta chirps rooms ?

[final hazel]

Chirp bot is retired, use the website https://suno.com/

Some helpful starting links: https://suno.wiki/
https://youtu.be/lj8utBd_IUM

[opaque pasture]

Why can't i make a song about the zombie apocalypse?

Why is "Zombie apocalypse" flagged? -.-

[hollow hare]

It got censored from ChatGPT (which moderates the lyrics in Suno) not liking something. Ask OpenAI's ChatGPT itself about "what would censor the following: [content]" and it should tell you.

Usual causes are using swear words or interpreting harmful situations. Even if it worked before, it could be a small change caused its interpretation to become different.

I remember people getting censored for using bones, ribcage, and skeletons because it made ChatGPT assuming some sort of horrific tale.

Use some phonetic spellings to help avoid the censorship if it is due to the imagery... so long as your lyrics still follows the ToS

[opaque pasture]

I'll try, i want it to rhyme and make sense in the end

[hollow hare]

Definitely phonetic spelling is the way to go and is needed I'm certain cases like live vs live (liv vs lyev)

[opaque pasture]

Kind of weird because i can make songs about hell and the devil and it doesn't flag stuff even tho it's also... a prompt for making unholy images via ChatGPT