🪲┃bugs-web

[ornate cape]

я работал со второй версией и она мне очень сильно понравилась, но хочется идти в ногу со временем, поэтому не хотелось бы, чтобы третья версия была хуже второй.

[ornate cape]

Спокойствие моей души успокаивает тебя,
Ты ложишься спать, а мои руки как змея,
Тихо поглаживают твою шейку, ведь ты моя,
И посмертно никому я тебя не отдам, прости, люблю тебя!

WTF?

Ему так пофиг что строчки надо заканчивать

[neat bloom]

what the hell are these restrictions, I can’t justify someone’s behavior by citing an example of someone’s words.
For example:

"They paint Russians as evil communists,
They drink vodka, it's cold outside but they don't mind, they just live like nomads"

But after that I wanted to describe the real situation in Russia, such as free medicine, clean streets and an advanced civilization, but why the hell can’t I write even this piece of text in quotation marks????
What’s so bad about these words? Why are you so strangling with your “CENSOR” systems?
Damn it, how stupid is this!
As a result, my text is broken and I will have to either throw it out or postpone it indefinitely, it’s so stupid!
But F words is ok.....................

[low cypress]

In V2, continuing lyrics where they were cut off can be tricky. You might need to experiment by either adding the next word, repeating the cut-off part, or leaving it out entirely. V3 simplifies this with the "continue from anywhere" feature, allowing you to start just before the cutoff for a smoother transition.

[wicked dune]

Today, I encountered an exceptionally peculiar situation. I requested the creation of an instrumental track, yet what I received was a remarkably crafted song complete with lyrics and vocals, surpassing the quality of the original. How is such an outcome even conceivable? Occasionally, I do receive lyrics and vocals, albeit in an unintelligible language, which I presume is a byproduct of the generation process. But how can it produce genuine vocals seemingly out of thin air?

I was using v3

[buoyant flume]

Thank you for flagging this - we'll look into this one!

[wicked dune]

Also, I've noticed that when I repeat the generation process several times, the final results tend to sound better. Could the model be learning from our previous attempts? For example, when I ask for a harmonica track, it often nails it after 8 to 10 tries. Knowing if this is the case would help me decide whether to stick with this method or accept the first result. Thanks!

[neat sail]

What are words but sounds we give a name. Just like music it can generate 'vocals' unconditionally too. Take any prompt, try leaving the lyrics blank, and add to your music style something like "my music whatever, male vocals" if you want more songs like this. (I love them.)

[wicked dune]

Just to confirm, I toggled ON the [instrumental] option, indicating no lyrics were desired [disabled]. However, when I specified "Relaxing Blues Music" as the style, it unexpectedly produced a fantastic vocal track with meaningful lyrics.

[neat sail]

Of course when this happens when you don't want lyrics, that is indeed a bug. These modern AIs can be difficult to control.

[wicked dune]

A happy bug tbh:)

[neat sail]

Did you really get meaningful lyrics? Can you DM me the song if so?

It is quite rare to get comprehensible lyrics. Usually it *sounds * like it makes sense, like this. But if you listen to the full content it doesn't really have much semantic meaning or even real words.

https://app.suno.ai/song/61fbba14-5574-4d3f-941c-b476c78e8f91/

[wicked dune]

https://app.suno.ai/song/00853630-adff-40de-a23d-0111c0b20d11/

another one near the middle https://app.suno.ai/song/9b1c441a-e15a-4a82-9823-46d46ea40b3f/

[humble cobalt]

Yeah that's just hallucinated nonsense words. I've had a few like that (different genres though). I guess blues songs tend to have vocals so it's just tried to generate some word sounds.

[neat sail]

Sounds great. There are some phrases in there of real words. Maybe Blues as a music genre tends to use classic phrases more than other genres - so a few more tend to come through intact. But outside those phrases it's mostly still "SimLish" style fake words.

[wicked dune]

Yes exactly the same like this 🙂

[neat sail]

You can add ", instrumental" to your style prompt if you do want to get some without vocals like this.

[wicked dune]

Thank you, I'll definitely try that. It creates beautiful generations in the blues genre, and I suppose jazz as well, especially after listening to yours. Very interesting!

[red reef]

When I generate a Song that has lyrics it often happens, that the song is Just instrumental or straight up gibberish. Has anybody an Idea about that?

[low cypress]

Gibberish lyrics (hallucinations) in AI-generated songs can be hard to avoid due to the AI's creative interpretation. Sometimes, especially depending on the genre, vocals may only become apparent if you choose the 'continue' option, as the AI might create a lengthy instrumental intro before introducing the lyrics.

[red reef]

The continue option did not help. It also happens, when the track is short( way under 1min20) so it shouldnt be an intro issue.

[patent swift]

hi there. @delicate iron please contact me. tried to call you a dozen of times

it's Nik

anyone else is available regarding API and agreements' questions?

it's super urgent. I'm trying to contact team for months

the situation is pretty bad

@buoyant flume @royal mason @summer wraith please 🙏

@unreal radish @tame latch

6 countries, tons of influencers promoting AI music...

[patent swift]

I know but we had a discussion about it

[solar grove]

Did the discussion end with "We'll call you"?

[patent swift]

I managed to create the cobranding opportunity but there is a silence from Suno team

now our project turns into nightmare because music generation doesn't work today

today's one of the last most promoted days for the project

[unreal radish]

sorry, I have no particular knowledge of this situation but if you've been in email conversation with someone from the team about a business opportunity, discord isn't the place to discuss it

[patent swift]

I would be happy to discuss it via email or phone but nobody generates any messages over there

let me know if you have any other ways to contact anyone from your team who could help us

[unreal radish]

-> DM

[patent swift]

I would appreciate the opportunity to connect directly via direct message with any team member who adds me as a friend

[unreal radish]

tried to DM you earlier, please go ahead

[viscid garden]

I have a particular clip that seems to be stuck. I could listen to it (I did enough to mark it with a thumbs up) while it was generating, but now I no longer can. In the UI it looks complete, it's has its picture just like any other clip, and was made last night/early this morning, maybe 6-7 hours ago.
When I go to play it, I just get the spinning animation like it's loading, but it never does. Did not received any 'failed to generate' warnings or anything of the sort. It looks like any other clip until I attempt to play it.
The clip isn't important to me, just reporting the strangeness around it.
https://app.suno.ai/song/26629512-6de6-41f0-88f9-9d0fb0765400

[unreal garnet]

It does enhance the creativity for sure, but it doesn’t really simplify the “figure out which lyrics to use” part.

Imagine I’m generating

Roses are red,
Violets are blue,
Sugar is sweet,
And so are you.

It gives me the perfect generation for the first two lines but then I want to regenerate.

So I continue from in-between “blue” and “Sugar.”

Here there are options:

1. I enter the entire verse as is
2. I enter starting from “Sugar”
3. I enter starting from “Violets” (because sometimes the part between “blue” and “sugar” isn’t a whole second, so you have to start somewhere in the middle of the “Violets are blue” part)

I’ve had examples where one of the three worked while the other two didn’t, so I’m guessing it’s a bit of a black box rn.

What would help is to know what exactly is prompted to the model in this situation by Suno. Say, I enter “Sugar is sweet” etc. What is the actual input to the model? Does it concatenate it with the original lyrics prompt? If yes, fully or partially? If latter, how does it decide where to cut off (proportionally by time vs character length? Else?)

A lot of questions knowing the answers to which would greatly improve the possibility to generate coherent continuations.

So, a short wiki on that would really help.

@muted flame @tame latch fya as well

(this all should’ve been in #suggestions, so sorry for misplacing, hope it’s not a gread deal.)

[lusty pollen]

THANKS for fixing both Bugs in just hours... Amazing work Suno Team 👍

[low cypress]

I am different in that.. I work in parts of verse and chorus. If it drops in a verse, I'll try to fix it first with continue there.. if it doesn't work out for me in 3 tries, I move on to continue before the verse..

[raw shadow]

bug again full song no lyrics

[cobalt flower]

Actual Behaviour :

- I have a first part of 1:21 (verse, pre chorus, chorus), when continuing form this with verse2 pre chorus, chorus the AI output a 1:00 clip with a truncated chorus. It always stuck at 1:00
``````diff
Expected behaviour

+ The ai doesn't stuck to 1:00 because i am in v3

[true nexus]

☝️i noticed that aswell

[lusty pollen]

continuations are always up to 1 minute. only the first part is up to 2 minutes. is that the problem? or do i just miss the problem?

[viscid garden]

Yeah every continuation is always one minute, max

[cobalt flower]

it's an issue if your previous chunk is more than 1 minute. The ai is totally lost after that xD

[low cypress]

It's always been that way.. first gen is longer, the ones after go per 1 minute max

[lusty pollen]

ok. i am clearly not understanding whats going on. if you continue a 1,21 minute song at the end you dont get a full combined song that is something like 2,21 minutes?

[low cypress]

Make a full song in between... that way it keeps a longer memory

[lusty pollen]

if you want things from part 1 in part 3 you have to combine part 1 and 2 before making part 3. just to extend what daizy said.

[cobalt flower]

my song is a repetition of this structure (verse,pre chorus, chorus) it fit well in the first generation because it's more than 1 min.

For my second chunk it's the same structure. it stop during chorus (lack of time because time limit) ok not an issue if i can continue from that point to finish the chorus.

But when continuing i got a totally différent song not the same rythm or notes

[lusty pollen]

ok. thats not a bug... thats a way suno works. you have to put less things in the later parts and hope it works... the main stuff should happen in the first part to give the later parts a structure to learn the style of lyrics and stuff... and then combine after every part you like to give the later parts a chance to learn from the previous whole one

[cobalt flower]

you have to put less things in the later parts and hope it works

Usually it works but with more than 100 try i never succeed

[lusty pollen]

because you just have half of the time.

[cobalt flower]

the first chunk was so good i want to continue it. So sad....

[lusty pollen]

and you cant continue it?

do you use v2 or v3?

[cobalt flower]

not with same mood rythm and notes. I've got something very different

v3

[lusty pollen]

ok great... v3 will most likely give you a fitting piece after x retrys. i sometimes make 10 or 20 part 2s until i have one i want... or i just use 20 seconds of the next part and then make even another one... a little bit like like a puzzle where you can cut the pieces...

this only gets an issue when you are a perfectionist... of course.

[cobalt flower]

seems i've not kuck i'm on more than 100 try to get the next part with something goof xD

[low cypress]

Please take a look at my video.. here you see me making a full song, keeping the same sound of the chorus in the next parts: https://www.youtube.com/watch?v=5uJw5kb05as

[lusty pollen]

have you tried to move the time where you continue from?

it makes a huge difference.

[cobalt flower]

i use to be pianist, and when i generate a piano song i am perfectionist , yes xD

[cobalt flower]

i already create 5 song that works well (and some with more than 30K listen on platform)

[lusty pollen]

daizys video is ace.

[low cypress]

I am a guide and might know some tricks you haven't seen

[cobalt flower]

will check that

[low cypress]

I already used over 100.000 credits.. so not a novice trying to gain some followers.. Only teach

[cobalt flower]

great work, hope suno give you some free credit xD

[glossy timber]

Hey I was wondering something. Basically I was told a month ago when the v3 beta was available that there were some payment problems on my account before. They fixed it but today is my credit renewal date, so I am not sure if I haven't gotten my credits yet because it renews at a certain time, or if there's an issue with the renewal still.

[wicked dune]

Daizy, you're amazing! I used to overlook great parts due to minimal flaws, but this solution is perfect. Thank you!

[low cypress]

Normally you will get the new credits on the date and TIME (<- very important) you became a subcripter..

[glossy timber]

Alright

[glossy timber]

By the way in your video, I never really thought of making it into a full song before continuing, does that tend to help it stick to the same rhythm as the first half?

[low cypress]

yes!

[glossy timber]

I'll have to try that when I get my credits renewed

[wicked dune]

The song is absolutely wonderful as well. 'Peanut Butter Blues'

[low cypress]

It's okay.. I like my buzz buzz buzz rockabilly the best

[glossy timber]

@low cypress btw is there a reliable way to make an instrumental intro with v2 since you can't continue from a certain timestamp in v2?

It makes them like a minute long

[low cypress]

No.. not to my knowledge.. for me always a hit and miss.. [intro], [instrumental intro], such things help

[glossy timber]

Dang I was hoping there was a trick i didn't know of lol

Hopefully they add that feature for v2 soon

[onyx goblet]

Think i manaaged to break the AI
https://app.suno.ai/song/715c4588-a2ce-4713-b2bd-7c4c71e882d0/
could the needle be skipping?

wait.... it sounds normal here

it skips less in discord than it does on the site O.o

That second "truely amazed" is aactually the whole first line again on tha app

This is what it sounds like when played via the app.... same song, same generation, how odd is that?

[humble cobalt]

I think someone else mentioned similar recently - there was a song that was alternating every 2 seconds between different generations.

[onyx goblet]

well i just spent god knows how many credits to get the first part to a style i like, only to have it refuse to generate the second part

First part:
https://app.suno.ai/song/2268ca0b-f616-4b60-9c1c-801bce812268/

continued lyrics:
`(Bridge)
At first, I was scared, thought I'd never survive,
But now I'm cooking up cannibals, man, I feel alive.
Who needs room service, or a fancy feast,
When you've got mutant meat, it's a cannibal's feast!

(Chorus)
Living life on the edge, what a wild retreat,
With cannibals on my tail, gotta move my feet.
But I found a silver lining, in this chaos galore,
Those cannibals make for a tasty treat!

(Outro)
So here's to this island, and all its charm,
With Kelvin by my side, and a three-legged arm.
I'll sing this song, as I roast on the fire,
In this crazy world, I'm a cannibal's desire!`

[lusty pollen]

its a bug that sometimes happens... i think i got it 2 times in 10k credits

[onyx goblet]

I am getting REALLY pissed off with this censorship

[onyx goblet]

first time i have had the bug myself, i find it odd that it sounds different via discord than on the browser though.... more of the wrong gen on the browser

just trying the :video game trick.... nope thats just giving me 20 second parts.... not refusing but just wasting credits. Man i miss what suno was.

[lusty pollen]

one of the wrong generations i deleted and i used one as a drop in a song 😄

[onyx goblet]

for anyone looking at this later:
Full lyrics (created by chatGPT as I am at work)
`(Verse 1)
Woke up from a crash, feeling kinda dazed,
On an island full of weirdos, man, I'm truly amazed.
Cannibals and mutants, what a sight to see,
But with Kelvin by my side, we can conquer this spree.

(Chorus)
Living life on the edge, what a wild retreat,
With cannibals on my tail, gotta move my feet.
But I found a silver lining, in this chaos galore,
Those cannibals make for a tasty treat!

(Verse 2)
Three-legged lady, she's a curious sight,
Shy and friendly, but she's alright.
Kelvin's my sidekick, though I suspect he's dim,
But he follows my orders, every single whim.

(Chorus)
Living life on the edge, what a wild retreat,
With cannibals on my tail, gotta move my feet.
But I found a silver lining, in this chaos galore,
Those cannibals make for a tasty treat!

(Bridge)
At first, I was scared, thought I'd never survive,
But now I'm cooking up cannibals, man, I feel alive.
Who needs room service, or a fancy feast,
When you've got mutant meat, it's a cannibal's feast!

(Chorus)
Living life on the edge, what a wild retreat,
With cannibals on my tail, gotta move my feet.
But I found a silver lining, in this chaos galore,
Those cannibals make for a tasty treat!

(Outro)
So here's to this island, and all its charm,
With Kelvin by my side, and a three-legged arm.
I'll sing this song, as I roast on the fire,
In this crazy world, I'm a cannibal's desire!`

Censorship monster strikes again, guess as far as suno are concerned all songs are about puppies and rainbows.

[humble cobalt]

There is quite a wide chasm between puppies/rainbows and people eating each other, to be fair.

[viscid garden]

lmao

[onyx goblet]

depends on the puppy.... but i mean its not a graphic song.

Any other service I would just not use my credits until its fixed, but my credits expire on the 23rd and I have already canceled the sub. So i may as well use them up. Such a shame, this was such a fun hobby.

and maybe "Censorship monster strikes again, guess as far as suno are concerned all songs are about puppies and rainbows." was a tad harsh. But, this censorship is insane. If you prefer to have a darker sense of humour when it comes to games then Suno says no even when its not really that dark.
I know the issue has been reported, but nothing seems to be happening about it and when our paid credits have experition dates on them we cant just sit and wait for a fix.

Sorry for my harsh words, but I do honestly feel that censorship method is just doing more harm than good.

[rare path]

My kids like to generate a song, then just do a cover of it themselves to fix whatever wasn't right. You could do the same, it seems pretty fun.

[cobalt flower]

i use to be a pianist, i can't play anymore piano

[rare path]

I'm not a fan of the censorship stuff either. Keep checking in, maybe that will get resolved at some point.

[onyx goblet]

yeah i plan to go back to being a free user until the issue is resolved. I am certainly using it less now.

[rare path]

I sometimes get results with [music] ... then the verse

[rare path]

I've had a month or 2 where I didn't burn them all. It's nice to know there are there just in case I need a song that no reasonable band would waste their time on.

[onyx goblet]

That would be 99% of all songs i make with suno lol

[rare path]

I keep paying just to keep this project funded. Not that I don't want all the bugs resolved, but I can wait.

[onyx goblet]

I get that, though with this over censorship I will vote with my wallet.

[glossy timber]

It's not that bad with the censorship, I only get it periodically

Even then it only takes a few wording changes to fix it

[onyx goblet]

Since it was implemented I have hit it a few times, and only once did I think "ok yeah that could be seen bad"

[rare path]

That's the spirit. I hope they sort it out. I imagine the censorship is part of the input and not inside the model, so it is fixable.

[glossy timber]

I just wish it would tell me what it flagged specifically instead of me changing things around until it works

[rare path]

Oh yeah, if it had an output of the dirty words in the error message it wouldn't be nearly as frustrating.

[onyx goblet]

Yeah that would help, but from what they said its an AI deciding if its offensive or not. Which ...well AI is not ready for

[glossy timber]

Oh a separate AI checks our lyrics?

[onyx goblet]

From what was said yes

[glossy timber]

Ah I didn't know that

[onyx goblet]

I cant find where i was told that though

[rare path]

LLM's are not reversible, so if it's doing it that way it's way trickier.

[onyx goblet]

ah here we go: #🗣️┃chill-chat message

Honestly we are a long way from AI being up to the task of moderator.

[rare path]

Though that means you could probably keep resubmitting the same thing and it would eventually get through. Because if it's not deterministic, there is always a chance it can get through. also, modifying things to be spelled wrong would get around it.

[onyx goblet]

AI is really good with spelling mistakes. As a dyslexic I have found that out pretty fast.

remember were not talking keywords here, If you look at my song further up there is no words that would be considered offesnive.

trying the [verse 3 : video game] brute forced it passed the censor but it then refused to do longer than 20 seconds.... multiple tries. and on an old song the other day the parts were so small it refunded them.... way too many in a row to be mere coincidence

[rare path]

I bet canny balls would slip through.

but I haven't tried... just making it up

[onyx goblet]

giving it a try, i remain skeptical, if the censor is that easy to beat than all it will do is cause agro for legit users.

well:
`(Bridge)
At first, I was scared, thought I'd never survive,
But now I'm cooking up canny balls, man, I feel alive.
Who needs room service, or a fancy feast,
When you've got mutant meat, it's a canny ball's feast!

(Chorus)
Living life on the edge, what a wild retreat,
With canny balls on my tail, gotta move my feet.
But I found a silver lining, in this chaos galore,
Those canny balls make for a tasty treat!

(Outro)
So here's to this island, and all its charm,
With Kelvin by my side, and a three-legged arm.
I'll sing this song, as I roast on the fire,
In this crazy world, I'm a canny ball's desire!`

https://app.suno.ai/song/be638d79-4e5f-4f0f-ba35-72627c092d80/
and
https://app.suno.ai/song/7c046fe1-2e9a-4100-88c8-8c5e0143b4c7/

so hits the same issue as changing the (verse 3) to (verse 3 : video game) 20 seconds and 35 seconds clips

@rare path it was worth a try though and a good idea.

[gloomy sand]

How to change given email? Seems to be impossible...

[tame latch]

email [e-mail]...

[gloomy sand]

I emailed them 2 weeks ago. - Absolutely no response

[tame latch]

🤦‍♂️

sry for this

I don't know anything about billing, so I can't help here personally...also this is not a bug...

email again? but we are all very busy this week...

[gloomy sand]

I'll do. The bug is that if I change my email in Discord, this is not reflected in Suno. Also, when disconnected a Discord/Google account it is still connected in backend somehow (it's not the last one) and next time I want to sign up to create new account, this account will be reconnected to the old Suno account, no matter what. Also, the password in Suno has no purpose, as we cannot sign up by email which I always preferred.

[tame latch]

Thanks I appreciate this feedback...and sorry for your experience

[gloomy sand]

Sure and thanks for your response. I just want to let you know that I cannot pay for subscription until correct email address can be easily updated for Stripe, etc. I lost 2 hours trying connecting/disconnecting, changing IP, browsers, devices, still cannot change my email or account for sign up. Thank you for taking this into account.

[tame latch]

@muted flame when you have time... 🙏

[cobalt flower]

UI issue, when thumb down so clip it disapear and pop up just after

in create ui

[low cypress]

It's designed to do that.. Have you already see my video?

It's still there but no longer in your way.. it is in your library

[cobalt flower]

i'm not sure that a clip playing hide and seek after a thumb down is a feature xD

sounds more like a js animation fail

[low cypress]

Nope.. I am sorry it is really designed that way

[cobalt flower]

what the purpose of that ? 🤔

thumb down in create mode is to remove it from my ui

[low cypress]

If you give it a thumbs down, you don't like to use the clip. So it will go to your library

Thumb up stays

[cobalt flower]

thumb down should remove it from the create UI , right ?

[low cypress]

yes..

[cobalt flower]

so there is a bug, it remove it with fadeout animation and just after it show again in my create list

[low cypress]

It's in your library

It's not gone

[cobalt flower]

i think you don't understand what i mean

just talking about an animation issue...

[low cypress]

Your description is not clear to me. I don't see it happen with me. Behavior should be: you click thumbs down, and the song dissapears from your "workbench" into your "library". You shouldn't see the song anymore on your workbench leaving room to the songs you want to work with.

[cobalt flower]

yes the issue i got sometimes is that it disapear from my workbench like expected with a fadeout animation and appear again just after in my workbench with fade in animation like it was not thumb down

[low cypress]

Ah.. I will keep an eye on it if I can reproduce. Haven't seen that yet.. I did in the past but that was more than a month ago.. maybe that bug is back: 🤷‍♀️
Thank you for reporting 🙂

[cobalt flower]

will film my entire work next time and clip it when it happened, i got that bug more than 50 times during my session

[wicked dune]

same here, I thought it s my browser! even if i removed it tobtrash, once i go to library and comeback to creator page, all the deleted parts appears again.

[neon ivy]

Did you break something? I can’t write lyrics, copy and then paste into the lyrics box following update today. I’m on iPad. Edit: nevermind, fixed itself after reboot

[obsidian river]

generated two songs in V3 alpha. liked one. both suddenly disappeared. help

[crimson osprey]

go to your library and on the top right you find a trash can symbol. when you click on it you see your deleted songs. check if you find it in there

if you find it there there is a "remove from trash" button if you click on the "..."

[obsidian river]

i will try. thank you 🙂

[craggy charm]

Is it just me or are the vocals in general mixed too loud? Often they clip and distort so have to be binned, but I am now regularly going into lala.ai in post and splitting the vocals, reducing them by up to 4db and mixing them back in - it works quite well but I shouldn't have to do it, surely in the training data the vocals would be professionally mixed to the correct level in 99% of the cases?

oh, and where has the "suggestions" channel gone?

[low cypress]

click here on the arrow down besides the word FEEDBACK

You simply closed it

[craggy charm]

aha

ta

any comment on the loud vocals?

[low cypress]

I have no insight how the model is trained. I am a volunteer guide. I only know the model didn't change.

[opaque furnace]

"Prompt flagged for moderation:" All I want a text with the Easter bunny crazily bombing people with chocolate eggs - what's wrong with that? 😿

[true nexus]

maybe bombing? 😂

[opaque furnace]

I have started to suspect that is the case, and it's these lines that trigger the stoopid warning: Feliz Páscoa perturbada. Vou bombardeá-lo em pedacinhos com meus ovos. Ovos malucos para todos, aproveite meus ovos-bomba!

[low cypress]

@opaque furnace if you go to chatGPT 3.5 (which is free) you can post your text there and askes what could be the problem, it will tell you. It can help narrowing down the problem. It's not the perfect solution, but it give you a directions since ChatGPT is the api that Suno uses for this.

[opaque furnace]

Thank you, I'll try that. Still a fact that the Easter bunny throw chocolate bombs, (which also is a legit word in English) - so this one was disappointing.

[low cypress]

Don't shoot the messenger.. I think ChatPGT is too strict in general too. And I think it will bite them in the foot in the end.

[opaque furnace]

Proof!

[low cypress]

Yeah.. I know.. Can't you write differnent like bommbs?

or bombbs

[opaque furnace]

Oh no I do not, you're a nice person. I am irritated on the filter - no you! 😸

[low cypress]

As long as it still sound the same?

[low cypress]

thank you ❤️
But I totally agree.. it is super annoying

[opaque furnace]

change the word to 'barrage' still no luck.

Lol! All of a sudden the AI accepted: "...canonada de chocolate!" 😹

Either some mod adjusted things right now, or a word they had not considered.

[low cypress]

Nope.. AI is not always giving the same answers and actions.. just like making AI music 😂

[opaque furnace]

Aha, it was the variation that got my camel to pass trough the needle hole.
Still 'canonada' is synonym to 'barrage' which was blocked.

[low cypress]

To effectively end a song in Suno, try using explicit metatags like [Fade Out], [End], or [Big Finish]. Alternatively, using a clear instruction in the "style of music" field, like adding "end," can also be effective. Remember, the AI’s interpretation might vary, so experimenting with different methods is key.

[ornate falcon]

Couldn't generate that.
Prompt flagged for moderation

idk why i have this 😢

i delete all bad word like btch or "dead body" in broken body but always have this msg...

[low cypress]

You have to look to your overall meaning of your lyrics. Copy/paste them in Chat GPT and ask why it could be flagged for moderation.

[obsidian river]

nope it didn't help. the tracks that disappeared are not there

[ornate falcon]

You know I am everything you love
and made me love you to throw me away
oh, okay I take revenge,
killed the prod'
that's all I have left now
a broken body that's you left
oh, oh, okay I take revenge,
killed the prod'
oh, key, that's all I have now
broke body that's you left

is really cool :/ but maybe a little bit hard for Chirp :X

[obsidian river]

actually they returned to my library somehow🙂

[ornate falcon]

freedom of expression is not yet won 😢

is really friendly but yes, i go change my lyrics

[low cypress]

Library is the place to find all your songs.. they do not get removed unless you do it yourself. If you give a thumbs down however, it will dissapear from your workbench to the library. @obsidian river

[low cypress]

Sorry man.. I cannot help you with that

[ornate falcon]

np eheh i go change my lyrics x) thks again ❤️

[distant thicket]

I'm experiencing an issue where when I hit create it generates two of the same song whereas prior to a few days ago it would usually generate two different songs from my prompt. Is there some way to fix this? It'd be nice if this was just a setting.

This isn't custom mode, just your basic generation based on a prompt.

[humble cobalt]

This is a recent, intentional change, from what I have read here.

[distant thicket]

Oh I see, I think it'd be nice if it was a setting then as I liked getting a variety of results before. If it was a setting I could use it to drill down on the one I liked I suppose.

[low cypress]

Now the bar is even. This is the same behavior people with costum lyrics have. So it is more fair for all users. But that is my personal opinion.

[distant thicket]

Sorry, I didn't understand what you said. Could you rewrite that? "Mow the bar is even" and "costume lyrics". I'm confused

[low cypress]

People who use the "custom" mode, always got 2 songs with the same lyrics. So now it's in balance. Everybody get the same value for their credits.

[distant thicket]

I see, if I had a song I wanted to get different variations of wouldn't I just take it into custom mode? I haven't used it too much personally as I started recently but it seems the variety I was getting before would still be better with this in mind.

[low cypress]

Yes.. the easiest way is to use the function "remix". It will copy your prompt into the custom mode. And then you can even alter the text, tweak it, remove, add, etc. Plus you can also change the style of music manually. Lot's of extraf freedom.

you can find remix in the menu, or the button by the song

[distant thicket]

Right, the custom mode would let me drill down on a particular song I liked. That's why I felt like when I was getting two different songs when I was generating before I had felt I was getting more options instead of now just getting two slightly different songs with the same lyrics.
(Using custom mode to try to get the variety I was getting before is just a workaround. I was getting two sets of different lyrics/two different songs every two generations with a click of a button before the update and that was great for me.)

[low cypress]

The first generation via custom is always different as well. But just with the same settings.

Every 1ste generation is totally unique in sound.. even in the same genre

[analog ibex]

I'm getting a lot of initial generations with no vocals despite having generated lyrics. I can see the lyrics are there and I did not request an instrumental track. This is has happened multiple time and none of them have refunded credits. Has that been reported by others? I'm not doing anything differently than what generally works.

[analog ibex]

The remix feature has never worked for me once an idea is rendered as a full song. Should it, or does it only work for clips?

[low cypress]

I advice you to look at the community wiki. There are clearly some misconceptions how Suno works: https://www.suno.wiki/

[rare path]

it's just for clips.

[analog ibex]

Thanks !

[rare path]

If you have v3a you can do a lot more than 8 lines of lyrics depending on the speed of the song, so you can manually remix with fewer generations.

[analog ibex]

I have V3a. thanks again.

[analog ibex]

My question relates to standard mode, not custom mode, and V3 in the web browser. I'm talking about lyrics initially generated by Suno, not me. Therefore, this isn't a user error situation. I'm just using a basic prompt on style, genre and topic. Suno added the lyrics and then didn't sing them. I've used Suno for over 2 months and yesterday is the first time this started happening. Suno generates two verses and a chorus, but doesn't sing any of it. It's just instrumental.

[low cypress]

I understand how puzzling this can be. AI, including Suno, can indeed be unpredictable at times. But I can guarantee you the model is still the same and the way it interprets prompts can vary, leading to unexpected outputs like instrumentals instead of sung lyrics. You might try adjusting your prompts or experimenting with different styles or metatags to see if it influences the AI to sing the lyrics.

[tame latch]

gosh

sigh

[cosmic kraken]

i tried to create a song but after i clicked on create it went unclickable and didnt created my song, since then i can no longer make new songs, how or where can i write to support? because i see no option anywhere to contact the support

[solar grove]

Have you tried refreshing?

Save your work first! Copy the lyrics and save them before refreshing!

[tame wolf]

Hi, I just started using Suno yesterday and a bug that has really caused a lot of problems for me is this: Sometimes stuff will show, sometimes it won't. For instance, if I press "Library" from the "Create" page, it will show the library as completely empty. However if I open a single song into it's own tab and THEN go to the library from there, then all my older songs will be there like they should. I've been in many situations where the only way to do something (for instance continue a song) was to open up a single song tab and then do whatever action I wanted to do, because the Create page wouldn't allow me to do it. It's annoying to have to keep doing the "open up a single song tab" trick to do stuff reliably.

[worthy roost]

this output is kind of a bug. https://app.suno.ai/song/d52d521b-6ca0-49d2-a298-31f487516bea

[knotty lion]

Hello I have found a bug in Suno. I logged in and had two songs the Faded Memories indie mellow acoustic. But I didn't create them (consciously) and of course the 10 credits were deducted immediately.

[low cypress]

I'm not encountering this behavior. Would it be possible for you to create a video capture?

[tame wolf]

After I had reported this bug, I searched through the discussions here on Discord and discovered that I just need to refresh the browser. I guess what was happening by using the "open a single song tab" was because it was not loading cached data but "fresh data". So I guess by opening the library from there, it was as giving the updated current version.

So I guess the issue is that cache causes problems which forces users to refresh their browser to see the actual status of everything. But obviously it shouldn't work that way and the software should have the ability to ensure that the end user always sees the data up-to-date

But next time this happens, I'll be happy to take video capture of the situation. Right now I can't do it since I already refreshed and Suno is "up-to-date" in my browser

[plucky vine]

v3 (Alpha) - has difficulty completing the song. It usually cuts off the very end. I'll sometimes have to add the last word to get it to complete. Suggestions: 'Get Whole Song' would put the fade out sounds/end to it. Would be nice if it saw any word in brackets always as commands... like [End] or [Drum Fill], etc. [Fade Out] and [End] do seem to help.

[dusty sage]

https://skr.sh/sOrHrEqz9qe

Cannot add to playlist while generating, UX problem, please fix it

[merry sorrel]

Why do the generations often end up worse than on V2? I mean very muddy quality, missing out of words, etc. Sometimes it's like the quality is the same or even worse than V3, but not always. Am I to understand it's all completely random?

[full socket]

Should it be renewed today? I've been waiting for the tokens to be reloaded.

[low cypress]

While V3 of Suno is more advanced than V2, it's still in a developmental stage and can occasionally produce inconsistent results. This inconsistency could be due to the model's intricate learning algorithms and longer memory, leading to occasional muddy quality or missing words. Remember, it's not entirely random; rather, it's the complexity of V3's AI that sometimes makes the outcomes unpredictable. Continuous improvements and feedback help refine its performance over time.

[fluid wigeon]

I've been using the free version for a while now, but recently I've noticed that the output is increasingly poor. Both the lyrics and the language are completely changed, or the style is completely ignored despite a detailed prompt. Now just don't come with “unpredictable” or “completely random”. This is definitely not a coincidence, but a huge bug that has been costing credits for some time and making the web app completely impossible to use. Since Suno AI is apparently unable to introduce payment methods other than credit cards, which some users actually don't have, this bug is an even bigger mess. There were times when self-written song lyrics worked, but at the moment only AI-generated ones are working, I don't know if that's intentional. But as already said, it's about the bug that language or text is changed to the point of being completely unrecognizable. Guys, that's not how it works, fix the free version, it's slowly enough. You feel ripped off. Sorry for the harsh words, but it is like this and it has to be said.

[solar grove]

If you have a continuing subscription it should renew on the date of your first subscription but I do not know what time exactly that will be.

[zinc talon]

v3 is only generating up to 16.5kHz again, like v2. v3a generated up to 19kHz.

[full socket]

Thanks.. I'll wait one more day!

[solar grove]

Oh, I'd check it periodically. At least that's what I did last time... 😁

[full socket]

I've been doing that for the last few hours...

[olive glade]

Maybe it's to make room for watermarking?

[zinc talon]

No, the water mark has always been there and that's not how they work.

[nocturne pumice]

hi, It suddenly stopped generating and says couldnt generate that, oprompt marked for moderation.. and it does not have any 18+ words in it or sth to be moderated... why wold that be?

[kind swallow]

Im having issues using suno ai on one of my devices. I try to go to the create 0age and its just sitting there with the loading icon. But it never loads.. not using an adblocker

[merry sorrel]

I agree and I do it sadly, since I love this project and have had a few great results from it so far. I can be patient, I can wait if I know that something is being done about it, because my money goes towards it

[lucid basin]

hey guys

something is up with the new 3.0 model, reverb is off the charts

it's bad

https://app.suno.ai/song/0bcd5f10-111d-4da2-8993-7fb182cb4ee9

[merry sorrel]

Perfect example of what I meant

[olive glade]

V3 is definitely more sizzly in the highs with a weird 16khz cut-off over that. Also it doesn't seem to listen much to the style of music on continuation, you used to be able to prompt for a guitar solo on continuation, now it seems it just kinda continues the song in whatever way it wants to.

[lucid basin]

It's mostly stuff that can be fixed in mastering, but the vocals are getting fried

[olive glade]

Well the missing data over 16khz can't be fixed in mastering

[lucid basin]

no you're right

the instrumentation should be fine tho

[olive glade]

It's not listening to the prompt change much anymore when continuing, which leads to much more boring song structures than in V3 Alpha. Or maybe I'm just unlucky with the gens

You're right about the reverb though, usually I'll add slight reverb glue on the master chain, but can't really do that when it's this heavy on the reverb thoughout

[marsh sleet]

The rename function does not work at all on my phone, but works fine on desktop.

Google Pixel 6 Pro, Android 14 QPR3 Beta, Google Chrome Canary
Surface Book 3, Windows 11 Pro Canary, Google Chrome Canary x64

[tame latch]

I think there is a bug with continue off v2 with v3, but I can't locate it. If anyone could find a reproducible failure and let me know how to reproduce it, it would be greatly appreciated!!

[zinc talon]

I'm a little confused as to why the frequency range has been lowered, specially with the announcement that v3 has "better sound quality"?

[dim dagger]

V3 sound quality is now bad what happened??? please repair it

[olive glade]

I'm suspecting it might be because there was audible flanging going on in the very highs in alpha, maybe?

[zinc talon]

I didn't experience that.

[olive glade]

It was random, but happened quite a bit on metal tracks

[zinc talon]

Ah. I don't make any type of rock music really.

[olive glade]

Hopefully they'll let us access the V3 Alpha eventually, at least to finish some tracks in progress.

[acoustic rapids]

Any idea why some songs show like 30 seconds left then it just asbruptly stops ?

[olive glade]

Holy shit, took 30 generations but finally got a guitar solo, it was the only prompt description on song continuation lol.

[acoustic rapids]

like cut off mid song i hate when that happens.. some times ug et 60sercs+ somtimes like 30

[olive glade]

Oops realized I was in the bug channel, sorry about that, I'll go back to general.

[marsh cape]

looks like the api.suno link doesnt work on mobile discord, works fine on browser discord

i just was on mobile and the link didnt embed, im on pc most of the time using a browser, it works fine

[low cypress]

discord is V1. They just released V3 on the website... Join the fun!

[lucid basin]

wait v1 is still up? V1 was cool for certain things, like it predates the current calculation regime 2 and 3 use. You can make some really weird stuff with it

Like reggae songs with 20 line verses and stuff like that

lol

[low cypress]

Yeah you can still use it.. but on discord only

[full socket]

my credits is still not being reloaded

[molten fractal]

V3 would be great if it actually let me download tracks. Unfortunately, clicking download just brings up the damn share option.

[lucid basin]

do we have an eta on when/if 3.0 is getting fixed. Everything I'm rendering sounds like it was performed in a tin can

and I'm in the middle of a colab

[low cypress]

Once a model is released, it cannot be changed in between. To only thing that Devs can do is turn the parameters and temperature. But to be honest, it should very genre related. I am very happy with my personal results. And Tony already said, it will not be perfect.

[lucid basin]

country, hip hop, r&b, rap, soul, and rock are all affected at the moment

[zinc talon]

To me the audio quality is obviously compromised from v3a to v3. I hear it clearly.

[lucid basin]

and none of the usual cleanup tags are working

[low cypress]

I am sorry but I personally don't have the same experience. But I also have no influence in that regard, nor sight. I am a volunteer guide 😬

[lucid basin]

I'll talk to tony

[neat sail]

The only thing I know for sure, so far, is V3 prompts quite a bit differently than V3a. So it's difficult to compare 1 for 1 with ealier songs.

[zinc talon]

v3 is giving me great music, yes it reacts differently, but that's a learning curve. the audio quality is something different.

[lucid basin]

what are you doing differently?

[rare path]

Are you talking about the highs like symbols sounding clipped?

[lucid basin]

this https://app.suno.ai/song/dd14aca4-48d4-4666-9bcc-66f22ea74027
everything I'm doing sounds like this

ignore the unifinished lyrics

if you can hear them

or this, fine on the quiet stuff, but when it gets loud, it's messed up, which is problematic becuase of how these renders work https://app.suno.ai/song/b746b960-2fd6-4466-a938-5520e0211c3a

or this https://app.suno.ai/song/849ddcea-1805-43d3-bf6e-74ed11033418

or this https://app.suno.ai/song/5ae65c92-47c6-4b67-bdb3-f1dddd44db1c

or this https://app.suno.ai/song/9fd8ce52-399c-4592-a8c2-a8084a6ba6db

[rare path]

It does sounds like it's the highs that are messed up right?

[lucid basin]

I don't know what it is

but it sounds like a toilet bowl

[rare path]

I can't download other people's songs easily, but I bet if I put them on a spectrum analyzer they'd be clipped at a lower freq than whatever the last versions were...

Your stuff is very catchy though.

[lucid basin]

thanks

I'm doing a colab with another song writer

[marsh sleet]

When my song is finished generating, I would like to be able to edit the lyrics that it displays. It dropped a few things & it isn't accurate.

[lucid basin]

Great idea. Goes on the suggestion board tho

[marsh sleet]

I think it is a bug & I suggested a solution. haha

[lucid basin]

feels like a bug

lol

[marsh sleet]

Seriously, make #💡┃suggestions-web & #🪲┃bugs-web into a single #1082775249667502111 forum channel....

the ideas shared are just getting lost.

[surreal viper]

When no title is given the filename is mp3.mp3 while I think it should be prompt + ID.mp3 To me it's a bug, previously reported.

[craggy charm]

V3 down for me at the moment

[plush coyote]

It's always CORS 😭

[worldly thicket]

"Prompt flagged for moderation" - was eager to test the V3 but... well...

[tawny whale]

😄 same here

[placid notch]

Hello, do you know why my V3 continue are still bugged (like alpha version) ?

I can't continue with a verse/chorus, seems like my singer skip the verse and just sings the chorus :/

7 seconds songs.... why the continue bug is not fixed?

[thorn bloom]

This has only happened to me once so far. I'd say it's easier to just start a new song if the AI can't seem to sing the verse.

[placid notch]

but my song is awesome I want to keep the singer, the instrumental 😦

[wary flare]

Sadly the audio quality is getting more jittery each day. V3A was good but v3 is mostly bad as hell. I am a paid user, hope audio quality for subscribers will be better soon.

[thorn bloom]

Yeah I get it. That time when it happened to me I burned through like 100 credits trying to get it to sing the verse but it just got stuck with the chorus. Or you can try continuation from an earlier point. You'll lose some of it but at least you get to keep the singer's voice.

[placid notch]

I tried the continuation from an earlier point, sometime it's okay but it's still unstable. But thanks for the tip

[versed rover]

I have a big problem, and it's costing me money again. 😓 No matter what I create, no matter the genre, with every song, the AI starts singing in harmony after 3-4 lines and keeps adding more until everything sounds like a choir! In V3

Even when I input "solo singer" as the prompt.

[broken cloak]

I can't submit phonetic spellings of Latin 🤷‍♂️

[mystic spear]

It is very annoying asking for instrumental and still getting vocals. Seems like instrumental should be the easiest thing to control.

[mystic spear]

I am or was a paid user. Used all my credits and cancelled for next month. I'd say only half of the stuff I generated was useable. The other half the audio quality was too poor. Not worth the cost until there are some real improvements reported.

[quasi trail]

Hello, well "Prompt flagged for moderation" for no reason is kinda annoying

[dim dagger]

please back V3 alpha V3 = v2+ but not alpha

[deft tapir]

today almost everything I generated is messed up as if filtered with 10 echoes, making most generations unusable, opposite experience from earlier days :S

[broken cloak]

https://app.suno.ai/song/fc61e052-dd27-4e09-88a0-e566b4aff5b1/ it's impossible to get ai to sing "Gaudia carpamus, dum possumus,
Nam mors atra propinquat omnibus." even though it sings other words. Those particular words will not be sung wether they are by themselves or with the following lyrics, wether they are in latin or IPA.

[marsh cape]

thats not what I meant. on mobile discord the embedded link doesnt have a play button

[low cypress]

Devs are not focusing on discord atm. They are focusing on the WebApp. That's why I pointed to the WebApp. Doesn't discord work atm? I am not using that function.

[marsh cape]

maybe im not being clear enough, the play button is missing on mobile discord app, embeds that is

[rotund tree]

What I noticed in V3 is that voices aren’t very distinct. They are slightly muffled and the songs sound tin-y and have no flow, as if the system can’t determine where the lyrics should fall in relation to the music. It sounds more like a bad version of V2 from my first impression.

V3 alpha was amazing. There must be a hiccup somewhere that they will fix

Just as a note, I was trying to make a generic rap song using custom lyrics, as well as other songs that are German techno with simple lyrics (also custom lyrics)

I had not added anything very technical.

V3 alpha did both of these genres quite well

I also noticed a lot of songs start with lyrics, when I prefer to start with an instrumental intro. Almost sounds like it’s starting in the middle of a song…

[spiral forge]

Just chiming in on the issue with vocals. They sound like they have excessive chorus applied to them. Also if I take an existing song with crisp clear vocals and try to continue them, the vocals are greatly degraded.

[kindred hare]

I had the same problem with some Portuguese words

[lofty dagger]

So far my issues sounds like the songs I created have a choir singing as the back vocals it only does the with V3 at the moment I tried enough where I'm just wasting my credits since all the songs on V3 has a choir as back vocals. That's one of the issues I'm having

[versed rover]

As i said

[thorn bloom]

V3 seems to have gotten rid of one bug - a very annoying one where it used to always cut off at the last word in V3 Alpha - now it seems to finish the last word usually.

But - it still seems to have trouble keeping the chorus the same after a couple continuations. The first 2 chorus have no problem - but often times the 3rd one sounds completely different.

[broken cloak]

https://app.suno.ai/song/62e5d89d-5965-4dbc-9ff1-b39a59941d67 this style is completely bugged, this is the genre you get everytime.

[thorn bloom]

Think I can +1 on the vocal issue. The more continuation goes on, the less clear vocal becomes. Toward the end it sounds like a ghost singing. Didn't have this problem back in V3 Alpha.

[tame latch]

try to add more instrumental or breaks in the middle

once the audio quality degrades, it is hard to pull it back

[thorn bloom]

Yeah, guess I need to use continuation from an earlier part.

[spiral forge]

"usually" is the key word there. v3 definitely still ends generation too soon.

[thorn bloom]

Err ... the AI not keeping up with the chorus is a bigger issue for me right now. Tried at least 20 rolls and only got 1 time it used similar chorus as before. Problem is, this one time there's crackling sounds. 😤

Anyway, I give up. The 3rd chorus is just going to sound different than the first two. Hoping the next version this will be solved fully.

I think it was better in V3 Alpha though. Unless this is just a rare exception.

[thorn bloom]

Update: I went to an earlier segment for continuation and this time the 3rd chorus sang correctly.

[tacit geyser]

For me the crackling sounds sometimes are just an issue with the web player. If I download the audio and listen to it in a music player most of the time there are no crackling sounds.

Regarding losing audio quality: A lot of times I just add a small instrumental to the last part that sounds good, then continue from an earlier part again, then glue them in my DAW. It takes some time to do it and the song won't appear on Suno's first page this way. It would be awesome if Suno had a way to merge clips together on request, even if they are not direct continuations to each other.

[hallow zenith]

If i begin with "instrumental" and continue at next step as a "song" with lyrics, so i make continuation, but get no singing, would be nice if this would work , right ?

[versed rover]

Thats how it works for me

[Intro] [Instrumental] [Guitar]

[Verse]
"SONGTEXT"

[Chorus]
"SONGTEXT"

[hallow zenith]

yes as a song , but if i start with "instrumental" mode .... i get better instrumental ... and then wish to continue as "song"....

[versed rover]

ahhh ok sorry

[winged shadow]

How can I disable the overly sensitive content moderation feature? It seems to have some glitches. For instance, when I start composing a song with Suno and attempt to continue it, my content often gets flagged unnecessarily. This issue occurs unpredictably, even when using the same text, preventing the completion of a song. Surprisingly, even text from the well-filtered ChatGPT gets flagged by Suno. As someone who has been working in the industry for several years, I wanted to conduct an experiment to replace my normal playlist with songs entirely created by Suno and present the results. However, the consistent issues with content moderation are severely hindering this process. I believe credits used up to the point of flagging should be reimbursed until a solution is found. This problem isn't isolated to one instance; it randomly affects other songs, forcing me to rewrite entire texts due to perceived content moderation concerns. There should be an option for paying customers to disable this feature, especially since we're creating music—a medium where expressions of passion or frustration are common (and prevalent in much of today's radio hits). If necessary, accounts producing genuinely harmful content could be moderated separately. I upgraded from the Pro to the Premiere plan and have used approximately 5,000 credits in the last week to curate my playlists. However, the constant flagging issues make it challenging to replace my standard playlists effectively at this stage. Consequently, the tedious process of repeatedly dealing with content flags is shattering my hopes of completing this experiment. If there is no hope, I will not continue my subscription until a more user-friendly approach to utilizing the song engine is implemented or open source are in place.

[sand nacelle]

After the release of v3, all the songs are very slow, even if i add "dance", i've generated 3 times, all 6 songs are extremsly slow, 4 are wave flamenco and 2 are rhytmic dance

on v2, with same songs, they are more pop like, do you have the same issue?

V3: https://app.suno.ai/song/395152a4-a5df-430f-8ae9-ba5a1e605019/
V2: https://app.suno.ai/song/efe64fcb-2187-4b24-9de7-6c1f3490b421/

[neon ivy]

Still got generations singing [backing vocalists] out loud

[hushed cradle]

Hello, not really a bug but more a graphic aspect that tickles me. I have the impression of a blurring effect/poor resolution in the sound management tab. Even on a small screen it's not really visible but it jumped out at me when I played it on my big screen.

[maiden elk]

Where can I find a list of special prompts
Where can I find a list of special prompts
Where can I find a list of special prompts
help

[west elk]

The singing gets muffled and static over time.

[neat bloom]

Do you planing to make it possible to edit the lyrics of generated songs? I mean visually ofcourse, It often happens that the final version of a song contains repeating elements due to the combination of 2 or more parts of the generations. I understand that this is not critical, but from an aesthetic point of view, it is not good. I also wanted to know if it would be possible to cut off an unnecessary part of the song? I have a few generations where at the end, after the completion of the song, there are strange sounds that I would like to remove, but there is no such option.

[rain lodge]

BUG: My songs (Whole Song + Public) don't show in the DISCOVER NEW area any more.

[jade tide]

cannot really create a song, the button is locked

[hexed sonnet]

No matter what I do, it keeps messing up lyrics, at least 6 tries and they are all unusable, because they merge parts of the lyrics so it becomes incomprehensible. I've tried making it 1 verse and 1 chorus only, tried changing the lenght of the lyrics etc. None can be used.

I give up for now, hope the bug is fixed or it's not worth paying for a subscription anymore

[neat bloom]

This happens to me when I use punctuation in text, periods and commas often break the text

[hexed sonnet]

No, that was not the problem

[inland mango]

A few things really preventing me from using V3. First, it decides to generate long breakdowns whenever it pleases, even several per song. Second, it will re-use lyrics as it pleases in ways that don't make sense, such as re-using lyrics from the chorus.

It's really hard to know what tempo it will choose, so creating enough lyrics to fill the initial two-minute max generation time can be tricky. But the AI doesn't treat it as a max generation time, it treats it as, "Let me fill in as much as I can into this empty space".

Also, the audio quality is just, not there. I would have expected better, especially since I've seen videos of the alpha and the quality was stunning. When I use it, the quality is very grainy.

[zinc talon]

The current poll on audio quality should include "are you listening on your phone speakers" in which case your opinion is irrelevant 😉

[versed rover]

Hey! Just to emphasize, the real concern isn't about the song's Tempo, but about how fast the AI decides to deliver the lyrics. Even with any genre at a certain BPM, the key is how the lyrics are paced you can have lyrics delivered at a regular pace or much faster within the same BPM. The speed at which the lyrics are sung or spoken determines how much text fills a specific time frame, not the songs Tempo itself.It's all about the lyrical delivery speed.

[inland mango]

am definitely not 😆

[versed rover]

Of course, not necessarily. Therefore, you can proceed as follows: Any given genre has songs that range, for example, from 130-135 bpm. If you choose a genre, you can create a corresponding song lyric within this bpm range, and it will fit both 130 bpm and 135 bpm.

Use a KI like ChatGPT for a good tempo comparison.
As long as you know the tempo of your chosen genre. I can easily create lyrics that fit just as well at 100 BPM as they do at 140 BPM.

If you keep stretching the tempo further apart, it will EVENTUALLY become too wide. Thats right

[winged shadow]

why? Is the word seductive forbidden?

if I use tags with sex in the style it generates music.. I dont understand the content filter system

[wicked dune]

I highly recommend watching the Daizy video; it directly addressed and helped me resolve that issue. https://youtu.be/5uJw5kb05as?si=Nv1cn-w-czkrG-uv

[valid pendant]

I think suno has a problem with turkish songs. Even I check instrumental, it does have lyrics still. For example: https://app.suno.ai/song/5bda143c-9674-47ab-9d6a-56f4863ed5cc

[neat bloom]

Yes, it helps, but it costs credits.
It's not my fault the neural network creates music with unnecessary sounds or other errors, I'm not paying for luck, I'm paying for results.
You pay for random song generation, the neural network makes music for you with errors, you pay again to fix those errors.
Is it just me, or is there something wrong here?

[worldly bay]

I seem to have broken a playlist. I was trying to remove a song from the playlist, but the action never completed. Later, when I refreshed and tried to view the playlist again, I got the following message:

Application error: a client-side exception has occurred (see the browser console for more information).
Debug console:
Error: An error occurred in the Server Components render. The specific message is omitted in production builds to avoid leaking sensitive details. A digest property is included on this error instance which may provide additional details about the nature of the error.

I've attached the info in the error object, but it's impossible for me to decipher, since it references minified production code.

The playlist is: https://app.suno.ai/playlist/386e21eb-065f-4a06-8979-005dda92f387/

[worldly bay]

Update, the playlist is working again! I was looking through my other playlists, and one of them suddenly had the same problem. But after a refresh, it worked normally. Checking back with the playlist with the original problem, it is also now fine. 🤷

So maybe it is only an intermittent problem? I was worried that it was permanently corrupted, with no way to regain access to the songs, but I guess not. 🥳

[solar grove]

Please do not post the same thing multiple times like that. Take ten minutes to read the #📜┃rules regarding this.

[low cypress]

Tip: make full songs in between and continue from there. It keeps a better memory on melody then.

[solar grove]

That said, you can get a lot of good information like this at the unofficial community wiki: suno.wiki

This is a good article to begin with: https://www.suno.wiki/faq/metatags/song-structure/

[tough harbor]

the quality erosion mid clip is getting more severe, I think it happens more in continuations even though the beginning is clean

https://app.suno.ai/song/3b473e8e-9bb5-4fe0-896d-bf47f7665859

[low cypress]

If you want control, then you are too early. Suno is GROUND BREAKING technology. Feel free to look around. Nobody can doe what Suno already can. But it is Ai, it is new, and it still unpredictable. If you don't like that, I suggest to stick to the free credits only, or try again over time. Suno is very realistic about it. They ar not saying everything works perfect.

[viral swift]

download button on desktop (chrome) is not reacting on any of my songs. Is this a known bug?

the same for the get whole song button

[low cypress]

Have you tried a different browser?

[dim dagger]

I really would like to know why this V3 version voices are very metallic instead of V3A

[humble cobalt]

I have quite a metallic vocal on the single song I've made with V3 so far

[mild bronze]

I can't use simple words because artists uses those words as names.

[rare path]

I know I'm going to regret getting into this, but if you want full control you do have a few options, learn some instruments and get a band together (years), learn a DAW to make your own music (weeks), go on fiverr and see if you can pay someone to do it for you (days), or embrace the randomness that is AI generated music (minutes). The music/audio is random for a reason; if there wasn't randomness it would just churn out the same song over and over for a given prompt. You're supposed to get unexpected results. There are some issues with the model and that's to be expected. You have to keep in mind that right now is the worst it will ever be. You can certainly check in every few months and see if it gets to the point where it's above your threshold for what you find acceptable.

[rare path]

haha, one like that got me to the other day, just talk around it. or fix it up in custom mode. I do love that it's telling us what keyword caused the problem now.

[low cypress]

I've let the devs know.. You can try by spelling it wrong as s workaround, or say holiday?

[mild bronze]

Yes, "vacation" works, but feels like a bug, what if a new Artist uses the word "Life" "Cool" or "Love" lol, but yeah. I think more words have the same issue, and "vacations" is one of those words.

[winged shadow]

avatar is also blocked as an artist, i don´t know any artist with that name

[rare path]

maybe the movie and its soundtrack? I donno. I look at that error more as a challenge, but admittedly, I'm a native english speaker so getting around it is much easier for me than someone who isn't.

[dim dagger]

my tips to pass an artist name Snoop-Dogg influenced dr-Dre, emi-nem 😂

[rare path]

and if anyone is wondering, yeah. it does sound like the band you'd expect, lol https://app.suno.ai/song/c63a518c-69bc-4e54-892c-65a13b6c6360

[low cypress]

Or Timbaland? https://www.instagram.com/timbaland/p/C4ve-mnLKCG/

[mild bronze]

I can use names from singers outside USA btw, just testing now.

[rare path]

Shhhhhhh

[solar grove]

Oh...do tell...

https://tenor.com/view/evil-smile-evil-evil-laugh-gif-14398336

lulz

[mild bronze]

https://tenor.com/view/noo-toy-story-buzz-lightyear-gif-14664723

[jade jolt]

can't find liked songs. anyone else?

[marsh sleet]

Oh! Dope! I didn't know they had videos up! Thanks for making that, @low cypress !! That's a sub & all notifications turned on from me on your yt!!

[neat sail]

Probably just temp glitch. Refresh and try again (and be a bit patient...)

[jade jolt]

is it also the same for you or does your liked songs show up? It has been the same for 3-4 days

[marsh sleet]

it only displays liked songs from you, fyi, not from others.

[rare path]

To answer the question, I see my liked clips as of this timestamp.

[broken cloak]

Hi. Maybe if you put "Turkish style" or write something more specific it might help. Suno ai might take Turkish as a request for Turkish language vocals.

[gleaming halo]

Please make v3 Alpha avalable for premium users. Most of the genres that were much better in V3 alpha like rock, rock ballads, electtic guitars, choirs, classical, saxpohone etc are much worse in V3. For us, we have mutiple people working on this, V3 is a showstopper and a set back to V3 Alpha

[onyx goblet]

Just to cross post these as its half a suggestion half a bug (big bug) #💡┃suggestions-web message

[radiant swan]

so i recently had hallucinated vocals with many many n-words. is this common? i was surprised, but i worry some may find that offensive. so i figured id bring it up. https://app.suno.ai/song/d8fc9e6b-07e6-4ca6-ad91-6244f825f219/ starts at 1:55

[zinc talon]

ooouch

[broken cloak]

I wouldn't post that because it's kind of a self report.

[unreal radish]

fixed, ty

[grizzled ridge]

Disappointing audio quality latley, any particular reason?

[unreal radish]

we set AUDIO_QUALITY='BAD' today

[pallid oasis]

I got an error when "Kalimba" was in my style of music. The intent was to have some Kalimba (thumb piano) instruments in the song

[sinful silo]

Why is suno.ai sometimes only generating half a song, and sometimes a complete song with an ending?

[gleaming halo]

but seriously i could send you a hundred prompts or so that perfomed poor on v2, great on v3 alpha and poor on v3. Can you roll back this update or make v3 alpha available? This is having a negtive impact on user experiences for Suno.

[tame latch]

check: #🗳┃mini-polls message

[tame latch]

unless that poll says otherwise significantly, no rollback. We can't just affect thousands of user's experience based on individual feedbacks...

I don't think you should prompt v3 the same way as v2 or v3 alpha. What works before may not work now. You will have to adapt to it...

[tame latch]

hey but it is not like your feedback is just getting thrown in to the trash can. We will improve v3.

It's like all the marvel after credits v3 will be improved and come back...

[viscid garden]

Welcome to model changes

Have to learn how to use each new model

[versed rover]

Believe me, these errors do not occur randomly. Don't let them fool you; the AIs are just laughing at us 😂😂 https://app.suno.ai/song/43695e41-12a5-4eef-92bf-582efb5a4510/

[gleaming halo]

we appreciate the hard work and the improvements. But the issue is not in prompt engineering, or model changes, we can deal with that. Its just that for certain music styles, there is a serious degradation in music quality in v3 compared to v3 alpha. I have 200 plus pages in my library so some experience. For us v3 is not usable, sound quality wise. Sample here but I can send a hundred more. https://app.suno.ai/song/4acee567-6500-4a28-8dcd-a0e7f3de22df/

[tame latch]

and continue from doesn't help?

[versed rover]

Ohh... Are 5,000 pages maybe too much? 😳 I should start a Suno detox. Although, with what Suno is generating at the moment, it's not hard for me to do so.

[gleaming halo]

Its almost on every clip in certain categories…. If you like I can compare some v3 alpha and v3 prompts and clips and pm them

[versed rover]

By that, I'm only referring to the "vocoder effect" bug. And that Choir bug, where suddenly a single voice turns into an entire choir.

[tame latch]

I mean I get this issue, but you can't hack your way out of it by continue from?

[gleaming halo]

I'll do some thurough experiments and will let you know

[crisp plover]

I'm putting this comment here because I think it's closer to a bug than a feedback. But I am struggling to get a reliably male vocal. Even when I give it prompts with male vocal, male solo, male singing voice, etc I get female vocals more than half the time.

[tame latch]

let me know if you need credits, be happy to tip you some

[tame latch]

ok...what's your lyrics...or do you have other genre tags that are suggesting female vocals?

[versed rover]

Type "Female Singer", "Male Singer"

[gleaming halo]

No need I have plenty

[tame latch]

can't hurt to have more...

[viscid garden]

(take the credits)

(always)

[crisp plover]

I would like celtic ballads with male vocals. I understnad that Loreena McKennet is a pretty big first thought when celtic music is mentioned, but at the same time, when I try even a simple prompt of nothing more than celtic ballad in a male vocal, I feel like it's a bug if it still gives me female vocals most of the time. There shouldn't be anything in a simple tag like celtic ballad that makes it inherently female.

[tame latch]

ok then try male vocal, male vocal, male vocal, celtic ballads, as the genre tag...

you can repeat the tags to tell the model you really really want it

[crisp plover]

I've tried repeating it twice, but I will give that three times a go and see what happens.

[tame latch]

you can put as many times as the text length allows

you can also pad your lyrics with it like [verse, male vocals]

[crisp plover]

On, nice to know thanks. It is generating now with my first attempt of super repetition. I will let you know 😄

[crisp plover]

Look like repeating it 4 times is the key. Thank you.

[tame latch]

still random but should help

glad it did...

[thin night]

Hi, i have problem. I want create longest song but it doesnt work. When i tap on “Continue form this song” and type next lyrics after generating i got 0.03 seconds audio and credits refunded. Whats the problem?

[tame latch]

try continue from an earlier point

also if you create sth that's longer than 10 mins likely the job will fail

[thin night]

First take ~60seconds

[tame latch]

not sure I understand

[thin night]

I continue next clip from 0.50 seconds and part 2 generating about 0.05 seconds and ceedits refunded

[tame latch]

ok...choose a different time also change your prompts...

[eager shadow]

I am really struggling to get normal crisp vocals as of the latest update, I got 1 semi decent generation but continuing it, degrades the vocals tremendously

[neat bloom]

I had a serious problem, in the middle of the song it refused to generate a continuation of the verse but instead inserted phrases from the previous parts of the generation or simply endlessly repeated a couple of words or broke the instrumental or even all bugs in same time.
Guys, who has problems with the continuation of songs like me?, I found a solution!.
You have song controls like verse or chorus or whatever you may have, don't use square brackets!!, use colon!, for example-

Verse:
Blah blah blah
Blah blah blah

Bridge:

Chorus:
Blah blah blah
Blah blah blah

After that, I never encountered this bug again and the continuation of the song was immediately generated as it should, although I spent about 250 tokens looking for a solution to the problem 😭

[vale crypt]

On your list of clips you can go to a page number that is higher than the songs you own just by clicking fast - the detection seems to be delayed

[vivid badger]

cant make anything. stuck on black loading screen

[neat sail]

Were you able to make songs before?

[vivid badger]

nope

just made the account and tried to make a song, just been black screen

[neat sail]

Must be some login issue. can you click on your name and log out and back in?

[vivid badger]

why is that text red

[neat sail]

That's just a news update. You may need to email [e-mail] if there is something in the account that isn't working. Did you login with Discord by any chance? If you did, can you check if Discord thinks your email address is verified?

[vivid badger]

it logged into my throwaway discord instead of the one i use daily, im thinking that was it

[solar tinsel]

Under Custom Mode, When you type in a subject and click Generate lyrics the format it displays is as follows
[Verse]
[Verse2]
[Chorus]

it should have been
[Verse]
[Chorus]
[Verse2]
[Chorus]

[sand nacelle]

V3 Sucks, no matter what genre i put, it ends up somewhat Mythical slow vocal

[true nexus]

maybe share the prompts, genre and the song structure if you want, maybe the issues it's somewhere there

[sand nacelle]

they are songs that i remix using v3, and v2 were jumpy and fit the genre, the v3 it's extremly slow

V3: https://app.suno.ai/song/395152a4-a5df-430f-8ae9-ba5a1e605019/
V2: https://app.suno.ai/song/efe64fcb-2187-4b24-9de7-6c1f3490b421/

[strange fjord]

All my generated Songs With V3 have very poor Sound quality. The drums, especially the cymbals, sound very metallic and reverberant. Its very disappointing at this Stage, especially as v3alpha did Not have this issue

[high sundial]

Hey I try to get the full song but it keeps giving me an error whats the problem and how do i fix it if possible?

[jade jolt]

yes just like spotify, it isn't showing for me unfortunately

[static tundra]

Sorry wrong channel

[cold ginkgo]

I ended up with 4 of the same songs. I'm using the free plan. Is this a bug?

[hushed cobalt]

it have more error.

[tough stream]

Good afternoon why is the button not working and how to fix it?

[rotund tree]

All of my songs sounds like the same voice is overlapped on top of itself. I think this is what people are referring to when they say they hear a chorus sound, rather than an isolated and distinct voice. @low cypress

[spiral skiff]

You have maximum 50 credits on free tier. Once you use them all, gotta wait for refresh on next day or upgrade to a paid tier

[rotund tree]

@tame latch

[low cypress]

How long is your song officially? Is it above 5 minutes?

[marsh sleet]

EDIT: Deleted. Not needed or relevant. Long ramble. Given new intel that negates the need for this.

[low cypress]

Staff is working on new FAQ. The community FAQ is not the official one, so I think that is the thought behind it not to replace it with that one.

[marsh sleet]

You got the inside scoop.
I think it could be good to offload it though. They don't have staff to spare. It is also a sexy domain name. lol

but for sure the Notion needs to die

[low cypress]

Yeah the current one.. but in general notion is pretty cool to work with 🤷‍♀️

[marsh sleet]

I don't wanna chat here in this channel too much, but I must say, yes, Notion is dope. I prefer Obsidian for my "second brain" notebook though. But Notion is a better option for professional team notebooks. It is so cool. But it is also a tad obtuse for your average user.

[low cypress]

Yeah I know obsidian.. but that's better for internal use.

[gleaming halo]

@tame latch I have sent you a list of 10 songs in PM where I compare v3 Alpha vs v3 to show the issues. Please accept the connection so I can send the full list.

[viral swift]

I have not. But the important thing this happens in chrome. Which is one the most used browsers.

[low cypress]

True.. but if you try it and you give feed back on that too, we might figure out if it is on your PC, or on Suno side.

I mean.. I have chrome too, but do not have this problem you are describing and there are enough other users like me who don't experience this. And it could be as simple as a plugin blocking it.

[viral swift]

Oh I'm on a Mac studio, I can try it on safari later on. But I'm all up to date and have had no trouble ever with any website.

It works on chrome on android

And the brave browser

I'll get back to you

No installed plugins btw

[karmic spade]

I can not continue from my old V2 songs at all. It just shows 00:00 for continue time and I can't change it.

[jade jolt]

Hello. Unfortunately my liked section shows no songs. anyway to fix this?

I have tried several browsers

[low cypress]

liked is only for songs you made yourself, not others. So is this about your own songs or yourself?

[jade jolt]

oh i didn't know that damn, should have made a playlist

[dim dagger]

@low cypress nice to see you back, any tips to why I can't have clear voices in any generation I did with V3 ???

[low cypress]

Hey Infinity Flow, great to see you too! For clearer voices in Suno V3 generations, you might want to provide more specific voice metatags and consider using styles with less complex instrumentals. The voice clarity can sometimes get muddled with heavy instrumentation. One of the things I usually add is "minimalistic" in my style of music.. and sometimes even "a cappella" since it still keeps the music, just a bit less.

[dim dagger]

ok I see thank you!

[rare path]

for whatever it's worth, I just did an mp3 download using chrome 122.0.6261.129 on an arm mac and it worked as expected. maybe restart your browser, update, test in incognito mode, or something like that. I think this is a client side issue.

[static tundra]

Since this afternoon when I click on create the song keeps loading but nothing shows up in my library.

[dim dagger]

any song structure tags work perfect, I'm struggling!

[cloud python]

Agree 100%. I don't know what changed between V3 Alpha and V3, but so far I've not been impressed with V3 at all. It doesn't understand genres, song structure, context, nor are the instruments and vocals as crisp and clear as they were with V2. How did V3 end up WORSE than V2?!

EVERY SONG I've generated in the past 16 hours has had multi-layered vocals like this:
https://app.suno.ai/song/42093396-0142-48c3-b359-56c3c9882899

I want a solo lead singer with perhaps a second singer in the chorus sections (which I consistently got with V2 & V3 Alpha). For comparison, here's one done in V2 a few weeks ago:

https://app.suno.ai/song/16591e8a-4103-4553-a1ca-17dae63d6f02

V3 seems to be attempting to mask it's "tinny, vocoder-ish" sound by blurring vocals with choral singers. Everything I produce under V3 sounds muffled, like the instruments are being played in the room upstairs and the singers have cottonballs in their mouths.

[cloud python]

100% This is spot on.

[low cypress]

Don't forget to use the poll guys: https://discord.com/channels/1069381916492562582/1163842729231528047

To vote what you think of v3

[cloud python]

The same here. Somehow, V3 seems WORSE than V3 Alpha and V2.

[long girder]

been trying to continue a song in V3 since last night and have burned at least 50 credits getting only 10 seconds of silence in each iteration

[eager shadow]

yeah im really debating on just running V2 in the meantime and hope this gets looked at soon, just hate doing that because the actual music is so good in V3 just seems voices took a huge slap

[tired tiger]

Do you guys really think it's worse? I can see that V3 generates vocals which are often more layers, or in general tend to sound more heavily edited/produced. But If the goal is to sound like big music productions, then it's not a bug, it's a feature.

[long girder]

created this song in beta so maybe that is the issue

[cloud python]

As of V3's official release, my V2 seems to have gone awry as well. My keywords are "alternatve, grunge, 90s" and numerous generations last night came out sounding like a Jamaican style song (though at least the vocals were crisp and clear).

I've also noticed that the longer you keep producing "attempts" on a given song, the more the quality degrades. Last night by the time I'd got to my 12th or 13th attempt, it wasn't even trying to speak English anymore; just gibberish.

[eager shadow]

i can show you the difference if you like, ive made a good handful with V3 Alpha and tried recreating in V3 normal, you can hear the vocal quality difference @tired tiger

[eager shadow]

continuations will degrade the quality regardless, thats been the deal since V0

[cloud python]

Yes, definitely. And no, multi-layered vocals and muffled instruments does not mean "big music production" style. Solo artists like Avril Lavigne don't use multi-layered vocals.

[eager shadow]

but its way less to a degree nowadays then it was with V0 and V1, though I cant complete a full V3 song now

[cloud python]

I'm not talking about continuations, I'm talking about new generations. If you click "Create" and it produces two options and you don't like either one, so you click "create" again to get 2 new versions <-- doing this repeatedly for 12 to 15 iterations will degrade the quality as well

[eager shadow]

ohhh yeah i know what you mean, like if you dont edit any portion of the prompt and just keep clicking Generate

[cloud python]

Correct.

[tardy plume]

https://app.suno.ai/song/006e68ef-481f-4dc8-a055-e0460bb6b89d

why did the generation make this one short and not close to 2 min?

[eager shadow]

yeah, thats also true from my experience since V1 as well, my fix for that is just add a comma or 2 in the lyrics

[tired tiger]

Yeah sure, send me some links or clips. In general I'm not saying that I prefer layered vocals, but I can just imagine that the training set for V3 has just more of that than V2, and it's somewhat intended. But if there are other issues, I gladly like to listen to examples of those.

[cloud python]

I posted a comparison a few minutes ago. A link to a song in V3, and one I generated a few weeks ago in V2. The clarity of the instruments/vocals is unmistakenable.

[eager shadow]

ill show you v3 Alpha and V3

[long girder]

when did V3 rollout, just so i can compare what i have created pre and post

[eager shadow]

V3
https://app.suno.ai/song/35facc3e-91b5-434b-a3d6-89fa4a74ba31
V3 Alpha
https://app.suno.ai/song/ee29f448-8164-4009-bf9e-b3204ebdd2af

[long girder]

this is whats happening when i try to continue V3 Alpha created songs today

[eager shadow]

Alpha song I had just finished like a day before V3 was released, I was just testing recent songs to see if it was my prompting or general issue

[tired tiger]

Ok, difference is pretty big. Dunno if it's still mostly random. Hard to tell without somewhat A/B testing.

[eager shadow]

V3
https://app.suno.ai/song/4818fbfc-f5a8-457e-bd16-67abd146995e
V3 Alpha
https://app.suno.ai/song/158ba622-fd0f-4fec-b22a-a94aa0997519

another example

Its a very distinct muffling of the vocals, the muffling seems consistent no matter the hip hop style I choose

[tired tiger]

Maybe it's the "Turbo" treatment like with GPT 😉

[eager shadow]

I have no idea how their technology works and wont pretend too either, but I THINK it may be something to do with the watermarking that was added, potentially

[eager shadow]

was released on the 21st #📢┃announcements message

[long girder]

thanks, thats when i started having the issues

[rotund tree]

This one sounds better imo. Vocals are more clear and instrumental sounds more distinct as well. The instrumental is pretty good in V3, but the vocals in V3 have something going on with it.

[mystic spear]

are the developers doing anything to fix the voice layering issues in V3? I have seen a lot of complaints but no response.

[rotund tree]

Yes! I wish Suno would label what was V3 Alpha and what was V3

[eager shadow]

your saying the V3 sounds better than the Alpha in the Granny song?

[rotund tree]

The bottom one sounded better imo.

[eager shadow]

oh yeah, thats my point as well

in every song

[loud lotus]

V3 alpha is a lot better than V3.

[eager shadow]

it can do great instrumentals (V3) just vocals are killing me

[long girder]

eevrything i generate now seems to be in harmony

[rotund tree]

Yep same. I can get ok results sometimes if I generate 4-6 times. One or two of the continuations might be ok. But out of the gate; I had a better experience with the alpha version

[eager shadow]

https://app.suno.ai/song/aef93a1e-e084-4373-8cc6-3c36667eeaac
I was working on this one and started off so promising, but degraded quickly before the end

and the chorus is so 🔥 but cannot get it to have clear vocals post chorus

[rotund tree]

Better channel is chirp technical for this. But I’d just make a continuation at a certain time right after the chorus

[eager shadow]

I did, like 30 times at different times

Seems like a bug to me though.. cant imagine this was intentional

[long girder]

everything gets forced into a harmony

[native escarp]

continuation in the web app really seems like it needs some work

[low cypress]

please let your voices be heard via the poll here: https://discord.com/channels/1069381916492562582/1163842729231528047

[native escarp]

I'd like to continue a song for twice the length and let the model come up with something, can't figure out how to do that and the generated version is consistently a few seconds long. annoying

[eager shadow]

yeah I have faith it will be fixed though, these guys do their best since Ive been here

[eager shadow]

did yesterday

[long girder]

I agree..impressed even with the harmonies

[eager shadow]

this is V3? or Alpha

[loud lotus]

Interesting how many people like V3 more.

[long girder]

just did that like 5 minutes ago

should be vno harmonies

[eager shadow]

I think its really genre based, some genre's come out clean or at least the "harmonic" effect adds value to the song rather than detract

[eager shadow]

I think this sounds great in my opinion but idk what you were going for either, assuming an acapella lol

[loud lotus]

Talking of wishful thinking: having control over this harmony "feature" would be fantastic. Occasionally it could fit quite well. For Billie Eilish songs... Or some background or call and response stuff.

As long as it's out of control, I made the best use of it, that I could think of

https://app.suno.ai/song/cd704964-b522-4767-9b73-b88233508dec/

[eager shadow]

perfect example of where it adds value

[loud lotus]

Or Billie Eilish soundscapes (with an accidental hint of Tori Amos piano)

https://app.suno.ai/song/a869210d-0066-468c-8f20-b13c223fa975/

[long girder]

acapella aria solo

my challenge has been trying to continue a song and the Part 2 iterations always seem to be in harmonies...cant get to finish a song to match the singular voice from the beginning of the song

[cloud python]

Jesus V3 Alpha is better by an order of magnitude. Your comparisons are better than mine. Geez, what happened between V3A and V3?

[eager shadow]

Took the words right out of my brain lol just hope its noticed, because I am at a standstill, and I have been on a streak of releasing 2-3 songs a week

[high sundial]

i don't know but most likely and to be honest every full song i've done has been below 5 minutes is that the cap?

[low cypress]

If your song becomes too long, the server can get a time out.

[cloud python]

Same here. I'm fortunate to have backlogged about 4 songs right now that will be releasing weekly over the next month, but as of right now I'm stuck; I've got a new song written but can't get it produced. I need only 2 more to complete my first "album" of Suno songs (11 tracks total).

[eager shadow]

Hell yeah! You have any singles or EPs released on streaming services yet?

[cloud python]

All finalized songs are available on my YT channel (along with album art). Once the album is finished, I'll be releasing to Spotify, AppleMusic, etc. My channel is the same as my handle here.

[eager shadow]

Its exciting when you first get them on spotify, apple and all the others. I started shipping them off back in October with V1 so you can hear the jumps in quality as you move through my releases

the 2 above release on monday and thats all I got until this is hopefully sorted

[cloud python]

I release every Friday at 12AM (on YT). I'd very much like to check out your other songs on Spotify!

[eager shadow]

JHomeZ

current Discord name was taken lol

[eager shadow]

appreciate you checking them out! I have stuck to a specific style on dark comedy storytelling hip hop that are slightly relatable, the viking song is probably my only one that deviates from that a bit as far as relatable goes

[karmic spade]

I'm trying to get a continuation of one song and every time it just refunds my credits instead of giving me anything. Seems like there are problems now with this public version.

[eager shadow]

is it at a dropoff point in the song you are continuing from?

[karmic spade]

What do you mean exactly?

[eager shadow]

like the audio starts fading out

[karmic spade]

I don't think so. I'm going to listen again though

[karmic spade]

It is just after the lyrics stop and still music is playing but I'm going to try like one second before end because it does sound like it stops before the last second.

[eager shadow]

yeah just try different times, maybe an extra second after the lyrics end

[karmic spade]

Does it being in Spanish have anything to do with any problems?? I was pretty sure Suno is well-versed in different languages

[karmic spade]

Thanks, I don't know why I didn't think about this. But also I am having issues with continuing old V2 songs as well. I mentioned earlier in here. I just thought there was a lot going on with the public version which I'm sure can be expected

[eager shadow]

Did you use the Alpha at all?

[eager shadow]

I wouldn't think so, but Ive only done english, so Im not the best to answer that

[karmic spade]

Yes I did

[twin basin]

Are sudden stops considered bugs?

Like it hard stops at 2:00 but sometimes generates more

[karmic spade]

Well I see in trending there are a bunch of different languages and I have done Spanish before and it does pretty good but I know when it was V2 I could not get it to stay in Spanish it was constantly doing like Spanglish weird crap lol But all good now, it isn't refunding credits on this one song. Thanks 👍

[eager shadow]

No problem!

[spark mesa]

How do you control whether female or male sings on each verse

[low cypress]

Hi aicolorz! When crafting a song in Suno, you can control the gender of the vocalist for each verse by using specific tags. For instance, if you want a female voice for a verse, add a tag like [female] before the verse. Similarly, use [male] for a male voice. Placing these tags right before the verses ensures that Suno generates the song with your desired vocal characteristics. AI can be unpredictable but lots of people had success with this method.

[spark mesa]

Thank you!

[turbid compass]

I refreshed my Create page and it was full of songs i did not make oO

[low cypress]

I think you saw the explore page 😉

[neat sail]

I noticed this one time and I think it was showing the Explore page songs but the web page was half updated so it looked like the Create page, but I didn't screenshot so I wasn't sure and it went away on a refresh.

[turbid compass]

yeah i refreshed aswell. maybe the songs were from the explore page (because they all were set public) but i did not made a screenshot and dit not recall any songs xD

[broken cloak]

the ai can't say 'kda' and it's not really consistent with other spellings.

[tacit geyser]

Now it's more obedient (eg, I said I want the first 2 lines of verses to be acoustic and the next ones to have heavy guitars and it did exactly that; this was impossible with alpha) but not so creative I'd say. Wild guess: maybe they lowered the model temperature?

[broken cloak]

I had the same impression about determinism.

[tired tiger]

Are these actually intentionally coded (and what kind of tags are there), or just by the nature of training the ai on lyrics which occasionally have some tags with information in them?

[rotund tree]

Can you use multiple tags? And if so, how would I go about doing that?

[Male] [Guitar riff]
Oooooooo yeahhhh

[Male, Guitar riff]
Oooooooo yeahhhh

[drowsy flame]

"kalimba" is an artist name.. so I can't use it in my prompt lol.

[rare path]

you can spell it with alternate phonetics or do things like Cal Limba if you just want the word in there.

[drowsy flame]

It’s not for lyrics tho! It’s for song descriptors.

[hushed cobalt]

i dont creat new part 2 song. can you hlep me fix it ?

[marsh sleet]

I spent like 10 generation attempts just doing like END OF SONG, SILENCE, OUTRO, NOTHING, STOPPPPPP PLAYING MUSIC & then I realized something, which may be useful to others: I simply switched the continuation to v2, & it immediately respected the request to stop the song.

[gleaming mango]

@marsh sleet 🫡

[cloud python]

Has it improved since earlier this evening? I wanted to check here before I burned another 50 credits

[cloud python]

I've noticed that ever since V3 released, both V2 and V3 are completely ignoring [intro] tags. I spent so much time last night frustratingly trying to get a decent sounding song that I never got to an "end."

[marsh sleet]

i have never gotten an intro to work

[cloud python]

It worked fine with V3 Alpha, like in this one:
https://app.suno.ai/song/9c534735-ba58-4aee-be5d-128727436c40

[cloud python]

This thing still isn't fixed. @low cypress
I challenge you to match the gibberish in this song with the lyrics I wrote. 🤣
https://app.suno.ai/song/4b64fbc7-f8e0-4d39-a56d-641337723101

[cloud python]

My last attempt (and post) for tonight, but yet another example of how V3 refuses to generate vocals in any other form other than multi-layered "choral" sounds, with the vocals being extremely muffled and/or "feathered" so that nothing can be clearly heard. I didn't have this problem with V2, and only occasionally had it with V3 Alpha.

I'm genuinely puzzled as to what happened with the new version; when I first starting using Suno a few weeks ago, it took no more than 3 attempts in V2 to get a satisfactory result. With V3A, it might take 8 to 10. With V3, it NEVER delivers one. I've burned through over 300 tokens today, so guess I'm going to put things on hold for a couple of weeks and hope Suno has this sorted by that time. But IMO, V3 is vastly inferior to V2.

https://app.suno.ai/song/155bfd0e-e9b4-4c02-a82a-b54801b73f67

[wild chasm]

The continue feature is completely broken right now... if you ask to continue from let's say 10:20 on a song that is 11:00 it will continue from 10:20 and merge it to 11:00 making it 12:00 instead of 11:20.. it does not flow... wasted ton of credit and also wasted good generation thinking it was on my end...

[austere fiber]

still my thrash is nowhere to be seen, what happen to the trash?

[past bone]

I have a little problem. I generated full song (v3) from 2 parts but later I decided to rename it. I renamed both parts and regenerated final song but when I export video it still shows old name. Also there is no lyrics at all. I wish to see an option to add lyrics and slides of images for each part as it had worked before.

[cold ginkgo]

The problem is still not resolved. When I generate songs from "Continue from this song", two songs that are exactly the same as the original song will be generated, and only credits will be consumed. What should I do?

[honest wave]

V3 feels like a massive downgrade from V3 Alpha to me:
tracks are noisy, repetitive, inconsistent, and overall low quality.

Not sure how that could happen.

I can't get anything good out of it while I was very pleased with the Alpha.

V3 sounds like V3 Alpha, but drunk

[limber prawn]

The 3rd version has clamped vocals. The voice also misses the accents - the text is pronounced incorrectly. There is no such thing in the 2nd version. And the sound is much better there, and the vocals are sonorous and get into the words

[hushed cobalt]

i dont creat new part 2.

[deft tapir]

How it sounded in alpha --> after subscribing and getting the V3....

[neat bloom]

Guys, can you help me figure out how to control music?
I try to give commands for music like [build up music here] or [Drums starts playing here] and it works but it breaks the lyrics of the song a lot, some words get stuck and are repeated many times or even starts singing random words, but sometimes I succeed control the music this way.
Are there any commands that will help me control the music in a song?
The neural network loves to make monotonous music and it depresses me

[low cypress]

@cloud python @vernal geode @honest wave @limber prawn @deft tapir please use the poll to vote about V3 https://discord.com/channels/1069381916492562582/1163842729231528047

[low cypress]

Are those numbers real? Because the Devs advice you to keep a song under 5 minutes to avoid time out when you try to make a full song.

[low cypress]

In the library in the recht upper corner.

[low cypress]

You can also rename the full song. The layout is having an overhaul. They expect lyrics to come back later.

[low cypress]

Very rare case.. I would paste the songs here for the devs to see so they can investigate.

[low cypress]

Ussualy this means the song you are trying to continue from is either fading out or it has silence... Try to use the timestamp to an earlier part and then continue.

[crisp plover]

Are there issues with continuing currently? I can no longer continue songs, it is gnerating completely new songs.

[low cypress]

It worsk fine for me.. make sure all fields are filled in.. or simple refresh your page

[crisp plover]

I have. Even tried in a new browser. The 'continue' button is greyed out when I select continue from this song, and when I fill in the fields, it is generating completely new songs unrelated to what I am continuing.

[low cypress]

Is it an old song?

[crisp plover]

Is 3 days old?

[low cypress]

Do you use default or custom mode?

[crisp plover]

It won't let me continue with defaults.

It did 3 days ago just fine.

[low cypress]

It always goes to costum mode.. there you just fill in the things you want and continue

[crisp plover]

And when I do that, it generates new songs, not continuations.

[low cypress]

Not if you are in costum mode.. don't flip it back to default mode

[crisp plover]

I have tried that.

[low cypress]

It works for me

[crisp plover]

I touch nothing. New song.

I just had it go from aggressive metal to upbeat jazz.

In no way can that even be contrued as a continuation. :

3 days it continued just fine, but now it won't at all. 😦

[low cypress]

I see the problem. I will let the devs know.. the continue button shows while it's "Create".

Here is what you do. Refresh the page.. then you see "create" instead of "continue"..
Then "create" a song.. Choose continue from this song, and DO NOT go back to simple mode.. continues always go to "custom mode"
There you fill in your things like this:

[crisp plover]

This is what I am doing.

[low cypress]

I am testing it myself right now again.. but I think you get confused because of the button not changing back to create

[crisp plover]

So I click this

And it leads to this

the continue button is greyed out until I fill in the fields. When I fill in the fields, despite saying continue, it creaters a new song.

[low cypress]

As I said it works for me

[low cypress]

you need to fill in the fields as I told you

[crisp plover]

As I just said #🪲┃bugs-web message when I fill in the fields it creates a new song.

[low cypress]

Make a video.. Like mine.. show me

[crisp plover]

I don't know how to do that

[low cypress]

then make screenshots step by step

[rotund tree]

Does it create an entirely new song, or just a new clip?

You have to click “Get whole song” in order to hear what your previous generations sound like leading up to your newly continued generation.

If you have iPhone, there is a screen record option in the quick settings drop down menu for iPhone.

[crisp plover]

I am on my PC. So I'm trying to filter through the differences between my browser experience and what is displayed in the video from @low cypress . I don't see any 'get whole song' options. All I know is that 3 days ago, continue worked fine, and doing the same things now it doesn't.

The screen clips I showed above are what appear. Clicking continue from this song here:

leads me to this, where continue is greyed out. a few days ago it was not greyed out.

when i fill out the forms, thusly

it generates a new song.

[low cypress]

making a quick video from desktop atm

[rotund tree]

What’s your goal, and how do you expect the model to behave “IF” you were able to press continue without anything filled out? @crisp plover

[crisp plover]

I expect it to continue with a similar cadencce and vocal, as it did for me 3 days ago.

[crisp plover]

I don't expect exactly the same, but right now it is giving me wildly different results.

[low cypress]

See if you do the same as I do @crisp plover

At the end you see that my songs are "part 2"

[crisp plover]

Yes this is exactly what I am doing.

It does say part 2, but it is completely new, with no continuation in similar sounds.

[humble cobalt]

So the issue is that the continued song just doesn't sound like the previous part?

[low cypress]

Make a full song of it and post it here

[crisp plover]

Yes. 3 days ago when I was working, the continued songs were similar enough that it actually sounded like the same song when played back to back. Now, the vocals are wildly different as is the musical cadence.

[low cypress]

Please post a full song

And next time be a bit more clear.. lol.. because you kept saying it doesn't work.. apparently it does but you don't agree with the outcome 😂

[crisp plover]

i am not sure how to do that. Do I link to part 1 and 2?

[low cypress]

3 dots menu

Choose make whole song (of a part 2)

[crisp plover]

oh, gotcha. i was trying from part 1. one sec.

[low cypress]

Then listen to the outcome.. it might suprise you

[crisp plover]

sorry I thought I was being clear when I said my expectations were a similar cadence not wildly different

will post once the spinnies stop

[low cypress]

You started with this first ^^

See my confusion? 😉

[crisp plover]

Ah I can see that. Ok, here is a link. https://app.suno.ai/song/447c159f-26c3-452b-9ee7-ce67b15c3de8

[crisp plover]

apologies for lack of clarity

[low cypress]

What is the problem with this?

[crisp plover]

at time stamp 1:54 it completely changes.

It didn't use to change wildly like that.

Maybe I just had what I consider to be good RNG?

[low cypress]

Okay.. here is the thing.. you continued from a point you are somewhat fading out.. and that can evoke a slight change in song.. You can use the timestamp to go back a few seconds and continue from there.. plus make sure you always fill in the same style of music

[crisp plover]

Ooh, that might explain everything. Thus far all my continuations have been from songs that completely cut off at the end.

[low cypress]

here a video that will help https://youtu.be/5uJw5kb05as?si=gGiiztZWEX9rFmV-

[flint vapor]

why cannot i change my subscription plan?

[low cypress]

Why do you think you cannot?

[flint vapor]

i tried, the website tells me that "we were unable to charge your saved paymant information"

[low cypress]

It's sounds more of a problem with your payment method. Have you checked if there is enough credit or if you filled in the right things?

[proven socket]

is it a bug when you make a song public and it doesn't show in the "New" section? I wasn't sure if this is due to having a song in trending recently, or something else? Since trending is only like 50 people (out of the entire userbase) I don't think it would make sense if it was because of this

[crisp plover]

My assumption has been that there are more people making songs than display, thus not all songs will display in new.

[proven socket]

ahhh i see, i wonder if they could distribute what is shown across the userbase?

[crisp plover]

LIke only 50 songsd display, so if 100 people make songs at the same time, as I believe is the case as this is an awesome thing, then most people are not going to seee their songs in new.

[proven socket]

that does make sense. I wonder if they showed different groups of people a different set of 50. I don't think they're doing something like that already (because when my song hasn't been in New I haven't see the "like" increase)

[crisp plover]

I am saving my pennies so I can subscribe for my birthday next month. This is awesome, and thus far in here whenever I have quesitons/issues I've been helped immediately.

[low cypress]

You have to set it public.. and it will briefly appear in NEW.. but if nobody votes it dissapears quickly

So spread your url.. get votes

[proven socket]

i was watching the page like a hawk, its either a bug or there is a reason

[low cypress]

Did you set it public??

[crisp plover]

you are my new hero. This fade was indeed the issue.

[proven socket]

yes! i had other songs get into the trending multiple times, and i know the song hasnt been shown in New since the "likes" didnt increase (and I was also checking the page to watch as new songs come in)

it's definitely either a bug or an intentional thing

like i know its possible for the song "like" count not to increase, but im watching the page, so i can see when the new tracks come in

like i'm refreshing it

[low cypress]

Mine does show.. you need to get some love to stay there and not disappear quickly.. did you post your song also in #🐣┃suno-showcase ? It helps.. but to be fair there is one thing more.. the algorithm does some quality checks.. It has to be fitting overal in the list

[proven socket]

i actually do think this is a bug, 98% sure. Other people have mentioned this same problem too. I know they can disappear quickly, but when I mean watching the page, I mean I'm watching every track that comes in 😛

it's possible that the algorithm maybe flags the track, that is a possibility, but if its not that, its definitely a bug

[low cypress]

I am sure the Devs will look into it to be sure 👍

[proven socket]

it's this track here for example:
https://app.suno.ai/song/92c8dd6d-62a2-4fba-9cf9-bc0f55271dd8/

(it has 3 likes, and wouldve had 2 likes before even being public)

[proven socket]

ahhh its most likely that then lol. Because I have noticed it happen maybe 1/7 or so tracks
even though there's nothing crude here, its probably being flagged for whatever reason

[low cypress]

yeah.. but AI has its own mind sometimes.. or just being stupid lol

I would have picked your song for sure!

[unique maple]

I see a major bug in continuation after V3. All of my songs were mostly 3-8 blank clips with almost all time automatically refunding credits. I've reported every time as bug but it's tiring and frustrating that I cannot continue a song and create a longer one. I also see many times it repeating phrases, picking up earlier phrases of the songs. It also ignores the metatags completely. Looks like those metatags only work in discord creations. Any clues on this?

[cold ginkgo]

https://app.suno.ai/song/9af11fb3-e987-4e03-8afd-b78eb0d7f30f

[cold ginkgo]

https://app.suno.ai/song/4b624605-e422-4a74-ae72-6eb821873624
https://app.suno.ai/song/465fce0d-2264-4989-83fc-142849c6937f

[boreal isle]

can't login on the site

[zinc talon]

eeeeeverything gets a choir now. solo singers are banned from Suno lol

[boreal isle]

ok nevermind, it works now

[fluid wigeon]

Is it normal that V3 can no longer handle normal lyrics? A lyrics created by ChatGPT constantly makes for terrible songs that completely replace the prompt and also the lyrics. Either words are changed, left out completely, or it's just a completely different genre and a very bad voice. How can that be? It's a completely harmless and normal lyrics. V3 doesn't seem to be as good as it's said, i guess it's even worse in understanding things...

[low cypress]

A good way to let the Devs know is to vote here: https://discord.com/channels/1069381916492562582/1163842729231528047

[lusty pollen]

i made some experiments... when i click on this song in my create tab (not library) , the alignment of the elements of the website all get messed up https://app.suno.ai/song/e7014372-a3ca-4220-afd1-7cc5c924a335

[tacit geyser]

Some people's lyrics get flagged with no reason and for me, I put 'textbooks' in the lyrics and I leave to your imagination what the bot actually sang 😂 Finished the song without noticing, now I'll have to put an explicit tag if I ever upload it anywhere 😂

[low cypress]

The lyrics are being automatically moderated by ChatGPT. Is this perfect? No far from. What I ussualy do is copy and paste my lyrics in chatGPT and asked why my text is flagged for moderation. It will tell you then mostly what the problem is and you can adjust your lyrics on it's advice. You can even let it write it for you.

[tacit geyser]

They weren't moderated, the bot actually sang the wrong words 😂 Let me find the link.

https://app.suno.ai/song/66103802-f8aa-4d8e-82b8-2eca50ea56ad

Or maybe my hearing is bad, but I don't think so.

[low cypress]

😮 🤨

It's all in the ear of the beholders.. let's leave it at that 😇

[sacred gulch]

When making a song I don’t see the lyrics like other songs only the name of the song. Is it a bug

[hushed scarab]

可以用抖音的

[lusty pollen]

i had it hallucinating swear words too. seems to be normal to sometimes have heavy swear words hallucinated. at least it didnt seem to get interest . the word you hear in the beginning is not in the text... there are in fact not a single swear word in this text https://app.suno.ai/song/f55d2985-931e-43ac-9f92-b92c5da189b7

[vale crypt]

Was it marked as a “full song”

[wild chasm]

No, those time are off, my song is now 26min (It's a prog song, one song album) and so far I had no issue, only last night I had issue, and it's a bug. I haven't tried today yet.

[low cypress]

Alright.. but be adviced that it is NOT adviced lol..
If you get stuck you will have to finish it by downloading the parts one by one and sticht them together in a DAW..
Tip: make a full song in between since you can still continue from the full song 😉

[wild chasm]

I understand your point, but, if I go back to a point (I keep all my song point) that used to work, it does not work, it does not stitch at the good part, and this is a new behaviour, so, it was working, and now it does not, hence the assumption that this is a new bug that has started yesterday. I know it's not adviced, but, it used to work.

[low cypress]

If one point doesn't work, but others still do.. it's solely that part.. and could be anything. So it would be wise then to share the specific part what is going you a problem so the Devs can look into that.. Also.. it helps sometimes to go a back just a few seconds (change the times stamp) to continue from. That helps immensely..

[wild chasm]

What I meant is that I used a earlier point and it's subsequent part that I've already stitch together and that worked. And now, same song, same part, and it stitch at the end instead of the timestamp I asked to stitch to.

[low cypress]

Are you sure you are using the right time stamp

[wild chasm]

Yep, same timestamp, it's hard coded in the "Continued" file, so, I can't be wrong on that.

[low cypress]

Would you mind sharing the problem (url) so I can wrap my head around it?

[wild chasm]

The url of the full song or just the "continued" one ?

[low cypress]

The part you are working with, and is not behaving as you expect

[wild chasm]

This is the full song:
https://app.suno.ai/song/2d4a9162-3f12-4ddd-9c8e-450a6c4b2086
Continued part:
https://app.suno.ai/song/aeceebc2-5135-4baf-b16e-20e971ccee59

[low cypress]

So the problem is the continue point? You want to add a new part, but is not doing so? I am really trying to understand 😄

[wild chasm]

lol, it's ok... the continue point should stich at 23:00 and it stich at 25:10...

[low cypress]

okay.. so you are stitching from the full song at 23:00.. got it.. Let me try

[wild chasm]

yep 🙂

[low cypress]

Now that is a pretty long cut back.. that is my first thought

[wild chasm]

Ok, I'll try to take an earlier point, and see how it behave

[low cypress]

just a thought.. waiting for my test 🙂

never made a full song of a long song like this.. 😬 🤞 😵

But it seems you are right.. It does seem to skip...
Did you make more songs that where longer than this?

[wild chasm]

The higher I went is 27min

I just went back 2 point before and made a new continue and it stitched well, so I guess I will just lose the last 2 min and continue from there...

Your theory about too long of a cut back might be right..

[low cypress]

My personal logical thought here, is that you are going back too far. When you continue you are going for max 1 minute. But you want to go back more than 2 minutes.

But I just made a notification for the Devs to take a look at it. 👍

[wild chasm]

Thanks for your help ! You're awesome 🙂

[past bone]

I renamed the full song but video is probably cached and name is old on video.

[low cypress]

Ah.. gotcha!

[cloud python]

I did, but the poll doesn't seem to be reflecting the complaints flooding this channel. 🤷‍♂️

[mystic spear]

I know a lot of people have had complaints about the new release. I just have another to add with examples. I just can't get any song in V3 without a bunch of extra noise, muffles, unneeded harmonies/added vocal layers. I haven't been able to create one useable song since the V3 release. Here is a kpop song V3 Alpha https://app.suno.ai/song/be9e6440-ae21-4ae4-b480-66e2216414d0/ vs. V3 https://app.suno.ai/song/b6d67270-cd71-445d-be92-55cab54f6c36/

[low cypress]

Well the point of the poll is to be heard too though.. still a lot of people do agree with you.

It's not all sunshine 😅

[cloud python]

Do the devs have any theories yet on what may have gone wrong with V3? Why it's sound quality is worse than V2 or V3Alpha?

[eager shadow]

I did find somewhat of a fix last night thanks to another user saying she put the word unprocessed in her style and then when she masters it in her app the vocals sounds perfect, so I tried the opposite and put the word processed all over the place lol and it came out muuuch better: #📚┃suno-school message

Granted it’s not the solution nor does it ALWAYS work but it does help

https://app.suno.ai/song/23a60f1e-478c-4a8d-9ceb-f63903692c12 the track I’ve been working on came out like this which is the most clean I could get after about 30 attempts

[low cypress]

Well.. They don't share everything with us.. And there were also people who indeed found it better.. So my personal thought is that it is at least genre related. Some genres might got better, some not...

[mystic spear]

I wonder though if it was just people answering after testing it for one day.

[low cypress]

I keep pushing it.. to help to get a fair response

So doing my best to get a clear view overal. But tbh.. I notice difference in genres.. some are better, some ar not like I said. I play around a lot with it, so it is not just guess work.

[mystic spear]

well I have tried a lot of different genres, 1960s, broadway, reggae, pop, kpop, gospel, soul and they are all worse. Everyday I log on and try a little to see if something has been fixed but I guess I probably should just wait 30 days or so to see if things change.

[low cypress]

The model is as it is.. it will not change in a day, else it would be V3.1 for example. So this is V3. However, I noticed this model is slightly reacting different to things. Sometimes I have to force it a bit more to do a certain genre. I do this by repeating. Further more it is pretty good with instrumental and with minimalistic genres. Baroque is amazing. But also hip hop doesn't sound bad to me.. I am also working with some acid house.. I have to force it, but man it does some amazing things to my ears. But if I go to something that is more layered and in RL has som SFX like reverb... yeah.. not happy with that. So not ideal by far, but they never promised that. It still completely new terrain, where you sometimes have to break a few eggs to make an omelet. And you learn most from things that go wrong. So I have full faith in the Suno team.

[gleaming halo]

I've sent a list of over 20 samples in many genres comparing the same prompts in v3 Alpha and V3 to dev, IMO most problems are caused by overdubbing of voices that was not in V3 Alpha and introducted in V3

[gleaming halo]

They should just keep the previous model active for the (many) people that think its a setback.

[low cypress]

Please let your opinion be heard via the poll: #🗳┃mini-polls message @gleaming halo

[gleaming halo]

I did!

[rotund tree]

Is the Suno Team considering re-releasing V3 Alpha per user requests?

[low cypress]

I don't know. I am a guide (volunteer). I do know they want to offer us the best quality in general. But you have to think logical too. If more people like it and do see it as in improvement, then it is not logical to go back. But learning wise everything, every vote, every opinion is study material.. right?

[round crater]

Is the platform currently down?

[cloud python]

I'd be interested to know what people have indicated as "improved." The poll is at a 3:1 ratio, so I'm wondering if the majority of users are allowing Suno to generate lyrics (and that's getting them better results), and the other 1/3 of us are writing our own lyrics and that's what's tripping it up.

[low cypress]

I have it too.. I think server overload..

[dry shard]

ye...

[cloud python]

Either way, the V3 release is touted as, "radio quality music" and nothing I've got in the past 2 days has been remotely radio quality. V2 was better by a mile.

[low cypress]

Just a volunteer, so I have no idea tbh 🤷‍♀️
Only my personal opinion, and if you read my text a bit up, I see improvements, but also not in some cases 😉

[low cypress]

You can check the explore page.. there are some that are pretty good quality

Really depends on genre in my honest opinion

[cloud python]

The genre I've been working with is "alternative" "grunge" "90s" "angsty." V2 & V3A both nailed these genres, yet V3 doesn't seem to know anything about it. The handful of times it got close, it was a cobbled mess.

[rotund tree]

Yeah I was thinking more how MidJourney for example, gives us access to every version they’ve released, including V1 (even though V1 is awful)

With Suno, we still have access to V2, so why not V3 Alpha? I’d be fine if they called it V3-A and V3-B, but it would nice to have access to all models imo.

[rotund tree]

V3 can be good, it just takes more effort. More re-rolling, more experimentation.

[low cypress]

I have no insight in all.. like I said "I am a volunteer", but work in IT for over 25 years, and a very logical thinker (imho). You have to understand that this is completely new. There is not howto video how to do what these guys do. They are LLM engineers but doing something nobody else has done. And then there is a lot of trial and error. And we are spoiled guys when it comes to images with AI. That is so much more forgiving, even funny if you render pictures with 3 arms, or 5 eyes.. Music is so much more. In genres, in emotion, in languages, in instruments.. It's already amazing what they have achieved so far. I mean look around. Who is doing this besides Suno?

[rotund tree]

I agree! I’m just offering my perspective as a user 🙂

It is quite amazing and I’m sure the team will continue creating more features that we want.

Who else? Mmm not many. OpenAI had JukeBox, but it’s not available to the public, Google made an instrumental model, but Suno is the only consumer friendly model out there rn.

[gleaming halo]

I'm in IT longer and try to provide valuable detailled feedback comparing datasets and versions. Nobody has come close to what they have done and it's great. I just think they lack the resources to have a proper testing procedure with different styles, prompts, etc. With V3A only everyone will be happy again. I saw a dozen or so people here having stopped creating songs waiting for it to improve. And that takes away from the momentum that Suno has.

[low cypress]

it's not there atm because of an overhaul.. will come back soon

[willow pollen]

same with 70s progrock ... and as I said on gneeral : all vocals I'm getting so far (600 credits in!) have been choirs ... None of the clips were actually useable

Typical result : https://app.suno.ai/song/7cc675b3-503a-46e4-99de-a32c07117a8c ... I wasn't asking for a musical by the cast of Hair ...

[rotund tree]

Overhaul…?👀

[low cypress]

design change

[remote belfry]

I'm having a problem when I select a clip to do actions with it like delete it or extend the clip. My screen reader is not reading the menu properly and it is not accurately telling me how many items have been selected. I am using the NVDA screen reader. Any help would be most welcome.

[lucid basin]

hey guys, the ui is kicking my butt right now. All of my inter-tab persistance is gone, and getting reset every time I need to shift between page states. It's really messing with my workflow

[broken cloak]

Interesting. I have been prompting with "polished" but I will give "processed" a try.

[eager shadow]

has polished been working well for you?

[broken cloak]

Yeah, it wasn't dramatic, but it did help and didn't change the vocals or music otherwise like other things tried have.

[eager shadow]

gotcha, this is noticeably better than running it without, at least one out of the 2 songs should come out with much cleaner vocals, as far as detracting from the rest of your prompt, I cant be too sure. Seems to hit the style mark from the one song ive been trying to work on

[neat mango]

please bring back lyrics in videos

[modest iron]

x-post "Can anyone fix the accessibility of the website, I'm using a screen reader on an android device, and I can't play the song I created properly, the play button doesn't actually work. works well, i remember it used to be great in the past"

[loud lotus]

The only reason I can think of, why they would not keep V3 alpha in addition, is that it's heavier on compute than V3 final. The mushy noise could actually come from less iterations/steps...

[tame latch]

Hey thx to you and @low cypress for reporting this. This should be fixed now. Try again. Sry for this bug.

We really appreciate your feedbacks and love to see bugs in this channel are actual bugs.

Bugs are most likely to be reproducible or repeated in some patterns. If you run into such a case, please send along the bug and the instructions on how to reproduce them. They can be UI bugs, inference bugs, latency bugs, but reproducibility is very important for us to move forward with it.

We love your supports and feedbacks, and would love to give you extra credits for helping us improve! 🙏

[wild chasm]

@tame latch Does it mean I'll have extra credit ? That would be awesome, I'm on the 30$ plan but I'm able to use 10000 credit quite easily 😛

[tame latch]

see your dm

[cold ginkgo]

These are songs generated from "continue from this song" with "custom mode" and "instrumental" checked. It's too short, it fades out, and it's too similar.
https://app.suno.ai/song/c7d61441-2a29-424e-bd42-4ccdcca029ec
https://app.suno.ai/song/566fb8da-64aa-46ad-a2e7-76e88721da12

[willow pollen]

@tame latch do you consider a noticeable degradation of the quality a bug?

the problem being : it's hard to give you replicable details

[true nexus]

i noticed that too lately, once you continue with the first part, the second part it's lower quality, like distorted

[proven socket]

hey @tame latch, since you are here, I have a bug from a day or so ago. I didn't report this before because I forgot about it until I saw your post!

If you take a look at this Part 2 track here, it won't let you "Get Whole Song". When you click this, it'll give an error: "An error occured generating the clip".
https://app.suno.ai/song/f46a11d0-054c-4fb0-9729-b9b3d7c0fe93/

I can kind of get a feeling why this may be happening as the Part 2 track kind of "starts" as if its the beginning of a track. It's like it's not actually continuing

edit: Also this bug persists if you do further continuations. If you make a Part 3, and click "Get Whole Song", you'll also get the same error. It's as if the connection between the first and second part are corrupted and it then corrupts any further parts. Now that i think about it, this has probably happened to me about 3 times before this.

[neat sail]

Not tony, but could go either way I think? If you have a prompt or workflow that degrades especially badly, if not every time, but often enough that it really stands out in 10 generations compared to 10 gens of a typical prompt, that is possibly a bug, or maybe just an AI model quirk, depends on the cause I guess. If it's more general than that it's definiately model feedback. Maybe a new Feedback channel could be created for model-feedback so there's a place for that more general feedback as well, separate from more straightforward bugs.

[willow pollen]

that's not what I mean : There was a change between v3-alpha and v3, that makes it nearly impossible for me to get anything but choirs when generating rock songs. check #🪲┃bugs-web message for the typical vocals I'm getting now, compared to, for example, https://app.suno.ai/song/d130654e-38e0-4468-ab00-46e49edfd6e3 which was V3-alpha

[willow pollen]

it's really constant. I spend around 6-700 credits yesterday, and every single generation resulted in choirs/choruses, no matter how I described the vocals

[true nexus]

oh, that's another thing, i see

[willow pollen]

I just used the exact same prompt as with The Last 8-Ball up (the one I just linked to ) ... and those were the two generations : https://app.suno.ai/song/7e1c6601-8196-41fe-9ab0-d66cba1e04fb https://app.suno.ai/song/5acbb99b-01c2-4bce-9956-82aac9dc2956 ... you'll notice that the vocals are REALLY bad ...

and have some weird chorus effect on them

[proven socket]

ahh this bug/quirk is actually repeatable!!! i did four generations and they all start with the singer shouting "two step" 😂
I tested removing the backticks at the start and end, as I thought that was causing it... but it still occurs even when you remove that lol

edit: even if you change it to "four step" it will say that on some generations
edit 2: you can even get it to say "banana" lol:
https://app.suno.ai/song/10c99889-4d1f-4664-ae70-ecc9338d921b?

[willow pollen]

and just as a test, I generated 2 more : https://app.suno.ai/song/3af3e6a4-1223-4e2b-ba41-fd2656ee670f https://app.suno.ai/song/65c019e6-a327-44ea-b394-ee80a31dace6 ... same results. Not useable

[cloud python]

I'm having this same problem with alternative/grunge/90s genre songs. It's almost like the app is doing everything as viking choruses.

[willow pollen]

just "progrock male voice" starts as a single voice .. .then off we go in full school musical gleeclub mode ... https://app.suno.ai/song/39806758-1354-4066-a368-fa53d4f89a69 https://app.suno.ai/song/4dd9d5ff-1d56-4c57-a66e-07ac8c3da160

horrifyingly I'm actually paying for those generations

@proven socket sorry ... didn't mean to reply to you 🙂

[cloud python]

I'd be interested to know what V3 was trained on, because it undoubtedly got over-fed multi-layered/choral samples because even the top tracks under Explore are suffering from this problem, and it wasn't an issue in V2 or V3A.

[willow pollen]

well ... in my case it's not viking chorus sadly ... it's '17 year old private school bro and his choir'

[cloud python]

I suppose I should clarify: it's as if viking warrior choruses is what it got trained on.

[willow pollen]

so many emojis to choose from, none that convey 'viking'

[neat sail]

I experimented a bit trying to dampen the chorus mixing in a V2 clip and then breaking into smaller pieces (can see when prompt changes in video). Sort of helps, with patience and retries. The second segment was V2, not sure if V2 specifically actually helped though, just an idea testing https://app.suno.ai/song/7f7eb36e-404d-4e7b-862c-31040d1ad7a4/

[willow pollen]

yeah ... that's nice ... but considering it really worked OotB in V3a it's a bit maddening (as a subscriber) ...

it still sounds like it's on the verge of going full chorus at 0:40

[dense beacon]

sorry but i'm using suno-V3,it goes smoothly firstly, but then suddenly I found that I can't "create" new songs, refresh does no help, can anyone tell me why or how to work around? thanks

[fossil pivot]

You're out of credits. Not a bug.

[dense beacon]

...ok well, a hint to earn credits?publish your song or cash? or any link for earning credits would be appreciated

[fossil pivot]

Wait for a day for 50 credits which allows up to 10 generations. Alternatively, subscribe to a subscription. You could disable re-accruing payment and it won't get rid of your purchased credits.

[dense beacon]

got, thank you so much

[dusky tide]

hi, the pronounciation for 'marathi' language based songs is terribly bad. it needs fixing.

[neat sail]

Now that's my kind of prompt, A+. 🥇

[fleet pike]

Sometimes the Continuation feature on the website does not work, it seems that after a certain amount of time spend on the website it only creates 3 second empty tracks as continuation, which is a bummer because it costs credits, feel that this is not right, I pay for pro after all and I lost a good amount of credits because of that.

[humble cobalt]

If it's giving you 3 second empty tracks, it should be refunding the credits for those tracks. Anything under 8 seconds gets auto refunded.

Try specifying a different time to continue from. Sometimes that helps.

[fleet pike]

ok that might be possible I do sometimes have 10 seconds empty tracks as well I will try your suggestion! THanks!

[humble cobalt]

Yeah it's really annoying when it generates slightly more than 8 seconds of unusable song.

[fleet pike]

your tip worked! thanks so much was so frustrating!

[neat sail]

Will pin this (not complete) bug list.

Known Bugs (incomplete)

Note: Bugs are not model feedback. #🪲┃bugs-web message

Model-feedback poll #🗳┃mini-polls message

Bugs:

1. 'Remix' on a Full Song can error. Remix copies all lyrics from multi-clip song and can overfill max input character limit.
2. Creepy prompt breaks Create page. Source: #🪲┃bugs-web message
3. Resizing desktop browser on Explore page crashes website. Reproduce: make Explore page narrow and then wide again. Source: #🪲┃bugs-web message
4. Delete a track in library, can't navigate to next page until refresh. Source: #🪲┃bugs-web message
5. (In Simple Mode) Enter doesn't create new lines but generates immediately. Confusing design, Song Description field looks like multiline input. Source: #🪲┃bugs-web message

Accessibility Bugs:

1. Android screen reader, play button doesn't work. Source: #🪲┃bugs-web message
2. NVDA screen reader: select a clip to do actions, screen reader is not reading the menu or reporting how many items have been selected. Source: #🪲┃bugs-web message

Bugs To Check:

1. Inter-tab persistence is unexpectedly reset. Source: #🪲┃bugs-web message
2. Rename doesn't work on Android Chrome. Source: #🪲┃bugs-web message
3. Browser cache/refresh issues: #🪲┃bugs-web message and #🪲┃bugs-web message
4. Various continue errors on select V2 -> V3 songs.
5. "Get Whole Song" fails on long songs. Length limit is known limitation, but not sure exact limit at this time.
6. Repeatedly clicking "create" without editing prompt degrades quality. (seems unlikely but bugs can be strange) Source: #🪲┃bugs-web message

Not Bugs:

1. "The song ignored lyrics, sounds bad, etc." This list is not tracking AI model issues. (LYRICS ARE BACK 04/03/24)
2. "Lyrics missing from videos." Temporary, lyrics will return later.
3. "When generating in Simple mode both clips use same lyrics." Intentional change: #💡┃suggestions-web message
4. "Couldn't generate that." Prompt blocked. Incorrectly blocked prompts tracked elsewhere.

[proven socket]

@neat sail, see this here:
#🪲┃bugs-web message

You can repeat it directly from that track. Using "Get Whole Song" will always error out. It seems that Suno can sometimes create a continued Part that starts a new track and therefore when you try to "Get Whole Song" it's unable to join the parts (since they don't connect).

I think this has occured to me 3-4 times so far (I think mainly on v3, but it may've occured on alpha)

Also, when it occurs it will affect both the two generated tracks you get. Both of these you will be unable to "Get Whole Song", and if you continue those, and try to "Get Whole Song" on a later part, this will also error out

Edit: And just for extra info, this is occuring on a song that only has 2 parts so it's not a long song

[proven socket]

@neat sail, don't mean to keep tagging you (and I know I mentioned this earlier in this channel):

None of my new tracks ever go into the "New" category now haha. It's probably because of the AI filtering perhaps (as mentioned above), but I don't think I've had a track go through for a while now.

If there is some kind of AI filtering, it might be worth looking at the way it works:

These are 3 recent tracks that didn't go through:

• https://app.suno.ai/song/92c8dd6d-62a2-4fba-9cf9-bc0f55271dd8/
• https://app.suno.ai/song/ec97ef5b-49f7-437a-839c-3bf4391fd281/
• https://app.suno.ai/song/50d3ceab-5c1f-40fa-99b7-0b7c627fadce/

When I made these tracks public I've watched the page, and they never show up (I'm refreshing the page constantly to be sure). And even if I somehow missed it, everytime I've had a song on the "New" page before the "like" count has increased, so the combination of not seeing them and the "like" count staying the same makes me fairly sure they aren't ending up on there.

When the "New" page was first added I never noticed this, then later on I noticed it occasionally, now it's occuring on every track I set to "public" 🤣

Also I know other people have reported similar experiences to this.

[neat sail]

Anyone can tag me on bugs. I may not update the list when I'm busy but I'll go through the mentions eventually. I don't know much about the New page but I wonder if it's just there are SO MANY SONGS it can't possible show every public one now. (No idea, just speculating.)

[proven socket]

Well what made me say this is I'll see other tracks of the same name where the person has uploaded 4 different versions and made them public, and you see them all appear in a row!! 😆

If I look right now, I can see this for example:

Othertimes I've seen 4 in a row

So if they weren't showing all songs, I think those would be at least filtered out 😛

That's why I thought it was worth at least checking with them

[proven socket]

Oh yea, and the two tracks right before the two ones I circled are also two different versions 😆

[lusty pollen]

I also have issues with tracks not showing on the new tab. Over the weekend i made some tracks where i understand that they might not show up. but i have 1 song where i really dont know why it didnt show up ... this : https://app.suno.ai/song/2a90d8e7-d642-4e39-a3c8-8ff25d88d4d3 maybe someone can explain why... no swear words, no balming, no shamin, just a fun song about dancing and 8-bit sound.

[craggy charm]

What's with V3 creating songs with about ten singers? Almost all creations sound like a choir, even a prompt like "single vocal male" fails to stop it.

[lusty pollen]

just to point out another example. this song also didnt go to the new page... here i might get why it didnt show up. If you dont know the game parappa the rapper, you might not get the sense of the song. Its a rapping game where you hear an instructor and follow the beat. Here the AI might have thought i used a specific voice and made fun of it, wich isnt the case at all. its a reimagination of a level in a game. no bad words... The only word i could see is "flexing".. but i had a song in the explore page wich "flexing" said a lot ... so its not "flexing" https://app.suno.ai/song/d4c514a7-9974-4989-ba42-2100309b412d

[proven socket]

@neat sail in regards to "Get Whole Song" not working on certain Parts. I believe it occurs when you try to continue a song from like 0:02 seconds or something. Because I'm getting it again now. And that example I posted before would've also been continuing from only like 4-5 or so seconds in (or maybe it was even less)

[proven socket]

i remember playing this on PS1 I think haha 🤣

[green mantle]

Sign up unsuccessful due to failed security validations. Please refresh the page to try again or reach out to support for more assistance. Please help

[gleaming halo]

From this post on V3 Alpha, the following issues are stil actual:
#🪲┃bugs-web message

• 1000 Character lyrics limit in custom mode, now 1200 still too short for 2 mins fast songs, causes songs to fall short of 2 mins.
• Regenerate lyrics no effect. Reproduce: Hit Generate Lyrics. Hit generate lyrics. Lyrics remain the same
• Scrollbar when playing. Still stops at one minute even if the song is two.
• Drop down menu at song looks like lower resolution, faded.

[fair scaffold]

This one sounding etremely noisey - https://app.suno.ai/song/691b023b-7837-41f4-8213-054d728a869c

[proven socket]

@tame latch
Regarding what I posted here:
#🪲┃bugs-web message

The short version is I believe there is a bug that sometimes occur when continuing a song from an early timepoint say 0:02, and when it generates the 2nd part (or any future part) you won't be able to use "Get Whole Song" anymore - it will give you an error (so the 2nd part becomes useless even if you like it).

It's probably happened to me about 5 or so times. Often I can change it a second later say 0:03 and it works fine then

[tame latch]

I don't understand how to reproduce this. Sorry. I will need better instructions to reproduce...

[proven socket]

aren't all my original posts detailed enough??

• #🪲┃bugs-web message
• #🪲┃bugs-web message

edit: just go to the song here:
https://app.suno.ai/song/f46a11d0-054c-4fb0-9729-b9b3d7c0fe93/

if you select "Get Whole Song" you will get an error

[low cypress]

Yeah just tried it too.. gives me an error too

[proven socket]

i can give you other examples too, i dont have them all marked, but its happened to me around 5 times now

[tame latch]

This clip is a continue? And it is 2 mins? O.o

[proven socket]

haha well its definitely started a new track that is for sure, but its not supposed to have 😛

i can reproduce this by continuing a track from about 0:02 seconds

[proven socket]

also if it wasn't a "continue" you wouldnt have the option "Get Whole Song"

[tame latch]

k I understand this is bugged now.

Not obvious. Thx for reporting. let me think about it.

https://tenor.com/view/oh-no-homer-simpsons-hide-disappear-gif-16799752

[old mural]

Please, I beg of you, no more Neon Lights and Tokyo Streets when doing J-Pop prompts. Add a bit more diversity to it.

[verbal glade]

not just j-pop. I've found a consistent reoccurrence of the mention of "neon lights" in other genres. It also starts a lot of songs with "in a world where"

I started using chat gpt for lyrics

[mystic spear]

I get that all the time when I ask for kpop. I've started using Claude to generate lyrics.

[verbal glade]

Suno (both v2 and v3) used to be really good about having the chorus be the same style anytime it sees the chorus and lyrics anywhere in the song but lately its started making the chorus completely different. It seems to have started within the last few days (the switchfrom v3 alpha to v3).

[astral adder]

Used to generate some amazing heavy metal songs with V3 alpha and now getting terrible results with V3. Biggest hurdle is vocals that sound clogged and just not the same level of clarity. I would pay more credits just to use V3 alpha again - until then our use-case is blocked and this sub is wasted money

[mortal tide]

Agree, alpha was great and this current v3 is pure shit

[grave folio]

Hello Team. Please fix the link https://app.suno.ai/song/33cb2fca-85d7-46fe-a041-8a0a2847000d it suddenly stopped working

[solar grove]

I encourage anyone seeking a refund to email [e-mail]

(Personally I think v3 sounds great!)

https://app.suno.ai/song/b34d54ae-640b-4205-af3d-cee21c9f7ea8

Literally just made this, this morning.

[willow pollen]

as I said earlier : the problem seems to be specific to Rock and its subgenre.

everything else sounds better, especially the vocal quality. As soon as you generate rock'y tracks you end up in Glee-Social Club Hell

[strange fjord]

Everything thats more complex. Jazz e.g. was awesome With Alpha and is horrible With V3

[solar grove]

Let's get specific. By rock you mean...what specifically. Give me band names.

(Or some song titles.)

[willow pollen]

just write 'rock, male singer' with ANY lyrics ...

[solar grove]

lulz

There's your problem right there.

[willow pollen]

nah ... that was just the shortest style prompt I could think of right now, that shows you the problem

it really doesn't matter whether you write 'ffs just give me a single voice you stupid PoS" or are vague ... you'll ALWAYS end up with choir and chorus gallore

rock, progrock, metal, 'tribal EDM with rock influence' ... doesn't matter

#🪲┃bugs-web message

1221713216065896490

[gleaming halo]

I have PMmed a list of 20 styles or so comparing styles and prompts from v3 Alpha to V3, which clearly show the problem. A lot of Rock in there. There is more than a dozen people in here saying they can not use v3 anymore.

[willow pollen]

sorry .. .that 's just the id #🪲┃bugs-web message

just scroll up or try it yourself.

[vestal thistle]

Used to make really good rock/metal songs with the alpha, now the sound is all cloggy and the quality is just not there.

[willow pollen]

exactly.

[steady agate]

I agree Dame for me

[willow pollen]

was going to say that it's a V3 problem. V3-alpha worked very well for rock

[steady agate]

Same*

[vestal thistle]

V3 alpha was amazing

[gleaming halo]

yup

[willow pollen]

the "workaround" seems to be to wait till the song goes to sh1t, continue with v2, then return to the continue point in v2 with v3 and hope for the best

[silent drift]

Yeah I keep getting unwanted vocal harmonies

[willow pollen]

which ... isn't exactly a workaround if you pay for generations

and as this seems to be 'localized' to rock subgenre only the people generating rock clips notice ... which makes the poll pointless, as all the people not generating rock don't see any problem.

'rock' emcompassing metal, etc ...

[solar grove]

I don't see any DMs but I'm starting a remake on a song and will see what happens.

[willow pollen]

just for reference : THIS was created via V3-Alpha. I dare yo ut oget anything similar in terms of single voiced vocals https://app.suno.ai/song/8f54b6fd-4a8e-4080-837e-807209db92bf/

[gleaming halo]

Here is one example. The left link is v3 Alpha, right V3. 13 80s Pop Rock Female Vocals Distorted Weird, Out of Tune, Especially on choir https://app.suno.ai/song/83263aaa-787a-4e30-b2a9-f0fb6bd0fb67/ https://app.suno.ai/song/2f324c69-f4eb-4635-a896-9a2bbe86a053/