#🪲┃bugs-web
1 messages · Page 6 of 1
that in a continuation
yes we will refund you -- email billing@suno.ai please -- let us know if you don't hear back in a day or two. Sry & thx
I sent the e-mail
ok just wait a day?
yes np
completely black screen
change your prompts
ok
How to upload my voice here?
You don't.
That has never been an option with Suno.
Is it not working anymore?
small v3 bug. When previously having selected instrumental, and then the option disabled, it still will generate instrumental with lyrics showing [instrumental]
still no lyrics on some tracks
v3 is taking liberties on the wrong parts, especially when continuing from a song that already has an established chorus, i' wanting it to be sung in that style each time the chorus is sung, also it doesn't seem to retain the vocalist and sounds like a different person, after continuing enough times
when starting a new song, the first lyrics are straight out of whatever i wrote in the style of music
Can't you scroll more down? It seems like you are not fully down
Under Title
Refresh, clear cache, try a different browser? If all fails you should mail them on support@suno.ai
V3 keeps giving me the same beat, over and over again. and same vocals, and it keeps refusing to do metal
Mine too.
I can't get it to show up even after clearing the data
Apparently everyone but us two lol
I keep trying to give a v3 song a 👎 and it will disappear from the create tab and then reappear without any rating.
What do the ratings even do
It is the only sorting mechanism available really. So 👎 gets it out of the way in the create list.
I had a 2 minute v3 song generation. It repeated the verse and chorus twice with solos in the middle and end. Too bad really as it was kinda nice.
The companion piece form the same generation was a 2 minute instrumental with no voice.
Yeah metal genres.. are not working well at all 😦
v3 works really great, but sometimes it just creates instrumentals when it should create songs with lyrics - plus it really struggles with the "Continue" function, doesnt work at all, just gives me strange things xD
Bug: Unable to generate a song with the word "Iyla" (the name of a family member). Error pop-up indicates that is the name of an Artist, however so is "Bob" but that does not get blocked. Any way to provide feedback on filtering for songs that cannot be generated?
Everytime i hit thumbs down, it comes right back i have to spam thumbs down
Maybe related to the 2 minutes songs: "pop upbeat, blues shuffle" 2 minutes; "pop upbeat blues shuffle" 1 minute. I can't waste credits testing, but that may be related.
Hey, sorry for the delay in responding.
Here is one:
https://app.suno.ai/song/480cedfe-df30-4bb3-b285-4fcdcccc43f4/
new wave catchy upbeat song. Listen and make your own with Suno.
There is plenty more in my library as well. This was made in V2 but before the update allowed both continuing and doing custom lyrics
This is how it looks to me, along with being grayed out, you can't click the button. If I clear it, the 'Continue' button lights back up and is able to be pressed.
also if you add text?
Can you DM me your acc info and someone will take a look?
Thanks, this a bug, we’ll probably patch this tomorrow. You can move to trash in the menu if you really don’t like it
We will look into this one, may be a few days but not intended
V2 had amazing metal capabilities.. now i cnat get anything thats not just grunts or unintelligible screams
For me, V3 with a prompt that i used in v2 is chipping the song giving me 30 seccond parts instead of at least the resonable amount to it to finish. In v2 sometimes i had that but it was rare, usually i would have a 60 minute part and sometimes it chopped. Now i mostly get 30 secconds chopped with simple prompts. Idk if i'm doing something wrong or if it is a bug.
Definitely feedback like this is helpful, we can still change a lot before it graduates from alpha
Not gonna lie, the sound in v3 is amazing for what i do, the problem is just that i lost the control i had in v2 to link the pieces
Mind sharing the prompt? DM fine too
ok thank you, its just insanely diffrent from V2. V2 was amazing for metal. now it just nothing i cant get it to make anything at all. deathmetal has pop rock instrumentals etc...
Yeah, I was calling it almost like a new guitar. Very different sounds. But definitely part of the reason of the alpha is to get feedback so we can improve weak areas before broader launch
Yeah i completely understand! metalcore is improved in V3 just the vocals dont sound right, and deathmetal / deathcore sounds like rock lmao.
sent
I dont want to spam about my question. Just curious about how to effectively provide feedback. There is more than one person named Iyla, just like there is more than one person named Bob. Just because its less common shouldn't cause it to be blocked as if it was owned by a single person.
Hi sorry
we are working on the moderation part. For now -- I suggest you add sth in between
like I-y-l-a or Iy--la
be creative
Thank you. That helps.
can i give you a example of some good deathcore i made on v2?
to help out
@gritty aurora I still don't have mine 😭
For me v3 continuation seem to be broke. It's not singing my lyrics but hallucinate.
Provided to YouTube by DistroKid
Darkness · Divine Echoes
Darkness
℗ 6436468 Records DK
Released on: 2024-01-29
Auto-generated by YouTube.
this is good deathcore i hope it can improve with v3, v2 was amazing with it
@gritty aurora can it make nu metal
i've made it in v2
I meant v3. My v3 isn't available for some reason
I made a whole nu metal album
I just wanted to see if v3 could
I'm finding v3 is repeating my lyrics, sometimes in big chunks. e.g. I'll have lyrics in part 1, instrumental in part 2, put new lyrics for part 3 but it re-sings the lyrics for part 1. Also it cuts off too early in the first part if it's got lyrics.
Bug: The beginning of vocals in a verse will sometimes get blended with a lead instrument like a saxophone or guitar. This was seen in the valentines day song generator as well.
want to DM me your account info and someone will take a look?
V3 is great!! a lot better just still the 1:00; 2:00;3:00... bug no silent
same bug from V2
Yeah, we've seen this one too. Honestly quite silly when it happens..
Sounds like a vocoder:
https://app.suno.ai/song/c42361be-66b9-43a3-bbbd-552607019466
upbeat blues shuffle song. Listen and make your own with Suno.
Haha. I had some that never got vocals, just the instrument singing the words if that makes sense
Yeah, we are a little stuck right now - looking into it
Just in case. Thanks!
Ah... thought it was just me and was about to report it. I thumbs-downed the first set that didn't complete generation thinking it was an issue specific to those, until the next set also got stuck 😅
The v3 amateur quality assurance team broke the server 👏
Most vocals seem to be too quiet in V3.
yeah theyre weird too
I am unable to make anything good with v3. ALL my continuations are broken and I get hallucinations instead of what I want, so even if something sounds ok I can't do anything with it...
All new tracks are too quiet or just bad, or with no lyrics sung at all.
Sure - quality is much better, but it doesn't work for me 😉
its alpha itll get better.
Sure.
If the continuation from any moment worked, that would be enough.
I loose like 90% of the credits because tracks end in a way that makes them impossible to continue.
All my generations are spinning except when there is no vocal, I don’t think that’s a coincidence- for the team trying to fix stuff
Had a couple good continuations and then v3 totally switched up the music to a completely different melody in the middle of the song on another continuation
Also it keeps switching back to v2 on a continuation when the clip I started from was v3
I was working on a full song with V3 for awhile before I gave up, not wanting to waste anymore credits; here are the issues I noticed (ignoring the stuck generations that are already known about):
-
I'm definitely still hearing the "sounds like autotune" quality in the vocals that I noticed in the Valentine's Day preview.
-
V3 seems more prone to skipping lyrics than V2. Other times, it'll duplicate syllables commonly -- so for instance "drift apart" gets sung as "drift-ift apart", or "fade from view" becomes "fade fr-from view". And more rarely, it'll go back and repeat entire lines in the middle of a clip -- for instance, repeating the last two lines before continuing with the rest.
-
The "features" section of the Announcement thread for V3 sounds like all clips max out at 2:00 now, but continuations still seem capped at 1:00 -- that said, it often stops continuations even before the 1:00 limit and before it finished the lyrics for that clip. Often around 0:20, but it varies. This happens so much .
The melodies and understanding of the genre prompt were much better with V3, but those bugs above made it unusable, so I reverted back to V2 and started over 😦
Hey all, system's in some pretty rough shape tonight. 50 bad renders in a row. My entire cue sounds like this https://app.suno.ai/song/37f7fb7b-567f-405a-96a1-b1199085f665/
top40, female, rap, hip hop, fast, orchestral song. Listen and make your own with Suno.
https://app.suno.ai/song/67e30afd-4d29-43d9-b6ca-d090d21258d9 this is what im saying with heavy metal isnt right on v3 hope it improves ❤️
love you guys and ur work!
Here's another one (this would have been perfect if any of it was in english)
https://app.suno.ai/song/cd39fe9e-0655-4258-b765-882cd9619399
top40, female, rap, hip hop, fast, orchestral song. Listen and make your own with Suno.
How I can finish a song with V3?
continue from clip and use [outro]
Each generation you plug gets you two songs, right? There MAY be something happening with V3 where a song just goes missing. I have a sequence of what should be 4 of the same generation but one is gone, I did not downvote it nor did I delete it. Just gone.
Yeah, this is related to stuff gettign stuck unfortunately
As long as you guys are aware, it's cool with me. Alpha!
Works a lot better, I used [End] before... Thanks!
I cant even download songs because it just had an infinite loading icon over the download buttons
It also seems like it's labeling parts incorrectly in V3. If I continue from a 'Full Song' it's labeling any clip after that as "part 2" when it is supposed to be "part 3", right?
50% of my songs are nothing but silence. Wasting my credits for literally nothing
So, a minute ago I was stuck generating when no one else was, and I think I have some insight into why… song would let me listen but not download and it was FOUR MINUTES long, then it finally stopped spinning and was only two minutes
regular songs or continuations? in any case, feel free to DM me and we can help you out
Seem to have fixed itself. Had a moment where one of the two songs it would generate was silent
Yeah, that sounds annoying. We're continuing to monitor!
@muted flame You made my night, It is working a lot better on my end sense couple hours ago... This is craz good!
When continuing .. clips are usually super short (~14 seconds) and it often cuts off the lyrics before the end, even if it's just 1 short chorus.
(v3 alpha)
If original clip is up to 2 minutes, the continues should be closer to 1 minute at least i think
Almost impossible to make a complete song using continues if you can't make at least 1 chorus or verse in each continue.
And it's frustrating because you finally find a good style / rendition but it cuts off the last word so it's unusable
Overall V3 is noticably much better though. I applaud the efforts and eagerly await future improvements 😎
Plus the more snippets you join together, the more the style drifts / becomes inconsistent
Anyone knows why the option for a V3 only appears only in one of my paid accts?
Those are both paid accts and the V3 option only appears in one of them
I'm not able to download longer full songs
clicking audio or download does nothing
Solved, chrome bug, worked in firefox
Im having V2 and Instrumental checked. However I still get lyrics in the songs.
for me it mostly works if i refresh the page and choose instrumental again after that
works best with V3
"Instrumental" toggle needs to be tapped on very precisely and the "?" don't do anything on Firefox, Android. Would be good to make it so you can tap on the text rather than just the toggle.
Instrumental does not work
Often generate hallucination or noise only like here
https://cdn1.suno.ai/f859c27b-cc5b-4794-9058-4c52caac7e45.mp4
why does it keep telling me "prompt likely copyrighted' every time i try to continue from clips i've made, and it's been doing it on different songs, for two days
We should have some fixes here in the next week
Suno V3: web interface: the Download and Report options in the "More actions" popup menu cannot be navigated to with a keyboard.
This is the second time I’ve run into a really frustrating bug. The song chops off shortly before the end, and won’t continue seamlessly unless I cut off all lyrics but the last few. But those few are flagged as copyrighted even though they were fine as part of a larger song. In this case I need it to sing the two words “for you” and it won’t do that.
What about "4 you" or "fore you"
Here is Deathmetal not being Deathmetal it’s so melodolic and sounds weird 🤣 https://app.suno.ai/song/fc93715c-1180-46d0-9561-261550d66d35
It keeps repeating verse one. Over and over
Here is another example in Deathcore I’m an avid DC listener and this I can’t understand at all. https://app.suno.ai/song/84aa6b9f-a95b-422c-acee-3a3a5eb1318c
Deathcore intense aggressive brutal song. Listen and make your own with Suno.
is it a bug when some users see "Playlist" in Library and some don't?
i clicked "continue from this click". V3. most of tracks are stuck in loop generation, some other is 8 seconds and some is refounded, what can i do?
Even new creations, I got stuck up to 30 minutes, some cases, they disappeared after that. Server load? for spinning but not sure what happened when it completedly disappeared, I lost credits and generated clips
generation will cost you credits when they are done.. I have some stuck too, after about 15 minutes I got a brief popup message they failed and weres not credited.
hmmm... I didn't realize that as I did not see the popup... maybe I was on a different tab.. but when I came back.. it's gone. I assume I lost my credits
I doubt you lost your credit 😄
ok thanks, that makes me feel better, I'll pay attention next time it happened
v3 seems to act like the valentines day song generator. If your lyrics finish in ~30s or less it starts riff on other lines or sing the chorus again in order to pad the song out to 1 minute. But this makes it impossible to continue to add more verses to the song.
use this option to continue where you want
That way you can skip the double lines it sung
I generated another and it is more than that, it is not singing the chorus as written. It is mixing up the lines and riffing on the lines in the chorus. So I don't have a solid base to continue from.
That happens now and then, I am sure they will make that better soon.. I notice that behavior you describe mainly with foreign languages.. english seems to be okay most of the time.. might also be genre specific.. 🤷♀️
It added foreign language as well to one generation. I had style "Folk, pop, Irish tin whistle" and it sounded like it sang the song in Gaelic and then again in English, but the Gaelic part was slow, whereas the English part was very fast.
I guess it has less training data then of what you wanted to create. I notice the same with Dutch songs..
i want to talk to someone at Suno please.
I'm having the same issue. It's also having trouble making continuing segments longer then 30 seconds
I've blown a lot of credits on V3 generating straight garbage
1/4 generations never actually finish it just spins eternally
those won't cost you credits...
V3 is also cutting off in the middle of a verse or chorus
However the sound quality improvement is noticable
I see your concerns with V3, particularly with the difficulties in creating longer continuing segments and the issues with unfinished generations. It's good to know that the sound quality improvement is noticeable, though
I agree with that
I'm testing out V2 continuation of my initial V3 generations
it's just an alpha version, it's very good for an alpha version.
Yesterday I had many 2 minute first generations for just a verse and chorus. Today I just got a 3:36 first generation. The verse was cut off part way and a completely different instrumental started playing.
Maybe V2 can create the continuations to complete a V3 start
I agree its revolutionary
It surpass all the other music AI generator.
Hi all
I think I should put some of my feedback in here. Noticing hallucinations really bad with continuing v3 clips past 2 minutes. Find switching to v2 for continues the best course.
renamed clips doesn't get renamed continues
Part 1 is singing...
https://app.suno.ai/song/5a7a5699-c3c4-4bdc-9634-b450fe80373a
Part 2 is rapping.
https://app.suno.ai/song/4c5545ea-a376-45a6-a4b2-788555d8c1c0
Synthwave, Big beat, Confident and angry male vocal without autotune song. Listen and make your own with Suno.
Synthwave, Big beat, Confident and angry male vocal without autotune song. Listen and make your own with Suno.
How can I download? I just completed my song…
The three dots on the line with a menu? Click download and audio.
That’s funny. iPhone users, sry for asking 😅
On an iPhone the three dots will show this menu
I made on Safari. So I uploaded to Google Drive and downloaded from there. Thx for the tip! 🙏🙏
the vocals often deteriorate after a few continues
V3 -
Pros: Well at least the sound quality is somewhat better.. mostly that is. 2 minutes playtime, Longer style description field about time,
Cons: V3 Dont care mush for descriptions, or it cares but not trained in the style like V2.
It ineed runs 2 minutes but most of the the time by repeating chorus verse x2 even though only promted for x1.
Hallucinations, like its on bad Psilocybin, Cocaine, LCD and all drugs between. Voice consistency none most of the times flippes between Metalcore and pop man/girl, If your lucky. All in all ,some progress made but some tweaking left.
you can only do that if you are a paid subcriber. You can do it then like this (see video):
thanks 😄
Bug: clicking on regenerate lytics for the second time in a song does not replace the existing lyrics
I find it's very odd about continues. I'm expirmenting but I find sometimes feeding it a bit of the last section helped it keep the plot. So either the last line or the last word. I wouldn't put a section tag with it, litterally just the last word it is singing, then the next part.
I somehow feel like V3 is set to Hallucinate way higher then V2. Like is SD when you crank up to high on Denoising strength. It can make magic but loose coherence. Most generations in V3 feels to Random.
very often it procudes a very short song singing just the music style and then a song without lyrics
WTF?
on V2, I put in part of my lyric in language other than English, and hit Generate... it often did a good job to generate the rest of it... but on V3, it always error out... please fix it, thank you.
Lyric generator might not be V3 though so Idk... giving me so much trouble this time
yes.. read this: #💬┃general-chat message
So they are probably tweaking to find the sweet spot 😉
Good rename update, but...when you click on the three dots to rename the track (most of the time) starts playing! You then have to stop it and click the three dots again before you can rename.
It keeps repeating my verses....
like it repeats the whole song 😦 https://app.suno.ai/song/5b0ff7ca-73f2-4a8d-9f64-f20295ac67f2/
V3: I'm getting ~1minute tracks and it's cutting off abruptly before the lyrics are even done. Is this a bug? Should I downvote when this happens, will that help train it? Or is the downvote more for how it sounds?
V3: continue track just generate 6 part 3 instead of 2 . Also notice it fades out the song even though i have a verse 3
if you don't like it -- downvote; if you don't think this is right way to finish -- downvote; unless, you like the vocals so much...you can try to continue from it
basically dislike anything you dislike, and like anything you absolutely love, if it is meh, just okay, that's thumbs down
V3 greatly overuses the word "neon" in song lyrics.
That's every Suno ever
in generated lyrics, right?
Yes
I noticed it repeatedly in songs I was generating + some of the example songs. It's just anecdata so probably needs to be checked. Maybe it was overrepresented in whatever training data was used.
yeah...our lyrics gen can be better
but you can click generate lyrics, edit and then generate the song
I think generate lyrics is free?
I honestly think it's a bug. About 50% of the songs I'm generating end up with the word NEON either I'm the lyrics, or in the title.
Everything ok?
They've been stuck for about 3 minutes now
V3 Model and Continuing a song
That happens sometimes.. servers are "hot" atm.. 😂 @gusty oriole
The spin off dead.. they are goners.. but they won't take any credits for them... you only pay credit for succesful renderd parts longer than 10 secs
when do they normally come back? :3
some are just lost.. sort of time out.. they don't come back.. got to create them again
ah yes, I meant how long till the server come back up
is this a v3 continue v2? sometimes that have issues
V3 continue V3, I tried again with V2 and then it worked
Oddly enough, the V2 part was mono... Not sure if that's expected behavior for extending V3 to V2
V2 did a good job too 😭 (after 52s)
For some reason the issue is still happening... Here's a link to the song that's stuck:
you are not alone.. I have it too now.. just growing pains.. try again in a few minutes.
k looking
v2 and v3 seem to be confused on genre now
just redo a continue -- that job may be canceled
shouldn't charge you
and gibberish is way up compared to a few days ago
mind sharing examples?
doesnt sound like gangsta rap
the hook is funny af tho
i would call this genre experimental medievel clown music
with a hint of LSD rock
Appreciate your effort👏👏
Waiting here to finish a Japanese anime song for my daughters
if you send me the suno link I will try to regenerate it
I don't listen to gangsta rap but: https://app.suno.ai/song/ba530dc2-a2b0-4804-839f-904350578f5d
this is happy pop
gangsta rap, fast tempo, gangsta rap, fast tempo song. Listen and make your own with Suno.
gangsta rap, hip hop, rap song. Listen and make your own with Suno.
I mean you can just continue off that ^ one if you like it right...
yes
i made a whole playlist with juxtaposition AI Gangsta Raps, things they would not be expected to do like baking a cake for grandma or a gangsta childhood teddybear
the best thing would be if we can make 3 minute
lol not 4?
3.5
but ok hear me out here, let's say we have that...would you be happy if it is 2x more expensive?
yes
the risk of generating sth bad is more costly now though
if that is 20 thats fine, 10 for 2 minutes, 20 for 4 minutes
if you put a disclaimer that the lenght might influence stability then its their choice
i can only fit this format now : verse 1 - chorus - verse 2 - bridge - outro
then stich it together in a video editor but i can hear the transitions, that have to be timed
ok noted -- ty for the feedback
its a great app though, its normal that new things have things to be ironed out. this is transformative technology
Tried again, no luck unfortunately
https://app.suno.ai/song/23065618-0c11-454a-90ef-e2575681c7c4
1:35 continue, but its confused and all over the place...
intrigue, sensual, smooth, groovy, electric bass, funk, soundtrack song. Listen and make your own with Suno.
original clip that you continued from?
i think that one's the original
all other 3 clips involved were stable/normal, just that one outlier
lol I can't hear the issue honestly, like it didn't stick to the prev melody?
lemme try refreshing, maybe I got a defective clip buffered
oh you might...
very interesting
if your prev job failed somehow, you may hear some streaming and then...start all over again
assuming you hear the 1:35 extension, at 0:25 you can hear it stop and then restart a different part, and then at 0:30 it just overlays other portions, like it's merged two songs over each other and alternates between them
which doesn't seem to be the intent given the other 3 portions that generated normally
I really don't hear the issue at 0:25, this clip? >> https://app.suno.ai/song/5279943d-066d-4066-8064-0244222f16d8/
intrigue, sensual, smooth, groovy, electric bass, funk, soundtrack song. Listen and make your own with Suno.
yeah, nevermind then... seems like i got a hallucination buffered
no worries, just figured if it was a technical issue, should report it, but I've seen other AI do that with drafts as well
Is it possible to do part 6 of a song in v3? Because I try to do it, and it never let me do it.
I was just able to make one but I think I've had issues with that in the past. Didn't do anything special, just selected Part 2 and continue and Part showed up at the bottom of the creation tab and created Part 3. I'm liking v3 so far for the most part but it's pretty consistently been giving me 5+ seconds of dead air between parts. It would be great if there was a way to cut x amount of silence when continuing a track automatically. I'd also like to see some sort of penalty for mentioning words in the prompt but that's always been an issue and you've have to exclude words like "and" from this rule.
Edit: Ah, I assume that's what the "continue from x point" is for. Great feature, as is the ability to create instrumentals without hacks
thanks
do you have credit discounts for higher volumes, or is the best but to just unsub and resub?
we are working on this...
Would be nice with a "relax" mode for subscribers, I'm at $30/mo and burn through half in 2 days 😄
yeah we are working on this
I think you can ask support to merge your accounts for now
Like sub with two e-mails and merge?
see DM
I'm getting a lot of errors that say prompt likely copyrighted when I am using completely original lyrics
you may trigger some words
try to edit it a bit like wor-ds
okay I'm just saying this is pretty original
beyond my wildest dreams
is a failure of imagination
you see how they made calculus
now let's put this in the equation
I know I am trying to help you...I don't know why this fails as well
if I get moded that's what I do, play word tricks
Hello, when I was using the V3 model today, I noticed that many of the generated songs are very short, and many songs are constantly in a state of being generated and cannot be produced. Is this a problem with the model?
I'm getting the same constant generation bug (and eventually they fail and disappear)
Same here. I’m trying to finish a song and is going straight to credits refunded. Is it because of Japanese characters?
Oh, it looks ok now. Back on track 🙏
the first part of a multipart song finishes more or less. but afte r that, it almost never does in version 3 for me. I think I got 5 decent parts. lost 100 credits
Instrumental isn’t working at all. I have half a dozen complete songs in beautiful Simlish made using the instrumental setting. It seems to work in simple mode, but turn off simple mode and then turn on instrumental and it gives a complete normal song with hallucinated lyrics. These songs also do not have the “instrumental” label like the ones made in simple. I know I’m not accidentally just not turning on instrumental because the lyrics field goes away.
If you add pirate metal to tags it will start adding Irish accent to songs screwing up numbers pronunciation
😈
Hallucinations is also off the scale 😆 eating up all my credits
V3 i guess its still in Alhpa but if any changes were made since yesterday, roll back, its on usable today its singing its own lyrics in custom mode
yesterday was fine
Might be another genre you are trying? My outcomes are great!
Its happens to me when i used own written lyrics but i solve this i sent my lyrics suno and tell issue and later dont have any issue
thumbs down only works when the song is fully rendered. If not, it keeps re-appearing.
Welcome! For help, check out Suno Help
Welcome! See Guides and Resources for help using Chirp and Bark. Email us at support@suno.ai if you have questions, comments or feedback.
/bark
Go to the #🐶┃bark-beta channel to use the /bark command, and check out our docs for step-by-step instructions for using Bark on Discord:
It's strange, no matter if I use Japanese lyrics or English lyrics to continue creating, the V3 model only generates songs that are around 10 seconds long.
I see you are continueing from a song.. that song might be the problem
Sometimes songs have weird quirks in the seed which prevents you from continuing.. I had it with V2 too
I signed up using my gmail like 5-6 days, I used my all 50 credits on the first day, since then i am waiting to get 50 credits but still couldn't get it
Oh that sucks.. you might have to send them a mail via support@suno.ai
ok i'll try
So, the problem probably is the first generation?
yeah, that could be. With me it was in the past
try a different song, or continue from a different place
Yeah, I’ll give a shot, thanks 🙏
[V3] I have a lot of trouble continuing the song. When I want to continue, I provide new lyrics that I want to be sung, and suno.ai suddenly makes a verse from the chorus and starts singing at a random place.
Yeah, I had a full verse, pre-chorus, and chorus and got flagged for the phrase "life and death" at the beginning of a line. I changed it to "life and deth" and stopped getting flagged. Kind of sucks bc you waste a generation or three (and credits) while you troubleshoot which lyric is triggering the flag (and I feel like the AI loses some of the feeling if it doesn't understand the grammar of what it's singing about) but that seems to be the best/only workaround atm. Tbh, I wish it would look for longer phrases before it flags (e.g. flag "I got my mind on my money and my money on my mind" but not "Money on my mind, always thinkin' bout them stacks").
V3 is a huge jump forward sound-wise, but Continue From Clip seems to be stuck on buffered lyrics from previous generations. E.g. I generated a 1:50 song clip yesterday with a VERSE1->PRE-CHORUS->CHORUS->VERSE2->PRE-CHORUS->CHORUS structure and it was great. Today, I tried to continue from it with a BRIDGE->PRE-CHORUS->CHORUS and 25-30 generations in a row have just repeated the VERSE2->PRE-CHORUS->CHORUS from the first clip (none of which are in the lyric field) and either (a) completely ignored the new BRIDGE lyrics and structure, or (b) tried to squeeze them in at the end of the repeated stuff but ran out of time and cut off a line and a half into the new lyrics (which do appear in the lyric field). I even thought, "Okay, it REALLY wants a third verse here" so I added a new VERSE3->PRE-CHORUS->CHORUS and it mostly either (a) ignored it and just repeated VERSE2 again or (b) repeated VERSE2 and then tried to do VERSE3 as a bridge. Thinking it might be a buffer issue, I left Discord and loaded back in several times, reset up the Continue From Clip, pasted the new lyrics back in... and same problem. The new 2 minute first generation is great but it feels like Continue From Clip is still trying to continue from a shorter generation or something.
impossible to contiue a song with new lyrics in v3 with trap upbeat r&b
Bug: New songs often use two lines of text twice. (sing it again)
In V3, if I have the first 2 parts in a song , then create the full song and try to continue it, it never sings what I give as input, but will always repeat what was the second part before merging. Basically the only way of having more than 2 parts is to continue part by part without ever getting the full song until the end.
The same here
Hi, I love v3, at least at first but now I almost only get nothing, 10 seconds or 30 something seconds, (or 1 minute), my genre is downtempo, did this good at the beginning but not anymore
k thx for your reports -- I think I fixed a continue bug and that should be deployed now; you should have better experience continue off long clips now.
If you still experience continue bugs, it would be super helpful to dm me 2 links + the timestamp:
- the clip you want to continue from
- the failed clip that you continued
I may not get back to you but I will try to look later today
Thanks tony ur cool as heck + then some
BUG: "Continue from this song": The style of that song is sometimes completely different. So you can't use this feature
did you enter the same style prompt?
Here's the link: https://app.suno.ai/song/2e6f1cee-8605-4e4e-ba5f-5a6e2bdb11bf If i use "Continue from this song" the style is completely different in all generated songs. (doesn't matter which continuation-time or entering the same style)
Italian classic pop song. Listen and make your own with Suno.
With V3 on the web interface I get 50% of songs with missing vocals while with V2 it was ≈15%. The lyrics are written by GPT4, they are shorts phrases w ith ad-libs, ≈4 by paragraph, up to 4 paragraphs with [Section] tag above, and almost the same prompt.
Hello, I'm still waiting!
Loving the quality of V3 Alpha. Just thought you should know I just generated a song and it came out gibberish except for several what might be considered 'profane' words. I thought it was hilarious but I think your AI might have been reading some ribald literature.
A woman spurned...
ive seen the bot spit gibberish, in metal it does it alot
V3 is producing clean and interesting tunes, BUT...it is ignoring the style input completely. My default style list is post punk new wave indie male vocal, yet every one sounds like som mid 80's American power ballad.
yeah... in metal its insanely wrong lol
Hopefully it gets fixed.
im getting upset with song constantly repeating verses, i made such a good song thne at the end it repeated a verse....
had that all yesterday very frustating i have burned 100s of creds trying different things, gave up in end today i managed to get a track out of it but was tricky
metalcore is the only one thats slightly right.
right now im doing emo punk rock shoegaze its not bad
i usually do a video of my own for songs. i was going for an incubus with this, it sounds pretty good to be fair which is why i kept it, but as you will be able to tell the bridge, verses and chorus prompts make no difference it does what it wants and adlibs lyrics. plus in no part did i ask it to rap, but it works so, i was like ok then lol
Thank you so much for your support 👍
Songs keep singing the style prompt, that's not great 🙂
Not sure this counts as a bug which can be fixed or a peculiarity of the black box that is Suno - I was testing to see if v3 would do madrigals which v2 will not. It does, but with a weird caveat. It absolutely will not sing the lyrics put in, in madrigal style, only nonsense. Or it will sing the lyrics but in a modern style, not madrigal style. I wasted a silly number of credits testing all permutations of this, including putting in lyrics of the actual famous madrigal April is in my Mistress’ Face to see if the lyrics were getting in the way, and it is 100% consistent. Nonsense plus madrigal or English plus no madrigal, never both. It clearly knows the style and can do it, but it doesn’t know how to do lyrics to it. Sometimes of the two generations from one spin, one would turn out one way and the other the other way. Here’s an example of Renaissance nonsense. https://app.suno.ai/song/8eec4301-3b93-4cb3-9ebe-1dcf65b39546 and this one is English but modern. https://app.suno.ai/song/57fa7d11-93a3-495f-9ff8-c450d49c67f6
College madrigal singing group, Renaissance, ren faire, complex harmony, acoustic, unaccompanied song. Listen and make your own with Suno.
College madrigal singing group, Renaissance, ren faire, complex harmony, acoustic, unaccompanied song. Listen and make your own with Suno.
mmm using v3 it doesn't put lyrics on content set to "epic movie cinematic, inspirational, motiviational, orchestra, strings, horns"
i can't think how to prompot it to force it - something that sounds like dead can dance, etc.
the bot seems to be singing the prompts to me on v3 lol
dm @muted flame
send the link to the singing prompts?
I think we need to get used to the longer 1st clip, we are so used to 1:20 that 2:00 is forever and we should chuck in at least a verse/chorus combo twice to ensure it doesn't repeat.
ran into situation where of the songs I generated, about 1/5 of them had no lyrics in them 😦
Mine are doing about 50% no lyrics, even when I change the prompts. or add the type of singer. It doesn't seem to make a difference.
Are you mentioning instruments in the meta-tags?
Crazy. So... why won't it work for me? It looks like you deleted the Style?
emm, should also work with style...
just luck...?
or maybe the model respects...creators of the model? 😄 (jkjk)
haha
I've tried Continuing from Part 2 (the 1:00 segment), Continuing from the combined clip (the 1:07 song), both with lyrics (Custom Mode) and with Instrumental, and various different lyrics prompts. I've also tried Continuing from a timestamp, instead of from the end. Everything just spins and fails. I also tried Continuing in V2, which works. ✅ But I'm going to try changing/deleting the Style, because I haven't tried that yet. Also, just more brute force attempts in V3, I guess. Thank you.
yeah sry sometimes it takes few tries to follow the instructions...
sometimes try to insert some meta tags [] helps
but if the model thinks the music should end...it is hard to tell it not to as well
v3 sometimes mixes two words together
Well, that's simple, isn't it? if (model.luck < 0.5) { Fortuna.smileUpon(model); }
Hi there, the new features of V3 Alpha say upto two minutes, but im maxing out at 45 seconds and even when i post just the chorus it doesnt finish it. My friend is using it and gets to 1 minute 48 seconds. What am i doing wrong? I've had to switch back to V2 just to get a full verse and chorus.
you are not wrong….
luck / change your prompt could help
Change my prompt? even if i just post the chorus it wont finish it. which is four lines. I'm not sure changing the wording will help.
this has been on every submission, not just one song. Over a 1500 credits trying to get it to play lomger than 45 seconds so far
Ok send me a link of 45 seconds?
This is the chorus, it gets to 15 seconds
V3 disappeared from my create menu, does anyone know how to fix it?
....I mean this is very short lyrics
see this: #💬┃general-chat message
Most cases it's that
I'm showing that it cant even finish that. New features say upto two minutes. So surely it should be ok to finish that? I can't get it to go beyond 45 seconds on anything
...you need to prompt more lyrics...
or just instrumental
i have, ive tried two verses, two chorus, instrumental and no matter what it gets to 45 seconds at most. then i switch to V2 and it works
i have, ive tried two verses, two chorus, instrumental and no matter what it gets to 45 seconds at most. then i switch to V2 and it works
and why don't you continue it...?
because it wont go passed 45 seconds
o.O
At the very most it wont go passed 45 seconds no matter what.
try prompting harder...
I have done that, ive filled the box and wont go passed 45 seconds. Glad it worked for you.
should really work...
i know, does for you and my friend. Which is why I was wondering what I was doing wrong. Its ok, ive burned through over 2k credits now. So will stay with V2 until v3 is released. It might work for me then. Thanks for trying to help and I appreciate the time.
V3 Instrumental only bug. (Suno is singing)
https://app.suno.ai/song/0961e91d-93ca-4e28-85a3-87ec58dd8ecd
honky-tonk country old-timey song. Listen and make your own with Suno.
See, i filled the whole box like you did and it did 15 seconds
Are you accidentally continue off sth? Or this is a fresh creation?
This one was a continuation to see if it could finish the song, but the ones before where I did the same were from the start
Sry try other songs 😦 can’t help you here
If you enable instrumental and type just piano you should even get longer gens
I don’t know what’s wrong in ur case
This is my 8th song, same thing. But thanks for trying
Okay, so here's a really interesting bug. In the V3 model, the genre prompt is sometimes overflowing into the music, it's happened twice on me now. https://app.suno.ai/song/7db8d258-1277-4ea1-aff9-c84a2b5bc13a
fast female vocalist, pop-punk, pop, melodic, hd stems, goes hard, 155bpm song. Listen and make your own with Suno.
refreshing the page and setting it up again seems to work
I ran into this as well...no idea why...
let me follow this up directly
I would audit the json that's being sent to the api. Chances are good it's probably some kind of simple encoding or deliniation error.
React breaks things all the time
this link is a new creation (not continued from), correct?
v3 is getting stuck on a prompt v2 can do. It has been over 10 mins, here's proof of 7 mins buffering but that was after sometime already.
link?
It's happening on both new and continue. Once it happens, it keeps happening.
What happened here 🤣 https://app.suno.ai/song/a168127a-09d4-4043-965c-3522f119080b
https://app.suno.ai/song/a3ce9452-16b8-4091-9dac-4cbe3eb00324 link to where I continued from. Now I will get link for buffering if I can.
Dark Atmospheric EDM trance song. Listen and make your own with Suno.
Just a question, but is there a way to regenerate before the chorus? Like re-generate the verse or is that not possible?
https://app.suno.ai/song/083d8616-620a-4226-9609-9a01cf601d2a constant buffering link
some jobs could get stuck
I also have sth that's stuck like that
did you check the continue time
I reproduced a bug like that
continue time was max, but I will try again
the continue time has to be less than your input clip
but if you continue and continue again, it doesn't reset automatically (which is a bug) check your duration.
Tony, just a question, but is it possible to regenerate the verse of a song and keep the chorus?
Duration up to the point I continued is 6:22
no the duration needs to be relative to the input clip, is your input 6 mins long...?
not yet
Ahh okay. But might be in the future?
everything is possible in the future
Very true haha. Thank you
The input is 0:42 but it won't let me continue from 0:42. I just tried continuing at 0:41 of the clip. I will report how that goes.
yeah...k this is tricky
if the end is 41.8 I think it shows you 42 but you can't actually continue from 42
typically if you are off by 1 sec the continue fills the 1 sec natrually
try again & let me know if it fixes it
I suppose I can give you the link but it's buffering still. EDIT: Failed to complete. V2 worked so this is specific to v3.
new v3 bug: Finally made it through a whole song. Here's a new one. Long dead air at the end of songs. It's normal for suno to finish a song in the middle of a note, so what I normally do it just extend it, and it gives me like two or three seconds that I need to finish the tail. V3 is giving me the same two or three seconds, followed by anywhere between 60 and 120 seconds of dead air on the V3 model
k send me the clip you continue from? (the original clip)
i have seen this before...rare but not sure why
Yeah, no problem. One moment pls
See, that's the cool thing about responding quick! I just made this one https://app.suno.ai/song/d82076bc-3d8d-413d-95fa-5e98f20350dc
fast female vocalist, pop country, pop, melodic, song. Listen and make your own with Suno.
This is the original song I continued from
oh... he wants the piece right before the tail?
duh, okay one sec
fast female vocalist, pop country, pop, melodic, song. Listen and make your own with Suno.
ok this one I get...if you conitnue from the end, you can hear the music is dimming out, so it will just continue to end it soon...
that's a typicall v2 bug where you can't continue off songs that it thinks it finishes
you mean the hallucination after the silence?
does one have a hallucination in it?
must be the wrong clip
one sec
I think this is the one I used. Just silence https://app.suno.ai/song/423ea82a-8e5e-4abb-989e-6e2f5cc9cf78
fast female vocalist, pop country, pop, melodic, song. Listen and make your own with Suno.
just dislike this -- this is why this is still alpha
many issues to be figured out
there is a clic at 28 seconds... 😄
hey question
what do like and dislike actually do? other than give me a way to catalog?
that's your user feedback to the generations
does it influence the cue at all?
we try to make music that you like more & dislike less
nice
eventually it kinda does
that's what I was hoping for
I've been mainly using it as an easier way to find the good takes
I haven't been using dislike much because I'm an organizational mess in this app
like everyone
we will try to make it better
dude you're killing it, and I appreciate you
oh, also, thanks for the v3 stability enhancements last couple days
lol I only fix bugs (mostly also introduced by me cough cough), FE magic is other team members
Well, in that case, tell them I love them too
@buoyant flume @royal mason
🙂
does V3 models get trained the same way as V2 as far as foreign language voices? I found in my experience... V2 foreign voices sound much better than V3, I'm working on a Vietnamese piece and V3 sounds a bit robotic in some word pronounciation... but some it sounds perfect... in V2, I haven't got a bad one... so that was what I noticed with a little time I have with Suno
like the nice sounding one and dislike the robotic ones and we will pick it up
would that be enough info to know why we dislike the clip? sometimes, I dislike it because of the music style or lyric
or certain instrument
or the voice in this case
yeah eventually it will be picked up
ok thanks
why won't suno continue from older songs? Is there a work around?
I have that bug as well and mc specifically it would be looked at this week
good to know, hopefully it gets rectified...
I'm the dumbest guy alive
I mistakenly wasn't using 'Custom Mode'
Hello. i have an issue. trying to "continue" a specific old v2 song it tells me "a system error occured" and doesn't allow me to proceed. The "continue from" box (where you put the exact timing) becomes red. Any idea?
This may be a bug. You could try putting in the duration of the V2 clip in that box (which is what should be the default when you put nothing in that box)
there is something on that specific song... i can't remix it also. 🤨 (and obviously it's one of my favourite song on my suno😭 )
Can you post the song link? (Or DM the link to me privately, if you prefer?) I am curious if it is a new bug or not.
yes, but i discover it is not the only one.. maybe i'm doing something bad :/
So good job changing the error message to say what the problem is when it rejects a prompt, but it’s still a bad filter. I was testing to see if I could do pasodoble in v3 and it didn’t seem to know it so I tested to find out if v3 knew the spelling paso doble and it said doble was an artist’s name. What? Not that I recognize, and not that Google knows either. So you can’t prompt a dance because of what now?
(V3) I think more then 80% of my continuations fail to generate last word, sometimes one last line; I even had multiple cases when 2 out of 3 verses were sung.
I'm changing it to 90% 😛
V3 hallucinates quite often than V2
Curious if everyone dislikes what the minority likes that would explain why suddenly all Synth-pop lately sounds like the soundtrack to trolls 3. I noticed that just about every prompt i made lately is unaffected, even though i type for example deadpan, minimalistic song i will get Beyoncé's vocal range for some reason?
No, not at all. I finally got it to work by adding the word 'song' to the end of the prompt. I also tried 'group' and that worked, too. It still gave me a couple of instrumentals, but not nearly as many as before I tried those words.
I don't understand the "continue" now... It's always bugged and impossible to get a correct song with "continue"... V3 or V2, what's happened? The V2 got nerfed ?
Can you post a clip (or DM if you prefer quietly) are these V2 - > V3 or V2 -> V2 clips?
V2-V3 = Bugged. V3-V3 = Bugged. V3-V2 Bugged
and V2-v2 is bugged to, since new update.
There was a bug where the the V3 continue time was being accidentally applied to V2 -> V2 continues, and improperly, so double check there is nothing in that field (even if you pick V2 the field disappears, but the value was still going through)
I have this bug in v3 - v3
whatever the moment I choose in the time, 90% of the time it's bugged :/
Is the clip a regular clip, or a 'get whole clip' clip? If it's a regular clip is it quite short?
Regular clip, mmh i choose 30 seconds, = bug, i choose 45 seconds =bug, i choose 17 seconds, it's ok
Before the update, the continue mode was alright, now even in V2-V2 it's full bugged, is it normal ?
I'm not sure if there are new V2 -> V2 bugs that came with the V3 continue_time feature, but it is not impossible. I will look into it some. (I was mostly testing V3, the new fun thing.)
I really can't continue my good old V2 song since the update 😦
V3 is alpha, so it's ok if there is bug, but V2 doesn't not be so bugged I think
The only V2 -> V2 bug I am sure about is the continue time feature. Continue time is not supported (at least not yet...) in V2. However if you pick V3, then type a continue time, then pick V2, what happens is V2 will continue from the end of the clip, but insert the audio into the custom time you typed in when you had V3 selected. (So probably a bad result.) So you have to make sure you clear the continue_time field when continuing with V2.
And when I want to continue my V2 song, i need to edit the continue time on V3, and I put "0:00" ?
Off the top of my head I am not sure if you should have it be 'empty' or if you should put the 'end time of the clip' (the duration)
But not 0:00
empty can't clic on the continue button
So I always need to put the end time of my V2 song ?
then reswitch to v2
Maybe, I'm not sure. There are V2 clips from different periods of time which sometimes cause inconsistency.
Ok so I need to make sure of the end time of my V2 song if I want to continue
I have an old V2 song, 0:45 seconds. I can't continue if i insert "0:45", i'm forced to choose "0:44" then reswitch on V2 mode
I've tested, seems alright
I used at least 300 credits to understand that...
thanks man
you gave me a way to continue my old v2 song at least 🙂
Can you send me the link to the clip in a DM message? I will make sure Suno knows about the bug. I think it is the same as the one I described, but just in case it is different.
Any tips for using the continue button in V3 ?
My prompt was flagged because the title was "Smack my Banghra up"? Really?
Ob this is very interesting: https://app.suno.ai/song/2be7befe-2d2b-4de1-bf54-1392a8505748 v3, I am sorry, the prompt isn’t meant to be racist, I am trying to push the generator to its limits. The prompt „Classic indian bollywood song about tikka masala. Indian song, hindu language”
pop song. Listen and make your own with Suno.
Continuations from a specific time seem to be getting stuck for me. 1 out of 4 has completed, the rest are just stuck. I then did a continuation from the end and that worked... oh except it's just given me 2x 12 second clips covering only the middle of the lyrics. weird.
Always check if you continue if the time isn’t stuck on the previous continuation. I noticed time doesn’t reset.
crazy ... Button 'continue' doesn't work, clip ist only 8 seconds ??? serious?
What the f.... permanent songs without voice - lyrics are just ignored - why... whats happening???
Hello, I buyed the 10 dollars plan and I have only 2 running jobs at once
V3 the music track is way louder than the vocal tracks compared to V2
Everytime you click Generate Song... there will always be 2... but what if you click Generate Song serveral times? I'm pretty sure many set of 2 will run at the same time...
Oh it was not very clear in the subscription sub, I thought for every generation you'll get 5 my bad then ...
Yeah I seem to be having he same issue. kinda spending lots of credits re trying or ill get 1 that sounds like its playing in reverse lol
try my solution.. I think that fixed it #🪲┃bugs-web message
Love v3 but sooooo many bugs. Most songs stop around the 50 seconds mark, don't seem to care about the rest of the lyrics. "Continue from this song" is very hit or miss when compared to V2. Also v2 used to make it much easier to continue. Now I have to figure out new lyrics, re-enter a title, re-enter the prompt... I seem to remember that I didn't need to do any of that before. Sometimes a song will play, then I will get 20 seconds of silence and get a different version in the same file. I'm still having a blast though, and I can't wait to see where v3 goes. I'm aware it's an alpha. I just reaaaaally wish "Continue from this song" worked properly.
yeah I tried that when I saw it, was like grayed out at 00.55 so i changed it to 00.54 and theres no red box around it, but still seems to be having the issue.
Need help plis ?
a V3 problem that has cost me a ton of credits - for some reason it is largely ignoring the genre and style prompts, In V2 if I put "post punk gothic male vocal half spoken" I would often get it, in V3 it completely ignores that, and I get a gentle ballad with soulful vocals, the opposite of what I asked it to do.
seems to be random, as I am not getting the same glitches as you (apart from a song finishing and then starting up again a few seconds later), V3 is ignoring style/genre prompts for me. One thing that has improved, the instrumentals are breathtaking, and the general quality has improved massively.
V3 is constantly skipping words from the lyrics. Sometimes the melody is great but it's unusable because of the skipped words.
Also V2 seems to now invent words after it finishes singing what was provided in the input, this didn't happen before, so maybe it has something to do with V3 or other recent changes.
im trying to continue and i keep getting clips pure silence
If I pant to continue the song in V2 it’s refunding my credits and gives a 3 second song with a click noise
@tame latch @neat sail I made some notes on my experiences generating 500 songs or son on v3. Let me know if there are any questions or if you need sample. Thanks. Suno v3
Biggest improvements
- Quality much better
- Out of tune issues for singing and piano fixed
- Instruments like Piano, Guitar Solo, Saxophone sound much better
- Very stable response to music keywords, very good for prompting a specific style
- Much better audio quality in continuation
Issues in order of annoyance.
UI Issues
- 1000 characters input for custom lyrics is too short for 2 minutes. Even on very slow songs its only enough to get to 1:30-1:45 or so.
- Regenerating lyrics in custom mode has somtimes no effect you need to clear the box
- Clicking instrumental sometimes has no effect in combination with other refreshes. It seems to use the old state, when you try again/refresh it will work
- Scrollbar when playing does not respect the length of the song. Is stationary after one minute for longer songs
- Since today only allows to have one tab open, message about single session only allowed
- Continuing song with fields full filled like keywords and artist - title create a really big screen for the lyrics part
- Generate lyrics in simple mode ony generates one verse and chorus too little for v3
Model issues
- Custom mode. Most times it will never make it to 2 minutes (See 1.000 chars limit above)
- Continuation in custom mode, same isssue, even if I just have a chorus and verse left, it wont make it to the end.
- Prompting for specific instruments like [Saxophone] works very rarely in custom mode, works better in simple mode.
- Using simple mode generates songs that are too short. In many cases a break/silence is added inside of the song that can take seconds. It then continues with the song after but also may make some other song
- Missing lyrics seem to happen more in Custom mode then simple mode.
- In simple mode only prompting music keywords between brackets [Rock ballad, Guitar Solo] will make the keywords part of the lyrics. This was better in v2
- Simple mode often writes songs about coffee. LMM's Frequency Penalty not optimized?
- Singing is sometimes in another (Unknown)language
**** Updates **** - I checked over 50 instrumentals which are all 2:00 on the mark. Short tracks are caused by (short) lyrics.
Bug: The song duration should be set and displayed properly when I play the song. Currently it often changes and loads as the song progresses. If I start playing a song and see the initial duration as 14s, I will think it is an error.
I have spend approximately 300 credits just for Suno to not include my lyrics, make random noises and on top of that just giving me less than a minute and skip parts of the lyrics
[20:41]
How do I fix this?
If I thumb down a track that is still generating, it pops back into the list and the thumbs down is gone.
"Since today only allows to have one tab open, message about single session only allowed"
I don't experience this. That would greatly hinder my work. I'm on Windows 10, Firefox.
This needs pinned - Great breakdown here of issues thus far without having to echo it and get this text lost.
And vice versa, you see 1:00 and it ends up as 0:18 in length you start to play it through then it gets to the 'actual' duration and stops, a lot of wasted time there.
and the credit just getting wasted like a str&p3%r Joint . At this rate i might as well go to a real one
interesting , so i wanted a continuation to a section of a track i started with V3, V3 constantly hallucinating and i have spent about 300 credits but switched the continuation to V2 and it got it right but obviously the sound quality is noticably different so cant use it. e.g track i was continining with V3 https://app.suno.ai/song/7218375d-721e-4ff1-b0d5-71b8636d903f. then the V2 version https://app.suno.ai/song/4116aa67-f7c3-44fd-ae7b-237e5e12da58 of part 3. Then same V3 version in part 3. https://app.suno.ai/song/85cfebef-e826-4ca6-8cad-017417195b60 and https://app.suno.ai/song/063b50df-9512-4c64-8cea-9b6b9f3d21da i have deleted loads . Its singing when its not supposed to . V2 just got it needed to keep rhyming . I also notice V3 generates quick short versions whereas V2 takes time.
slow rap gritty Brooklyn black song. Listen and make your own with Suno.
slow rap gritty Brooklyn black song. Listen and make your own with Suno.
slow rap gritty Brooklyn black song. Listen and make your own with Suno.
"bluegrass" tag seems to reliably hallucinate vocals in Instrumental mode
try to type in instrumental and the instrument you want in there to supress
its doing it in v3a "instrumental mode", I'm not sure how to discourage vocals in the style prompt
You can still do instrumentals in V3 the same way as in V2, using "instrumental" in the style tag and other prompts that create them.
ah, k... I would have thought "instrumental" would already be baked into instrumental mode
you can always enforce it (even multiple times)
It's ultimately still just AI prompts and sometimes the AI intereprets things in surprising ways
yeah, seems like v3 is more sensitive to vocal/instrumental reinforcement in the style prompt, heh
BTW, if you *want * hallucinated vocals, try adding ", vocals" to your style prompt.
bad guidance! lol
what about say... backup vocals ?
It's SO GOOD though, really not bad! https://app.suno.ai/song/f2064d32-e131-4797-a1dc-2e0d0049af47/
indie folk, vocals, vocals, vocals song. Listen and make your own with Suno.
In an instrumental? That probably requires a fully custom lyrics prompt using [instrumental] but I'm not sure off the top of my head how to make that work
cuz on day1 i did a bellydance instrumental, and it had some hallucinated backup vocals at one point, and it was really nice, but it gave up on them and I couldnt think of a prompt to encourage them
I think there probably is a prompt like "vocals" that may encourage them, but it's not something I have seen yet. But you can do it with prompting in the lyrics field and using "instrumental" in the style field and [instrumental] tags in lyrics. And then writing out (oh oh yeah yeah) or whatnot - though not totally reliably.
In V3 if you click on the down thumb the song vanishes from view! I had to dig into my library to locate it, and even then it refuses to appear on my work page and I have to keep going back to my library page to work on it.
"nerd" is blocked in the Style of Music description. I get a warning it would be an artist name. :/
Is this the right place for Alpha feedback?
- I'd like to thumbs-down without hiding songs.
- I would love if the create page list of generations would scroll as far as a like. Currently just showing the last 10-12 generations.
3.. When starting a new song it feels strange that the "continue" feature remains checked and you have to remember to turn it off. - My kingdom for stem separation.
- I feel it's a bit harder to generate weird surprising songs in v3. Could there be a chaos setting?
- A nested folder structure visualizer for the generations would be so nice. Even something just like a windows folder browser would be amazing.
- great minds, just posted about this above.
- Also that the library page and the creation page were in the same order, currently library is top down and work page is bottom up - very confusing.
- I noticed that, thought I was doing something wrong.
- I can do that in another app. Don't need it.
- I love "hallucinations" etc, V3 does seem a little safer, which oddly, is a disappointment.
- Not bothered about visualisations, in fact if they dropped the whole "video" thing I wouldn't care. Folders for finished songs etc would be great.
- Trash is in chronological order whereas Current and Library are in reverse chronological order. Make it make consistent.
I can't get anything that I can use using V3. It frequently skips words or pronounces them funny. I've wasted more than 500 credits trying to find any song that doesn't skip words, but so far, no luck. I would've preferred 1.20 minute songs for alpha since the chance of my songs skipping words would've been much lower than the 2 minute songs that V3 creates. Maybe Suno should offer 50% credit off, if not more, for those creating songs using V3. We're literally paying to find bugs and report them. 😦
That's why it is alpha... it's up to you to test it.. you can still use V2. They've been very transparent that it is not ready and can expect behavior like this... @still leaf
Okay, here's another related one
https://app.suno.ai/song/ad650d83-85da-439a-947b-4d551e6d5aee
female vocalist, southern rock, energetic, pop counter, rebellious, confident song. Listen and make your own with Suno.
a shame too, because it's like 99% of a perfect take
in v3, when trying to complete the song stitching by placing the last part of the missing lyrics with [fade out], the system comes back with error and potential copyright infringement, no matter how many times you attempt. Perhaps the system doesn't understand what to do with there's short lyrics.
are we supposed to get the credits back as part of their alpha tester?
Everyone got extra 300 credits to play with V3 alpha, then you're on your own if you still want to try V3
My credits disappeared off my account, I had over a thousand and now it’s down to 40.
I don't know if this has been mentioned before, but if you load the web app and don't switcht the 'instrumental' switch on and off again you'll get an instrumental even if the switch is off by default (v3)
At around 0:30, this track hallucinated lyrics about female anatomy...
https://app.suno.ai/song/5d09e1d1-0672-4c96-bf00-b0781de19161
Trap, hip hop, confident male vocal without autotune song. Listen and make your own with Suno.
Yes it is included here: #🪲┃bugs-web message
It seems impossible to recreate the beginning of a song without losing its style. So, imagine you have a perfect song but a bad intro. If you choose to continue from 00:01 (since starting at 00:00 isn't possible), you end up with a completely new song. Could this be solved by allowing us to recreate the song with a slightly different seed?
not enough credits
- Remix doesn't sound like it's going to regenerate, or re-roll. It sounds like something existing is going to be transformed as in a remixed song, or that multiple things are going to be combined (mixed). Feels like a better word would help here? "Re-generate" or "Re-roll" or even "Again with random seed".
- Some way to see what's alreay been downloaded would be great, as well as some way to see when a full song has been generated with a link to it?
V3 is having an issue ending a song. I mostly do instrumentals and I've had to uncheck "Instrumental" and spam the custom box with [end] about 20 or more times, and even then it is hit or miss if it will end a song.
- a case for stem separation-- I'm speaking of native separation. I've tested a dozen tools for stem separation and all of them even the best have a cost in quality (for one thing they all seem to create a shelf on the high end, losing fidelity.). By the way, I use lalal.ai and "gaudio". Anyone have better?
But during the suno process, if there's ever a point where things are separate under the hood and mixed together. ...even just vocals and all else, maintaining that is super valuable (could even be something i run as a post process, but again, only useful if it's giving us some under the hood, native separation without the usual nyquist hit.
Check if your payment failed.
I agree (I use lala), there is still some way to go before separation apps (at least those available to us mere mortals) become useful in a practical sense
So today I've got this annoying v3 issue that every clip I generate is being padded out to the max clip length by repeating (sometimes random) lines from the chorus or the verse. Including the end of a song that went all the way to silence, then started back up to play an instrumental that went to silence, then started up again to sing part of one line. It seems that it's absolutely determined that it's going to deliver the max length clip no matter what is indicated by the lyrics. I've managed to splice the song together by using the "continue from" to chop the 2-5 seconds of unwanted lyrics off the end of each clip, and I'll manually edit the end of song to trim the 45 seconds of nonsense... but this is different from what I've seen in the past (random early terminations of clips before max length and before the lyrics are done)
It happens to me often, not just today, the bot will just repeat lines from different verses until it reaches the max length. Regarding the end of the song , going to silence and then start back up, I also noticed this today for the first time and it happened to basically all my song endings. I will have to cut them manually in my DAW.
Do the credits not carry over?
No, not if you are on a monthly plan, because it is like it says.. monthly
hmm .. can't seem to get any clip longer than 1mn now and always ending in the middle of the last word?! (v3)
fun ... the last one worked ... 2mn ... only voice, no music 😉
I have been a premium user from close to day one. V3 is better in one area. Clarity. All other areas it is worse than V2. Making it a step down from V2 in total. After spending over $200 in total I have come to the conclusion that I will not be able to recoup any of that money. Not due to sound quality but due to people not wanting to pay for A.I. generated music. They love it but being so much music is free they just do not feel the need to pay for generated music. Unless you have thousands of dollars to advertise and push it out to a wider audience. Social media by default will not suggest music or video to others due to computing costs. Images do well due to not costing much to push too many accounts. Hence the major push in socials for images. The consolation prize is I now have 150 music videos I can enjoy and share with friends and family. Thanks, Suno team for your efforts.
People pay for the whole package. Not just the music. You need to become the producer, show your personality and have a presence on social media. You need to have a consistent sound so you will get fans. Then they will pay for your music. 😉
Of course it is my personal opinion.
It's so trite and feels lame to say but 3Daizy is absolutely right
People don't buy talent
They buy personalities
Wait...you need to have a consistent sound?
😦
As I said, it’s my opinion. So I might be wrong
I don't really agree with the consistent sound part if your ~personality~ is strong enough
and as we move forward the entire concept of 'an album' gets weaker
Producers are usually known for their specific styles
In the traditional music place for sure
I don't think Suno is there though, I think it's in a different, much more unconstrained place
This is casting way into the future though and I'm just some jerk on the internet
We are all learning to prompt and talk to the models.. that’s a style too
Not just music styles
Yeah so I can see a 'producer' in this space to just be someone that can manipulate the machine well
The consistency becomes that ability rather than the style people can hear
Right
It's real fun to be on the bleeding edge though
Related to that, it sure is fun to get your stuff on the 'explore' tab
I'd like to think I'm not so vain but uh
I'm not 
Feels good
and related to that now I feel I do have to keep a consistent style because it's what my brain says I should do
Prompt: Create a bossa nova lounge song about a fictional place "Isla del Fuego"
Copyright infringement: Fuego is an artist.
Suddenly the island name is 'fwego' 
I thought i had it figured out, on how to make V3 work again but it stopped doing good parts again =/
I can make a good first and seccond part if i forced [Build] in the end so it would not chip the Chorus, but later it stopped helping with that
Can't you continue again and it'd finish the chorus? I ran into cases like that and I continued and it continued through the Chorus and finished it.
How on earth can I get a copyright prompt on a part 2 when all the lyrics were in part 1 (and worked!)
I think if people are using suno in the hope of making money then 99% of them will be disappointed. I am a premium user but use it for enjoyment, I have been making music since 1965 but only made any money for a brief period, for decades it has been relegated to a pastime, a hobby if you will. Yes, I plonk it all on Spotify hoping someone (anyone?) will listen, but having had only one royalty payment from Spotify in 10 years will not pay the bills.
That's why you build on.
Music videos. Movies. Try to make it with other stuff. Create virtual bands. (Granted without a way to reprompt the same vocals this is harder than it should be.)
It's never just the music. It's the music plus all that other stuff. Eventually I hope to get good enough to make a short film using my songs.
But as much as we love the music it's almost never been just about the music. It's the lifestyle. It's the aesthetic. It's the artwork. It's the whole illusion.
Yes it would be nice to retain the same vocalist, in the absence I am tending to create themed albums (jazz, blues, reggae, goth, post punk etc) that might appeal to fans of those genres.
Yeah I'm leaning towards compilations now myself.
I thought I could get away with multiple singers in a band and trying to aim close "enough" to pretend it was the same person. I even think I sort of did it with some of my stuff. But at the end of the day it's a collection of one hit wonders without a way to have continuity.
My music generation is failing now, endlessly spinning or ending at 10-20 seconds.
The ones that are endlessly spinning return this when trying to go to it's page:
Application error: a server-side exception has occurred (see the server logs for more information).
Digest: 3496197124
Is this a known bug with v3 alpha?
Build a musical act then have "ft. (person)
Afaik I'm still being charged for those generations
That way all the singers being different, who cares!
I saw a YT video of an AI that -only- sings with an AI vocalist... you have to submit either a vocal recording you want it to replace, or a midi file with the tune + your own lyrics. It can't produce music on it's own like Suno... but... if you REALLY are hung up on having consistent vocals, you could take a song from Suno, feed it through one of the AI vocal stem splitters, feed the vocal to that other site, then merge back that vocal to the non-vocal stem... it's a lot of work and how much quality you're going to lose from the manual stem split could be an issue... but it's a way to get completely consistent vocals
You get charged when they finish
The ones that do finish only finish between 10 to 20, ending early and not finishing the lyrics unfortunately
Any attempt for the third part, because the chorus wasn't finished in the second part, altered the whole genre for some reason
I had to change "kill" to "hurt" and it passed, lyric didn't seem right. It's artistic vandalism.
Phonetics can be your friend.
Some violence is permited as i've seen for example, but not much. I changed died for fallen for example more them i could count.
Hey everyone
Do you have any problems with V3 Alpha
I dont see it suddenly in my account
But my friend does on his account
any ideas?
no problems here, have you tried the drop down menu where it says "v2"?
there is no drop down menu
i used it before nut it desapeared
i've seen people complain about it, and they sent ticket to the team to help them trhough. On account you have premium, as far as i know you need it to have access to V3 Alpha
i should definitly
whre to send the ticket?
something like this, seek a moderator or something i think
I Don't think this is intentional
I think moderation system is going haywire again I did have this word in my songs in the past
It's generally jumps between words so much
I mean the filter
yeah
V3 vocals seems to have quite a bit of distortion.
@buoyant flume @royal mason @summer wraith @delicate iron
I recently discovered that you can do vocal chops with certain types of input under the lyrics section. It's one of my favorite features that I discovered. I know it might not have been intentional but I am just reaching out here in hopes that you guys will not remove the capability to do this as you guys make further improvements. It's so wonderful.
example (look at the lyrics + listen to the correlating part) - https://app.suno.ai/song/f5b2556d-811a-4cde-af20-2149230d5a61
Devs did you do anything in the engine just now? updated it or something?? it went down for 3 minutes and now everything is out of tune!!
Sometimes it sounds like the music or vocals get muddy. How can I ensure crisp music tracks for mastering and avoid that sound or distortion?
Nobody knows...
If you find the answers, you will win a million dollar
Muddy isn't new, but out of tune just happened now right after it went down and came back up
Every new generation builds up bit more distortion etc.. growing pains
yea, the longer you make it, the more it goes off track but I'm talking about first versions sounding muddy or not crisp or clear
If you keep songs minimalistic in style, it helps..
Tennatively I see this happen when trying to mash really different genres together
Exactly what 3Daizy said yeah
can't believe it happens so soon in V3 life span. In v2 it took months and never got fixed, hardly got anything not out of tune for about 45 days. now it suddenly happened in V3 exactly when it came back up, I generated tons, everything was fine. engine went down for 3 minutes, comes back up, boom, out of tune. that's why I ask if it was some update that can cause it
If nobody else is mentioning it
They are still tweaking under the hood.. so it's good to stay vocal about these things 👍
well, peopele didn't mention anything about V2 being out of tune, but if you listened to what people uploaded, 95% of it was out of tune
this is just an opening music I was trying to play with to get a beat
True, but people like garbage so 
and it sounds muddy I think
depends on the style.. like with alot of problems
My issue with v2 was not able to picking he time to start from and now, with v3, v2 sounds so muddy, fuzzy, compressed almost
it happened just now as I was generating, same style, exactly, went down, back up, all out of tune
I try everyday certain styles etc.. and I do see improvement. The first day I tried it, it refused to honor verses and chorus.. that is better today
But I am sure sometimes it's one step forward, and two steps back
@summer wraith shot you a dm
Something is defenetly wrong, or they just updated the engine and it's "starting new". cuz I got 3 generations almost e3xactly the same music, all first ones, never seen that, I thoguht I listened to the same track by mistake, but no.
and it's not a 2nd generation (part 2)?
my clip was a 1st generation with simple instructions
no, ill show you, check this out, all 3 sounds exactly the same, all 1st generation . I didn't cut one song to sound the same, all first 5 seconds of 3 out of 4 different generations together
All 3 of those have different instrumentals
They aren't vastly different, but they are different in the notes played
small difference, but start exactly the same
Agree there
i don't know for sure since this could also be a faulty download... I might say something stupid but I will say it anyway.. If I were you I would clear my cache and relog and see if that helps
never heard 2 so similar, and those 3 made in the same minute.... what are the odds
For me feels as cache
feels like engine restarted somehow
I don't have it
So that theory seems not fully right
What I am trying to say.. it's best to first rule out everything on your side
Well I also have been having failed generations. Even still, they either fail or stop between 10-20 seconds. Something is going on
This was 9 parts of V3 mashed together, it's not terrible but definitely can use some TLC.
I will say the V3 "start from" was a life saver for all the words that were cut off, even only 4 lines for a verse would cut off the last word repeatedly
This isn’t deathmetal https://app.suno.ai/song/8445e7d3-9993-4102-829a-ab944f40efd2
You might need to paint the music by using parent genres and other terms to mix together something which works.
I now repeatedly (3 out 0f 4 times) get that no vocals are generated even though I have added lyrics in the custom mode. It almost seems like the new 'Instrumental' toggle was selected but it wasn't. And i get this on both V3 and V2 songs. It started when I was trying out V3 but that may have been a coincidence.
I just generated a future bass pop song with lyrics and all that came out was 1:20 of instrumental and no singing at all. I hope this bug gets fixed. It doesn't generate singing even though i put "male vocals" in the prompt.
@halcyon violet exactly what I was just typing. Had this for a few days already actually.
Yeah, this didn't occur before. I just want to generate a short intro for the song before the verse and it just comes out as a whole instrumental with no singing at all. It's annoying sometimes and a waste of credits.
Hi on my side i have the feeling that V3 want to always creating something new and doesn't take into acount the song it self.
V2 understand the song and suggest verse with the same rythme. V3 want to add new rythme on a verse that's kind of anoying when you try to continue a song.
Also V3 keep in memory the previous text writted even when we change the lyrics. so doesn't know what to do.
V2 understand better the lyrics and how to handle the new lyrics. even if they have been changed during the process.
V3 consistency is not good enough to be honest.
Also lots of bug with V3 generate to much song of 08 - 12 sec without any meaning
Never needed to with v2 it was amazing 😭
lulz..I started doing it with v2 to make it more amazing...
¯_(ツ)_/¯
Different workflows for different people?
@summer wraith can I schedule a brief call with a team member? Is this possible? Huge potential use case, no bugs. I am a producer. I just want to have a brief back and forth for a few mins.
I'm something of a producer myself
is there a refund process for failed generations (I'm assuming it's failed at this point, it's taken longer than any other... icon spinning forever no play/download)?
under 10 seconds and unfinished generations are not charged.
I made over 150 music videos of the quality no one else can produce. And I am on social media with a following of people that love my music. None of that matters one bit. Videos not just music. The whole package. You can be a cheerleader all you like it will not change the facts.
i am paying my subscription but i cant find the option to change to v3, does anyone had the same issue?
I keep getting a prompt is likely copyrighted error and it is not! I'm just pasting the lyrics from the chorus from the beginning of the song to this second generation and it keeps saying that my prompt may be copyrighted which it isn't. I'm using the output that the lyrics gave me and it worked the first time when generating the lyrics for the first chorus but it won't generate the second chorus. Any fix for this or is this just some stupid copyright detection bot messing up?
lmao so hey check out this 3:50 generation I just had, single shot:
https://app.suno.ai/song/53b1bc43-0560-4405-9198-33a96c1b14ce
synthesizer virtuoso, complex melodic, electronica, downtempo, blues, trip hop, sophisticated, Japanese synthwave song. Listen and make your own with Suno.
It's weird as heck
pat this may or may not be important at all.
It kind of sounds like two generations layered over each other?
single shot as in it's supposed to be a 2:00 minute v3 generation but for some reason went to 3:50
how?
No idea, only the devs could answer that
Someone the other day said it happened to them, didn't link anything
What did you write that made this happen?
It's in the link
"synthesizer virtuoso, complex melodic, electronica, downtempo, blues, trip hop, sophisticated, Japanese synthwave"
Instrumental, v3
Yeah it just happened, I didn't do anything different than I normally would
It's a pretty good thing in my opinion.
How can I switch from v2 to v3
Are you a paying subscriber?
If so it's just a little box underneath the title
If you are on a free plan then v3 Alpha isn't available to you
the "continue from [timestamp]" box needs another background. 1500 credits in and I only found out about it because someone mentioned it on General
I wonder if V3 skips words in English too. I was working on a Romanian language album when V3 came out so now I'm only generating in Romanian. Most songs are really nice, clearly an improvement over V2 but it skips words in like 49 out of 50 generations. It's basically a waste of credits at the moment. Does anyone else have this issue?
DEVS please please check what happened yesterday when engine went down for 3 minutes, cuz since it came back singing is out of tune. exactly like the V2 issue. something clearly changed. all was fine, engine stops for few minutes, comes back , and strange bugs happen : 1) I reported yesterday I got 3 generations very very similar out of 4. also added examples. almost identical tune. happened right after engine back. and then 2) singing out of tune even now. please. don't let the issues that broke V2 happen so early in V3.
Same, but I only "wasted" 800 credits. 😂
Still had a blast doing so 😉
Not sure if this is what you mean by out of tune (probably not) but I also get a lot of parts where after two or three verses the singer will start slowing down, the instrumentals will fade out and then suddenly total silence.. for some seconds and then it all starts again (voice, instrumentals).
Examples:
https://app.suno.ai/song/c0d0dcb0-a2fd-417f-b023-4a751f7fd8b6
https://app.suno.ai/song/56c8a53a-1179-43e2-8774-4f90896b3ccc
They are not in English but the problem is easily spotted.
pop-rock gothic slow piano ballad female vocals song. Listen and make your own with Suno.
pop-rock gothic slow piano ballad female vocals song. Listen and make your own with Suno.
maybe they corrected the AI chipping the end, Yesterday i was doing really big prompts with good results
Are you sure you are having multiple parts? Because you only see it then.
With only 1 part it is not an option. And I see it. So it’s still there
oh no
it's me
nm I'm stupid and haven't had my coffee yet
damnit, that was a really good take too
cries now
lol
Aww sorry to hear that.. enjoy your coffee 😉
I feel like I'm spamming with bug reports, sorry for that. I've actually posted this before but I hoped it got fixed in V3 - it's still there. Question for devs: what is the context of the bot considering the new feature to 'continue from' certain parts? I mean, in V2 the bot was only aware of the previous part (so basically we had to carry over the chorus in every part) but now if I continue from the first 5 seconds of a part, I expect the context to be longer than those 5 seconds, otherwise the whole style of the song gets messed up. I expect it to be something fixed and hopefully at least 2 minutes, so no matter how many parts I have, if the chorus can be found in the last, let's say, 2 minutes, the bot will be able to replicate it.
Suno begins to sing in another language unknown to me, not according to the text. Part1: https://app.suno.ai/song/8097a7e6-f686-42d9-b751-2833d96cd3f3 Part2: https://app.suno.ai/song/50405091-66d0-47ab-8b44-44bbe04a4bff when you re-create the continuation of the track, the problem remains the same.
EDM, Rock, Dark, Emotional, Female song. Listen and make your own with Suno.
EDM, Rock, Dark, Emotional, Female song. Listen and make your own with Suno.
After a cache clear it worked, not sure if it's just a coincidence or not.
it is some stupid copyright detection bot messing up, sorry. we expect to have a major update on this soon.
My "Create" button is disabled and I can't do anything. I have lots of credits. Seems to only be in "custom" mode.
Please , for the love of god, check your engine, something went wrong exactly after it went down yesterday. 3,000 credits in V3 all is fine, engine goes down, come back up, and this is every generation, exactly like the V2 issue. and exactly like the V2 issue, it happens when there is music (so just beats or lower music there is little issue with out of tune, but with rock, drums, louder music, singing is** out of tune**)
Seems to work again, not sure what happened
@unreal radish i discovered like a few different potentianally unintended things that are god-tier features and i really don't want you guys to remove them going forward
main thing is, PLEASE do not remove the ability to use vocaloid vocals in instrumental mode. it implements vocals in a MUCH more natural way(for good production/vocal balance) imo, atleast for this genre. its insane. here's a couple examples
https://app.suno.ai/song/228eb14d-0c10-4592-9e92-f5182e3ef782
https://app.suno.ai/song/a27639eb-c27f-410d-8b05-afc4c5e6aabb
i also found a way to make vocal chops through custom lyrics in non-instrumental mode which is so good (mentioned this before in another channel)
https://app.suno.ai/song/f5b2556d-811a-4cde-af20-2149230d5a61
V3's most persistent problem is the lack of "intros", no matter how I word it, the lack of any intros is almost 100% (unlike V2), it always starts with the words, often acapella, and it is not always possible to bolt on a start in post if there isn't a suitable passage later in the song.
is there any chance of v3 ever being able to continue from v1? I'd LOVE to remaster some of my old songs.
Also, a feature request-- could we have a current song drop down? And then when you're working on a song it would only list the parts for the one song in the create and library tabs? If you've got lots of songs going this could help a lot.
yeah no plans to remove these, are you by chance also the author of: https://app.suno.ai/song/af9e4aa7-3607-4a3a-98ea-cded0776647f/
if not I feel like you guys should hang out
definitely a rare bug, if something retries on our end this could happen and then the two gens get layered
we're thinking about how to address this / make this clearer, but you'll want to use "Get Full Song" before using continue from if you want something to have more history (especially when using continue from)
a workaround while we tune the alpha is - use instrumental mode or something else to get an intro, then use continue from to add lyrics
Why is my song continuation getting "flagged" for moderation over and over?
Ive changed the lyrics 4 times, changed the continue point, nothing works
Glad to see I'm not the only one who's noticed this... and that it's NOT just a psychological thing, as was suggested by someone...
Not only that, but I've made enough tests myself to notice that the quality is off overall
Example 1 above
Example 2 below
Example 3 below
The labels for these two was swapped on purpose by me
Voices sound out of tune, the track seems muddy, and it's just not good as it was a while back, really hoping it goes back to normal
Another one who agrees
LOL nah, i would love to give u access to my page though - it also seems like Arabic vocals might work better in instrumental by the way. Seems to integrate them more naturally [might be the case with most languages, im still digging around]
also awesome!!!
Great use of music!
Recently v3 came out and I was using it well for a day and then suddenly the version change disappeared and I was left with v2
(I’m using premium)
Also, I can‘t get up to 10 jobs and it stops creating at 4
V3 is currently only rendering 2 to 5 seconds and says refunded but takes 5 tokens each time with nothing to show for it.
Seems the issue with V3 is continuing. Forgot to add the part to continue from and it worked so seems no go for adding the last part 4
When I create instrumentals with v3 it sometimes puts in language. There are no lyrics to read and I'm not sure if it is a known language but they sound different from version two. You can make out words and it's kind of beautiful
@unreal radish https://app.suno.ai/song/b82a5d0e-e7d9-4080-a610-cfb15c386b14
another example. god tier imo [lyrics in instrumental, arabic this time though]
I tried to generate Style Paso Doble Intrumental but it won't let me, saying Style contains singer name 🤔
Doble
ty, will be fixed shortly
a silent clip... should get a refund right?
https://app.suno.ai/song/752b5b38-b564-4589-ae54-bbab4ccdc6bb
Gangster Rap, High Quality, Male Singer song. Listen and make your own with Suno.
you can "get whole song", in this case it'll take it as the context. But yeah that's annoying to be doing this on every generation
Thanks. Actually I was doing that and it didn't work (tried over 10 times). I then reset my browser cache, don't know if it mattered, and it worked after.
Ok so out of 2500 credits purchased, about 500 were worth it so far, i tried different words and shorter genres for each song i generated, say i did them in groups of 10, about 4 generations were usuable for each song. I reported them and i have yet to see a single credit back. Sending an email, but I feel like this is more than a bug for me. This didnt happen on v2 for me, theres no way i should blow through 100 credits a day just trying to do a continuation that doesnt hallucinate. I even tried leaving the genre and lyrics empty and just using the instrumental, same issue.
For me it's at least 9 out of 10 that are hallucinating. On one hand it's alpha, basically we're testers when using it, on the other I wish there was something like an alpha program for testing so if we apply we get like, I don't know, say 5000 credits to spend testing it.
When I started doing genre fusionslike "indie rock" or "punk soul" it was 9/10 then i just decided on doing art pop as the gen and that works beautifully until it loses its mine. I agree, I also think for us testers the thumbs up and down button should completely delete the song if it wasnt downloaded, and refund the credit. I feel like that would not only get rid of the bad ones on the server quicker, it would be easier to weed them out from the good ones with no problem.
Not sure whats the issue but its stuttering just to playback previously generated songs, and I AGREE WITH AN EARLIER POST about how easily you can blow through 100 credits just to get something USABLE in V3. And what happened to putting instructions in the “[ ]”? *so helpful before almost invisible now 🤷🏾
huh, interesting that clearing the cache worked.
in my case, it stops working when it has a significant change in music. E.g., if you have an intro, a verse, and a chorus, it works fine, but as soon as you veer it off into a bridge or something, it forgets some or all of them.
Perhaps that's some sort of "memory" the neural net has developed internally.
I also had a bridge in between. Maybe clearing the cache didn't help, maybe it was just luck after multiple generations.
Idea for the devs (me and my programmer mind): give us the ability to save parts of the song in variable like structures (eg, have some parts of the songs in like curly braces and give them a name) so when we want to replicate them we just place the variable again in the lyrics. This doesn't even need to be interpreted as is by the AI I assume, could be a simple procedural parsing and some logic (directions to the AI) over it.
This would support for example two different kinds of choruses (something that can be found in real life, long ballads) or to finish the song in the same style it began (which is impossible right now to tell to the bot).
Last night v3 performing very well. But today, it's terrible. Is there an issue now? 500 credits were used, and we still couldn't get the song we wanted. The more we generate, the worse the song becomes."
It's working for me. Try closing the browser and reopening it, clearing cache, etc.
And keep in mind.. if you are using V3, that it is Alpha.. You can always switch back to V2
If you flag lyrics (for whatever reason ... the bar seems to be quite low) it would be REALLY nice if you didn't also charge for the failed generation
Oh so maintainance work in middle of generating track?
you don't get charged
Oh ... Weird. I would have sworn my credits went down. Anyway, I modified the lyrics
If it comes up with an error message, that kicks in before any credits are taken off.
When V3 released it performed flawlessly, something definitely happened on their end
If you scroll up a bit, you'll see what I said
Quality has decreased, the AI seems less capable
Hoping they'll fix it
It seems to me to be so. I also generated an amazing first part of a song from the first 5 generations. No skipped or wrongly read words. Amazing quality. And then yesterday I used around 1000 credits to get an acceptable song (after some editing).
Exactly, you could be hyper lazy and the AI would spit out amazing results
Man, this "Tribal Eurodance" thing
The prompt used was
Sussy Reddit Blackpilled Suicidecore It's over tribal eurobeat
So yeah, something did happen, I think devs are aware of it
I typed in absolute nonsense, and it delivered
Now it won't, it just sounds muddy and muffled and weird
We'll have to wait and see
why is my billing say my plan ends the 29th but i upgraded the 15th
shouldnt it be the 15th??? so im losing a bunch of points for what?
@solar grove
I will highlight this in the moderator's section for visibility but you should send an email to billing@suno.ai about this right away.
ok
thats the point they want your monie
what are you talking about?
about nothing dont listen to me
mods need to ban idiots in here smh
u are rude
hi, your subscription should be prorated for this month-- sorry, we will make this more transparent. please feel free to doublecheck your billing for next month and email billing@suno.ai if anything doesn't look right-- we'll make you whole if not
this makes no sense . my points are gonna go away then i have to pay again and lose my points ??? i paid on the 15th
this is an issue with how stripe does billing but it's not very intuitive and it's on our roadmap to iron this out
???
so you are telling me my points that are unused are gonna go away ??? i pay for premier , this is not good support . i emailed and got a generic response
your charge next month will be discounted according to any unused credits this month, so that you're not paying for credits you didn't use this month. Apologies for the confusion, we're working on smoothing it out.
i still have about 9000 credits i want able to use yet
if you don't use them this billing cycle, we'll essentially buy them back from you in the next cycle. Again, please feel free to email billing@suno.ai if your next charge doesn't look right.
Thanks for bearing with us as we iron these issues out!
we'll fix this for you - DM me your account email
Generations are currently enabled for Pro and Premier subscribers while we work through more maintenance?
This means subscribers (who pay) will all be able to continue... while they are still working on free accounts
I'm not a free account
then fill in your prompt and start cranking out some sweet music.. just because you see that banner doesn't mean it is not working for you
just did
Hello, i am tryaing to sign up on the page on my browser but it keep saying that bot action is detected and i should contact support. Any suggestions?
Are you using a VPN?
Could you let me know if now that only Pro and Premier members are the exclusive users of SUNO, has the quality improved?
I've often thought that the instances of generating low-quality songs might be because many people use SUNO. Now that it's restricted to Pro and Premier users, it would be good to confirm this theory.
Quality was insane when V3 released, it's bad right now
I'd say wait until it's fixed
Scroll up a bit and you'll see how it changed
I disagree.. I am actively testing all kind of genres for a tutorial and they turn out pretty sweet... so feels more in your genre @sly heart or just bad luck
I'm afraid it's not bad luck at all, Suno was consistently spitting out banger after banger for me and a friend, but something did happen.
Check out what I sent above
that happens sometimes.. I had it the first week too.. it comes in waves.. it's just like that..
So not reliable?
Yes.. I saw above.. but still stand by my words 🤷♀️
There are more people making good music.. and some are complaining..
I heard for some that clearing cache helped.. please try that
Ok, I will
Even though it's been said by multiple people, I'll see if clearing cache does something
One sec
count how many have posted their songs.. they will beat the few above
Would you be willing to run a quick test for me? And input the prompt "among us CHINESE cyberpunk synthwave happy folk music trap dance beat shoegaze" in the style box to see what the AI comes up with?
Instrumental
that's a mouth full.. lol
okay.. let me see if I got some luck on my hand
Ok
oh sounds pretty sweet.. letting it render
o
And synthwavey
Yeah well if you are complaining not doing exactly the style, that is something that can shift a little from time to time.. they are actively working on how to communicate with the model
This is what this prompt gave out a couple days ago
But you responded on someone who asked about the quality.. and the quality is better than with V2.. and yes it is very alpha
You can hear the Chinese, the Synthwave, the Folk Music, the Trap drumlines and the Shoegaze nearing the end"
https://app.suno.ai/song/cf18978a-84d0-478d-8f96-41ba17747f8b
This is today's. Very cyberpunk, but that's it.
among us CHINESE cyberpunk synthwave happy folk music trap dance beat shoegaze song. Listen and make your own with Suno.
yeah well still, like I said, you are really narrowing it down on your own wishes, and of course they want to accommedate you.. but so many people so many wishes. It is not just dailing a knob left or right. I am sure they have seen your remark but have to find a middle way.. It is still alpha, things will still change.
sometimes it's one step forward, 2 steps back 😉
I hope it does come back to how it was before, because i'm more than willing to dump a truckload of money into Suno, I've seen what it's capable of and it's more than amazing
I totally understand..
It actually left me speechless, the day V3 launched
Hm... I like them.
First the style with my usual prompting:
https://app.suno.ai/song/629e3674-e615-4f36-9583-6123d674595d
https://app.suno.ai/song/0a661810-7e6d-4708-a39e-0015ab792b88
Then with the instrumental slider on:
https://app.suno.ai/song/1250483d-fcc8-4ca5-8bdd-ab03642cf908
https://app.suno.ai/song/fa9c8b84-96eb-4798-a09d-4eeda956c44c
One comment I will make is that I used it as listed. No commas to separate the requested styles, effectively making it this huge runon setence of s style....
among us CHINESE cyberpunk synthwave happy folk music trap dance beat shoegaze song. Listen and make your own with Suno.
among us CHINESE cyberpunk synthwave happy folk music trap dance beat shoegaze song. Listen and make your own with Suno.
among us CHINESE cyberpunk synthwave happy folk music trap dance beat shoegaze song. Listen and make your own with Suno.
Don't give up.. give them some time
Making my best guesses at where comas would be:
among us, CHINESE cyberpunk, synthwave, happy folk music, trap, dance beat, shoegaze
Using just the style and the Instrumental slider:
https://app.suno.ai/song/7085aaa9-19dd-4972-9de9-236e369486bd
https://app.suno.ai/song/c03e9b6d-16f6-4e6c-b23c-35668dbc9e4e
I will, I trust them
Me too.. this is going to be EPIC 🤘
The original one didn't have commas either, it was just a very long sentence
.
My usual prompting with the commas:
https://app.suno.ai/song/49d3d7cf-82a4-46af-a7d2-729396b54cc6
https://app.suno.ai/song/33422d37-faf4-47c8-baaf-b53192be42ca
https://app.suno.ai/song/ef8d0cae-a5d2-4900-9844-1fb0171ed0b1
(One was blank for some reason.)
among us, CHINESE cyberpunk, synthwave, happy folk music, trap, dance beat, shoegaze song. Listen and make your own with Suno.
Well did the commas make a difference?
That's another data point.
Not sure, it was generated when V3 launched and there's no way to test it as it was
v3 sounds worse now. much worse then yesterday. out of tune is crazy. also i dont see what clearing cache have to do with it. does suno generate on my browser and not on his server?
v3 isnt giving me any instrumental intros at all
You know we have an instrumental option now right? I use that quite often to generate the best intros.. and continue from where I want
i dont need a full instrumental i just need a instrumental intro instead of it throwing the word instantly
Yes.. that what I am saying.. I generate an intro via instrumental, cut it after a few seconds and tell it were to continue
WOAH i didnt know you could choose where to start
Now you have way more control on your intro and song in general 😉
thank you!!!
The continue from @ time doesn't seem to be working. Took 10 credits and run 1 was a different song and run 2 was a few seconds and no refund.
In that case I advise you to post the parts here or send to support@suno.ai so they can look what is going wrong
I gave a thumbs down which I thought with V3, sent it to them?
That's what I've been told as well.... wish I can add a comment to a thumb down
I'm trying to generate a song with lyrics I created 100% my own and it keeps saying I can't generate music because it's copyrighted... is there any way to fix this?
I stopped trying as i'm not happy that we have to pay to test their new version just incase this happens again. There should be testing credits and they would get a better testing and make it so we can't download anything V3 until it's considered "ready" ish.
@whole bay i've seen others with this issue. depends on the word. sometimes spelling it fenetically or changing letters helps.
ok but i still get to keep the copyright of all my songs, right? (paid plan and own lyrics)
thank you so much for taking the time to answer me, drew!!
just find which word it breaks on. cheers
Tried a continue from in V3 at the end instead of the middle and the second run was only 10 seconds and unusable but still took my credit. I'll leave testing to others who don't mind losing credits and wait for it to be workable.
I forgot to mention that a continuation in V3 seems to change the vocal to a country style regardless of the style and doesn't seem to look at what it's continuing so they don't fit in a mix
Hi. does anyone know, what means elevated usage? I got that warning text, when I tried to generate a song, and I have that 50 credits.
and also this text: "Generations are currently enabled, but only for Pro and Premier subscribers."
It's only availabe for paying sub atm due to server issues. See >> #💬┃general-chat message
bug: on the create tab, when I click thumbs down, the song fragment I just downvoted still appears in the play order if I click play on the song fragment above it
you can't see it anymore, but it still plays
Two in a row of a continuation seemingly becoming just a brand new generation completely unrelated to the song I'm continuing from. It's abrupt and not blended together at all
Almost like switching the radio
Both of these, right at 1:50
https://app.suno.ai/song/f1d3f8f6-186f-4ca2-86d4-599085bd6ca6
https://app.suno.ai/song/7a306f2d-7a90-4a84-b479-3ab88c6d0d7d
New York, synthesizer bass, sophisticated, Japan, 80s, vaporwave, Island, Jungle song. Listen and make your own with Suno.
New York, synthesizer bass, sophisticated, Japan, 80s, vaporwave, Island, Jungle song. Listen and make your own with Suno.
Ok it seems like one of two of the continuations of this particular clip always result in one of the generations being completely unrelated
Ok I think it's because the particular clip I'm continuing from is confusing the model at the end. or rather, confusing me since I think there should be a smoother continuation of the song but the model probably does see a smooth transition. There is some sharp waves at the end there and it's probably continuing in a pit of silence I can't discern because my dumb slow human ears
Alright nevermind, even pushing the 'Continue From' time to a much more musical, not-quiet part of the song I still got a hard lack-of-transition with a continuation. Heard here at 1:44
https://app.suno.ai/song/ff267744-2024-4035-b40e-8305f06a39be
New York, synthesizer bass, sophisticated, Japan, 80s, vaporwave, Island, Jungle song. Listen and make your own with Suno.
Ok so it's happening with another, different clip. Since both of the clips I'm continuing from are from before the maintenance today I have to assume something has changed with the v3 Alpha which has this behavior. As heard here right at 2:00 mark https://app.suno.ai/song/e8360914-18fb-482a-a274-bfbb6d49b2bb
sophisticated, polyrhythmic vaporwave, chill, virtuoso song. Listen and make your own with Suno.