📚┃suno-school

[charred merlin]

"more expensive" is debatable in the long run after you compare the cost of taking a more direct route vs how many attempts it takes to get the ideal song as a "one-shot generation"

[unique stone]

True. I should probably word it better: It seems more expensive but it probably won't be.

[golden plover]

Still waiting for Suno to be able to change lyrics while retaining the melody of recording. By not having this is just sad and pathetic.

[charred merlin]

Like the hamster'd pointed out, using instrumental extensions can be very powerful in that sense -- instead of using [instrumental intro] in the lyrics and hoping that it lasts 20 seconds, you could generate an instrumental track and make an extension where the lyrics start after the 20s mark

[humble phoenix]

Let’s say an original generation of a song is owned by Suno, as defined by “rights and ownership”, then you become a pro user and do one of the following

1. If you extend a song, does the “full song” now belong to you?
2. Does a “remastered” version belong to you?
3. Does a “cover” belong to you?

I found versions of my songs that I made before subscribing that I liked and wanted to keep them, but needed a newer version in order to publish them later down the road

[charred merlin]

If the song was made on a free plan, Suno retains ownership of it -- extending/covering/remastering it doesn't mean that you now own it any more than I could own your songs by extending from them.

If you've subscribed, you can email [e-mail] to request the commercial rights to specific songs you made on the free tier 🙏

[humble phoenix]

Thank you!

[fathom siren]

Yes few tricks to it. Intersperse some fake lyrics in the long instrumental (inaudible: hum), you can also have luck with telling it how many bars you want it to be [Breakdown: 24 bars, ..]

[fathom siren]

You can achieve that, just replace, persona, cover, etc..

[solar karma]

the website keeps defaulting and it didnt used to do that.... what did i do

to 3.5

[golden plover]

It gives a completely different melody and then the whole song is different. What am I supposed to do? Put it in a daw and manually tune the pitch of the vocals? If I do that it will sound terrible and at the same time the way the song is sung is completely different than the original.

[fathom siren]

Just cover it, and fix the word, remove all the existing tags and styles except the lyrics you are fixing.

Or do the same with replace

If you don't care about it being on public suno even easier, as you just generate like 10 of the same cover, and cross fade in external app the one word fixed.

[charred merlin]

[8 bar intro], etc?

[fathom siren]

As with most things, with the prompting you are influencing the outcome, not purely directing it, but in my experience if used sparingly it works (obviously not every gen).

[charred merlin]

Not to say it cant be done, but I've yet to be shown an instance where they worked reliably.

[fathom siren]

Yes but they are re-inforcement training using suno at this point, and we are writing things like that and clicking thumbs up and down etc.

[golden plover]

Nope. Doesn’t work. I am still getting the old lyrics even though I edited the lyrics.

It’s obviously that the features which are extremely crucial are broken

[fathom siren]

If you are doing it right, you should be able to cover the song, and get nearly the exact same song.

[golden plover]

All I am doing is editing one lyric and all it’s doing is generating a different arrangement of the same exact song

[charred merlin]

The thumbs up/down do help train the model, though it doesn't provide personalized results

[fathom siren]

First edit lyrics, remove all tags, just leave the words that are said, if they don't match the existing track you first need to align them include every word you hear spoken.

[golden plover]

I did that, and it’s not changing the one lyric line

[fathom siren]

That is okay, are you able to get nearly identical covers now?

[charred merlin]

No one is saying they're adding user-generated data into the training data -- that's not the same thing at all. We're talking about using the existing premium training data set, building a model based on it, and getting feedback as to whether the current settings are working or not.

There is NOTHING to prove that Suno is feeding user-generated content into the training data -- if anyone is actually claiming that, please share the data -- harly made a lot of big claims, and has spent absolutely zero time showing his work.

[fathom siren]

with all the words remaining the same?

[golden plover]

When I cover the song, I am getting the same song but the new lyric is not being applied. It’s just one word and it’s refusing to change it!

[fathom siren]

If you highlight that portion in the lyric editor does it highlight the correct portion?

[charred merlin]

It's sort of a moot point unless there's anything we can reference -- we can speculate, but they haven't shared these kinds of details regarding how the model is trained.

[golden plover]

Looks like I have to sing it myself and use kits or some voice changer to make it sound like the AI voice. But what voice? It’s impossible to find these voices!

I used the replace section and it’s giving me a completely different melody and variation

Changing lyrics using the cover feature does not change the lyrics!

[uncut hawk]

Wish they would fix the Trash can process.

Wish they would watch what people actually click see keep hitting the same dam buttons, and wizards easy to predict an how to fix the UI.

Wish for click all and or flags all to be delete, then un-click thumbs up, click trash can. or even better, click the thumbs down, and click move thumbs down to trash can. I stopped cleaning.... Trash Not going emptied any more.

[fathom siren]

Yes can take many generations to get exact melody, or ensuring you select a large enough portion it can understand the melody. Often times easier to fix with cover, anyway once you have a lyrics alignhed and covers generating the same, you can try changing the one word by writing it phonetically or a - before it

Can make it change word when it's being stubborn

[golden plover]

I also wish you could seperate the vocals from the instrumentals naturally let’s say I generate a song 2 seconds ago, it is straight up criminal to have it generate a song and then use a built in voice separation tool which is something lalala ai can do

[ashen carbon]

anyone figured out a way to get rid of the "shimmers"

[charred merlin]

@golden plover https://bandlab.com/splitter is free

[ashen carbon]

ugh but the vocal quality comes down quite a bit in 3.5

[golden plover]

All it is is a in built vocal splitting tool within the AI. You hear a lot of bleeding from the instrumental. That is completely unacceptable

[ashen carbon]

maybe 3.5 + v4 Remaster? or does the remaster introduce shimmers?

[ashen carbon]

yeah i don't like the built in stem separation compared to demucs

[unique stone]

Sadly, remaster can. But it also can sometimes generate stuff that doesn't have them.

[fathom siren]

I mean it's free or $10 a month lol

[ashen carbon]

i don't follow

[unique stone]

v3.5+ isn't an official/documented thing, to be fair

[golden plover]

It’s an unnecessary feature which should be removed

[charred merlin]

They refer to covers/personas as a different model, but calling it v3.5+ is confusing

[charred merlin]

The occurrance of shimmer has definitely improved since v4 launched, and a large number of my remasters have saved a lot of songs that were previously "lost" to shimmering or a progressive trebley-hiss, but these are still artefacts present in v4

[fathom siren]

I can just never get a remaster with the same volume or bass as an original. Always pulls back

[unique stone]

It was never intended to be used in the way that a few of us use it 😅

[charred merlin]

In future instances, It would help to explain what you mean by v3.5+ if you want the notion to catch on 🙏 Or to link to your workflow post where you explain your theory

[woeful yoke]

Yeah things are disappearing here

My whole new post got deleted under v4 feedback the other day.

[woven sail]

suno is written on the thumbnail

[charred merlin]

dude... furrybeans went over this so many times.

This is either willful ignorance, or you just really don't understand.

[woeful yoke]

Sorry I haven't been able to keep up with everything all the time. I guess I have to go and find it then.

Thanks, I guess

[charred merlin]

#1318532031545610290 message

Here's your post -- furrybeans explained why the automoderator removed the post. He's also followed up a number of times when you've talked about the post being removed to reiterate

[west yew]

Extending an instrumental is something I have done, but prog rock instrumentals tend to have a feel that has problems with lyrics. But now you've got me thinking of angles I will try! Thanks!

[charred merlin]

The mods should be able to comment on what happened -- just like with Harly's post, they should be able to look into why it was removed, if it was an automated thing

[hollow hare]

Please use this form for the Suno team to clear the issue from your account:
https://sunomusic.typeform.com/to/zGsJSzsf

Do you recall who the author was?

[untold briar]

Extend, and also Replace used similarly.

[charred merlin]

I don't see it there anymore, but even if it timed out presumably you would still be able to see it -- one of my older ones was locked due to inactivity but I had the option to unlock

[woeful yoke]

The threads are usually jumping around depending on the newest comment activity,. though?
Sometimes it's so annoying to try to find a thread again that you just had seen or commented on recently 🤪

[polar spade]

🎶 ... it wasn't me.... 🎶

[rough pelican]

i'm kinda frustrated. regarding lyrics. i know there is ways to make it emphasize words or pronounce them differently using "~abc~" (Tilde) or simple "(xyz)" parenthesis. And i'm also aware how the effects of these are. But is there any way to make it say words in a long way? Say "gooold" instead of "gold"? I tried this a few days ago with other lyrics and it just chopped up my lyrics, so is there any other "secret" method?

nvm now writing "gooold" instead of "gold" worked. RNGesus trolled me

also adding an empty line after that word seemed to have helped too

[fallow musk]

is there an issue with the replace section !, iam trying to replace 15 second only, but its geeting me 2 mints

[past glen]

is there a way to change the lyrics in a created song without it changing the sound/music - just the lyrics?

[ornate roost]

Why is every song made using v4 suddenly cut of at 3:16? (using persona)

[rough pelican]

if you want to keep the music exact, probably no. if it's part of the lyrics in the beginning of your song, i'd say it's not possible. If it's in the middle, you have to extend your song from that part on and hope you get a good result but it most likely will change your music aswell.

you can try the replace section feature, but in my experience that is extremely unreliable

[untold briar]

Replace feature can do this, if you mean a line, not the whole song. Watch the tutorial video though, you have to be sure to have your song lyrics match what is being sung in the song, for the whole song, not just the prompt in the replace section. And then make sure the red highlighted lyrics match exactly the words you hear in the selected section that you highlighted.

To test if your segment matches, pause playback, move to a time before your segment, then hit play. The player will skip to the start of your selection. Look at the red text in the right column and make sure the red words 100% match the words you hear, not missing any, no extra. If you have no red words, then click "Edit Song Detail" and add a blank space to the lyrics, save the lyrics, then try again. This should fix it. Maybe helpful to @fallow musk too

[pallid ibex]

I bought a premium subscription and what I get are only songs that last until 03:03 on the v3.5 and on the V4 only 03:07 even extends me and repeats everything but doesn't continue the missing part!!!

I paid 30 $ for a 03:03 Minutes Song?

[untold briar]

If you want to keep the whole song the same and change all the lyrics, this is actually possible with Cover feature, but it is usually not easy to do, will require patience and trial and error. But you can probably get something very similar without too much trouble.

[grave wraith]

What do you expect how long it should be? 🤔

[pallid ibex]

According to the description it can last up to 4 minutes, my lyrics were designed for 3:30 minutes... the last 2 verses are not sung and with Extend it starts again.

[torpid thicket]

It seriously depends on your lyrics

[grave wraith]

One thing I learned: YOU can not design. The AI designs 🙂

But that sounds a little like you have a tag in between that makes it think it should end? It also depends on your genre... if it sings faster it is shorter i.e.

[torpid thicket]

But if it's just Instrumental, don't expect a full 4mins

[pallid ibex]

No matter which song lyrics I use, it only takes 03:03 minutes.

Yesterday it lasted up to 4 minutes

[ornate roost]

My issue is that ALL songs using a persona are cut of at 3:16, no matter if there are instructions or not in the lyrics.. I plan for 3:20 to 3:50 and suddenly have to extend meaning I get much worse extensions than if the song had finished..

[torpid thicket]

I think SUNO is working on something, because I've had a few people complain about it's taking awhile for Generations to finish

[grave wraith]

If you want you can give me your lyrics and prompts and I will test how long it will be on my end 😄
But really it could be it's just the style

[pallid ibex]

the lyric is on german

[grave wraith]

Kein problem

[pallid ibex]

I've send a frienbd request

[grave wraith]

Here you see...changed 2 words between the first and the second and the third is even the same lyrics as the second

[ornate roost]

Style and lyrics altered for almost every song, yet this..

[unique stone]

This is currently a limitation that affects personas/covers. Apparently it depends on the length of the original song a persona or cover is made from (longer original = shorter new song.)
With that in mind, planning your lyrics to fit this time, and adding any further lyrics in an extension would be the way to accomplish a longer song using persona.

[ornate roost]

Oh.. darn.. hope they will fix this.. thanks for info.

[unique stone]

Me too, it seems a frustrating limitation but there's likely some technical reason behind it.

[lean crest]

Then I can't cover a 4min song?

[unique stone]

Not entirely, unless it somehow makes the tempo faster to fit the lyrics in.

(though again, extend can help here)

[near timber]

Anyone know how I can get a song to sound similar to this

https://music.youtube.com/watch?v=mX4H_rco9oc&si=AY59zuQlVzB0eIf0

[lean crest]

Good tips, I'll keep that in mind next time.

[polar spade]

Just to split hairs. . . . " up to...." doesn't mean "exactly" . So it can vary from 1:30 to 4 minutes. "Up to" means it won't go any further than 4 minutes. Songs in real life don't all end exactly 4 minutes. If u just have 1 stanza of lyrics, for instance, Suno just might go 28 seconds and end thinking you are writing a jingle. The general consensus over the year seems to be members don't want Suno to repeat verses and throw in interludes just to take up space and extending songs near 4 minutes. I think over the past month, Suno has therefore become better at being succinct. But most want them to at least end over 3 minutes, which is about normal.

As for all song ending the same time, that's an oddity that's not fully explained. 3:14 was a popular time for Covers I think. Certainly try to change something. Oftentimes , cutting your prompts and just sticking in one odd style to reset or wake suno up from repeating it's previous stats and style seem to have an effect. Then paste back in your style to see if it changes on the next gen.

Many of us, I included, like to extend after each verse or section to have better control over the outcome, so 3 or 4 minutes is often never needed. Experience tells us an average song will initially cost about 600 credits. Changes and tweaks after that will cost more, of course.

And anything that repeats should go no further than two clicks before stopping and trying something different.

[polar spade]

As a general observation, the sounds and even vocal seems doable. But what exactly to use is difficult.

One prompter says: Melodic Pop,Expressive Vocals,Acoustic Guitar,Piano,Love Theme,Intimate Production,Smooth Style, Commitment Mood

Another: chillwave, synth-pop, down tempo, contemporary R & B, g-funk

It's a mid tempo piece. You may want to browse persona's to see if there's a vocal that sounds similar.

Your lyrics should reflect the same or similar theme as Suno will pick up on that. You would likely NOT get an Ortiz like vocal quality by writing a Viking Epic Story as lyrics.

[proper glacier]

Is it possible to generate the same exact melody but change the vocals to all female, for example? I thought cover might do it but it doesn't seem to be having an impact on four attempts so far. Not sure I understand the cover feature well enough

[untold briar]

Cover can do it, just might take a bit of trial and error. Try keeping the same style prompt as the original song, but adding at the end "female, female, female".

[proper glacier]

I tried adding female vocals to the front of the style and excluding male vocals. I'll try this sounds good

[untold briar]

You can try Exclude as well, the reason to add it to the end is that you want the song to be as close to the original as possible, so try to use as much of the original style prompt as you can. I'm not sure if Exclude will change the Cover too much, just haven't tested that, but Exclude in general is fairly week so probably not.

The other thing to try that will almost 100% work is Cover a Cover instead of the original. This will increase chance that the song itself is changed in other ways though, so it has some downside.

[proper glacier]

Hahaha. It made it a deeper male voice 😂

[prime anvil]

@woeful yoke mate, are you there?

[runic stream]

Hello everyone, what does "song radio" mean? Could you provide any learning links?

[polar spade]

Try it out. See for yourself. It costs no credits.*

• It s just a list of members public songs which might be similar in style to yours.

[runic stream]

I see~ thank you Dasvon~~

[indigo sandal]

does anyone know how to make suno fade out at the end of an instrumental song? it keeps making the end of my songs abruptly stop: https://suno.com/song/0a0fcab8-8c5f-4bcd-9925-4069d9c969f3

very frusterating

[fathom siren]

Try something like

[instruments fade in layers]

[final fade]

[indigo sandal]

ok ill try

[indigo sandal]

your a god, thank u

[opaque ingot]

Do you write this in the style section?

[fathom siren]

No you write it in the lyrics at the end of the song @opaque ingot

[opaque ingot]

If its an intstrumental though, it removes the lyrics box. Or do you just not check the intrumental box, not add lyrics, and just type that in.

[fathom siren]

You can still do pure instrumental with the lyrics box as well

[opaque ingot]

Ahhh okay. Thank you.

[fathom siren]

Things [stuff here] inside brackers genereally gets interpreted but not sung, so if you just don't put lyrics you can control the instrumental composition

[opaque ingot]

Oh thats wild, okay thanks.

[naive sedge]

How are you accessing the menu shown? I don’t have options for cover or persona or stems and have pro subscription. Is it only an android feature?

[fathom siren]

The three dots on the song.

or if you are putting in songs, switch to custom

at the top of app I belieev

[maiden lantern]

Anyone know why all my songs got cut off at 4:00 suddenly? They used to play through fine

[barren coral]

been with suno a while, thought about getting the pro plan now i can afford having another subscription, wanna ask, though, when it says access to V4, does it mean a certain allotment a month seperate to the regular credits? (akin to free users having 10 'v4' credits') Or is it just available in general with normal credits?

[fathom siren]

4 minutes is max length for version 4, unless you extend

[barren coral]

4 minutes is the max

[maiden lantern]

I had already generated the song (v4) and listened a bunch of times it was like 4:30 probably not longer than that. But suddenly it cut off now like they went back and chopped it

[leaden lark]

It uses normal credits for v4

[barren coral]

yummy i love normal cream :P

[leaden lark]

I hate autocorrect

[barren coral]

lmao me too tbh, funny though! thanks

[maiden lantern]

Ok so wth still doesn’t make sense why V4 generated a 4+ minute song and i had it and could play it, then suddenly it got trimmed and the lyrics display speeds up at the end like it glitches out. Do I add the lyrics that are now missing to the extend lyrics and edit the original lyrics to not have them? This is weird

[fathom siren]

I've occasionaly had songs that glitched past their intended length

[maiden lantern]

It was good though like everything was perfect

[fathom siren]

Try downloading it, see if the file is different

[maiden lantern]

It’s the same unfortunately it got cut too

[unkempt fractal]

mine as well

[normal estuary]

I really wish there would be a pubic api to use

[cyan holly]

wtf, why on earth am getting 'Your card has been declined' while I'm trying to pay for my monthly subscription. And yes, before you ask, my card behind the account has money on it. This is annoying AF. I wanna slap someone right now..

[rapid needle]

Yoooo every time I try to make any song it says “something went wrong”

And the remaster songs option has disappeared

[cyan holly]

Also, on the Android app, it tells me that my subscription is activated and I have 500 songs left this month. On the PC/web, it tells me that I'm out of credits, but I can create 10 v4 songs for free, lmao

Also, the 'Verify your phone' doesnt work, as I've been waiting for 5 minutes for the code sent to my phone that is yet to arrive.

I'm gonna try another browser just in case..

lol, so according to people with the same issue on Reddit, some countries are now labeled as 'high risk countries' and payments from those countries are rejected?! THE F*CK!?

@umbral radish @sharp oar @grizzled prairie

I've been effing using your service for 6-7 months, never had this kind of problem and treatment

Pathetic

Just made a subscription to Udio with the same card, just to see if it works - everything worked perfectly well. So the problem isn't on my side.

[stable mirage]

Can you help me?? So on thursday I bought the pro plan of Suno AI from my phone but till now I haven't received my credits and also I'm not able to use any of the paid features on both browser and phone, I also emailed Suno AI about it but they haven't answered yet. Please help me.

[rapid needle]

Getting a something went wrong message?

Because same

[stable mirage]

no

I haven't recieved my credits

[tame eagle]

Hello i have a question. I see the monthly plan has a 2k song limit. is there anything more extended? i would need to create 6k songs

[graceful condor]

Please, message Mike B

I remember you having the issue before

[lapis loom]

how do I switch genre mid-song??

[graceful condor]

Extend

[opaque canopy]

i know im crazy but i want to make a song that has death metal verse .... opera chorus....... and how about a blues outro......

[lapis loom]

doesn't seem to work

I wanted to change from afrofusion tribal to sea shaunty

[graceful condor]

Have you changed the styles?

[graceful condor]

Should be possible, but tough

[lapis loom]

yeah I erased it and inputted the other

[graceful condor]

Replacing gives you the same output?

[lapis loom]

yeah

[graceful condor]

Try adding as a tag in lyrics as well

[lapis loom]

mmmh good idea

how do I evolve a 3.5 song to 4? it isn't allowing me

nvm

found it

I got a feeling that v3.5 is better at soft vocals

https://suno.com/song/f8a6eee1-cd34-4565-aafd-bc620bbdb054

as you can see, audio here in v4 is clearer but less soft and calm https://suno.com/song/0ff7b225-e282-4feb-aae8-8206f6e35976

mmmh, it makes me like v3.5 more in this song due to the chill vocals

and why does it struggle so much to end the song anyway, even if I write (End) in the lyrics

it just wants to continue the song on and on

[sonic salmon]

Hi all newbie here on Suno, a.i , I am trying to make Afrobeats songs, is there any prompt I should use to make the vocalist sing with a Nigerian accent as an example..forgive me if this is not the right place for my question.

[karmic pasture]

trying to add heckling during live music. ANybody got luck with that?

[proper bluff]

Hi guys!

I’m trying to do a jersey club song but instead it gives me a techno song. I need help! (3.5 version)

[karmic pasture]

define jersey song

[proper bluff]

Jersey club you don’t know that?

[karmic pasture]

never heard

[proper bluff]

https://suno.com/song/6fd488f6-0245-4f71-a76c-2b210b5b6f99

This is jersey club

That big boy I wanna use

[polar spade]

I don't know Nigerian. Certainly, you can add [Nigerian Vocals] and see what you get. That would be a first path. Even for afrobeats, you could nudge a dialect by throwing Mbube at the end of prompts to see if it helps.

[proper bluff]

Someone help?

[karmic pasture]

ok, my experience is more often than not the AI will not be precise with specific genres, especially if you dont trigger with good prompts and music themes common to that genre.

[proper bluff]

Well I did pretty good

Suno should fix this

[karmic pasture]

try adding genres and prompts to flesh it out

[proper bluff]

Which ones

jersey club is dance music but whenever I put it on there it just uses the original dance music but I want it jersey club

[polar spade]

A quick cheat is just search jersey club on suno Home page to see the prompts used , when you scroll through the songs and find one's that have the vibe you want. I do that all the time.

[sonic salmon]

@polar spade thank you I tried that no real joy…I tried jamaican patois that worked fine for a reggae but struggling with African accent, and black African american vocalists

[proper bluff]

I got 10 credits left so wish me luck

[karmic pasture]

i get a lot of synthwave when I try making a mallwave, and I was never successful at making mallwaves or good, decent vaporwave. I am as clueless as you in very niche styles. Suno still is hopeless at deciphering niches right away. gotta waste lots of credits

[polar spade]

If I were acting like a pro, I would have searched the Home page and run through members prompts until I found a good combo, then send you as if I was clever enough to know good Jersey Club. But all that takes no credits anyway. Yea, Suno is random and 10 credits -- living on the edge. Do your best, mate.

[karmic pasture]

Suno is great with particular popular genres, but struggles with niches

[proper bluff]

Wow… the song dosent work.

[polar spade]

Are there any old 50s American artists who were into an afrobeat vibe that might influence a accent. Even Miriam Makeba sang Xosa, Suno might not know the difference and give you Nigerian instead.

I'm just shooting in the dark. I wish you luck in finding the right dialect with the afrobeat groove . Sounds like a hyped vibe and good mix.

[proper bluff]

I mean like

[graceful condor]

V3 is better with niche genres

[proper bluff]

It gave me v3

[karmic pasture]

thats what I use!

[graceful condor]

Are you trying to make mallwave?

[karmic pasture]

yeah, never had a good one, it goes to jazz or synthwave

[graceful condor]

My respect

And condolences

We've tried many times

Not enough data on mallwave

I've tried breakcore many times as well

[proper bluff]

Well. I failed

[polar spade]

I'm still pushing for a credit donation/gift option. I think its a good marketing idea. But it may just feed us addicts for asking for more. I'd become a pusher of credits. Lol

[karmic pasture]

Suno should tell people to remove shadows and neon lights if telling AI to create songs. AI is obsessed with those words.

[polar spade]

We have a list... haha.

[karmic pasture]

lots of music are polluted with shadows and creepers

[old edge]

why does one of my songs from may say its in v4

its from before 3.5 was available to all users

[karmic pasture]

I guess it is free for mobile users, or you paid for credits

[old edge]

i never paid for anything, i have only used the mobile app in this month

my song is from may its before everyone could use 3.5

but its in v4

[karmic pasture]

thats why, mobile gives you v4 for free

[old edge]

i only used mobile in this month

the song is from may

thats over 6 months ago

[karmic pasture]

no idea, some glitch

[old edge]

maybe i have the first v4 song? 😎

[sonic salmon]

Thank you, I will keep trying

[graceful condor]

https://youtu.be/WIZlU0WIkYI?list=PLPgyBkASWM8zQfeMIkKWst-jJEDK7H3ky

[graceful condor]

Could be

[old edge]

yohoo

[graceful condor]

It's an interesting genre, for sure

[karmic pasture]

lol, I am already subscribed to that channel.

[graceful condor]

Best case would be to experiment with vaporwave to get mallwave

Mix and match genres

[karmic pasture]

yes, synthwave or jazz standard will try to butt in though

[graceful condor]

Exclude styles doesn't help?

[karmic pasture]

I will focus on that, for sure. Keep here posted

[graceful condor]

Great, thank you very much

[karmic pasture]

in my attempt at vaporwave the other day SUno made a synthwave at best

https://suno.com/song/599fa810-c26f-4ceb-8656-1c2250fe054d

[graceful condor]

Interesting

I really like the sound

[covert patio]

hi here. I'm trying to create bollywood-style song, but nothing helps from "Disco 80s", "bollywood", "Indian Cinema Music" in prompt . Hindi vocal is good enough, but music in most cases moderd, with a lot of beats.

[unique stone]

I have one like that and have heard of this before. It affects songs made just before v3.5 was available so my guess is that some songs were actually v3.5 which didn't have a proper version number at that point so it defaults to showing whatever the latest one available is (would've been v3 at that time, but is now v4)

Either that or they were testing an early version of v4 🤔

[covert patio]

I'd like to get something like this https://www.youtube.com/watch?v=1jf5kuvScJc , but all my attempts finish with something like that https://suno.com/song/edc8d97d-d649-4309-9f92-dfa27b818337

[daring lava]

Hey Mstr 👋 Thanks for reaching out about this! Took a listen to the track and played around with a few options and found this might be what you are looking for. Hope this helps

Live performance, Late 70s, early 80s,Bollywood disco-pop fusion, lively female vocals, danceable funk bass lines, reverb

https://suno.com/song/1b36e6d1-5c6e-4a44-90d0-45361d466fad

[cunning siren]

Hello all ... finally we did it - SUNO is broken. No matter what kind of style prompt I use I always get like Disney-esque B$ princess songs 😄

[dark spire]

Yeah it's really bad. They need to roll the whole thing back to pre-v4, including v3.5 because I can't produce anything good on it now, either.

I'm going to unsub until they fix things.

[oak flame]

New here. When will the Shimmer Noise be fixed?

[hollow hare]

I know for a fact that nothing in v3.5 has changed, a dev confirmed that to be the case.
Chances are the prompts you are using may not be the best for what you intend to get.

Please show your prompts and what you want to get and we can suggest what may need to be changed

[hollow hare]

It's not like anyone can predetermine that. It is like asking "please let me know when the next 3-digit lottery will have 123 as the result"

That being said, shimmering has been reduced significantly from what it used to be

[dark spire]

Using the same prompts that brought good results in v3.5 before.

[hollow hare]

Aside from RNG, not sure what else I can say. A dev in the Team confirmed that the 3.5 model is the same for months 🤷

[dark spire]

I'm not alone in thinking that v3.5 has degraded as well.

[oak flame]

Is it better to use v3.5 to avoid the Shimmer? Is there a particular instrument causing it that can be avoided?

[dark spire]

The shimmer is in v3.5 too but less obvious. They did improve the shimmer in v4 but it appears other issues were caused in that improvement.

[hollow hare]

All models have a chance of shimmer, though it is more noticeable in 4.0 compared to the other models.

Heck, I listened to someone my gens from 3.0 after it's released and picked up on the shimmers due to the Frequency Illusion.

For 4.0, Usually I notice that shimmer used to happen with cymbals in the past (esp. with Rock and Metal songs). Not sure about it lately

[hollow hare]

As for avoiding 4.0 ... hmm, if you do not care about the vocal clarity and/or dealing with Metal or Rock, sticking with 3.5 may be best until it is announced by the team that they will freeze the 4.0 model.

If you are seeking for an unpopular genre, you will want to generate an instrumental in 3.0 (the best for that) then extend into 3.5 or 4.0 for the better song structure

[covert edge]

are personas broken? I wasted a lot of money trying to make songs from personas and a huge majority of them were ruined by shimmer and persona duping, I tried creating new personas but no matter what I do, the songs just repeat the base song word for word with almost the exact same background music with each song, and the songs that I try to make outside of that end up being a shimmer 80% of the time, I wonder if there are means of giving back to community members that pay for no results

I also keep getting spammed with pop-ups telling me I have no credits but I feel a better use of pop-ups would be to let people know their credits are running out either by the fact it's running low or that their month is almost up

I feel like this isn't designed for community members that support this

and on top of that, people who dont pay get 50 credits a day to try and even help the A.I learn from users but paying users essentially getting nothing in return, some users even argue that you're paying for a service but I'm curious at this point what service I'm paying for

I feel like I'm paying to help the system, which I'm fine with; it's modern Early Access rules, people don't get paid to beta test, we pay and test but I feel like it becomes a punishment when there are clear separations to what matters

[oak flame]

They seem buggy to me. I tried using a persona and got loud pops and other garbage sound effects in the beginning of songs.

[covert edge]

I would love to keep supporting Suno, I top up from time to time but each time I top up it's getting disappointing each time, I feel like if I pay for the monthly sub, I want to use all my data so that I'm not throwing money away but I've brought up this issue before and it seems like the community is fine with it, I personally can't accept this level of disrespect

[covert edge]

Yeah I had a similar issue in the beginning, and manage to 'randomly' bypass the issue but now my personas are just stuck and I've wasted probably around 500 - 1000 credits just trying to get something to work. I go out my way to vote all songs up or down to help the system even recognise what I want and even get some credits returned due to voting but I'm starting to feel like I'm working and getting nothing out of it at all.

it just feels much better as a free user right now, you use 3.5, you get 50 credits a day and there's not much issues

negativity aside, I love Suno; I want it to do great but I believe it starts with the community, just my take

[hollow hare]

I don't use persona much so I cannot offer much advise other than, iirc, changing genre should decrease the chance of repeating happening. This is a known issue with persona (I also recommend searching this channel for persona from other members for helpful tips)

[hollow hare]

Subscribers do get 50/day as well as soon as their subscriber credits are exhausted (since mid-August: #📢┃announcements message)

[flint haven]

🎵 Credit System Overview 🎵

🔄 Free Plan Users:

• Daily 50 credits refill at 00:00 UTC+0 (<t:1713222000:t> standard local time).

💳 Subscribers:

• Credits refresh monthly at the time of subscription (2,500/10,000, depending on the plan).
• Once subscription credits are exhausted, subscribers switch to daily 50 free credits at the same refresh time as free plan users.

📈 Bonus Credits:

• +5 credits for generations under 8 seconds or failed generations.
• Daily credit refresh will roll your balance down to 50 credits if your total exceeds 55 credits.

⏳ Carry-Over Credits (Non-Refreshing):

• Gifted SUNO credits 🪙
• Top-up credits (require an active subscription to use).

🛠 Credit Usage Priority:

• Subscribed/free credits are used first, followed by carry-over credits.

[hollow hare]

I thought the same @charred merlin But English is not my first language so I didn't say anything about it.
When I suggested this /credit command, I had that category listed as "Other credit sources"

[covert patio]

Thank you for sharing, it's closer for what I want to get as a result, but more modern, I think, it's more 90s, than 80s or late 70s. I'm afraid, model wasn't trainer enough with this kind of music, so with my prompts it can't catch right sound But your prompt gave me new vector for experiments 👍

[finite geyser]

Friends, Suno transforms the audio of a song into the same rhythm that I sing.
I want to download the audio of the song, but with the rhythm and my voice.
And Suno only plays the music in the background.

Is there a way to do this?

[daring lava]

When I get home I’ll throw a couple more tries to see what I can get. For me, experimenting is half the fun in this uncharted territory. Will share my findings if they look to be of any help 👍

And very welcome!

[daring lava]

Hey Ricardo Lima 👋 thanks for reaching out! Could you explain what you’re looking to do a little differently? Just want to make sure I answer the question properly. Thanks

[jade mist]

Why are all my uploads uploading forever? Anyone else experience this or found a solution?

[daring lava]

Hey Marcogarattoni 👋 Thanks for flagging this! Just did a quick testing my end and everything seems to be working fine. Could have something to do with the audio file or maybe connection issues. If you’d like to share the audio file (DM if you prefer privacy), I can give it a shot. Additionally, maybe trying on another browser to test if that could help

[wintry fox]

i cant generate

suno is down?

[hot trout]

Any idea why, having just renewed my Premier membership, I cannot use ANY of my usual features? Not even the basics, like crop song?

[daring lava]

Hi Lars 👋 Thanks for reaching out. Just tried on my end and everything seems to work. Would you mind trying with another browser?

[daring lava]

Hi Mixtrix 👋 That’s very unusual. Could you try logging out and logging back in to see if that helps?

[hot trout]

It did... sorta. Suno was still hiccuping in very weird ways, but it seems to have sorted itself out now after two more tries.

[daring lava]

Glad to hear it’s worked out! Thanks for reaching out 🙂

[wintry fox]

worked!

[daring lava]

Awesome!

[finite geyser]

I have a song that I wrote for my daughter with my voice and musical style.
I want to download the audio of this song on SUNO.
And after that, the app would understand the lyrics and the rhythm made by me and make the song.
Does this exist?

[woeful yoke]

Not entirely sure what your issue with downloading is but I think you're actually asking something else.
You can upload a clip with your song (preferably both a verse and a chorus part), after which you can extend your upload or use it to make a cover song. For both you can then write additional lyrics and it will generate the whole song with your melody and voice.

[verbal crown]

All good

[woeful yoke]

Contact the customer support: [e-mail]

[daring lava]

First and foremost, that is beautiful! Way to go being an awesome parent ❤️ If you upload the song you made it will understand structural elements (such as the sound of the voice, music, things like that) and that will be demonstrated in the extensions built off of that upload. You would need to include the lyrics that you’d like Suno to generate alongside that extension and try to use meta tags that match the music you supplied (eg if you provided a rock song, include “rock” in style of music”) Please feel free to reach out with a link of the upload if you’d like help with the next part (either here or in DM’s if you’d prefer privacy)

[velvet hornet]

Anyone knows why I get a pop-up to verify im human, when im generating song right now? also the quality get terrible.. :S

[torn spoke]

I'm experimenting with chord progression

[cunning siren]

I guess this is one step to prevent bots from mass-producing ... sad they did not find a technical way to secure their own APIs

[torn spoke]

like

[Chorus]
{F - G - Am - Em | F - G - C - Am}

It surprisingly works for me

it's giving me the same melody

got a song now where it sang them, first in 15 generations

so it's not perfect

but it's giving me a lot of consistency

[velvet hornet]

sadly it affects the quality.. it gets real bad.

[cunning siren]

Well, cannot imagine how this check would affect the creation. Maybe you suffer from ear-fatigue 😄

[velvet hornet]

Haha.. sadly no. ^^

[finite geyser]

I sent an email to support.

I downloaded the audio of the song I made for my daughter, the SUNO app created the song's melody, but it cut the audio and played the verses randomly, not following the lyrics correctly.

I would like to keep the music made by the app, but without changing the sequence of the lyrics or cutting them. to fit the song.

[torn spoke]

You sure you downloaded the audio instead of remastering it?

try to download the song using the website

[unique stone]

Have you verified the chords are the ones you have prompted for?
Or does the prompt just produce the similar output to the previous attempt? (i.e. similar pattern = similar results)

[torn spoke]

I am extending from a previous instrumental intro, so could just be that, but test it too c:

[safe tartan]

Hi, can you exclude instruments or use negative prompts? For example, if I want a bass line in a certain style, key and tempo (but only the bass) - can we do that?

[sharp sundial]

due to Illegal downloads from the music industry without my permission my profile has been changed
https://suno.com/@jay_j_brike

[past glen]

Any idea why I keep getting male vocals even though I specify female?

[sharp sundial]

are u using [Female Singer]

[past glen]

the frist prompt is for female vocals

it's never not worked before

[sharp sundial]

the 2nd use [Male Singer]

or [Verse 1 - MALE]

[Verse 1 - FeMALE]

same for the chorus

[past glen]

I just want female vocals, not a mix

[sharp sundial]

ok

theys worked for me

this also worked for me [Verse 1 - Low, rhythmic delivery- FEMALE]

all female male all in caps worked for me

see if it works for you

[naive quail]

In Custom Mode, there's an "exclude" toggle that allows you to add instruments/styles to exclude.

[sharp sundial]

what prompt are you using can you show me

[naive quail]

There are a number of things that can affect this, making it more difficult for vocal styes and tags to override - other styles and the lyrics themselves, for example. Can you share an actual track link that is failing?

[sharp sundial]

righr

right

[past glen]

I've deleted them all as they weren't following the prompt, I'll try again and see if it keeps happening - the lyrics are non-gender specific so shouldn't be affecting the choice of vocalist

[naive quail]

You could hop into the trash and click on one - the links are still sharable, even from the trash.

"Trash" is basically just a separate folder.

[past glen]

actually it only happening when I use V4

[past glen]

yeah I know, I permanantly delete stuff from there

[daring lava]

Using the “get whole song” function after you generate the initial extension will retrieve and reattach the upload. The problem you’re bumping into could be prompt-related. Would be happy to help with your song if you’d like to share more about it 👍

[hollow ingot]

Hi, I bought a monthly subscription and I have 2500 credits left, but I can't create music in either v4 or v3.5, it says there are no credits and offers to buy a subscription, but there are 2500 credits on the account, did they burn out or something like that? Or will I use them when I buy a new subscription? And will they give me more credits? I would like to spend the hundred I have, but I can't, I don't need more credits, I want to spend what I have

[graceful condor]

Please, message [e-mail] @hollow ingot

Oh, you've sent a message here, great

Saw it in showcase

Log out and log in back to your account first

If it doesn't help

Message the email

That usually depends on the genre

[Studio Quality] in Lyrics helps usually, but it's all random, unfortunately

Shouln't have that much distortion with those genres

Great intro

Holy

It is great

I think it's v4 being v4

It lowers quality for many songs

Adding noise

[leaden lark]

Reminds me I should work on this one https://suno.com/song/6a882a9b-6875-4fbc-a92e-ef89f100c688

[graceful condor]

Great as well

[leaden lark]

That was just me testing out a persona I made and gave me that

[sterile inlet]

why has metal and rock generation quality dropped so much lately? (v4)
but i gotta say, the vocals have become, if not flawless, then much more alive

[hollow ingot]

how long to wait for a response?

[graceful condor]

Around a week

Too many emails

How long have you waited for so far?

[sweet wigeon]

U mean this? https://suno.com/song/5a31c80a-ec2f-4384-801f-11ff4f47404f [Instrument#1: Classical Guitar Admira Málaga 7/8 solid wood cedar, Sapele 7/8 - cedar top solid, back sapele. Nylon strings.]
[Instrument#2: Violin]
{Chorus}
[Melody: Play tremolo with vibrato on the end of pure note.]
[Tempo: Largo/Adagio.]
[Notes: F - G - Am - Em | F - G - C - Am]

[tiny falcon]

I can't tell a diffferent if it's distortion or the style of the music. it sound fine for me. I don't think the listener notice?

[graceful condor]

That's the thing

[tiny falcon]

I am not sure if this is accurate but I ask gtp which part of the prompt make suno ai generate distortions, and here's the result:
The following keywords in the prompt could contribute to the distortion you're hearing:

Brostep: Brostep often involves heavy basslines, aggressive synths, and sharp, distorted sounds. It may inherently generate more distortion.

Psychedelic: This can introduce swirling, layered effects and modulations that might sound distorted, especially if not balanced well.

Wonky: This term implies an off-kilter, irregular rhythm or sound design, which can sometimes result in distortion if the AI exaggerates these qualities.

Heavy: Suggesting heaviness often leads to amplified bass or compressed sounds, which may cause distortion.

Sampling: If interpreted literally, it might add a chopped or granular quality to the sound, potentially introducing distortion.

Suggested Adjustments
Replace brostep with chillstep or melodic bass for a smoother sound.
Use psychedelic sparingly or clarify it with terms like dreamy or ethereal to reduce over-processing.
Avoid or refine heavy by specifying rich bass or deep bass for clarity without distortion.
Swap wonky with offbeat rhythmic elements for controlled irregularities.

[graceful condor]

Sound correct

[tiny falcon]

This is what I asked in gpt "I hear a lot of distortion from this prompt in suno ai. which part of the prompt might generate distortion: brostep, futuristic, melodic, melancholic, atmospheric, modern, instrumental,psychedelic, halftime, wonky, heavy, sampling, party, rhythmic".

[woeful lake]

It's been impossible to make music since version 4. The remaster never makes the complete song, and the replace doesn't fit the song at the correct time. Version 3.5 is also different and never finishes the songs. I hope they return at least to version 3.5, because I want to stop subscribing at this point.

[eternal topaz]

Ok so this is a new issue Ive never had before when it comes to replacing sections

replacing 10 seconds

not only does it redo the song entirely but extends it by 2 mins

Update: Found a somewhat of a fix. Had to select far more than I wanted to and it worked then. so I am assuming the replace thing doesnt work with anything under 15 seconds or so 🤷 so theres that

[soft bronze]

How? I tried editing the time stamp both beginning and end but it always gives me “it got stuck” error

[eternal topaz]

are you editing multiple sections or the entire thing?? cause what you have highlighted is the whole thing from what I see

[soft bronze]

Entire thing

[eternal topaz]

the AI probably doesnt like editing the entire song just like it didnt enjoy me editing only 10 seconds XD did you just like the musical score but not the lyrics?

[soft bronze]

I did it with lyrics but wanted to edit some of the words in the lyrics

[eternal topaz]

your best bet is to choose the sections of the lyrics you wish to edit and do it one at a time. not ideal but it should work. this also might work by selection right when the lyrics start and right when they end instead of the entire song. that way it understands youve selected something

[soft bronze]

I’m replacing a word in the chorus but when I edit the first part, it completely changes the melody of the 1st lyrics and not the rest and I’m afraid every time I replace the chorus, it will have a different melody

[eternal topaz]

yeah sometimes it will change the melody. also make sure you have the box "Make same length as selection checked off as it does help maintain the melody a bit better when you render

[soft bronze]

Wait where can I find this checkbox? I don’t see it in the replace section

Oh I have to scroll down. It’s checked

[eternal topaz]

cools

its sort of a crapshoot if the AI renders out the exact same melody but I do hope you are able to pull it off 🙂

[foggy sonnet]

I didn’t see a suggestions channel for the overall platform itself is there a place I can make one that’s not tied to iOS or android or whatever?

[soft bronze]

I did try to do a cover and updated the lyrics and it COMPLETELY CHANGED the lyrics LMAO

[eternal topaz]

lol

[soft bronze]

IT’s only that one result but the other one captured the song but not the updated lyrics.

But I liked it more than the one I’m planning to release

[rigid gorge]

I dont see a sign in button for the Pro account, does anyone have any suggestions, I've been on Youtube trying to figure this out!

[untold briar]

If Replace seems extra difficult, double check all the lyrics of the song and make sure they match what you hear. For the whole song, start to finish. Even if they do match, try adding a space to the lyrics and saving them, so you get the "aligning lyrics" message again. Just in case something was buggy and not fully up to date.

[fierce phoenix]

v4 sounds like an FL Studio track with EQUO slapped on top, harsh

[plain surge]

i m trying set a song as public but is not working tried many times already
cant vote songs
anything to solve this ?

it happened today like 3 hours ago

public button not working it say is public but ppl cant see the song and my vote dont work cant get alerts they r bugged too

[hollow hare]

I assume it is an uploaded audio? As the footnote says when uploading, if vocal-like sounds are detected, then a few functions are disabled (like disabling the derived song from going public).

This is to help prevent issues by users doing this like making songs using copyrighted material or third party vocals (who may not have consented)

[plain surge]

nope it happen with my latest song

tried make a new one but still bugged

both of them

the only way to set as public was playlist

[hollow hare]

Can you link to the song? You can DM it to me

[plain surge]

cuz the button dont work

https://suno.com/song/0da4e065-0b0a-4493-8a04-721d7a715aa4 https://suno.com/song/08ac1d09-9e5f-44a5-ac54-51cb96ae9895

it happen with both

i set as public but dont work

the notification its bugged too

[hollow hare]

I tried on my side with Extending your song and it worked and had no issues going public. This is so strange

Try logging out and back in, see if that helps

[plain surge]

it happen on my main account

the secondary its working

[hollow hare]

Hmm, I think I might know what is causing it. But you will need to log out and back in if it is what I suspect is the cause

[plain surge]

did already

not working yet

[hollow hare]

Hmm, a head scratcher then, ugh. Sorry I cannot be of much help, I will see if some other people may have an idea

[hollow hare]

Make sure to report it in #🪲┃bugs-web if you did not already so someone on the Team can also investigate

[plain surge]

mmmm

okey

[opaque hamlet]

Does anyone have experience with extending music and avoiding degredation of instrumentation, vocals?

I've tried 6 times now on one particular track and I'm not getting anywhere.

I've been working on a new series thats going to be 12-segments long. Here are the first two: https://suno.com/song/a1b27082-cfae-4791-8470-5cf74db1be77 and https://suno.com/song/ba2ccd54-a255-4100-9378-020914ea151a

Here are parts 5 and 6, where it is most apparent during the continuing segment of part 6. https://suno.com/song/db7fa62f-45df-4933-84cd-7fcf8473bc8c and https://suno.com/song/83008ba6-58e6-4ce3-99bc-453fdb2d7b01

[small dirge]

I don't know where to ask this. But my payment was declined because I had forgotten to transfer funds to that account. Now, I transferred the funds but it won't let me resubmit payment. It wants me to add a new payment type. The billing links do not cover this issue so any idea how I can get in touch with Suno to fix this?

[opaque hamlet]

They provide a billing email address for corner cases.

[small dirge]

Okay, I'll look again for it. Thanks.

[opaque hamlet]

[e-mail]

[eager condor]

Wow. So a few days ago, I got some help with getting classical music to work. I mentioned uploading Hall of the Mountain King from archive.org, and after spending 1000 credits on countless iterations, genre changes, lyric rewrites, etc.... I ended up on this. Wanted to say thanks to everyone who helped

https://suno.com/song/632797f2-799a-4642-8340-efcaa0257c17

[opaque hamlet]

The Generative Model seems to go off of the tracks when leveraging existing audio as a source (for Covers, or Extends) - and it seems to me we should be able to request a "return" of the track and have our credits refunded. I realize its not ideal from a compute perspective, but it isn't fair to place it entirely on the customer.

[royal palm]

I am creating a web app that I want to stream the music I create using the URL of the song and storing the url to be called from airtable. Is this possible?

[opaque hamlet]

I'm having to generate on average 4-5 attempts for the second part of a song, sometimes more, whereas original generation can nail it within 1 or 2 attempts.

Does anyone know if Product Management joins these forms?

[eager condor]

I have to agree with this: I'd estimate between 35-50% of my gens came out as just remasters of the original uploaded audio. I reported the bug and deleted it, and it would have been GREAT to get those credits back

[opaque hamlet]

It seems ike they do have some auto-credit capabilities when the gen completely fails, so the feature capability is there, they just need to extend it to user request. I don't believe they have any evaluation post-production to determine if there are quality issues, and I believe this should be done.

For example, OpenAI is constantly assessing the output quality of their LLM generations, and in fact will block the output if it fails statistically to be within their AUPs.

[unique stone]

If there were a button for a user to press to have credits refunded, you can guarantee that would be heavily abused.

[opaque hamlet]

Furry: Not if it requires that you fill in a form, explain why, etc - if you're recursively doing this they can algorithmically block you with a cooldown timer, and ban you if you do it repeatedly. RIght now, its us the users that are getting the short-end.

Once I get to 40 hours of use time, I think I'm going to reach out to their CPO with feedback. I haven't had this much fun in a while, but it def has a long ways to go.

Looks like SUNO has a new CPO who came over from Snap. I hope he takes them to new heights.

They've been openly stating that they need help 😉

Post-Production analysis should solve this problem, they really should be able to tell mechanistically if there is a problem - it will cost them more compute (say 2x) but until their foundational models (like v4) get to GA quality it probably makes sense to do it - or at least for a sampling percentage.

I also find it annoying that we can only extend by up to two minutes. I'm starting to wonder if thats because they're unstable that far out.

I've noticed it from time to time even in One-Shot generation attempts towards the end of the track I get really strange artifacts that sound like compounding noise.

Do you guys know the etiquette here? Is it OK to Tag SUNO employees on the chat who are online?

[charred merlin]

We don't @ the devs 🙏 The only staff member regularly active in chat is the head of support, Mike B

[opaque hamlet]

K, Its also holiday season...

I wish Suno would release their older models into O/S or continue Bark. I realize only Meta is doing this right now, but it would be nice.

For those of us who have A6000's, or higher series GPUs it would be nice to kick around locally.

[west yew]

Here's a question for you all...for those of us who have experience writing real-world music, have you ever tried Band in a Box? It's very expensive, but it also understands chords and uses real musicians. I'm wondering about the possibility of taking some of my Suno creations and trying to re-create them via Band in a Box, but I'm thinking it's not going to give me a whole lot of similarity, and I'm not sure it's worth the expense.

[opaque hamlet]

@west yew Its 99 bucks on Sweetwater?

Why not start with CakeWalk, its free?

I'm just realizing something.... If Suno goes the Agentic Approach, they could write separate audio tracks for each instrument in a time-series thats quantized - they could present the tracks in an advanced DAW edit mode for Pro users. I'd actually pay for that.

Right now, it seems like most people are using one-shot generation, and thats it, but if they're really going to get serious they need to start thinking about that.

[west yew]

The basic Band in a Box is $99, correct. But that gets you a very limited set of musical styles. I'm working in and around various kinds of prog, and the version of the program that includes all the RealTracks is $469.

Cakewalk does not actually providing backing tracks. Cakewalk is a pretty standard DAW. I've been using Ableton Live for over 10 years now.

[opaque hamlet]

Then they could build plugin's for existing DAWs as well so real musicians could work in their studio setup.

[west yew]

Noise-free separate audio tracks - I would pay $30 a month for that.

[opaque hamlet]

I thin they could work to reduce the noise by focusing their generative models on one segment of the music at a time.

[west yew]

And thus something like Band in a Box wouldn't be necessary. But right now, there's way too much noise.

[opaque hamlet]

It gets really complex when the models are having to handle multiple tracks, plus voice, etc.

It can be done, but it doesn't mean it should.

[west yew]

Absolutely. Part of the issue going on is fast generation. I would wait 5 minutes for a song to generate properly. Most people want their song to start in under 30 seconds.

Computing power plays a part in this.

[opaque hamlet]

Their voice and instrumental audio separation is really rudementary and does not work well. There is no reason for it if its generated in parts.

My guess is they need to migrate their models over to different hardware, and not on GPUs.

[west yew]

I can get better stems from Moises.ai, which aren't great, but better than the ridiculous stems currently provided.

Ultimately, the tools aren't in a place yet to truly pull apart tracks well. Vocals can be done very well -- just look at the last Beatles song -- but instrumentation is much harder.

[opaque hamlet]

Lets see how long Groq lasts, but they've got hardware that does generative AI at very fast speeds... I suspect as long as NVIDIA allows them to be around. NVIDIA just released some super cheap generative hardware for robotics this week. But if you look on Groq, they can run OS models very quickly on their ASIC hardware.

Its interseting because I'd think that human voice models are actually as complex, or perhaps more complex than intruments.

[west yew]

Yes and no. For all the complexities of the human voices, it's easier to pinpoint them in the wave spectrum.

I mean, when you arrange a band, you're constantly trying to make sure that each part isn't stepping on the sonic area of another instrument. But voice is very different from that.

[opaque hamlet]

I don't know enough about the music theory to really speculate, but I suspect that Voice really has a broad range, but that they're typically focused and few in number relative to the instruments being played - and that they also very different in audio dynamics. I'm not a musician

I'm just in awe of them 😉

I've been hyper-focused on trying to get the generative model to do things I did not originally realize it had capabilties to do.

[void salmon]

anyone know any details on API relesed soon or ever?

[opaque hamlet]

The API exists, but I don't believe its exposed or documented.

@west yew See this track I generated. It goes way beyond how I was using SUNO originally. https://suno.com/song/a1b27082-cfae-4791-8470-5cf74db1be77

I have this sinking feeling that even Suno doesn't know how it behaves.

[void salmon]

so you don't know if they developed an APi

[opaque hamlet]

Its not documented anywhere, but their Mobile App and Web App both use it.

[hollow hare]

Suno does not have an API released for the public. Further inquiries should be directed towards [e-mail]

I shouldn't need to say it but the "unofficial" APIs you see are against Suno TOS and you will face consequences for using it as people have been finding out

They do have some sort of API setup for their partnerships, which AFAIK, is currently just Microsoft

[opaque hamlet]

I was looking at the java code this week and saw it pretty clearly.

You mean for the CoPilot Plugin?

Funny you should mention that, I haven't been able to find the plugin since Microsoft nerfed coPilot.

[west yew]

I don't think it can know. I have had great success with significant prompts, and others where it ignores them. The biggest thing seems to be if Suno understands the prompts within the framework of the genres it has chosen to work with.

[opaque hamlet]

Jason, but I mean -- beyond the genres -- it has intelligence that is much much deeper.

[west yew]

I mean, I have a 19-song sci-fi prog rock concept album that's pretty great if it weren't for the noise

[opaque hamlet]

I've been using the lyrics to actually steer the generation.

[west yew]

There are prompts that -can- tap into the inner workings of Suno, but Suno also has the choice to ignore them.

No, I read your song, I get what you're doing.

[opaque hamlet]

I've been much more successful than in the genre prompt window.

[west yew]

I've done the same.

Some things simply do not work. Chord structure does not work. BPM does not work. But moods absolutely do, and instruments within the range of Suno's understanding of the genre can absolutely work.

[opaque hamlet]

My guess is they need to provide some LORAs for accomplishing some of the tuning we're looking to do.

[west yew]

It took me a few times to get a song with harmonica, but when it did, it was perfect.

[opaque hamlet]

I noticed its entirely hit-and-miss

I've gotten BPM to work in some cases, but not all.

I had one case where it went off the rails, and starting going up in BPM in the middle of the song and then crashed out

[west yew]

I have told folks that AI is like an extremely intelligent child with ADD. It may do precisely what you tell it to do, but it also may ignore you completely, or only do part of what you want it to do. And yet, sometimes in its creativity, it does far better than what you initially imagined.

[opaque hamlet]

Thats true in the early models, but as they develop and improve the models, the controls actually can get quite good.

[west yew]

Reading up on LORAs, that makes sense. It would be helpful to know what Suno does and doesn't understand, and how we can help it understand things that are not currently in its training.

The question has less to do with the controls, at that point, and more telling users what the controls really are.

[opaque hamlet]

If you look at Open AI, and now Gemeni 2.0 that actually exposes their thinking you can see that a big part of the challenge has been getting the AI to understand exactly what you want from it, but once you do, although the Specifics of each generation is not predictable, it does act in predictable ways within boundaries for specific requests.

Also, I would imagine if we could be exposd to more of the meta data, like the seed, etc we could have more flexibility. I suppose that is what the "personality" is.

My impression is that all of this is going to be flushed out. They have foundational knowledge in AI models, now they need to bring it out into the product

What is not clear to me is where they're going with the product. I listened to their founder discuss SUNO, and my take on it is they're mostly developing for the Spotify/Listening community not nec. for Pro users or musicians.

I hope this is wrong, and that they will be focused on the larger community.

[west yew]

One of the toughest things about the creation situation is that we can't reference actual musicians. I mean, I get that, because a lot of people just want to make Dark Side of the Moon 2. But it would be useful.

[west yew]

I think there is such antagonism from artists of any sort regarding AI that it's harder to chase that market.

[opaque hamlet]

Tough: yes. Its not legally flushed out

[west yew]

I mean, I think it's great, and I've been writing songs for 32 years. Helps my thought process. Hoping to use a reworked version of one of my songs in church in the next couple of weeks. But many don't like it. I kinda understand and kinda don't.

But no one wants to admit that their hook or chorus was written by AI

[opaque hamlet]

None of this has been resolved in court. I'd say as long a they stay clear of reproducing work in prompts, they'll be OK, but the moment you can specifiy styles they'll start to be challenged more anda more in court. I think there is already litigation going on now.

[west yew]

Absolutely. Suno and Udio are both in court over this.

[opaque hamlet]

But if you own the material, and its yours, the model should be able to pick apart the tracks, and analyze each section and reproduce whatever you like.

They're already running the tracks through audio finger-printing and not allowing users to share commercial audio generations.

[west yew]

Thing is, with a program like Band in a Box, that's what you can do -- choose your chord pattern, choose your genre, then take what you like from it, and mix styles to your heart's delight. The onlly issue is, you have to have a strong musical base to do so -- meaning that you need to understand the basics of music theory.

[torpid thicket]

Is that why SUNO V4 sucks right now?

Seems like they don't want to fix the problems that everyone is mad about

[west yew]

I think V4 was released too early in order to capitalize on Black Friday.

[opaque hamlet]

I don't think V4 was intentionally nerfed.

[west yew]

They have definitely worked on the major sonic problems that rendered most everything unusuable at first.

[torpid thicket]

I think they released it to make us Alpha testers

[prime thistle]

By the way does anyone alse have problem downloadig hes music or is just my app broken?

[west yew]

My songs were awful in V4 at first. They are still not as good overall as v3.5. But they are getting considerably better.

I'm primarily doing variations on prog and stoner rock, and V4 has serious problems with those. Other genres, not so much.

[opaque hamlet]

There are always going to be some regressions in new models, its just the way it is. But they need to update their models will more rapid frequency -- the thing is, its really going to drive the user community crazy if the behavior keeps changing.

[torpid thicket]

V4 still degrades horribly with each extend

[west yew]

But you can hear the difference between V4 on launch day and V4 today.

[opaque hamlet]

Yeah, I've had really poor experience with the extends.

[west yew]

V4 does degrade on extends...sometimes.

[torpid thicket]

All the time

[west yew]

I've had a couple of successes...but I am still in agreement with you. V4 still needs a lot of work.

[opaque hamlet]

I think there is an issue there and the reason why they only go out to 2 minutes in length.

[prime thistle]

Help please 😭

[torpid thicket]

It's probably your Phone's settings

[west yew]

No problems here with downloading. What are you using?

[opaque hamlet]

You need to give a little time for the content to settle, and then you can download.

[prime thistle]

I dont eaven get the option on app to download it

[torpid thicket]

Don't use the app - Go to the site on your phone

[opaque hamlet]

Also, if you don't have credits / subscription you may not be able to

[prime thistle]

Ok, let me trie

[opaque hamlet]

I noticed that when My credits dropped, I was no longer able to grab WAV files, only MP3s.

[torpid thicket]

Yea, only WAV files can be downloaded with a Sub

But, yea - 3.5 is still the superior Version; V4 has good voices, but you can't get the sounds that you get from V3.5

[prime thistle]

Ty so mutch, it worked on website

[torpid thicket]

Great~

[opaque hamlet]

SUNO needs to go in tighter with early testers and get more feedback before they release. There are areas that V4 is better, and many areas that it feels like a regression.

[west yew]

I made a prog stoner rock song on v3.5 yesterday, then covered it but changed it to an acoustic version with harmonies, djembe and cajon. It was brilliant. But I couldn't have created the original version on v4.

[torpid thicket]

If they just take the Voices from V4, and combine it with the music of 3.5 - That'd make a better V4 lol

[west yew]

Well, at first V4 voices sounded like bored cover bands. They are getting better. But V4 doesn't understand exactly where vocals should be placed in the mix for rock songs. We need to be able to understand the lyrics a little, but they shouldn't be so clear as to be annoying.

[west yew]

I think Suno is plowing full speed ahead to get as many subs as possible to deal with lawsuits and other expenses. I hope they can keep the whole house of cards up, quite frankly, because it's a lot of fun, but I'm not certain.

[opaque hamlet]

Yeah, TBD. Not clear.

They have a new product owner, and he's only been there a few months.

[torpid thicket]

https://suno.com/song/dcb3a4c7-9b62-47d7-96d4-6d0573f93718

Made it sound like the Shinedown voice xD

[opaque hamlet]

After listening to SUNO releases for a number of hours it feels like the AI voices continue generating in my head -- There is a strong distinct Suno signature to the sound, and its not pleasant. I never have that experience with human music.

Then there are cases where the clarity is really quite good, but overall its mediocre

[torpid thicket]

Yea... When I listen to real bands now, I'm like.. "Was this made with AI too?" xD

[opaque hamlet]

AI is everywhere.

Real bands do use it, but in their DAWs through plugins, etc.

My feeling is that more and more you've got technologists with an ear for music vs classically trained muscians, etc.

[west yew]

What I have told friends is, Suno is a curating experience. The best thing you can do is curate the very best tracks you generate and not settle for anything mediocre. I've heard too many people putting out 40 songs in two weeks, most of which sound exactly the same. Excellent curation means it isn't all the same and doesn't have that problem...at least, not as much.

[opaque hamlet]

Personally I can't listen to music after the 90s, and I think that has something to do with it.

[west yew]

It's not so much AI as it is plug-ins that help enhance the sound, which we used to do with analog equipment back before the late 90s.

[torpid thicket]

90's and the early 2000's had the best Rock/Metal songs

I still blast Disturb when going to work

[opaque hamlet]

The whole idea of a genre is coming to an end

[west yew]

Is it genres that are going away, or people's unwillingness to listen to genres outside their own preferences that are going away? Good discussion on Rick Beato this week about that.

[opaque hamlet]

Yeah, I heard what he had to say about it, and I tend to agree with him.

[west yew]

This is true - you don't have to be a great musician with great equipment to make great music anymore. I use what's known as an IR on my acoustic guitar, and it replicates the sound of a much more expensive guitar properly miked. It's stunning. My $500 guitar sounds like something usable in a recording.

I do have some classical training in music theory, but I'm mostly self-taught on guitar. And yet with the helps that are out there, I can sound halfway decent.

[opaque hamlet]

I'm really curious how large the training data is

It shows up when you've got it doing multi-lingual

SOmetimes it latches onto a single accent and uses it throughout all of the pronounciations but on the other hand it has the reading down about 80%

I did one song with over 100 languages just to see how far I could push the envelope

[eager musk]

Hello Everyone, I keep putting the title of the song in the lyrics sections so the AI thinks is part of the lyrics and sings it. I have deleted it but the song still have has. How Can I remove it. It's driving me crazy.

[west yew]

Agreed. Because what I've found is that bands that have a very specific sound that not a whole lot of other bands have is very difficult to reproduce. You can stumble across certain bits, but it's really a question if Suno knows the sound I'm looking for at all.

[opaque hamlet]

Here is an example: https://suno.com/song/cc01f110-de39-4771-9447-1ff41c7e1878

I believe that they're not tagging the artist names

(in their data set)

So they probably keep auto-generated genres, but delete or anonymize the artist names

[west yew]

That's possible, but it's also true that certain bands have very few "sound-alikes" that could get in the mix. Very few guitarists sound like David Gilmour of Pink Floyd. Very few bands sound like Tool. These are huge bands, but unique. You can enter in prompts that you can find on rateyourmusic, but it doesn't mean you're going to get anywhere. And quite frankly, Suno risks more lawsuits if you can create a unique band's sound.

[west yew]

Have you tried replacement, or covering the song?

[opaque hamlet]

I remember meeting Tool in berkeley 25 years ago

[west yew]

I am surprised that Suno picks up languages so well. I have a character in my prog rock epic with a Hebrew name, and even though I put it in English characters, it pronounced it correctly.

[opaque hamlet]

It pronounces Hebrew quite well. And Arabic amazingly well.

https://suno.com/song/4a3b464e-56f3-4ff9-9d99-8a0834412a5e

[west yew]

I like some of Tool's sound. I am definitely not good with the occultic thing the drummer has going on...but that's another story entirely. Definitely some of the most unique prog rock out there.

[opaque hamlet]

It does make some mistakes in Hebrew tho, I've been noticing that even if the hebrew is vowelized it can make errors

[west yew]

Meanwhile, it would still be helpful to understand the underpinnings of Suno so that particular sounds could be more easliy created.

[opaque hamlet]

I was going to spend some time to try and jail break it, but I haven't had the time

If theyr'e smart they'd just not include it in the training data set, but where I expect there is an issue is not in the initial prompts, but in the main body

[west yew]

Well, without vowels Hebrew is a guessing game for most of us...well, those of us who've studied some Hebrew. 🙂

The question also becomes how often the training set is updated

[opaque hamlet]

I"m not sure. It costs money to retrain, and they're probably doing it at least once a quarter.

Interestingly, they're not only training on music. They're using a large corpus of text as well.

[west yew]

For lyrics? That would surprise me, given the ease at which you can spot AI lyrics still.

I mean, not really, but still...it needs to get better. Although I will say the better the prompt for AI text, the better the text turns out.

[opaque hamlet]

https://help.suno.com/en/articles/3599681

Thats where the corpus of text training is used

[karmic pewter]

I have the shimmer sound burned in my brain. When I listen to anything else I expect it and it screws with my head

[opaque hamlet]

Dezy - yes, its a known problem we're all suffering from

[west yew]

I'm just glad the loud plate rattle is going down dramatically

That was just killing me

So...many...artifacts...

Alright...thanks for the conversation, Golanie...have a good night (or day, wherever you are!)

[opaque hamlet]

Shalom Shalom

[regal jay]

Shalom and happy holidays, chatfriends. Can anybody point me to any technical documentation about teh suno?

[junior stone]

I keep getting errors, songs not completing then they disappear tonight.

[opaque hamlet]

Have you checked the Kowledge Base? https://help.suno.com/

[graceful condor]

#1294418290348921007

[regal jay]

ty but i meant stuff about api and logical operators/syntax in prompt engineering etc

[graceful condor]

No Official API

[regal jay]

😦

[opaque hamlet]

Ya, we were discussing that earlier.

Many of us have exp with our own prompting, with hit/miss results.

[regal jay]

pls i am trying to form an anime girl band but i need to automate

(just kidding mostly) thanks for your time, everyone

[opaque hamlet]

I've had some luck steering the music in the main body of the music regions, but it does not work consistently

[zenith depot]

Need the ability for Udio and Suno just to 'remaster' the music file where it doesn't change anything, but brings out the full-fidelity of what the song has or better in 24-bit or more so the clarity should go up?

Also, maybe have a prompt where we can add instruments, swap instruments out, only play certain instruments of a song, etc.?

[fluid sentinel]

Hey everyone. I'm trying to use "replace section" to just replace ONE word in the lyrics (from "as" to "of"). But the new generations refuse to change the word at all. I've tried changing the lyrics for the entire line but it still says "as".

Any tips?

[drifting vale]

yes, I'm having this same issue as well!

[drifting vale]

totally agree on this one, many times, the song generated would just start immediately with the vocals, even though you may have specifically mentioned that you wanted a piano instrumental intro, etc. it just wouldn't adhere to the commands! and sometimes, you'd like to add maybe an interlude or a break in between, it's just not happening!

[twilit wind]

I totally agree. The remaster is not a remaster. It’s more like a re-do. They change too many things that generally make it worse. It’s basically wasted to do. Over and above that I find the remasters remeber to thinner overall as well.

[twilit wind]

Me too. I’ve probably spent 1000 credits in the last 30 days to fix stuff like that. It’s should not publicly available as of right now

[vapid nymph]

Anyone else have v4 default to like the most generic voice possible regardless of style/genre prompt?

[opaque hamlet]

Been having a challenge with this one choosing the right voices... Its hit/miss mostly. https://suno.com/song/722ce2e8-4ee5-497a-8a3b-64a89260fe4e

Sometimes it just nails it. Like in this case: https://suno.com/song/08009da1-011b-45a8-a12c-b9d3c4005d65

[thin marten]

Hey guys, I'd like to have a song created with my own lyrics. A duet though. Now ChatGPT has told me that I can use parameters like [M: Deep Male] or [F: Soft Female] to differentiate between individual paragraphs and verses. Suno also changes the first two verses (1st woman, 2nd man), but after that it's only a female voice. Am I doing something wrong?

[grave wraith]

Is something wrong with that audio upload tool? I am singing into it and the sound recorded is full of artefacts?

[paper quarry]

]

[raw tiger]

Is there anywhere to get a master list of genres?

[molten ledge]

I’ll lure you to my lover’s nest,
And lay your head upon my BREAST.
I’ll softly whisper words of love,
Like an angel from above.

breast was flaged for moderation, had to change it to chest

sounds ridiculous

but this ... no pb that's alright "And with my butcher’s carving knife,
I’ll cut your throat and take your life."

[molten ledge]

it's a poem from A H Shacknofsky, "beware her smile"

[dull bronze]

emmm

[naive quail]

ChatGPT will just make stuff up when it doesn't have an answer and it will confidently present that made up information as fact. People put waaaaay too much faith into that thing.

There aren't any hard rules regarding duets (or Suno prompting in general), only things that other users have found to be effective (to varying degrees). You can search the discord for duets and find a ton of discussions on it as well as a bunch of tracks that you can look at for some examples.

My two cents: #📚┃suno-school message (you can scroll up AND down from there for more of the discussion.

Also: https://suno.com/song/547b778f-4662-435a-9eb1-1f8aed34888c

[molten ledge]

What ?

Oh ... these are troll accounts ... 😆

[lost jolt]

what do you guys use to make Suno songs sound more realistic - like some Audacity settings etc.

[molten ledge]

@lost jolt Not really to make it more realistic but rather more audible on smartphone: compressor to narrow the dynamic range

[opaque hamlet]

Does anyone have bots following them on Suno? I have three accounts that constantly like and follow/unfollow me all with similar sounding names. I can’t figure out what the reason is.

[bold pecan]

Hello,

I had more than 1000 credits left before the switch to the new credits and instead of the 2500 new credits being added to my remaining credits, I had these credits deleted!
I paid them, I ask for them to be returned to me.
Who should I write to?

[polar spade]

First, what are you using a browser on Mac/PC or iPad?

[opaque hamlet]

Try emailing [e-mail] and cc [e-mail]

[charred merlin]

https://help.suno.com/en/articles/2524993

https://help.suno.com/en/articles/2550209

I’ve heard of people reaching out to the support email because they weren’t aware of that, and received a one-time lump sum of the credits they lost, however.

[polar spade]

Well, it can probably solved here. Let me ask. When you say, " before switching to new credits" I think you mean --
Before your renewal date, you had 1000 credits left.
Now your account renewed and you only have the new 2500 given at each renewal?

Remaining credits are voided/lost/forfeited at renewal and you only get the refresh credits due of 2500 on a Pro account.... and 10000 on Premium.

[bold pecan]

This is not logical 😡
But thank you for your answer

[bold pecan]

Like I said to @charred merlin It's not logical
Thank you for your answer

[terse tulip]

Any reason why constant errors are occurring when trying to create a song?

[formal depot]

Having an issue myself here.
I am trying to downgrade since I have so many credits I didn't get to use last month and I cannot change my current plan. I tried doing the 10 per month.
Can't use any of my credits or purchase any, nor a new plan.
Can't make songs, can't do anything.

I hit renew as well

welp

[rose wadi]

same here. haven't sent a new email. just followed up in the same thread and still no response

[formal depot]

Guess I'll try and emailing again .-. Had the same.

I did.

[rose wadi]

icl at this point just binge all the credits and never use the platform again.

[formal depot]

like what in tarnation >:c

I came here because I am not getting support emails back lmao

[polar spade]

No worries. As Harbor said, your email may get responded to by giving you courtesy credits .

Many phone companies still don't roll-over their minutes. It's more about a choice of marketing than being logical. 🙂 I understand the frustration.

The alternative is to say Pro accounts get 10 (no credits) unlimited days to use Suno before charged again and Premiums get 25 (no credits) days. 😊 But I bet no one would like that either.

[formal depot]

Are we all in the same boat here?? :0

^ You're not wrong

[polar spade]

We're enablers. We keep widening the boat to keep it afloat.

[formal depot]

I find this to be a newer problem all of a sudden. Which is so weird since I've been using Suno since release. But there is no fix to it and it is slightly frustrating, especially when not recieving any emails in return.

But waiting until january 21st? I just got new credits.. ( 2 days ago) now I cannot even use them or change my plan.

[polar spade]

I didn't follow.

I know if you simply cancel the subscription, you have full use of your level, credits and extra credits until end/renewal date of your current period.

You're saying you canceled and so have NO access to credits or means of creating outputs?

[polar spade]

I've canceled my premium before and let it run to the next month, so I could decide if renew, let it roll to a free account or renew at the pro level. No problems.

[formal depot]

I tried to make music before I canceled since I was already trying when my plan renewed the 21st. Didn't work, then I even realized I was paying 30 per month. Decided to see what was up, it was paid.. So I tried to change my plan for next month for 10.
I tried to make music but it says I have no credits..Which I do and paid for.
Then I tried to cancel and renew with the 10 dollars. Still no change.

But I have a plan so it is weird.

But even before I did anything, this error showed up so I checked.

Nope.

I never been able too actually lol

It's always had this

So I never had been able too xD

Not like I needed it with the 30 dollar plan anyway

Never heard of it, I'll have to see it.

[fluid sentinel]

Win 11 PC, Edge Browser.

[formal depot]

Currently using Opera.

[polar spade]

It's weird. I agree. Something maybe support can only work out then.

Yea. If you "cancel" pro or premium, it simply drops one to Basic. It will even say Basic is your current status upon renewal, but as long as you dont update at that moment, then you can just ride out your pro or premium account with all credits intact.

If you upgrade/downgrade , at that moment , the account date restarts that moment (literally at that time) and old credits disappear (except for extra credits, and your upgrade/downgrade level new credits are tallied onto your account.

I know that doesn't help. But just saying it's good to ride out the account until the day before renewal.

[naive quail]

lol. A quick peek in the channel... all support issues. Suno really needs to rethink how they have this server setup.

[formal depot]

So weird..

[flint haven]

I concur

[formal depot]

I was informed on this. Yes, but it is just giving so mnay errors when paid so that is just weird to me.

Basically

[polar spade]

It seems you did all this at the very day-ish that it was renewing. So something glitched, for sure.

[formal depot]

Maybe it is an error on the site itself currently?

Also I did this 2 days after my renewal.

[polar spade]

🤷‍♂️ It's nice when support just hits a button and it's resolved.

[formal depot]

So idk

[formal depot]

Yeah I understand lol

Well I'll just write an other email and hope for the best then. . .

[polar spade]

Your account still says you have 12850 credits..... 2 days after renewed?

[formal depot]

I had previous credits from the month before.

[polar spade]

Somewhere you had extra credit (i.e. bonus or bought credits)

[formal depot]

Maybe bonus? I never could buy credits. Not too sure than. Unless they gave out free credits and I didn't know. Not too informed on what suno does lately.

[polar spade]

Premium would renew at 10,000 and Pro at 2500. So 12850 included, likely, 2850 extra credits and it renewed and re-set you with 10000, as you said you paid in $30 on the 21st.

So

[formal depot]

Huh. Yeah that is weird then. I am wondering if then it is something else? Do credits save after the new renewal?

Maybe then that would be it.

[polar spade]

So if you canceled ..... then today changed sub levels to Pro, then your new 10000 would drop and you'd get a refresh of 2500 credits instead, for your new Pro level.

What are your credits showing today? "O" or 2500 or 5350?

[formal depot]

I don't think it actually updated or changed when I did the change of plans. One moment let me see.

[naive quail]

It's definitely extra credits outside of any subscription allotment. The billing details screenshot shows Extra Credits Left: 2840

[formal depot]

Extra credits left: 2840

How is that even possible?

[polar spade]

No credits saved. You get a refresh lot of whatever credits due based on your new level. But the "extra credits" would travel with your account no matter the subscription level.

[naive quail]

You may have been carrying them over for a while

[formal depot]

Gotcha. But how the heck did I have extra credits?

Didn't even know that was a thing to have "extra" Yet, I never noticed it if it were there.

[polar spade]

Who knows. They come from things we do and not know it. Enter a contest? Had some short gens, you get 5 credits for each of those.

[unique stone]

Wild guess: If you upgrade from pro to premier you can get the 10000 added to the 2500.

I wonder if downgrading to pro from premier added 2500 to your 10000 and tried to bill you for the new subscription, but maybe that payment didn't go through.

[naive quail]

I will gladly accept them in order to restore your balance...

[formal depot]

No, maybe I shared with a few friends in the past when it became a thing so maybe that lol

[unique stone]

Typically if you have subscription credits but cannot use them, that'a an indication that there was a billing issue.

[formal depot]

That could be it.

It went through 100%, says in my bank. And suno. Idk it is just an error lmao

[polar spade]

Are you in Australia? Your subscription date is in the future there. 🙂

[formal depot]

No I'm in the netherlands :D

[polar spade]

What's that... 7 hours ahead of Suno time. So still. Should be the 22nd that you updated.

So , anyway, you now show only 2840 credits total. As in, NO subscription credits at all? Each day, if empty, you should still get a gift of 50 credits (only) a day - not roll-over either.

[formal depot]

Yes, I figured the time thing.
Well this so confusing but I suppose yes.
It may just be an error on suno's side with certain payment issue's and upgrading. But It's fine, I will see if they email me back to see if it was an error on any part of it. Especially the "extra" credits as I have no idea where they have came from.

Though I am not complaining about it lol

If it is a billing issue, and they reset my subscription, will I still have to wait the next month for things to work?

[polar spade]

Yes. Often it just take a new day to see what changes. Things could have time to update itself. It was Sunday after all yesterday and not a bank business day. Some computers need their day off too.

[formal depot]

That is understandable.

[polar spade]

Suno probably has an agent set up in a maintenance closet who's only job is to verify account renewals, and they are are lagging behind. 😁

Ok. Good luck.

[formal depot]

I completely understand that, They do a lot.
But thanks for the help!

[velvet hornet]

Hi! So.. im having the watermark/audio issue when generating or extending songs. I filled out the type-form for this problem, but does anyone know if these kind of issues take long to fix? Also, if refunds for the wasted credits is possible?

[opaque hamlet]

There doesn’t appear to be any training data for nature scape sounds. Has anyone else experimented with trying them?

[harsh turret]

hi there, dont know if happened the same to you. I have use replace section a moth ago and worked properly. Now is just burning credits. More than 500 credits and Suno do not do what have to do. In fact replaces the words but also errase next part of the song. Any tips?

[naive quail]

They seem to be backlogged with issues ATM. I'd try reaching out to Mike B via DM here on discord. He's the official community support person and may be able to poke someone on the team to look at the issue. If you've already mailed them about it, I wouldn't mail them again - you may end up bumping yourself to the end of the queue.

[naive quail]

There are definitely nature sounds in the training data. v3 seems to tap into them the best. There's some discussion on it around the time of this post: #📚┃suno-school message

[polar spade]

You mean sounds of nature.... rivers, wind, rain, thunder and so on?

[opaque hamlet]

I haven’t tried them yet in the text body but in the genre/style section they seem to be ignored. I’ll play with this a bit now.

[west yew]

Right now, extending is mostly broken on V4. You can get good extensions, but mostly the bass half of the spectrum peters out. No official ETA for a fix, and after scanning here and Reddit I don't see any credits given for it.

Unfortunately, for long songs I'd go with v3.5 and then remaster, but that's a crapshoot too.

Many of my prog rock creations would not exist in the current v4 version.

[west yew]

BTW, found and following on Suno. Soul on a Mission's vocals are unexpectedly brilliant for the genre.

All it needs are record scratches.

[dapper sluice]

how does SUNO "learn" ? I am trying this prompt and wonder if SUNO will keep the volume signal strength down? Prompt: keep music volume within clipping parameters
will SUNO get this eventually or is it a dead end? (not for shimmer just uploaded clips that may have a stronger signal strength that normal)

I am a sound hobbyist not a proffessional so my clips can be "loud"

I am learning but I take it all seriously

I use Mixcraft 10.5 as a DAWQ

DAW

[velvet hornet]

It's not just bad extensions. Every extend or new generated song comes with the watermark, and also affect the quality of the sound. It is a massive difference. Like not even close to the quality of v4 normaly sound like.

[west yew]

Hmm. I was generating over the weekend and I finally wasn't getting the noise I had been getting on generating prog rock/metal/stoner material. But realistically, the genres are all over the place, so your mileage may vary.

I hear your pain. I have one worthwhile song on V4.

And it's a cover of a V3.5 song.

[west yew]

Are you getting clipping on your tracks? I've found everything I've imported into a DAW is well within clipping parameters, usually -6db or even less.

[polar spade]

I mean.... they probably exist..
Its just about how to root out the sounds... and I don't know yet.

https://suno.com/song/446140de-25bf-4ebc-a964-6b45fd3fd495

[velvet hornet]

yeah, I was working on a song and suddenly i get a pop-up to verify im human, when i was "extending". Along with that the watermark issue started. I cant generate new song or finish the one im working on.

[naive quail]

Suno doesn't learn in real-time; it's manually trained by the devs. The Thumbs-up/down are for the devs to review for future tweaks to the engine.

[naive quail]

I think you two are talking about two different issues... The watermark issue being discussed is an intentional change that suno makes to the audio files and it will repeat things like "This song was made on Suno.com!!"

[west yew]

After reading the next post with the pop-up, I concur.

[velvet hornet]

This is the song im working on: https://suno.com/song/9b09957a-1c2c-42df-9130-0a57adc8f41b
This is an example of the quality i now get: https://suno.com/song/98c9c2b8-08e6-499e-b751-73990a587ef7

its downgraded

[opaque hamlet]

Morning — Not following exactly?

[unique stone]

Yeah you have SUNO over the picture which indicates that the service thinks you might be using a fake Suno app/site.

If this started after several popups appeared (captchas), it indicates that an incorrect response to those captchas was given several times.

[velvet hornet]

ok!.. so how do i fix this? xD

[unique stone]

Did you get a link to the form to fill in already?

[velvet hornet]

yep!

[opaque hamlet]

Are you using a consumer VPN, or web access in privacy mode settings that mask your IP (private relay) by any chance?

[unique stone]

Just a case of waiting for that to be picked up, can take a few days unfortunately

[velvet hornet]

Nope :S

[west yew]

I thought you posted a couple of songs late last night...Heaven's Funky Dilemma and Soul on a Mission...

[velvet hornet]

ah ok.. How to I know if they fixed it? Do they contact me, or do i have to spend more credits to see if its fixed?

[opaque hamlet]

Usually there are step-up human authentication steps taken when the IP or device you’re assigned is associated with potential malicious activity. It’s common in consumer VPNs/proxies and when using private relay.

I’m also suspicious of browser plugins that modify the browser sessions. There are some really nefarious ones out there that steal your data or rewrite data for stealing referral revenue.

[velvet hornet]

Yeah i got the "captchas", but im pretty sure was giving the right response..

i was not using a VPN when i was working on the song.

[west yew]

Question...what are you hearing that's different? I'm listening to both halves and I'm not hearing a downgrade in quality, nor shimmer. Is it maybe where you're listening? I'm really quite confused. (I'm listening on studio headphones, BTW.) While I didn't listen 100%, the only place I heard noise was in the ending of the first version. But it also sounds like The Iron Titan is starting mid-song.

BTW, I like the style a LOT.

[velvet hornet]

there is a massive downgrade in the quality of the sound on the "extended" clip.
thanks btw 🙂

[opaque hamlet]

Suno seems to be using Cloudflare it may have been Cloudflares detection mechanism that promoted you for a captcha. It’s interesting because their team wrote a whole article about how captchas are dated and why they shouldn’t be used.

If you go here how does it rate your IP https://www.ipqualityscore.com/free-ip-lookup-proxy-vpn-test:

[naive quail]

With my audio gear, it sounds like the kind of audio degradation that we often hear in the last half of tracks

[velvet hornet]

I dont know, it sound like its using another version.. not the v4. The sound is not as crisp.. more muddy

[west yew]

OK, for a minute the parts were flipped for me

[naive quail]

Is it like that with every extension? Sometimes, you can get a cleaner extension by shifting the start point a few seconds forward or back in time. Extensions can definitely pick up.... and make worse any existing noise/degradation. It may also just be related to the watermarking thing - I could see them intentionally downgrading the audio on those flagged clips.

[west yew]

I hear it now doing a side by side comparison. The parts being created aren't the problem; in fact, The Iron Titan has some Morse Code/reverb cannon going on around 2:09. But The World She Craves sounds sonically different. The Iron Titan sounds like it was made across a much broader soundscape, but The World She Craves is more boxed in, with less ambient fullness to the sound.

It does sound as if The Iron Titan was made in V4 and The World She Craves in...I dunno, V3.75.

I didn't catch it at first because I'm so used to the sound of V3.5.

V3.5 doesn't open up like that in sound unless you throw it in a DAW and do some mastering.

[velvet hornet]

its like that with every extension or new song i create. Been trying quite alot.

[naive quail]

Yeah, I bet it's related to the watermarks. Sucky

[velvet hornet]

yeah, it started when that appeared.. :S

Too bad.. i feel its a banger of song in the making xD

of a*

[west yew]

BTW, followed, even though you don't have much public yet, because you're on to something sonically.

It's got a surprisingly good vocal for metal

[velvet hornet]

Thanks! I will make more of my song public soon. I got maybe 20 or so that im satisfied with

this is my latest "finished" one: https://suno.com/song/ae96299d-3603-427f-8b99-38bdb2723761
Im going to see if i can mix the sound quality more though.. not public yet

[opaque hamlet]

• Suno: Raised $125 million.
• Udio: Raised $10 million in seed funding.
• Sonauto: Raised $500K in a pre-seed round.
• Loudly: No specific funding amount mentioned in available sources.

Suno (@sunomusic) on X

We’re excited to announce that Suno has raised $125 million to build a future where anyone can make music, and where technology amplifies our most precious resource: human creativity. We will use this funding to accelerate product development and grow our world-class team of

Twitter · May 21, 2024, 2:15 PM

[velvet hornet]

Anyway, thanks for the help, got to go. Merry Christmas all of you. 🙂

[lavish brook]

hey can someone help with one question about distrokid?

[verbal crown]

When I use a persona with a new text, it still generates it with the old text. Is that normal?

[west yew]

Rarely with V3.5. Often reported with V4.

[verbal crown]

Ouch, I’m on V3.5… and I’ve generated 20 so far and still the same issue :/

[hollow dew]

anyone know how to make own voice personal ? any trick to do ?

[opaque hamlet]

Surprisingly this small 2 person team at sonauto has created some clean sounding generations. I just stumbled on it now. My feeling is that these audio quality issues will get solved and may have something to do with the size of their dataset

[west yew]

Ouch. Do you have any examples to look at?

[west yew]

Sonauto...link? Info? (I'm feeling confused...)

Nevermind...just found sonauto.

So a few generations on Sonauto suggests to me that their instrumental generation is TIGHT. Very good, no artifacting, and strong instrumentation. Their vocals are struggling...not always on key, but interestingly that's not a huge factor in one of the genres I tested. Not quite figuring out what to do with certain words and pronunciations. But still...the instrumental is radically cleaner than Suno.

[dreamy temple]

suno we are adults why do you censor some words

[soft lake]

https://suno.com/song/1b0d5b49-2648-417b-b20d-f1871fb1dff5

[west yew]

Because from what I've heard of Suno's output in general, most of its users are not adults

[graceful condor]

Suno uses ChatGPT for lyrics moderation

[stable sparrow]

did the content filter on V4 become more strict? I had a song that I want to remaster from 3.5 to 4, but its not working because I now get "Prompt flagged for moderation"

I have songs that I cant rerun on suno now, because they get flagged.

[charred merlin]

https://help.suno.com/en/articles/3198209

If you're unsure why certain content was flagged, feel free to reach out to the team at [e-mail]

[stable sparrow]

but on older version it was ok and working

only with the latest update the moderation seems to be more strict now. Its weird since even chatgpt etc now is able to produce more explicit songs lyrics

song is not public and I think its only because it has some sexual language

few month ago I could generate it on suno

[verbal crown]

How can I do that..?

[gentle quartz]

Yo yo, new here. I'd like to report that the v4 and v3.5 model is way too stiff now, very hard to steer.

[charred merlin]

The moderation seems to have been updated, then. 🙏

The article I linked above has information about why content could be blocked, and what to do if you're not sure why it's being blocked.

[opaque hamlet]

Sorry I’m in flight and had to hack the WiFi to get on 🙂

[gentle quartz]

The model modifies my lyrics pattern and ignores my instructions.

[graceful condor]

If you have v4 feedback, please send it in #1308499132356362383

[gentle quartz]

nor did v3.5 work though.

[opaque hamlet]

V4 feels like a v8 engine with no steering wheel 🙂

[gentle quartz]

ikr, it can only generate music in one pass, extending or modifications will result in garbage.

extending with v3.5/v4 will modify B section melody dramatically.

[graceful condor]

If you have created songs, go to Library and click on the third icon next to thumbs down to copy the URL

[opaque hamlet]

Yes the one shot generations are where I’ve had most luck, sometimes requiring two or three- but in general I’ve had far less luck with continuing segments.

I’ve been notating all the music with bug reports.

[gentle quartz]

Suno is becoming more like an amateur thing now.

[bold pecan]

Perhaps, we will see

[charred merlin]

Not sure what I'm doing different for my extensions to come out cleanly, but it feels like hyperbole to suggest that they result in garbage -- I don't doubt you're running into issues, especially if you're extending from extensions, but I'm Team Extend 💪

[gentle quartz]

Suno models are having more and more limited creativity.

[graceful condor]

I've been mixing deathcore and orchestra

[gentle quartz]

V2 was random but very creative, you can try and debug.

[graceful condor]

That doesn't mix

But they do

Deathcore, Orchestra and Vocaloid

[gentle quartz]

Now songs are mostly the same when you generate with same lyrics.

I'd rather compose songs myself now.

Sorry I've just want to complain a bit.

Yes, costing more. I understand Suno has to generate income, but not in this way.
I want to tune temperature, and more parameters.

[charred merlin]

v4 responds to prompts differently than v3x -- sometimes it can be a learning curve using them differently.

I'm sure the OG users took a bit to adapt their prompting style from chirp to Suno.

"Different" can subjectively mean "worse", but it's a very powerful tool in itself, it's just unfortunately not functionally the exact same as v3x just with improved sound

[naive quail]

v4 is much more stubborn when it comes to voice-acting stuff, what with its constant inclination to break out into song, but even that gets better once you start figuring out how it responds to tags & such... still not as good as 3.5, tho. 😦

[gentle quartz]

GPT style music generation is not so promising, and generating together is bad. Suno should consider generating different tracks separately.

So vocals and BGM can be more clear.

Get some professional singers and record some music, finetune your model.
Suno was trained using prob DEMUCS and the vocals are interfering with BGM.

[opaque hamlet]

Yes I believe that this is true- we’ve been discussing multi track generation here — eventually the model will become smart enough to handle it well but for now I think we’d see a benefit in tracks and segment generation on a syncopated time system

[gentle quartz]

Yea, @opaque hamlet that'd be a hit, but require a full retrain.

Suno would have to gain a new investment probably.

[opaque hamlet]

I don’t think so- i suspect model today has the capability it just needs prompt engineering and a new interface model.

Nah they’ve got bank.

125 million is a lot of money for a start up. alot

[gentle quartz]

Not for an AI model.

[opaque hamlet]

They’ve got a competition with only 500k in seed

[dreamy temple]

it's shit are there any better ioptions out there that don't moderate

[opaque hamlet]

Let’s do some math - 25 million / year in salaries headcount, 100 on compute generation and inference - plenty of room for growth with their current funding

[gentle quartz]

No idea, I'm not a business guy.

[dreamy temple]

@charred merlin But you got our subscription right and cant you set up like a photo uplaod of our passport to prove we are of legal age so you dont moderate usd.

[opaque hamlet]

Google trained palm 540B parameter model for 12 million and even gpt4 cost under 80 million to train

[gentle quartz]

And they got plenty of commercial copyrighted music to train their model.

[charred merlin]

Red names aren't developers, just testers 🙏

#💡┃suggestions-web is the channel the developers set up for general suggestions about the platform, if you wanted to share your suggestion there.

[gentle quartz]

I wonder if using copyrighted music to train a model but limit its ability to repeat the content is legal or not.

Anyway, really hope there could be a developer mode, to adjust temperatures.

[charred merlin]

That's currently playing out in court over the last half year or so -- I stopped following the updates after Suno released a really sound statement on the matter shortly after the case was announced.

[gentle quartz]

It's very difficult to prove unless dataset itself is revealed so...

IMO, the ICL capability is not fully utilized at all.

[gentle quartz]

Any results? Who won?

Cover, singer, replace section, are like ICL but there are so many other potential uses.

[charred merlin]

afaik it's still playing out? I haven't heard any major news one way or the other after the joint statement in response to the RIAA sort of kneecapped them

[gentle quartz]

what's RIAA?

[opaque hamlet]

https://www.courtlistener.com/docket/68878608/umg-recordings-inc-v-suno-inc/

[gentle quartz]

Dang it's long.

In US, is gathering evidence using illegal ways count? Like spying?

In my country, using illegal ways to gather evidence will be considered invalid.

[opaque hamlet]

No. You can not gather evidence illegally and have it admissible.

[gentle quartz]

Then UMG can ask court to force Suno to reveal their training dataset.

If that's possible.

And if what's left are very old cracky recordings, that'd be funny.

But anyhow, Suno wins, even laws progressed and banned training with copyrighted content, Suno will be the only company who has a model trained with those data.

[opaque hamlet]

Everyone knows that the dataset was trained on real data. The question is how far can fair use defense take them.

[gentle quartz]

I'd say: "Humen are also like models trained with those data. "

You study to compose music by listening, and your output is considered yours.

The problem is, can an algorithm be compared to a human?

Fundamentally it's a philosophy problem.

Shiet sorry I made it more complex than it was.

[opaque hamlet]

The case, filed in June 2024 by major record labels including Sony Music Entertainment, UMG Recordings, Inc., and Warner Records, Inc., alleges unauthorized use of copyrighted sound recordings to train Suno’s AI models.

The court has conducted initial proceedings, including the assignment of Chief Judge F. Dennis Saylor, IV, and the issuance of summonses to the defendants. As of October 29, 2024, a status conference was held, during which the court referred discovery issues to a Magistrate Judge and scheduled further status conferences.

The case is currently in the pre-trial phase, with both parties engaged in discovery and pre-trial motions. Specific dates for future hearings or a trial have not been publicly disclosed. Given the complexity of the issues at the intersection of artificial intelligence and copyright law, the proceedings are expected to continue into 2025.

In response to the lawsuit filed by major record labels, Suno, Inc. has acknowledged that its AI models were trained on copyrighted music. The company contends that this practice is protected under the fair use doctrine, asserting that their technology is transformative and enables the creation of entirely new musical works. Suno’s CEO, Mikey Shulman, compared the AI’s learning process to that of a human musician drawing inspiration from existing music, stating, “Learning is not infringing. It never has been, and it is not now.”

The Recording Industry Association of America (RIAA) disputes this defense, arguing that Suno’s large-scale use of copyrighted material does not qualify as fair use. They assert that the company’s actions involve “stealing an artist’s life’s work, extracting its core value, and repackaging it to compete directly with the originals.”

As of December 2024, the case remains in the pre-trial phase, with both parties preparing their arguments for court proceedings. The outcome is anticipated to have significant implications for the application of copyright law to AI technologies, particularly concerning the use of copyrighted materials in training AI models.

[gentle quartz]

If Suno loses, will the platform go down?

Or they pay a huge number and continues to run?

[opaque hamlet]

See: https://suno.com/blog/future-of-music and https://cointelegraph.com/news/suno-ai-training-fair-use-riaa-lawsuit?utm_source=chatgpt.com

[gentle quartz]

Thx.

[opaque hamlet]

It’s all speculative. I think they’ll be permitted to proceed.

[charred merlin]

There used to be a legal expert on this discord, then he got a job doing actual lawyer stuff and couldn't spend time making music and hanging out on discord anymore 🤷‍♂️

[opaque hamlet]

No one actually knows what will happen yet.

Anyone who says they do is just speculating.

[gentle quartz]

So currently, take time and use copyrighted stuffs to do AI stuff.

Before it's banned and become rare.

[tiny talon]

Can we give a shoutout to @charred merlin for the tireless effort of dealing with our constant BS? Not many people have the dedication, or patience that he exhibits here on a daily basis 👏

[gentle quartz]

@charred merlin Thanks buddy! ❤️

[graceful condor]

Agreed

[gentle quartz]

Shiet I got to do work stuff, ttyl ppl. Happy creating.

[charred merlin]

🫵

Thanks tech-crew

Love hanging out with y'all 🙂

[opaque hamlet]

Most or perhaps All foundational AI models today is trained in materials which include copyrights. The alternative is a much smaller dataset or generated synthetic data sets which are far from ideal.

[daring lava]

A BIG S/O to @charred merlin for all the time and energy spent here helping so many! Such a privilege to see him here contributing and putting in so much effort to lift people up and give them understanding of Suno tools and member support! Truly appreciated 🫡 ❤️

[opaque hamlet]

When a budding artist emerges in the real world they’ve all trained on other artists music. The argument of fair use is that it’s the same here with AI.

[gentle quartz]

Again it's a philosophy problem. 😉

[daring lava]

Additionally, running through today's feed, S/O to @naive quail, @polar spade, @cedar thistle and all the others that contribute here to help others learn and grow!

[naive quail]

Now rub Daddy's belly...

[gentle quartz]

Can i give you a big kiss on the cheek?

[daring lava]

How about a cookie. 🍪 😅

[naive quail]

I'll settle for the cookie!

[charred merlin]

It's edible, but it's an an edible.

[gentle quartz]

A quick question to the tech crew, is it possible to join the R&D team?

[naive quail]

https://suno.com/song/145a3df7-523f-4153-86dc-e015283d14a5

[gentle quartz]

Bachelor degree on Music Composition, but self-taught Deep Learning.

[naive quail]

lol

[gentle quartz]

The sound was cut off?

[naive quail]

only in the video...

[gentle quartz]

Yo im serious.

[charred merlin]

https://jobs.ashbyhq.com/suno if you wanted to apply

[gentle quartz]

Yea i mean I don't have a CS degree at all.

[naive quail]

I was laughing at the "Rub daddy's belly"

[charred merlin]

There's a general application form at the bottom 🙏

[gentle quartz]

Thx, I think i gotta be brave.

[gentle quartz]

O gotcha

[opaque hamlet]

It’s primarily a legal and ethical argument with philosophical undertones.

Here are the legal arguments: Suno argues that the RIAA’s lawsuit is legally flawed because training AI models on publicly available copyrighted materials does not constitute infringement. They compare this process to human learning, where individuals absorb styles and patterns to create something original.
• Suno highlights the similarity of its practices to those of major AI developers (e.g., OpenAI, Microsoft) who train their models on open internet data without direct copyright violations.

[gentle quartz]

The latter point is true, but OpenAI still faced many lawsuits.

[opaque hamlet]

Here are the ethical arguments: The company frames its mission as ethically aligned with innovation and democratization of creativity. It emphasizes enabling access to music creation for everyone, arguing that its platform encourages originality rather than duplication.
• Suno positions the RIAA as resistant to progress, portraying its actions as driven by control rather than the broader benefit of the music industry or society.

Here are the philosophical : Creativity is not copying; it’s about synthesizing and building upon existing knowledge or styles to create something new.
• The platform challenges traditional gatekeeping of music production, arguing that art and creativity should be accessible and not restricted by monopolistic control.
• The case raises deeper questions about the intersection of technology, creativity, and ownership: Who “owns” the inspiration derived from a body of work? Does learning inherently constitute theft, or is it a natural, necessary process for innovation?

[gentle quartz]

very true

brain bomb exploded

[opaque hamlet]

Record labels are way too powerful

[charred merlin]

I've often spoken about how great of a guy you are, but I don't often comment on how well you strike the balance between community moderation and community development.

You speak softly, carry a big stick, and present yourself very earnestly-- glad to have you here with us, man 🙂

[gentle quartz]

But they do own the songs.

[tiny talon]

shit's fucked

still dont have that in you database, eh woodhouse?

[opaque hamlet]

Weird I just got a language warning for mentioning the new administration

[naive quail]

did you say "bridge"?

[daring lava]

Aww thanks @charred merlin, very proud of what we all do here! We have a great community and a great server staff that put in a lot of time and effort to make this place enjoyable and safe! Glad to be here, and just as happy/proud to share space with so many that contribute to making it awesome!

[opaque hamlet]

No I said Tr*mp administration .

[gritty badger]

Do you kiss your mother with that mouth 🙉

[charred merlin]

They'd updated the server rules recently -- some candidate names were banned at a certain point during the election cycle, and instead of updating the list as time went on, they just asked people to not talk about politics because it usually lead to some kind of kerfuffle in a music-forum

[naive quail]

We CAN talk about Harber's hairy feet, tho...

[opaque hamlet]

Bots don’t understand context. The new cabinet has technologists who understand AI well and I think this will bode well for AI companies.

[tiny talon]

I thought "Harber" sounded awfully close to "Hobbit"

[gritty badger]

Get a room guys 🤭

[naive quail]

He's a Harbit

[opaque hamlet]

David Sacks, recently appointed as the White House “AI & Crypto Czar” by President-elect, has not publicly commented specifically on the issue of AI fair use. However, his general stance on AI development suggests a preference for minimal regulatory constraints to foster innovation. Sacks is expected to advocate for policies that provide clarity and support for AI and cryptocurrency industries, potentially influencing discussions around AI training practices and the use of copyrighted materials.

[charred merlin]

I immediately know who I'm casting in the role of the sage wizard addicted to Old Toby

[naive quail]

::sigh:: ... bringing a hurdy-gurdy in with an extension, even when the track STARTED with a gurdy, is proving to be a challenge.

v4 just want it to be a violin. v3.5 is like "But, I'm happy with the guitar, you should be too!"

[opaque hamlet]

The reality is that as large as the music industry is, the AI industry simply dwarfs it in size and scale and value

It’s heading to 650 billion this year vs 30

There is no way the us will let it stagnate against the likes of China

When I went to visit china in 2018 I listened as the ceo of huawei committed over a billion a year in the education sector alone to compete with MITs innovation labs. If we turn in the direction of Europe we’ll destroy our innovation against this backdrop and I have a feeling we won’t.

[west yew]

While I do think the ethics of using copyrighted material in training AI is dubious, I also think that it's ironic that the music industry itself continues to work in the interests only of its biggest artists that bring in major money. When it takes close to 50,000 streams of a song to generate $100 in revenue, you recognize that it's already impossible for any minor artist to make money except through live performance.

[opaque hamlet]

Also think the music industry is in dire need of a major disruption

At the end of the day the musicians who the industry claims to represent are the ones getting the shortest end of the stick

[west yew]

The industry is only signing proven artists. Artists have to go it on their own to develop a following before they get signed.

[opaque hamlet]

There needs to be a direct to consumer model and I think they should cut out the middle men who’ve leeched off artists for years.

[west yew]

In the current industry situation, Rush would have never hit. Nor would have Bruce Springsteen or Billy Joel. Or The Cure, or countless other bands and artists who had to make several albums before becoming an "overnight success."

[opaque hamlet]

The technology is there. And I think most people are good and will want to contribute to the artists in a way that they will be able to make a living doing music again

[deep nova]

distrokid uploads to apple music and Itunes makes you put in a artist and producer now ... O opted not to upload to them but what would I enter for AI music?

[opaque hamlet]

Let’s see what happens. I noticed that Musk is staring a gaming studio. Maybe this year he’ll develop some ideas in the music industry for leveraging X as a distribution and discovery platform for music.

[west yew]

Ultimately the way this happens is through live music performance and the selling of merch. The actual music doesn't make anyone any money, and probably won't again. Unless it becomes licensed for something else, like BGM for a movie or show...and there, I think the AI argument has a point. AI music could potentially replace BGMs for shows that don't really care.

[opaque hamlet]

Yes this is what I’m hearing from my friends and family who follow the industry more than I do.

Also i wish the largest artists would give back more than they do.

[west yew]

For my own music, I put myself in as the lyricist and myself with Suno as composer. Again, I'm very clear that I'm using AI music, because the monitization possibility of any song I write with Suno is so close to zero as to be zero.

[deep nova]

I just don't know what the legal ramifications are ... they look for a real name so I am not sure SUNO AI would be an option ... i'll try it out and see

[west yew]

The problem with being a "big artist" is that there really aren't that many "big artists" out there. Pearl Jam attempted to fight Ticketmaster back in the day. Taylor Swift showed a model for releasing your music when you don't have the rights to your masters. But a lot of these things are way beyond the level of artists themselves, who very rarely have an understanding of the legalities involved.

[west yew]

If you made the music on the paid version, you own it outright, and you don't owe Suno anything. So you are free to put yourself in that spot. You created the prompts, you produced it, you paid for it, it's yours, and Suno agrees.

Suno says it owns the rights to anything you made on the free track, but given that you can't copyright AI material, that absolutely will not stand in court, and Suno has a whole lot to lose by trying to sue users who monitize a free song to make $7.

[opaque hamlet]

No way is Suno going after their consumers. It would be a nail in the coffin.