#๐โsuno-school
1 messages ยท Page 106 of 1
I would love to see the day where we can get stems for all the parts that are generated and being able to generate individual stems and then put it all back together.
techinical disscussion more like website disscusion
also im a moron when it comes to computer systems and what not
I think that's the main drive for the different prompts as well as the upgrades. So that A.I. music making can be more intentional.
With this V4 shimmering issue, as I dig through my old songs, watch what others do to "fix" it, I'm becoming more convinced that the effect is something that was introduced in an earlier version and V4 just enhances something that was already there. I took another song of mine and remixed it and it also has no shimmer.
I got this while using songwriting.
Alex, you legit need to wait 10 seconds between each link.
You can't bribe me with cookies. Nice try though.
could be issue be that microsoft is updating something
ChatGPT was down last night and ReMi still seemed to work
If ChatGPT tried to chastise ReMi for not taking lyric writing seriously, ReMi would say something akin to "you just don't understand me, old man"
Suno =/= OpenAI
From what I gather, ReMi is a small-scale LLM made in-house by Suno
So technically Personas should mimic only the voice, right?
It seems very likely the simple mode uses openai, and presumably the default lyric gen from custom mode might still be tied to openai.
Not meaning to counter what Alex had shared, just noting what furry had mentioned that ReMi wasn't impacted by the openai outage
Persona is just creating the same song over and over again tho
That's why i am confused
Same kind of quirks that we can run into with covers, tbh. Some personas are better than others, it seems.
re: ReMi not being impacted?
I've seen a couple of times where it seems that Suno just get's stuck and generates things that are almost exactly the same. when this happens, I'll generate a song that is completely opposite of what I'm trying to do and then when I come back to the one I'm working on and generate a new version, I'll usually get something different.
Technically I changed the genre from Symphonic metal to Power metal and added a few more stuff
I used the lyric generator back on v2 and I didn't like it so I've been using ChatGPT and I keep refiniing it till I get something that is consistant and I like. Then I ask ChatGPT to consolidate all my "instructions" into a single document and then it gives me one I can download. So then when I make a new song, I can just feed it that document and then the song.
I've also taken what ChatGPT has given me, uploaded the song that was created with it, and had ChatGPT analyze the result.
What happen V4?
Original without persona:
https://suno.com/song/a41c0e06-f5b6-4650-bcfd-3f750a747fc1
Symphonic Metal, Vocal duo Male Female, Guitars, Dark and Love vibes, song. Listen and make your own with Suno.
https://suno.com/song/0e1ab9cb-3d92-4942-b486-f2d052f5f5e2
One with persona
Power Metal, Vocal duo Male Female, Guitars, Dark and Love vibes, powerful and nostalgic vibes, doom eternal, doom 2016, mick song. Listen and make your own with Suno.
Basically the same dumb structure, and cover doesn't modify it a lot either.
I've just generated 4 new songs with V4. In all of them I have "Clear Vocals" as one of the prompts. NONE of them have a shimmering issue. I'm wondering if the "clear vocals" might have something to do with it?
Wait till you put cicada in excluded styles :))
Nope. I have been using clear vocals and V4 clanks away. It's not the Shimmer anymore. ๐คฃ If a song is longer than 4 minutes and you extend, its quality deteriorates. Badly.
Dude.. I'm dying here.. Can you imagine if the best fix for the V4 shimmering is just a -cicada
It was at the beginning
Now we have hissing issues
Ok.. who broght the cat to work? ๐
The devs who made us spend money to beta test their bs XD
I renew in a day or two and I don't have enough projects going to use all my credits up before it renews so I was having fun and trying to see if I could get a positive result.
I churn through 10k credits like a snowplow
At the end of the month I usually end up buying 30k more
.. and I thought 10 pages of music was a lot. lol.
Yeah man, it gets cluttered real fast -_- they have no sense of organization for us that try 100 times hoping of a different result
I don't even know how many pages I have.
Based on what I know of web programming, it might actually be easier for Suno to craft some kind of music library section that is separate. Essentially a stand alone area that is custom designed just to manage your library of songs.
Would be nice.
I just think that trying to integrate that level of library control within the existing layout and structure of the current site would be progammically more complex than just creating a stand alone area.
Two things I'd note about song structure and tags. If you are using tags inside your lyrics to control the song, one of these might actually be guiding Suno towards a more specific outcome. You might try removing all prompts just to see what happens.
Yeah. That's exactly what I try to do. Unfortunately I can't just take that into FL and isolate everything and have just the vocals as a cover and Suno completes it because of the 120s limitation. I mean, if the songs are 240s long, why not put the same limits on upload smh
The other thing is that it does sometimes seem to get stuck and only by generating something completely different and then coming back, have I been able to shake it loose.
Well, for me it seems it has to do with persona. Not really a Suno issue. I know how to get different stuff.
My guess is that the limit is more of a technical thing of how much the A.I. engine can actually handle. The amount of power these things use to generate a single song is quite a bit so upload file size restrictionis may have a different reason.
I've also noticed that with the persona's they can bring over more of the song the persona was based off of, instead of just the vocals and prompts. I've actually had more luck killing the persona and just using the original prompt from the persona.
Everything sounds more like an RNG than actual stuff.
I made a persona off an instrumental piece I uploaded with no prompts at all and the end result was pretty good but I could tell that the original song was heavily influincing what I was using the persona on.
Hows the best way to change a song mood? I started with something melancholiac and slow, but now im extending and adding heavy dark riffs tags, but doesn't seems to work, it keeps adding more positive music
try Uplifting Vibe in the prompts and make it the first one
This is what I eventually made off that uploaded track and persona: https://suno.com/song/dfcc8f1f-1cc9-43f4-99fb-f4b40cbb598d
acid-swing, glitch-metal, dubstep, dirty electronics, electric guitar solo's, strings, symphonic, choral ooo's and ahhh song. Listen and make your own with Suno.
Try adding *feeling" vibe
As far as my knowledge goes, persona was my last hope =)))
Ya know.. if -cicada worked for the shimmering in the beginning, I'm wondering if -cat might fix the hissing issue. ๐
you mean as a tag? Feeling dark? or something alike
Something like "dark riffs, powerful vibes"
You could add something like this to the top of the lyrics: [Mood: Reflective, Uplifting, Transformational]
or whatever mood you want
I might try to do that. I am afraid that the way this works might produce Acapella only
But imma try it.
Thanks for the idea.
I had that issue when I was generating one of my Christmas songs. No matter what I did, every song sounded exactly the same. I eventuallyhad to go generate something different and when I came back later, I started getting different results.
I think that many of the issues that show up while generating probably show up during high traffic times. I think that sometimes the servers are a little overloaded and the end result is taking a hit. So I guess it's possible that just wating till there is a time when the service isn't under a heavy load, might actually yield a better result. But I'm not 100% sure of that.
Ok, legit question. How do you get the Christmas vibes? I have no idea what genre these have usually lol. I've been producing music for the past 12 years and I still have no clue what genre is Christmas carols and music generally lol
Really, it's all about the sleigh bells and etheral chorals. That's basically it.
Hmm
I actually used Christmas Carol Of The Bells mashup in one of my prompts
That's interesting
I'll even do Christmas Vibe and it will pick up on Christmas song titles. I actually had Suno re-create one of the songs, exactly as the song is actually sung.
It completely ignored every prompt and just re built the song as it originally was, mostly.
Best way to be tried is in incognito
Even if it IS a cache issue, then just clearing the cache would fix it.
@lean flicker i see. Thanks for the advice.
Also seems suno has a problem with phonk, dubstep, brostep and other genres like them. Electronic genres in particular.
@naive vigil question for some reason the songs that i make suddly get cut off when they are near the end
๐ค
I only had this issue once, I disliked it and since then i seen no issue like that
I have no clue what to tell you in this case.
So V4 is gone?
it's a feature on the app version that u get from the app or download the pc version
om so what was your work around or did u generate the same song?
Are you subscribed?
I should be let me checl
Tbh it was gone by itself
check
omg, what the hell is happening ๐
I just discovered that the neural codec can also remove the distortion from a distorted song... what the hell have I done? ๐
I am not, โ Thank you
how
You are very welcome ๐
I have no idea man, that's why I have no idea what to tell you. I heard liking and disliking stuff helps
What if they are prune cookies?
What?
anyways, who wants to hear, this was encoded from wav to neural codec's tokens at only 6.7kbps, and then saved into a 312KB small file, and then I read the file and decoded the contents back into audio. And this is the result I got before I started screwing around with fixing suno inputs/outputs ๐
it's just mindblowing what quality suno could obtain if done right ๐
damn, I've spent 8 hours straight tinkering. time for a break
Hello, I'm having issues with my subscription, I subscribed through Play Store and it doesn't reflect on my account. I verified through the app and the website, same thing.
k will try that tomorrow
Thank you so much ๐๐ป๐๐ป๐๐ป
That's actually the second person that's reported an issue with suno not updating the credits
ReMi is killing me ๐คฃ It literally went mental on randomness
[Instrumental Intro]
[Verse]
Give me land lots of land under the starry skies above
Don't fence me in
Let me ride through the wide open prairie that I love
Don't fence me in
Let me be by myself in the evening breeze
And sing me some lullabies
[Pre-Chorus]
Life sways away
towards an endless maze
call my mind in tomorrow
taste the splendor of sorrow
[Chorus]
Where's our ticket
Was it a violation
Unborn the sons of men when planets revive
Admiring your splendorous form
[Interlude]
[Verse 2]
Impure is the earth under my footstep
Let me plow the plain
And sprinkled her with disgrace for a while
Tell the stars tonight
We were born to be masculine in this world
Or they have to perish
O daughters of gold dust
Imitating the universe
[Pre-Chorus]
Life sways away
towards an endless maze
call my mind in tomorrow
taste the splendor of sorrow
[Chorus]
Where's our ticket
Was it a violation
Unborn the sons of men when planets revive
Admiring your splendorous form
[Soft Break]
[Bridge]
You were asked who were the sinners of this earth
All the spherical suns of your idolized solar systems
Demanded who were the plagues of other worlds
The far-off daughters of the moon inquired
[Chorus]
Where's our ticket
Was it a violation
Unborn the sons of men when planets revive
Admiring your splendorous form
[Verse 3]
Persephone asks and Hades
The prayed Goddess Gaia and Karma Sutra
Who is he who's outraged the meaning of Bhumidevi
Asked the breast-clad Brahma and the black scarves
The sun himself wants to know
Is it our time to go to paradise
[Breakdown]
We no longer desire the earth
It doesn't belong to humans
Bowing their heads
they beg for mercy
[Mellow Guitar Solo]
[Verse 4]
Her soul was pulled from her and gambled
Smote the earth with my step
Until your earthbound body falls
So too does the sunrays' witchcraft
what was this song supposed to be about?
it would make a good poem but I'm not sure it makes a good song
I bet you ChatGPT could fix that
Struggles of Life ๐คท
Christmas Songs can be any genre. I have a Playlist with different songs all about Christmas. Some more classic than others.
The song came out amazing actually. It's the senseless lyrics that makes it outrageous ๐
Rumba Metal song. Listen and make your own with Suno.
Links to what? I've been literally hacking a NAC and for the purpose of getting high fidelity out of it. Accidentally also erasing shimmering and other artifacts is a happy accident here...
So far, everything is looking like this (in addition to python code too)
I'm having a small issue with right channel's phase and synchronisation but I'm getting close to fixing all that too. It's an iterative process ๐
my main point is that I think I've identified the codec SUNO is using. That's the only reason why I accidentally was able to fix the artifacts of suno-generated music withput even training an AI yet ๐
I was hacking the codec to see how it works and what it does, before even deciding what kind of AI system to build. Or perhaps I should talk to SUNO devs to finally get us on the right path... after all, my implementation here doesn't start dynamically drifting
somebody please point me towards a badly shimmering track somewhere?
My song above shimmers, will you be able to fix that as an example?
Remastering it now ruins it with the extreme highs it delivers. Don't know what to do with it
Maaaaaaan, I would give a kidney to have that stuff to fix my songs with ๐
Yeah, I wanted to see what this thing does... the funny thing is that I haven't even gotten to building the real clean-up tool that I have in mind ๐
Let me test your song
are you gonna at least give me an mp3?
Keep your kidney, I think he has a solution to all of this issues. Maybe the fix is nearer than we think it is ๐
mp3.. wav.. you name it
I have no preference... it's just the audio in the video is at very horribly low bitrate
It was a joke, but yeah, I wish I could test it on badly distorted songs.
wav it is then
Tha "it" ain't even ready or anyuthing... I'm still discovering in what many different ways it has an effect on music, depending on tiny configuration changes ๐
gimme 40 seconds
oh ok hang on
Well, keep at it man, you are doing an awesome job.
๐
Sounds so clean
damn, it's so fast.... NVIDIA RTX 3090TI baby ๐
these are for the second wav you sent
Woow that sounds so different ๐ฎ
Ok, you convinced me, can you try out one for me as well?
If it's not a huge disturbance.
I'm absolutely impressed with the results ๐
bear in mind, my main objective was to explore the high fidelity side of things... I discovered that the low fidelity stuff is also affected purely by accident... I had an idea that it might... but I wasn't yet so sure if I can get it to do something like this without training ๐
let's jus say that when I finished my job at 5pm, I had nothing... just an idea and desire to help suno team somehow or something.... yet... here we are now, 8 hours later... ๐
shoot
maybe I shpuld spin up a temporary web site for people to abuse my GPU with those quick fixes ๐
while I sleep ๐
This is pretty distorted and fricked
It would probably make sense to write code for result file name generation too.. byt hey... everything is just 20 lines of random, ugly code at the moment anyways ๐
Would be nice
it actually takes less than 7 seconds to process the whole song and return 2 variants... it's just starting up takes a bit time for GPU to reload everything, as I don't keep the script hot and ready to process as soon as I drop a file
That code does womders
Once again, I haven't checked...
all it does is basically encode SUNO song into "AI" format and decode it back into an audio file. There are just a few tweaks... I'm sitll in the discovery phase
Anyone having an issue where all the songs are ending at 3:23
If you're using a cover or persona this is normal
Doing neither
mu real target is actually also all the high frequency hissisng and so on.. but for that I think I already need to do a bit of training
Hmm. Audio upload?
Nope
Normal generation on V4
Can you link an example?
yes
earlier
Cinematic, dark epic, grand, emotive male vocals, haunting soundscapes, whispered vocals, and thunderous bass drops, Chilling atmosphere, brostep song. Listen and make your own with Suno.
You're doing gods work here my friend. I highly appreciate your effort to find a solution here. Lets hope Suno takes these ideas and implements them into the system as well. Can't thank you enough for the quick fix to my song. I literally thought it was a lost cause ๐๐๐
maybe somebody from suno could allow my discord bot to join this server here... so we could, in principle, have a "repair" bot ๐
That's using a persona
Hmmm, output 2 seems a bit cleaner than original, while output 1 is clearer somehow.
I just realized that was still clicked
Never mind
lmao
Thanks @unique stone
actually... I've known what's wrong with SUNO audio quality for a long time... it's just I'm a nobody, so I work around the issues instead of helping them solve the root cause
Suno is the reason why now got back into AI and ML stuff after a couple of years of hiatus
@woeful yoke and that was a 3.5 song
on the other hand, suno is the reason, why rebuilt my music studio too... even bought new gear
To be fair, i never encountered loud shimmer in v4, but the cats are annoying me.
It's hissing like hell
It's good that you're able to put this kind of stuff together. If it's possible to fix up shimmer or other artifacts with something that can specifically target them, that's probably going to be more useful than having users trying to massage the audio with other/external postprocessing.
hello guys with what prompt can i tell the v4 to make a high tempo song ? cuz high energy, high tempo, high bpm , energetic song, dosent work and every generate is a slow melancholic song and its allways a "soft" song and not have that high energy
Sometimes not even a tool like that is gonna fix stuff, sometimes is better to learn splitting it into stems and fixing stuff
well.... in all fairness.... SUNO teeam needs to build an audio clean-up tool themselves anyways... because as things stand right now, they can't continue generating low quality audio which would be used to train a new model that's going to be even lower quality if trained on unclean data.
The biggest problem still is the distortion and "clipping" - that's one of the main problems that has now turned into a huge snow ball
upbeat,edm,...
I mean, it's good that someone outside of Suno is able to provide some kind of workaround. It's better to have this kind of utility than to not have it.
throws speedcore
Gets a ballad about a granny named Gretchen
The amount of work arounds I tried, just to get rid of the shimmer or try reducing the increasing highs, is so annoying. Splitting stems didn't do, Remastering and tweaking the highs and low on LANDR didn't do... I was literally at the verge of giving up on Suno till a proper fix came out.
Well, if they finally come to their senses and tackle the root causes, and fix stiuff, then we woudln't need to fix and complain about bad audio quality. I know what neural codecs of 2024 and beyond aer capable of, in terms of audio quality... everything that sounds garbage isnt' because of technological limitations but rather because of some bad choices
Try showing it to them maybe you would get hired lol
a model trained on HIGH QUALITY data CAN NOT generate LOW QUALITY results.
Errr.
Yes. It can actually. By giving it bad parameters
I'm talking about audio tokens right now, not whether it's good listening and so on
Ah, i see. But still, needs to be listenable, it would be useless to make high quality triangle noise only or high quality noise.
I once prompted skskskksksksksksk
And it gave me a masterpiece
We keep paying though.
want to hear that ๐
What about "Shimmering" fx
I'm paying for "that" moment to be honest
Here have a listen
in all fairness, it's absolutely capable of using your one-shot uploaded "nonsense" audio clip too ๐
Hahaha this exactly why I pay Suno ๐คฃ
Add some vocals, and let's see how that fails
I'm playing this on xmas morning pal ๐
Brothers, it's the clearest shit Suno gave me lol
anyways... anyone knows who should I sleep withm to get someone from the suno team to actually start listening to some of us who also know stuff? ๐
We shouldn't have to keep paying for V4 that is in BETA version. I got enough credits that shimmers so well
They tend to respond on the feedbacks section. May start one and explain a detailed description of what you are up to about fixing the issues Suno has
weird... did they skip the alpha version altogether?
Sounds they skipped in-dev as well xD
I was part of alpha testing.
I was a beta tester for Corel Photopaint years ago. All the software was free until the beta was completed.
Should've been the same here too
nice... I'm occasionally "testing" new, upcoming Meta Oculus stuff and so on
We suffer of something called "Enshittification"
It's gonna get worse.
๐ค not sure where you heard that but I'm not aware of that.
after they fix v4 im happily pay for it but right now i just burn my credit to generate songs that not follow the propt that i give, and just stuck on styles or idk
Ah. Hm. Somehow I must've missed that.
Well let's not announce, praise, and push a version like it's something is amazing. I paid so many credits, thinking I was at fault with prompts, and trying and thinking, and than finally seeing V4 has a "Shimmering" Effect
Well take it down
FIX it and stop taking our money
You are under no obligation to use v4 beta. v3.5 remains available.
I am not even that active here lol. This is the most active i have been. I usually use Suno to make metal and rock music and sometimes to goof off
I don't care about 3.5. 4.0 IS HERE! Better, pay now to get access
well, either way... they must fix the audio renderer... and also the component tht handles audio input into models... right now either side is completely off, or perhaps both...
The same code that I ran my experiment today, would, by the way, fix it...
I can use the output of the decoder as input for the encoder over and over again an nothing changes... no drift, no volume increase or decrease....
I even made it to properly handle the loudness.... meaning that if too loud song goes it, normal level comes out... and too quiet will get lifted to the appropriate level too...
In suno, this has been screwed up since v2
Tbh I like what Suno outputs most of the time, but mostly because I know I can fix them if I try, usually I don't care about the quality of AI stuff.
I still complain about it tho xD
in all fairness, it took me a couple of hours to come up and get the code to work which willa utomatically correctly hadle mixed audio files .... with various fidelities and dynamic ranges.... you can't just use the loundess measure.. you must also consider peak to RMS ratio and then also the true peak - for getting all things to sound at the same volume... it's especially important for future training...
How you do that?
Suno devs can do a lot of AI stuff, but when it comes to audio processing, they're just noobs
i'm sorry for being critical
Don't be sorry about being right.
100% agreed on that. V2 had and still has a far superior quality compared to the newer models. It's just to robotic and messes up the flow of the songs most of the time, but the audio quality was ten times better than the others. Such a shame they couldn't build on top of that thing that was already good, instead they made it worse ๐ฅฒ
Ah i see. Already been doing that but not like that.
here's an example from today:
Top is mp3 from v2
The bottom one is wav from v2
and the middle one is the automatically correctly levelled variant of either one... I managed to get auto-loudness to work for real... and how it should work
I was just using normally generated stuff until it gave me a better version. Will try to do it with your method and see what I got.
the middle is, by the way, 100% per standards of our streaming platforms these days
I saw your thread and tried it. But the results sounded more like V3.5 rather than V4 to be honest, wasn't 100% convinced it allowed a V4 post processing on top. Maybe I'll try a bit more on Saturday when my credit cycle resets.
Well, guys, imma go to sleep. Thank you all for all this new information.
Sleep well!
A thing of beauty
So bottom looks like the left and right are very uneven
Don't let the Suno bugs bite pal ๐ ๐
and by the way... it doesn't blindly crush anything that doesn't need to...
suno should have this kind of thing as a separate component... so all audio tracks get fed through that, to reach some kind of consistency and standard... and also.. it doesn't screw up good audio for training data ๐
At the end of the day looking at waveforms without the LUFS means nothing to me. If clipping is occuring when those waveforms hit the peak, like on the top of the photo, that one is clipping.
๐
Oooh I see now what you meant. I'll give it a go tomorrow โ๏ธ
actually, they weren't
although it sit's at -0.1 dB peak, it would probably be bad with true peak measurement... but either way.... I set the output lpudness according to the standards we have these days... max -1dB True Peak with -16 ... -14 LUFS... 90% the time this results in most of the tracks having at leeast 3dB headroom...
also... the code first adjusts the volume according to the peak level anyways.... it doesn't start with LUFS target.... LUFS target is actually tetriary...
anyways, I'm gonna head to bed too
The image you provided shows no measurement. Just waveforms that look bigger on each instant.
Sleep well!!
and don't pay attention to the top and bottom grapth... I'm only concerned about proper levels that go out in the middle ๐
I'm not even sure what I'm even building... or if I'm gonna even build anything... but it's handy to have some code ready for future already
If I was going by my eyes without hearing it, or measurement. The middle waveform looks the safest to me
The middld one sits at -15.7 LUFS with -2.8 the loudest sample. so it nicely comforms to spotify and apple and other streaming platorms "recommendations" - I even tested with a coupld of online tools that confirmed everything was up to standatd ๐
for me it was more like a coding exercise rather than an actual need
anyways... see ya ๐
Awesome, makes sense. Sleep well!
Good night and thank you again for all your effort pal, you gave me hope today ๐
You're very welcome. If a rando like me can do it, Suno should be able to do it all too ๐
my biggest concern is more anout whther they actually care enough... as v3 unresolved fiasco is haunting us to this day
I'm also about to head to sleep. So tired ๐ can't really think what to write so I'll summarise as just: I hope good things come from all of this effort, one way or another.
anyone can help me, i already record my voice and become song, but the songs like trimmed
Hello my personas disappear from my list and when I find them they doesn't sound the same is it normal ?
Suno accepts my lyrics (old poems in public domain), but often refuses to do the covers even with exactly the same lyrics
Is it really in the public domain though? Chances are that the lyrics are but the audio recording may still retain copyright (though it depends when it was recorded)
- Under the Music Modernization Act, many of these recordings will begin entering into the public domain on January 1, 2022, when all recordings published prior to 1923
- Recordings published between 1923 and 1946 are then protected for 100 years
- Recordings published between 1947 and 1956 are protected for 110 years
- Disc labels that are more than 95 years old are now in the public domain and are free to use and reuse
https://www.loc.gov/collections/national-jukebox/about-this-collection/rights-and-access/
Based on USA laws, may be different where you are but this is the minimum needs to be followed as the servers are in USA
I wasn't talking about covers of old recordings, but covers from a Suno output (Using the beta Cover Song).
Ah, I see. Hmm, is there a persona involved? I think I remember there being an issue with that (I may be wrong, I rarely play around with covers and personas)
Not always, but it happen also if I use a persona from that output. Cover is a handy way to keep the same tune and style.
i wonder when Suno is gonna get all this stuff(legal) under control
What do you mean? If you are referring to the lawsuit, it likely will take a few years (and questionable if it will be settled out of court or stay the course to make a precedence in regards to AI)
Okay,.I finally listened to all those generated files I sent you. Holy crap, I'm surprised now even myself ๐ฅน
dadgum years?
Hi, not sure if I should be asking here but recently I've been unable to make suno sing with a male voice despite trying multiple keywords. Is this normal? How do I fix it?
Same model โversionโ during beta but generated a month apart.
Sometimes the genre influence is so strong asking for male vocals has a low chance of successfully overriding the genreโs expectations.
ROFLOL, I'm dying.
I tried to get Elevenlabs to sing ๐๐๐
The result is freaking hilarious on top of suno song
Different codec too, I believe
Hello everyone! Someone knows how I can get my own Suno API?
V4 is quite happy to make women sound men when generating a cover song ๐คช even if you don't want to
SUNO has an official API? ๐ค
Idk thatโs why Iโm asking ๐
I'm pretty sure if that was a thing, they would actually advertise it and also would have developer documentation and so on....
I saw too much โunofficialโ api btw
That you can get from cookie
Companies usually don't hide an API they'd like to sell
Maybe if you bought premium you can get it? ๐ค
I have the premier account. It just blows my mind that you think that a company would like to sell their most expensive subscription by not advertising everything you would get for that.
The answer is no
I can't use suno for a week now because spam is constantly inserted when generating music! I pay for proplan and I can't use it! Give me an answer on how I can continue using suno.
The answer is no
I have been using the pro tariff plan for a long time, I pay for it regularly. And now, having paid for the tariff plan, I am not able to use suno
You still haven't really clarified why you can't use it. I'm on the highest plan and I've never had an issue that I can't use.... It's just right now I don't feel like using it much. So in other words, I could use it but I don't want to.
But I've never seen or heard anyone talking about Suno inserting spam into... where exactly?
I'm so confused
Foreign text is automatically inserted during generation
this text has been automatically inserted into any generation for more than a week!
Ah, that... You can always use older version that doesn't have this issue.
how do i fix this?
V4 is a beta model. Its behaviour may change and probably will change from week to week.
Don't use the newest version of the model
this text is also inserted into models 3.5
Without seeing your actual song and the prompt how you generated it, I can't do any comment because I never encountered that issue with older mother versions unless you are not prompting correctly.
It might be that the model was a little bit more forgiving earlier when dealing with lyrics.
generated just now 3.5
Based on my experience, prompting with the correct structure will solve those maybe that's the reason why I've never cultured that because I've used the good prompting since day one. If I want the model to start hallucinating and making up shit then I'm usually being really unstructured with my prompting.
I really don't care about listening to your songs I'm more interested in seeing the prompts. So why don't you share the link to your song URL
hard rock ballad, melodic, male deep sad voice, slow, dark, emotional, blues song. Listen and make your own with Suno.
hard rock ballad, melodic, male deep sad voice, slow, dark, emotional, blues song. Listen and make your own with Suno.
ะ ะฟะพัะตะผั ั ัะตะฑั ะฟะตัะฒะฐั ะผะฐะผะฐ ะฟะตัะตะด [verse 1] tag?
Give me a few minutes because I need to listen to those
[Instrumental Intro]
ะะฐะผะฐ!
[Verse 1]
ะะฒะฐะดัะฐัั ะปะตั ะฟัะพัะปะพ ะบะฐะบ ะดะตะฝั ะพะดะธะฝ
ะะฐะผะฐ!
in this version, unnecessary text is still inserted
Jesus Christ you would have started by saying that you are getting in this error message saying that you are not using the right app.
ะขั ัะผะพะณ ะฑั ััะฐะทั ัะบะฐะทะฐัั ััะพ ั ัะตะฑั ัะฐะผ ะฝะฐ ะผะตััะต ะฟะตัะฝะธ ัะฐะบะพะต ัะพะพะฑัะตะฝะธะต.
Well prompt doesn't matter really. Did you at one point use some kind of an app or website besides Suno's official one?
no! only by suno site!
What bothers me is that the image preview UI is in a square format but then it scales and crops your image into a portrait format.
Well for some reason the suno system thinks you used some kind of an unofficial app.
Either way you must contact customer support because right now you're not getting a single piece of your song but instead suno is singing about the same issue.
I can read and write and understand the Russian language. That's not your song ๐
And while you're already contacting the customer support, please also give them a suggestion that is kind of error message should sing back in the same language as the person making the lyrics. This error message song should have been in Russian language.
What does it mean: This is not your song? These are the poems of my friend Alexander Dudarev. I am making a song for him.
Yeah that is a bit annoying indeed
Tech support hasn't responded to my requests for over a week now.
I try to include relevant type in an image and try maximise the legibility. So every pixel is important.
Do you even understand the single word of English language? Or are you just using a translator? But I was saying that the song that you heard was an error message in the English language informing you that you are using an unofficial app. That's why I'm saying that the song you heard wasn't your song because it didn't sing back a single word from your song.
Lol
I can definitely see that from your profile icon that has a transparency index ๐
Or at least it looks like this... I'm on the dark background so I can't really tell
Yes, I use a translator. It is clear that this is not my song. That is the problem.
No the problem is that you didn't understand what the song was saying to you. It wasn't even trying to sing your song. And that's why I'm saying when you contact the customer support also tell them to make those singing error messages language-aware
Anyways I hope you get this sorted as there's nothing else we can do for you here.
You pretty sharp to notice that. I am a professional graphic designer. So the image is just as important as the song from my own perspective. Although I get that many don't care about that.
Yeah I used to do a lot of graphics stuff in my past life as well ๐
If you're using the suno.com website or the official Suno app, this shouldn't be happening. The extra lyrics at the start (and the "SUNO" watermark on the picture) are meant to only happen for accounts that are using a fake app or trying to hack the Suno API.
Please fill out this form so the Suno team can get this resolved for you: https://sunomusic.typeform.com/to/zGsJSzsf
Suno adds this text to my text. Part of my text is generated
Okay now you're getting annoying already. We gave you an explanation. We also gave you instructions on how to get this problem resolved. Why are you still talking about this?
Please use the web link above.
That one
Thank you. You are very generous. Sorry for taking up so much of your time.
ะฃะดะฐัะธ ัะตะฑะต
AI imagery is everywhere, but AI can't really do typesetting or typography. So that then becomes an area of differentiation. But it does require pixel level detail, so Suno's cavalier approach to image control is frustrating.
Anyway I just raise the issue every now and then in the hope that the Devs eventually get around to looking at it.
Or perhaps I can make a simple app that addresses the issue.
I seem to have a technical problem with a particular song, I have been trying to create various iterations, covers, remasters etc of a particular song, because I love the song but the lyrics were prompts that got mixed into the song and I have not been able to change them. and editing the song to manually change the sections fails completely. I created alternate lyrics but no matter how many changes i put it through, it keeps the same original accidental lyrics it started out with. can I get any help with this problem?
How do i choose how long a song should be? My suno is making 4minute songs that ends wierd
Iโm sure thereโs a better way to have phrased that.
There might be a language barrier.
When they were complaining about getting the watermarked audio, you advised they use v3x without taking any time to gain agreement with what their actual issue was.
I talked to him in Russian too yet he didn't seem to want to focus on what's being said. It's not like I can talk Russian. I can actually overcome the language barrier but he didn't want to.
I'm not sure about how to sharp end a song while extending, every time generation is 2 full minutes and credit being deducted, it's very hard to generate less than that. It's really annoying...
Hello everyone, I'm trying to create a duet song, but it's not coming out well. Sometimes the whole song is just the woman singing, and there are parts where instead of the man continuing to sing, it ends up changing to the woman singing instead of the man.
I'm sorry if that might upset someone here but we, Estonians, have low tolerance for bs. We're quite direct sometime ๐ฅน
Oh, and also, I'm well aware of how well English-Russian translations work these days. I'm going to reject your language barrier argument, if you don't mind.
You might notice how incredibly easy it is to remove yourself from a situation when you feel like someone isn't contributing to productive conversation.
is this v4? which one was generated before and which one was after?
I am experiencing a serious issue with my account. All my profile information, payments, and recorded compositions have disappeared. Here are the details of my situation:
Issue Description:
- All profile information and payment details are missing.
- None of the compositions I recorded are visible in my account.
- Despite having a paid subscription, it shows as if I have no activity and only 50 credits left.
I urgently need assistance to resolve this issue and restore my account to its previous state. Please let me know what steps I need to take to fix this.
Thank you for your prompt attention to this matter.
Just a moment
Yes , but another account dont use
@spice badge does everything appear the same if you use the official android app?
I will try to do this. I hope it works out.
If the account looks the same from the mobile browser & the android app, or on 2 separate devices, it's more likely to be something specifically related to your account, which would require support@suno.com to properly investigate or resolve the matter ๐
The discord has helpful users, but we're unable to get into "The nitty gritty"
Gmail
v4 doesn't seem to respond the same way to methods we relied on for ending songs previously -- I've seen people swap to v3x just to use "end" as a style prompt, which of course brings the song quality back down to v3x, but you can remaster if you wanted to "smooth things over" after the fact
Hi, I used the free version to create a short song for trial purposes. I really liked it and since I planned to use it for commercial purposes, I purchased the Pro plan. Now, I want to recreate the same song with my Pro plan, but even though I enter the same lyrics and keywords, I cannot generate the exact same song. Is there any way to recreate the song I previously generated or how can I use it for commercial? I would greatly appreciate your help.
Email support@suno.com, let them know what song you made on the free plan that you would like the rights for -- if you're not unreasonable in your request, they've been known to grant the rights to specific songs ๐ช
Make sure to provide the song link so they know which song~
I cleared the cookies in the browser through which I accessed the Suno program, but there were no changes. The account remains empty, all information is still unavailable, and it looks like a new account.
thank you for your help. I already sent an email but they said it might take a long time to reply. I also upload the song again and remastered it but I am not sure that I can use remastered song for commercial use.
I'd run into a similar tangled mess of accounts when I thought my tester account was wiped, but there was a separate unused account in the mix associated with the phone number
I'd imagine while this could very well be what you're running into, can I assume that you made at least 1 public song, or had song likes/plays/fans? @spice badge
If that's the case, and suno profile is showing nothing anymore, it might be a bit more than a simple error of having mixed up sign-in info
I still can access the old account through the browser history used to log into the Suno program, which allows me to listen to my songs but not download them. However, when I do a standard login through Google Mail, there is no information on the account about the created tracks and monthly payments.
No, I am using the same information I have always used. I log in with the same email and enter the same data, but for some reason, the program does not recognize me.
I have already written to support. They received my email and responded, but nothing has happened so far.
It is strange that when I log into the Suno program with the same email, I do not see any tracks or previous information about payments, the number of credits, or the created songs. After that, I go to the browser history, click on the link, open it, and the song I created and other tracks appear. I can listen to them but cannot save them, and it is also impossible to see the track creation history there, only the listening option is available.
What should I do in this case? Should I keep checking everything, as I have already done, or wait for a response from the support team while they resolve this issue and find a solution?
Until today, I have always been able to log in without any problems. However, starting today, this issue suddenly appeared. The program's algorithm was working fine before, so it is unclear why this error and problem occurred suddenly today.
Just a question here as I don't know where to ask else. Is Suno supposed to mix existing music into creations ?
Thank you very much. You helped me solve this problem, and now I can quickly deal with it. I will check the account and restore access.
It's not impossible for Suno to regurgitate training data, but it's not intended to do so.
Anyone had any success at getting suno to not do guitar solos at all? It's especially bad when doing instrumentals.
Both V4. The bottom was generated in the first couple days they made it available. Top last night.
Almost every song I make has some kind of "thematic descriptor" trying to get Suno the chill tf out
Hey to those who had the same issue with โfake appโ in lyrics and SUNO on covers after purchasing a subs. The next day I emailed to support@suno.com about it with screenshots. I was replied in 8 days - like whether you use VPN or if there are problems with the block of sites in your country (I dont use VPN on SUNO and its not blocked). After another 8 days, the support emailed that like we removed the flag from your account on the backend, here's another 100 credits. After that, watermarks and strange lyrics didnt appear in the songs anymore. Also I filled โfake appโ form twice, but I think it helped more to email the support :). Eventually, the issue was resolved 16 days after the 1st email. The subscription time was not returned ๐ I hope this helps someone else
https://suno.com/@golden_fishy have a look at this guys djent songs... Literally the lyrics are "Happy halloween" and [solo] over and over and yet he gets no solos lol
Dear Suno,
I have a Pro plan for 1 month (2500 credits), but after only 5 days, Iโve already used up all my credits. To my surprise, the next day, I received an additional 50 credits. With these credits, can I create a song and monetize or commercialize it? Please provide information as soon as possible.
Thank you.
Songs made with extra top-up credits, subscription credits, or these supplementary daily credits are all granted similar ownership if they were made with an active paid account.
Ah if I weren't going to reconstruct this thing I'm working on in a DAW I would totally add [squeaky dog toy] to the percussion solo I got recently
**** โข Boreal Barracuda โข ****
[Season 1]
Participate collaboratively in the Event.
Event Date:
From December 13th to January 13th
I was thinking and had a great idea: how about we team up for a music project?
I want to build lasting partnerships.
It will be important and fun to create helpful and productive music content for the world and the platform.
You choose the date and time.
The will is when you want.
When can we start?
[#* โข How can we make something special together? โข *#]
โข We write lyrics together: Letโs mix our ideas and create the perfect blend of our inspirations.
โข We decide the details together: We pick the music style, the main theme of the song, key words, slang, and the emotional tone we want to share. Everything will be planned to show the essence of both of us.
โข This teamwork will be a true mix of experiences, creative sparks, and different strengths. Together, we will explore new views and fit each part together to get the best result. It will be a lively process full of creative energy!
[#* โข Lyric language: โข *#]
โข How to Prefer
โข Parts mixed with other languages โโare allowed.
โข All Current, Local Slang, and edecetera, edecetera.
[ โข And at the end we mention authorship, and publish it simultaneously on our profiles โข ]
Example using FT. Or part. Or & or - or โข, or however you prefer.
[#* โข Mention in Acconts: โข *#]
For my account:
[Option 1] v1
Name of Song โข [Your name] ft. Genho Akadion
[Option 1] v2
[Your name] ft. Genho Akadion โข Name of Song
In your account:
[Option 2] v1
Name of Song โข Genho Akadion ft. [Your name]
[Option 2] v2
Genho Akadion ft. [Your name] โข Name of Song
[#* โข Where will it be available?* โข #]
On Playlist:
[https://suno.com/playlist/4f39a0eb-01e8-4c82-951c-f0f090171e3c]
My sincere Kindness
I'll wait for you there!
I'm counting on you there!
Think with affection and cuteness
I looked in your suno ia help center and it says that the credits are recharged daily, but today at the time of writing this email, my 50 credits in the free plan are not regenerated to return to 50 credits on my account?
@woeful crypt
/refresh
Refreshes at 0000 UTC+0 (<t:1713225600:t> in your localtime)
No idea what you're talking about...
https://suno.com/song/f662d0ec-7ed4-465e-b325-b937c0d99f09
Djent, halloween, Djent, halloween, Djent, halloween, Djent, halloween, Djent, halloween, Djent, halloween, Djent song. Listen and make your own with Suno.
@woeful crypt You're probably aware of this already, but in case you're not - Discord has flagged you as a spammer, so that wall of text you posted earlier shows up like this to the rest of us:
omg
Still a wall of text to me. Odd.
How do I revoke this procedure? ?
Is there a setting for "possible spam"
Interesting. Freakin' discord.
Maybe. I've never noticed one. It seemed to be an appropriate flag, tho ๐
@open sleet oh thank you, but any strong healer to support me?.....that also helps. don't you think?
Anyone else having trouble downloading WAV?
working here
I think I'm one to like to have all the words on one page, so I can review back without scrolling.
Then again, I butcher my own rule when I type a novel in here sometimes.
@polar spade because it is...... controversial and volatile, human beings are just like that. can you believe....right
I like when I see people enthusiastically make a plan and endeavor. I hope you can make it a go.
For me it's just old age. I need things simple, more often, lately.
Three chords and the truth. That's enough for writing songs. ๐
@polar spade yeah..... bro so I want to turn some of those sky lights into reality
Does anyone know why Suno is suddenly asking for CAPTCHA-style verifications?
@hot trout I think there are more robots than humans
Of course. We're not allowed to have nice things.
From toasters to light switches to cars. Robots everywhere..
It's something that they added a few weeks ago. Probably in an ettempt to cut down on the bots, but it bites regular users periodically. I only seem to get it during periods where my connectivity to the Suno servers seems to be struggling
@polar spade yes.... they were scared by the robots from dynamic robots, robots from tesla, and clone robots
hey.. there's a memory. I still have my Amiga 4000 Toaster. Anyone remember that one. Wonder if I could use it to make a video now
I mean, I can see why my default generating behavior might LOOK bottish... I generate songs in batches of ten at a time, weed for the best version, then extend/edit from there, so it might look suspicious... just surprised I am getting tagged with it now if it's been going for a while. Oh, well. It's not that big an inconvenience.
I often do that. I'll rapid-fire several gens, pick the one or two that came out decent, make a prompt tweak or two, then rapid-fire several more.
This is the way. ๐
@hot trout We can't move, otherwise their earth will spin. We are big force here, right
I BURN through credits that way, but it doesn't really matter to me; it's more important to get a song I love.
Are there, like, 5 separate conversations going on here? Lol
@hot trout True, you're right, I have 12 accounts, for me to get a result I burned 7 accounts and the 8th worked.
@polar spade Because of what I can't pay, there's a crab in my pocket
@polar spade nop
frequently 
so heres a question. Im one to clean my trash on suno. had it all cleaned. in a recent update or something all the trash songs have returned. twenty something pages worth and they cannot be perm deleted. issue on the server side? or is it something i can fix?
okay
Damn. I haven't run into that, but that does remind me that it's time for another trash purge. Can't let that beast build back up to 1700 pages again. lol
lol that sounds like a lot. yeah its weird that it happened. after the glitch i just stopped trying to clean it out XD
That's an odd one. So to go over the steps... we can 1) delete or "move to trash".
Then... from the TRASH , we must "delete permanently" to remove permanently. Yet you did that , with a pop up notice they were deleted, and still they all came back?
yep did all of that. i have a weird habit of deleting all the things in the trash. idk ocd i guess XD but yeah all the songs came back. i should specify if this is something but its all the songs since i got membership. the ones i rendered free before that point didnt
Let's see if these reappear at some point
That's a "who knows" response, I guess. Best tested by giving it a day and try a few more to see if it happens again.
Hmmm, it works with clips under 1 min, but not over. I cannot download Video either. MP3 works though. Frustrating.....
Other than wondering if you're looking at a second browser page you still have up? And you're looking at the cached version until you refresh. Silly thought, I know, but hey. Who knows. Right.
Hmm, do you have "SUNO" watermarks appearing in your inages? Or audio messages saying to use suno.com?
i checked desktop mobile and app and they are still there. i could try deleting them again and hoping it sticks this time but thats 29 pages XD
You are only allowed to have 1 free account.
If you have more than 1 free account you are breaking the Terms of Service.
Just so you are aware.
Yea. Just do a page, so you can keep track, then see in a day what happens. If good, do 28 more.
OCD is a bear sometimes, can't leave things half done. So 1 page might tide it over to test.
specify explain
You are not allowed to have 12 accounts
as? please show me the section where this is proven
One moment
I also have a dumb side question that probably cannot be answered but my wife made an account using the code i got provided. (supposed to grant 250 free credits) but I never got the credits. does it not apply to pro members? I mean its no biggy but was curious as ill remove the code from my videos I post if it doesnt work
"You may not create more than one account to benefit from the free tier of the Service"
Is there a guide online for providing instructions in lyrics like for making it a verse, chorus, bridge, or add instrumental section in? been trying to find out on the website
hmm... do you have to live in the shadows? Someone to help me with finance, because I work, but the budget doesn't cover it
What choice do I have, I'm forced
Only to say I don't know: I do read a few comments regarding the referral program rewards thing, but no one posts of their follow ups if a resolve.
Nobody is forcing you to make multiple accounts.
I mean no offense but bruh you literally get free credits daily. Just wait the time??
The invitations don't work, I did the test with a friend (created a new account), he is a witness, we are friends, we decided to make the invitations, but it doesn't work
@eternal topaz If I wait, I won't achieve my goals quickly
Itโs important to understand that no one is forcing you to violate our Discord rules. Spamming and creating multiple accounts are choices youโve made, and they disrupt the community. This behavior must stop, or we will be forced to take action.
oh hey daisy ๐
the suno system, this term, is because the terms are for an account already logged in. if the person accepts, how can they know, each operation is individual
Your argument not only fails to justify breaking the rules, but even now you are advocating for continued violations. This is unacceptable, and further actions will lead to a ban.
Terms of Service can be read without you needing to sign up or sign in.
I've waited too long, in life, at work, in practically everything, then a program like this arrives and doesn't help with time, forcing me into this dependence
How do you manage to throw away 802 clips at once? My trashcan shows only 20 clips per page and there seems no way to show more at once
@final hazel Yes, that's not the conduct, they use these terms to make people dependent, even more so on their system
ikr a mass delete button would be a blessing LOL
in the desktop version, you can use the windows multi-select modifiers - SHIFT and CTRL. It will remember your selection across multiple pages as long as the first election on each new page is a CTRL+Click
like seriously though what kind of business are you running that you need these renders like today. also if memory serves without a pro membership you dont technically own the license to the renders (idk if thats accurate)
@eternal topaz man, I really like it, but I have high expenses, this goes beyond the spending ceiling
@eternal topaz and the little I have is not enough
Hey @woeful crypt letโs please end this conversation, this channel is meant for technical issues and help. Feel free to reach out to me via DM to continue this conversation if youโd like to. Very interested to better understand whatโs going on and how we can help. That being said, please know that having multiple free accounts is against Sunoโs TOS
Omg Bless You!!!
@daring lava ok, agreed
Yep pretty much. Suno owns stuff you make on the free plan.
ok, agreed
Thank you
No, no watermarks... it doesnยดt even work on old songs. It just says "Download failed, please try again."
Hmm, try downloading it in Firefox if you are using a chromium-based browser like Chrome. That is one solution that often works
I'll remove any non-related comments.
@polar spade We are, true, but we can change, and you know that, I just used a monologic way, where I had to go forward with this choice.
But we understand, good morning everyone
I felt that way about Ferrari's F80, but I didn't go out and steal one. 
Done.
I have been away for a week or so, has shimmer been resolved yet?
@naive quail I found it funny. But sorry friend, I'm not exaggerating, I'm not going to that point, that point really is theft. There is no way for a comparison of this level to coexist, in my case it represents a less pronounced content. In this case, their government has nothing to do with public or property security.
Still getting shimmer here. Sometimes I think it's not as bad... then I get a bad one. ๐ฆ
@naive quail In this case it is much more punishable, the level is much higher for you to get hurt
unfortunate, I hope they can fix it
My current PREMIER subscription (copy of invoice attached) contains an e-mail error. I am sending you this 2nd e-mail to inform you that my file has disappeared from your site along with 1000 credits. There is no way to restore it, and no response from you despite my first e-mail !!!! And it's impossible to cancel!
If you don't reply as soon as possible, I shall block my direct debit and report this problem of silence and malfunction on your part on the social networks.
Best regards.
Doesnยดt work ๐ฆ crap....
Um, there is no official support here, this is a community ran server. You want to email billing@suno.com
You should be able to cancel subscription by going into the subscription page and changing to the free plan.
Still can hear it sometimes but not as frequent and prominent as before so they definetely improved things
i can't create anything no matter what...system error
Shimmer is better, audio degredation in the second half of songs is now a bigger issue than the shimmer ever was, IMO.
It's at least possible to extend before the degradation. Shimmer is just everywhere in my songs
It will still degrade.
Hello, I was wondering about the subscriptions. If I get them, do I get all the rights to the songs and also to the ones created before and do I keep the rights even if the subscription expires???
Hello, please check out #1294423389515284552 first ๐
If you have any questions afterwards, do not be afraid to come back and ask ๐
@final hazel I understand that you have a position here, I respect you, and you are doing it very well, you have nothing against me, it is your profession, and I agree that is being a good natured person.
@final hazel I'm going to solve all of this, all of this, even things will get better and I'll be a premium member
@final hazel Please, may you reveal my attitude, I know I'm wrong now, to this whole multitute
would you mind sharing the suno link to the one you made last night? because my experience with v4 has been the complete opposite, if you were able to achieve such good quality with v4 last night i'd love to hear it!
Would it be possible, on a previous log in you inadvertently created a new account with you email this time instead of phone number, so you're looking at a "dud" free account with no credits. If it shows no credits, maybe not, if it shows 50 credits, maybe.
That's what I did... and others.
Can't extend the music for two days, please help!
What browsers have you tried this on?
Microsft Edge
Can you try Opera GX quickly?
Hi Suno team, any news on the official API?
oh they are short
Yea, and I just tried it, and that tag works.
What is that? sorry
It's another web browser
We need to check if it's related to your browser settings
Your original maybe go silent enough nearing th 1:11 point to cause extension to be short. Is 1:11 the full length of the first output? It's always suggested to back up the extension point a second or few.
Try 1:10... or 1:07.. or any few seconds previous to the end that has noise/music
I only have chorm
short songs usually appear if you extend from a time where suno thinks its time to end
Chrome can also be tested on, months ago I had to swap over to another browser when I experienced similar issues
Browser should not be an issue for this case since it is being processed server side
and "jingle bell" is maybe a bit like a song name in the style...
Not even third party cookies, scripts, adblockers?
That would be more of an issue when downloading and whatnot, they are trying to extend.
Usually those things would prevent a submission from being sent if they were an issue (or client-side issue like downloading a song or cannot access the site)
Let's try this, make a brand new song @half spire and then extend it?
@half spire. Are all attempt using 1:11 as your extension point?
Quick question So... if I'm here how do I crop from here? I can only move a little bit. I mean just a little bit left and right. No up and down. And if I tilt my phone It goes back to the homepage.
Might be it, that the song has ended, and they are trying to extend into nothing?
So I just guess On where I'm cropping from? And when I click I just move one line back and forth
I tried the drag and click that didn't do anything. Hmmm Cuz I can't see anything other than what's on this screen
You probably right, the song name including Jingle bell Can not be extend, even I switched to a different browser
It takes me about 10 minutes to get to this spot
How long is the song in total, and what point are you extending from?
You probably have your display settings small enough to show the crop screen, but not scroll on your phone. You mightcheat it by hold finger down on a word to highlight and slide finger down to see if it highlight-scrolls down enough to see the bottom areas, but that's going to be frustrating.
not sure how it exactly works on mobile but should be similar to pc.
Okay, I'll try that. And the most frustrating part is I have to keep the phone completely sideways. Otherwise it goes back to the homepage.
i just select a part in the waveform and click crop
And if I try to edit with it upright it just goes to the homepage cuz it clicks through all the tabs
Crop and replace are formated for desktop and tablet only... but an iPad and my fold 5 can emulate a tablet enough to use the features.
mobile website is a bit of a pain... yes ๐
And it's not on the app yet unfortunately.
Correct.
The worst pain is it clicks through the tab to everything behind
you would not have all these issues in the app... ||because they cant do that anyway ... not even extend ๐ ||
yes. of course... at the moment they are bare bones and get perfected until good and then expanded with more features
yay I have hope then. ๐ฅน
I have two songs, one is 1:11, one is like 2 mins, both can not be extend
even if you change the continue time a bit and dont use the song name in the style?
You are missing something, which we need.
- What point are you picking to choose the Extend from?
Until then, set both page format and font to their lowest settings (split page for layout, if possible) and see how much you get on the screen.
Well. I emulated a phones aspect ratio and couldn't even get half a page of crop. So it only works in tablet format for me.
Ok, same issues even though I no longer use Jingle Bell and changed to Chorm browser, is it because it's Christmas music?? Why I cannot extend Christmas music??
Its a shame, I really like the music, just wish it can be a little longer for my weekend Christmas party playlist.
Well, more to point, is it being extended from the default ending or backing up a few seconds?
How long is the original song?
2:16...
Try putting 2:00 just for testing
Yep. We keep trying to ask. ๐
@half spire As others have mentioned, the problem is that Suno already thinks 02:16 is a good time to end the song. If you try to extend from that point, Suno's just going to add a few more seconds and end it again.
You'll need to pick an earlier point in the song to extend from by changing that "Extend From:" number, like what Drew suggested.
It looks like the song actually ended at 215 not 216 . So it's just confused on its own numbers.
You'll probably have to extend from 2:15 or lower
That make sense, I never try that, will give it a try, thank you
Everything is so tiny now . Lol My phone is like 1 in away from my face now. ๐คฆโโ๏ธ Hopefully this works haha
We've seen remaster and cover repeat a , say, 2:15 song and note as 2:14, on the timestamp. It's fantastically conspiracically imaginable that that last second might not be consider as "sound" ..... so good to just extend a couple of seconds back, if, after 2 generations (4 outputs) the results don't change.
Yeah it looks like that's what kittens problem is
All right I'm going to go try my stuff again. Thanks for the help and hope. ๐
รznur merhaba, I'm Turkish as well, but I'll write in English for moderation purposes. I believe the song you've generated while a free user can't be used for commercial purposes, even after you switched to a pro plan. But yes, do write to the support email and explain it to them. They should allow it for you.
Lol... same for mine... but I couldn't get it to work on mine with a cell phone aspect. You may have better luck as you're at least getting half.
It's working! Thank you thank you.
man my eyes hurt now. Lol
Woo! Woo! ๐ค
Time for reading glasses... or magnifyers
Suno magnifier specs! Now available at Walmart.
Hey. Can you mention what model phone you're using that will give you a workable landscape version.
I use a Samsung Fold 5, which is default considered a tablet in its open position.
Guess I can put them over my 4.35 and 5.05 prescription glasses
Woa.... and it works in portrait as well. Cool.
I think I'm stuck in this mode now though. Lol ๐คฃ I might have to go back into settings and adjust something.
Lol. Well, we gave it a go, right?
Hey, progress is progress in at least I can crop stuff now. It's a win in my book
but do all of your songs come out really small now?
Yeah but its fine.
Somebody having Issues with Inviting other and not getting the Credits?
Hi everyone, do you still have the shimmering noise issue? Have you heard/read any updates about it?
Shimmer is better but now extensions are terrible quality.
I still notice it in 9 out of 10 songs. Most of the time, the singers sound "thin," and I canโt quite explain why.
Any Mods or owners currently online? I have a question regarding my account
the discord server is for users to chat, it's not an official support channel -- you can contact the team at support@suno.com ๐
can i somehow pay for subscribtion without an credit card? because i cant find my german bank at "bank"
When you create an account, you automatically start with a Basic (free) plan. Our Pro and Premier plans will give you benefits such as more credits, c...
well i have google play, but there it wants an credit card as well, does it work if i cash 10โฌ on my google account? or it still needs a credit card then?
https://support.google.com/chatsupport/?hl=en Questions about wallet balance for your google account would be better directed to google support, unless another user has personal experience with that -- or billing@suno.com for an in-house response on the matter
Official Chat Support Help Center where you can find tips and tutorials on using Chat Support and other answers to frequently asked questions.
hmm ok thx..
is there an api for suno yet?
i have found a few unoffical ones but i dont wana put that into my stuff then it becomes defunct
I canยดt find the tutorial but I think I did read that if you get stems and download in .wav and remix them again you get rid of the "shimmer" issue. true? also if you get stems and mix them you get the original song back? or stuff goes missing?
I wonder if I could get the team managing this discord server to invite my bot into the server. I could then have people use it as a tool to quickly clean up shimmering in their songs ๐
Suno don't provide API access. Avoid the unofficial ones because they will either result in your account being banned or restricted.
Maybe you could contact them directly at support@suno.com and let them know you've developed some software that the devs are sure to be excited to use ๐
Beyond the bugs & suggestions channels, the devs don't really spend time hanging out in the discord, so just trying to make sure that all of your hard work gets in from of the most relevant eyes, in a way that could be impactful.
yea its so wiered on every first song generation the mobile app generates a hcaptcha pass token even through it somhow never shows the hcaptcha box its realy hard to reverse:
Q. When tried to send a link of a song it always turned into the song/pic itself. Anyway to send a link so one can add the link itself?
I'm not interested in getting anything in front of their eyes right now.... At least not until it produces better results but I'm mostly interested in doing something for the community and with their help and feedback to achieve the best result. I'm not really keen on violating discord TOS and doing it as a self-bot using a modified discord client ๐
I'm not interested in getting anything in front of their eyes right now...
But you were just talking about wanting them to implement the bot you've been working on? Either way, resolving these audio issues is clearly something a lot of people care about, so I wish you the best of luck getting it to a place you're satisfied with ๐ช
If I understand, you want to send the link to your song through discord without showing the video?
https://suno.com/song/197cffc1-da6e-4fb6-afe7-362b62f0c84b like this?
Are you trying to reverse-engineer the app?
yes reason is that we add diferent links in one place as a birthday present
I"m not sure what I"m doing wrong. Why isn't the "Recreate section" button becoming available for me to click??
just the link not the picture of the song
You need to select a test 10 seconds of the audio
Meaning that's your click on the waveform and drag, a range selection box appears and when it's wide enough the button becomes active you will also see the lyrics that you are about to replace getting highlighted in red.
You can right-click or long-press the message afterward to choose "remove embeds", or you can place angled brackets before and after the link to prevent it from being embedded <www.google.com>, like this
Ooooh ok. Thank you!
pls help! https://suno.com/song/5d9069b1-7a96-4e95-98b1-1b09d4c39dfb And so with every song! I did not violate anything, I use Google Chrome address https://suno.com I filled out the form twice https://sunomusic.typeform.com/to/zGsJSzsf but there is no answer. I have a paid PRO
futurepop song. Listen and make your own with Suno.
Turn data collection into an experience with Typeform. Create beautiful online forms, surveys, quizzes, and so much more. Try it for FREE.
Ok, so I tried something odd. I created a song in V4 and it actually wasn't bad. I extended it.. the extension was bad. So I thought.. hmm... swapped to 3.5, got a warning about quality, but it rendered a 3.5 extension that sounded "normal". So I combined that with the v4 song and got a 3.5 full song. It didn't sound bad so I decided to push the envelope and took the now 3.5 and remastered back to 4. To be honest, I'm not sure how to judge the end result, or if this tells us anything. I will say that the shimmer on the second go around is more obvious. The original song was the full 4min so the extension starts at the 4min mark. There are some silences where you can clearly hear the shimmer as if it was a badly rendered echo or reverb effect with a long tail.
Good afternoon. Yesterday I purchased a Pro subscription. But for some reason nothing changed, and I was left with the basic version. Please tell me what to do in this situation?
contact suno support. Your the third person I've heard of this issue over the past two days, so it's probably best to contact them directly.
Did you use google to process the payment?
Hi I was wondering if anyone had ideas thoughts on a particular issue Iโve been running into when downloading the stemsโฆfor example I prompted a cover that resulted in the perfect beat/insurtumental I was looking for but not vocals. When I downloaded the stems to secure/isolate the instrumental there is still residual vocals in the background of the track. Anybody have any ideas on how to deal with that? Really want to keep that instrumental but the leftover vocals are ruining it.
I think that's what the other two people did as well.
I have the problem I described above, I've been waiting for a response from support for a week))
yes
I had the same issue with the stems on 3.5. I think it might be an artifact of a process like reverb or echo and your hearing that process that was added to the vocals. In either case I tried using audacity to pull out the extra stuff but never had any luck. I've see other people in here use third party programs or other systems to generate stems from the songs so you might need to go outside suno to get good stems.
Based on what I've seen from the three people that have had this billing issue, it seems that using google as the payment processor, is the common link.
I wrote to them, but they are still silent
I have no idea how to get this to Suno tho. you might try posting it in one of the bugs-X Feedback channels.
thanks, I'll try to write there
" message afterward" not sure that understand
You can click this to remove the video
On mobile, you can press and hold to remove the video after youโve shared it
This song is Arabic yet It didn't reach Arabic trending, why?
Arab Nu Metal, heavy tuba downtuned Rhythm 9-strings guitars, heavy guitar interlude, Arab woman soprano, female vocalist, heartfelt pads, driving Oud, Heavy Rhythm, furious vibes, HD/3D stereo song. Listen and make your own with Suno.
okay if It is not possible to have a certain chance like others then I want to say to fix the problem if It is an known error.
because It happened a lot of times
Hi, I paid for 2500 points on November 12, and on December 12, I lost all my points even though I still had more than 1000 left. Why?
Unused credits don't carry over after the end of your subscription month.
If I remember correctly, devs used to chat here back in Bark/Chirp days. It remains a good place to focus on discussion of prompting or specific usage techniques of Suno.
yea we all just realy need an oficial api why this still isn't a thing yet
ahah it's a scam
omg
Hi, is there a way to download the autogenerated cover image of a suno song?
Haha sorry found it!
like you need soo many requests just to generate a song corectly why suno makes all this effort and their service probably slower by adding al this mesures and affecting even users that don't use api instead of just providing an api lol
It tells you on the subscription page that the credits "renew" monthly. Also this
I'm not sure what benefit there would be to them offering an API.
Besides app developers just reselling the service with a different skin over it. At least, that seems to be mostly what people are using the hacky APIs for.
i personaly forexample just prefer using an api over a website and think about the posibilitys of forexample a chatbot directly generating a suno song ik bing exists but imagine claude 3.5 sonet generating songs automaticly way more freedom
I use ios app and I swiped up and got notified song reported. How do I know which song got reported and what shall I do?
It's unlikely to cause a problem. As far as I know it doesn't automatically do anything, just logs it for the team to look at. I expect several reports have to be made to avoid it being abused. Wouldn't worry about it.
It happened accidentally. I was playing one of the song while accidentally I swiped up and got notified.
@unique stone so my created song is safe I suppose?
Hmm. This doesn't really resonate with me as I prefer having more control over the creation process. If anything I'd feel it would take some freedom away from me. For me it'd save a copy/paste from the rare occasion I want an AI to generate lyrics.
There would probably be some very innovative use cases for an API, but like I suggested, I expect most common use case will be: acquire API, build app that makes songs, publish app, profit.
Should be fine. Even if someone does look at the song, provided it's not overly offensive or anything they'll probably assume it was an accident.
Thanks! That was helpful.
If you want a chatbot to create the songs for you, you can use simple mode ๐ช
Hello, I don't understand this... However, I have definitely deleted the audio that I am trying to put because I put it at the wrong time.
What were you trying to upload?
A music. But the first time it worked before I deleted it to reupload it a second time.
Is it music you have made?
No ๐
That'll be why then.
It's against Suno's Terms of Service to upload audio you don't own the rights to.
This is a filter to try to block copyrighted material from being uploaded. It sometimes misses things.
I find it a shame because we can do some pretty funny things. ๐ฅน
Oh no I just tried to make another song and it's distracting this way . I have to put it back cause it's too distracting. Wish I knew how to make a toggle button.
Holy crap. Quick pivot to generic AI news. Meta dropped a paradigm-shifting paper today. They figured out how to get rid of tokenizers in a way that's also scalable. This will reduce the computational cost by 50% during inference (as compared to regular tokenizers).
What you're doing is the creative use of Suno, there are no right or wrong ways. I mix the models and ignore the warning (I take that it means ยซdon't complain if it breaksยป, not at all ยซit won't workยป, this isn't troubling).
I combined that with the v4 song and got a 3.5 full song
This doesn't happen always. When I use V3 to cut the song which refuses by itself to end, I "extend" from 1โ3s prior to the desired timestamp with V3, no style and only but the[end]in the lyrics box. V3 is more agreeable. The 4โ10s ending extension doesn't turn the whole song into V3. Larger extensions do (sometimes).
I do certain little things differently. In my mind, this is common.
The original song was the full 4min so the extension starts at the 4min mark.
Try to start the extension 1โ3s prior to the point which lends itself to extension musically. E.g., if you extend from the middle of the final coda, you'll nearly always get the same coda in the extension. Think where it is musical to steer direction, step back 2โ5s, and extend from there. To extend from the very end of the clip works (most of the time) only when it ends abruptly. The silence which, if you don't want, suggests to move the extension point back a few sec. I do a number of extensions, and 1s makes a difference. Vary it somewhat back and forth, ยฑ2s is a good starting variability in my experience.
push the envelope and took the now 3.5 and remastered back to 4.
Not much push here. I believe it's common to do. But the first pair of the remasters may not work, do a few and compare them. I'm looking at 14 remaster of a single, if longer and more complex, piece which I pulled into Cubase to switch solo between them easier, not losing the timestamp. Concisely, don't expect the first remaster to be perfect. Suno randomises them to a degree.
On the lighter side, I start my day with two quad espressos and hanging on Suno's Discord skimming the new papers feed of ArXiv. I almost can't remember a day without a holy-crap-paradigm-shifting paper. Just sayin'. ๐
(Srsly, I'd've indeed paid more attention to the equal of Meta Researchย ; although it's more commonly the authorsย / the collab, not the institution.)
I find it a shame because we can do some pretty funny things
The judge could disagree. ๐ฒ
Yes unfortunately ๐
I rip the downloaded .wav with RipX Daw Pro, which separates the vocal stem really well, usually giving me 6 or 8 stems to work with. RipX is available as a 21-day free trial. Worth trying.
That's maths and business. Assume 2/3 of all credits over the whole of Suno get unused. That's realistic. You're offered the plans for โฌ8,00 and โฌ25,00. (Don't miss the fact that โฌ8,00 used to be the price of the lunch, but now it's the price of a salad with water dressing.) You allow credits to roll over indefinitely. Now, you must increase the price 3-fold because you want to eat at least three times a week. This comes toย โย ah the sweet maths partย โ โฌ24,00 and โฌ75,00. Another thing is predictability of the income and the load on the servers. To compensate for the spare capacity, you can't do anything but increase the prices even more. Add another 15โ30%. If the credits were to roll over, you'd have to pay some โฌ30,00 instead of โฌ8,00. Perhaps you'd call it honest and be happy, but then the user @random loom would call out the price increase the scamโฆ
To say it simple, you're subsidised by all those who lose the unused credits and subsidise them when you lose your credits at the end of the month. This is simply how money (and many other quantities in Nature, FWIW) works when the scale substantially increases.
Is thria tool to break down generated music to instruments? Sometimes I have a need to remove some noise without regenerating the entire section or song.
Try a tool like Steinberg's Spectralayers and do the stemming in it. You can get a month-long trial. Sometimes it's possible to fix. I think the latest version has a vocal removal model, but can't recall with any certainty.
If it's a synth instrumental foundation, I'd just replay the instrumental using some real synth and a DAW. Suno mixes are far from even average real-life mixes, not stellar at all, and stemming in general works fine if you're remixing them back together, but not so good if you aim to subtract something from the mix. Unlike singing, you can literally retype an instrumental using one finger on the keyboard into your DAW at however slow rate. One doesn't need to be able to play fluidly to do that, just have some ear for recognising music.
I don't know your situation, you can't do that to a complex guitar solo. Just sharing my experience. Or thinking aloud.
Having an issue with Suno on Chrome for Android. Trying to create a cover. Goes through the motions, but when I hit create it just whirrs away and beach balls. Hmmm.
Try downloading the wav and run it through Fadr. I think it's better at stem separate than the Suno solution. Or try Rip X
Just started using suno in October, so I've only seen the one new version release. Maybe a longer term user can shed light on about how often there was a new version or new beta access released?
@cedar thistle Thanks for the info. Figured it would be something like 6 months or quarterly. ๐
Yeah you can definitely see the output changing with tweaks. The shimmer was terrible before. I hope when they button up the issues with v4, there'll be an easy way to fix past songs without having major changes to them.
okey i think i found a weird bug
songs diapear from some of my playlists but is in other playlists unsure what to do
okay... so...
The v4 remastering as of today has gotten less noisy but...
one of two generations has totally wrong voice and also the melody occasionally plays wrog chords or the vocals sing wrong notes.
and the other one sounds kind of the samer as the original source, though through a cleaner yet less sharp top end that's also shimmering occasionally
so yeah, one mastering usually has OK vocals, even if the voice timbre is different... but usually melody is off... especially songs with low stabs
What is the max song length in one shot for V4? I thought it was exactly four minutes? I just got two songs that cut off in the middle of choruses at 3:34.
depends on the song , lyric and genre some genres are faster then others. same with the length of songs and rhymes. i prefer if the song doesn't run right up to the limit as they flow better
Agreed, but how do you avoid the limit when you don't know what it is?
Maximum is 4 minutes.
If you're using covers or personas it is usually less.
oh wow, my tool from yesterday which today got 2x sound quality increase can still remaster suno songs by removing noize, shimmer and distortions (if any), yet not changing anything else like v4 remastering does....
I've found the original remaster (or master if original in v4) is fine enough. But if you have to extend? Woof.
You don't have to find the exact correct prompt for a music genre to generate that genre. You just need to something vaguely like it that works well enough to point Suno in the right direction, then use multiple steps to get there.
prompt->pick closest->make persona or cover->crop closest part of persona or cover (usually quite short, occasionally longer)->prompt again with persona/cover->repeat until the songs aren't getting close to what you want. Personas and Covers are used as one-off intermediary steps. And probably add a touch of [style tag spam] into your lyrics once in awhile, you'll get hallucinations, but these gens are just middle steps, you can remove it for the final gen.
I feel like not enough people prompt iteratively like that, it was super useful even with just Extension feature (though more difficult).
was aiming for something in the ballpark of the Killers. bit of shimmer city, but https://suno.com/song/37194f99-61aa-4a15-ab6f-761be7971071
A โ2x sound quality increaseโ sounds like a significant improvement.
Would you mind elaborating a bit on the methods or metrics youโre using to measure this improvement? For instance, is this a subjective evaluation based on listening tests, or can you share how youโve approached measuring the noise reduction? It would be great to understand how youโre quantifying the difference.
so you dont provide public api, but when the devs go and make one themselves, you ban them, lol. What a stupid idea
API access is simply not allowed.
then provide people what they want
people pay for a service to create songs, intentional disregard to your userbase by closing accounts
makes non sense
more effort to block than to make the api public lol
It IS against the TOS and the consequences of developing or using one is mentioned (see the pictures of the excerpts at #๐ฌโgeneral-chat message)
Depending on how intrusive the use of the API is (i.e. if hijacking the service would constitute as effectively stealing the resource), it can be against laws in USA (See Computer Fraud and Abuse Act Section B.1, and specifically 18 U.S.C. ยง 1030(g))
And, honestly, a lot of the API usages does sound like does violate that law
so i follow the same flow to create a song, the same way you do on the website
but my way violates the law
?
๐
you press a button, i dont
only difference
It is what it is.
How would Suno lose it if people are using someone else's product that hijacking Suno's services to make money that, mostly, Suno never received?
amount of people that switch based on no api access
Cool, goodbye then
99% just wanna create songs, rate limiting is the solution
you rate limit api and these people cannot scale
or create copy cats to suno
but devs can create songs if needed
๐คทโโ๏ธ
Suno are aware that there is demand for an API (it is often suggested/requested)
They do not currently provide one.
Until such time as they do provide one, API access is forbidden.
If there's an alternative music generating AI offering an API, great, people can use that instead.
To be clear: the LIMIT is 4 minutes. So song lengths are up to 4 minutes amd not over. But may be, as any song you may hear radio or streaming , anywhere in a range of 1 minute up to 4 minutes.
Do you guys think they'll ever fix the "replace section" feature or is it going to remain like this forever? It's horrible but when it works, it's good. But it eats tons and tons of credits and doesn't usually work so it's probably designed to eat our credits and not work so you keep hitting the button and keep buying more credits. It feels like gambling.
Suno is practically rigged against the users.
You should work for Suno and work on their audio processing end ๐
suno seems to think every song should have a faux end that starts back up after slowing down
I hate how paid services like Suno, Midjourney, are so against providing paid API access. Guessing they are worried that people will just repackage their product and sell it
smart lads
It's the easiest to spot the difference between a constructive discussion and, ahem, its exact opposition is the use of the fallacious ad hominem argument. You know that you lost the dispute, they know the same. There's an abyss between "thank you for your time. I'm still unconvinced. Maybe one day, after I think deeper about the issue, we could return to this discussion?" and "you are [insert an insult here]". The former doesn't close the door, the latter blows the bridges and leaves one with fewer options in the futureย โ thus not only illogical by itself, but also illogical to use. How that expression goes, "to miss both birds with the stone"ย โ do I not butcher it? Besides, it's considered societally savageย โ and was read by a couple of hundred people. They have had thoughts along the same lines as me. Being nice to the others, at the other side, doesn't hurt.
Guessing they are worried that people will just repackage their product
- It's not worrying inherently. More of the product's sold without the front-end maintenance. E.g., OpenAI has an API.
- If there's a reason, the easiest way is to prohibit the repackaging of the front end/or the API itself is using the contract/TOS.
Probably the best is to look elsewhere. Maybe there just isn't enough demand, or possibly it's not yet known. Someone has to design it, implement (the initial cost), then maintain the software and the servers. A considerable demand needs itself for it: such things pay for themselves (and then some more) only starting at a certain scale.
what in the chatgpt
all i said, make a user friendly api with rate limits and charge high pricing, stops people selling it as external service and gives devs what they need to code their own inhouse stuff. there are plenty of wrappers out there that do it anywa
You said #๐โsuno-school message
smart lads
and #๐โsuno-school message
clever clogs
Maybe you said something else, but that was enough for me to lose all interest of it.
The pleasure is mine!
๐ซก
Made a song with a persona. After all the section swaps and extends, it no longer shows within that persona. Is there a way to manually move songs within a persona?
Hi guys ! I bought a Premier subscription for a year on December 1 and on December 5 I installed the Suno app on my android and it didn't open, the splash screen at the entrance to the app froze and that's it...I installed the official app on my android and it had just come out then..... I uninstalled it ...after that on my computer in the browser on suno.com all songs started to be created with the words โThis is a fake app let's go to sunoโ ...also now Cover, Persona, Remaster functions don't work properly ...it became impossible to use Suno ...I wasted a lot of credits these days for nothing. ...I write every day to support@suno.com and there is no answer at all...I wrote here and filled out the form 3 times https://sunomusic.typeform.com/to/zGsJSzsf and also no answer from suno...I did not use vpn while working in Suno and I do not use it...it's been 9 days since the failure...how can I still get through to Suno support ? https://suno.com/song/84ac59ae-937d-435e-a2f0-ea4167450646
Turn data collection into an experience with Typeform. Create beautiful online forms, surveys, quizzes, and so much more. Try it for FREE.
dance pop disco male vocal 80s song. Listen and make your own with Suno.
dance pop disco male vocal 80s song. Listen and make your own with Suno.
Hello everyone! Today I too, like many here, got the fake app problem. I really hope for a quick solution from the developer, before it becomes an epidemic.
Can anyone explain this? Did ai have a conversation with itself? ๐ฑ๐คฃ
How can i choose how long instrumental songs suno creates? My suno keeps creating songs 4:00 minutes with abrupt endingsโฆ doesnt matter what i type in. โCreate a song 3 minute longโ โcreate a song with a smooth end at 3,5 minutesโ still creates 4 minutes songs and it ends in the middle of a beat..
I think the longest you can get is 5 min if you use extend
because that's exactly what will happen
I dont want longer, i just want a normal ending. Im satisfied with 3 minutes
Hi all, dose anybody have problem with upload function?
Hey guys. I think I posted my message in the wrong place. How do I get the vocalist to scream speak some lyrics?
the search in playlist is incredible uncapable sht.
i thought thats one of the easiest part to build on a website, but looks like there not
if would be some option to change how much song displayed on one page i would can use effectively the browsers built in page search, but ofc that not possible too
more i use suno, more i disappointed
for example, get stems: actually it makes worse separation than a random website for voice extract, meh
but this search, seriously... have to browse one by one all from 1000+ song, because playlist search cant work well... cant believe
If you're using the suno.com website or the official Suno app, this shouldn't be happening. The extra lyrics at the start (and the "SUNO" watermark on the picture) are meant to only happen for accounts that are using a fake app or trying to hack the Suno API.
Please fill out this form so the Suno team can get this resolved for you: https://sunomusic.typeform.com/to/zGsJSzsf
Can you be more specific? Two things come to mind from the usual complaints:
- Cannot make it public? Suno detected a voice-like sound in the upload which disables the Public switch (and I think creating persona?) in case of copyright infringement
- Cannot upload due to trying to upload audio detected in ContentID. I really should not need to state that this is against Suno's TOS because it is something you agree to twice: https://media.discordapp.net/attachments/1069381916492562585/1267223169979256913/sunoaudiouploadtos.png?ex=675e98cb&is=675d474b&hm=f1dcee96ba5bd56e4e3ecfae620632f6fa9c9c0a23406128f2911aa2cec5ffe4&
but it happened. I didn't use any third-party apps, only the official website and the iOS app! And why would I need that if I paid for the service?
Because something on your end is triggering the automated services -- they gave you a link to inform the devs that you're being impacted so they can look into WHY you're being flagged and try to resolve that.
How can i choose how long instrumental songs suno creates? My suno keeps creating songs 4:00 minutes with abrupt endingsโฆ doesnt matter what i type in. โCreate a song 3 minute longโ โcreate a song with a smooth end at 3,5 minutesโ still creates 4 minutes songs and it ends in the middle of a beat..
But how to solve my problem then? My suno only makes incomplete songs 4:00 long..
That's bang-on in regards to the implementation, but instead of guaranteeing an end to the song, I consider the second-long extension to be "the terminator", as it attempts to eliminate the context from the earlier clips -- it's intended to hardcut between styles, but I wouldn't be surprised if that could be used to help end the song as well, with some experimentation for what you use as "the pivot"
Sound advanced to create a song.. and takes alot credits compared to normal
If you just use Suno's basic mode and never venture into custom mode to experiment with tags and other prompting methods that are frequently discussed here, you're completely at Suno's mercy - stuck with whatever it gives you, with very little actual control from you.
might be fun at first, but then it'd get boring ๐
Agreeably so, but there's clearly some stuff that should be more straightforward -- like asking to fade a song out at the very least ๐
and maybe automatic guitar solos for every 30s of non-guitar music? 
Good morning everyone. Is it possible to add musical interlude or instrument intro before the lyrics start singing at the beginning?
I had a good song but it just doesn't have much intro at all.
The best way to do that is probably to start with an instrumental; no lyrics at all. Once you get an instrumental intro that you like, extend from that (tweaking the start time) with your lyrics in the extension.
If the song is already generate, it'll be too late right?
Or maybe get the sterm out and chop that part somewhere that it sound like the intro and add it in the daw, i guess?
Here's the song:
https://suno.com/song/cae94070-67cd-4ffd-b808-7a0030ba9cc7
Emotional gospel ballad with a crying, raspy female lead vocal, rich choir harmonies, and dynamic backup ad-libs. Soulful delivery, heartfelt lyrics, and a steady, uplifting progression. song. Listen and make your own with Suno.
Likely. If there's some space at the beginning, prior to the vocals coming in, you could try to use replace on that space and include something like '[Instrumental Intro: Playful, Percussive]' as the replacement text. That can definitely work, but not always on the first.... or even 5th attempt.
We really need a Prepend function to go with Extend. Or at least the ability to select much smaller replace sections.
Yeah. We can extend the middle but no option to fix the begining eh.
In future gens, if you include some prompt noise after the [Intro] tag and before the lyrics start, Suno will often give you a longer instrumental intro. Prompt noise is basically anything that Suno won't sing - I use slashes for that, like //\\//\\//\\/\/\\//\\//\\//\\ But you can use pretty much anything except letters.
So something like:
//\\//\\//\\/\/\\//\\//\\//\\/\/\\//\\//\\```
Cool. Thanks for the tips. I normally only use [intro][musical interlude] and then the lyrics. Maybe that's why I could'nt make any nice intro.
Here's an example using an intro and your first verse. It's not a very long intro, but it does give a bit more space at the beginning. The length will probably vary from one gen to the next. The best/most repeatable way to get a longer intro is going to be to extend from an instrumental, but this can still definitely work. https://suno.com/song/5c222348-809d-4c9f-9883-d3f3cd776685
Emotional, gospel ballad, crying, raspy female vocals, rich choir harmonies, dynamic, ad-libs. Soulful, heartfelt uplifting song. Listen and make your own with Suno.
Wow. That sound really good.
So the trick to gen a new song is basically right a short intro with the first verse, pick a good one and then start extend with the remaining lyrics, right?
dumping too many lyrics into the song at once, we might not get good intro, I guess
if you want an intro that's 10s or less, you can probably just add the intro section (with some prompt noise) directly above the full song. For longer intros, I wouldn't even include the first verse; I'd just focus on the instrumental portion and then add the first verse and whatever other sections you want in the extension(s).
I only included the first verse just to demonstrate the idea - it should still work with the full song added below it.
Cool. I'll give it a try. Thank you very much.
...and the prompt noise doesn't have to be slashes. I just use those out of habit now. Some people use emojis ๐
anyone can say me if anytime a subtitle tool (like crop the sub of the suno videos) comes to suno? i tryed some out and all are bad (i hope its not from the suno songs)
and other ascii/ansi symbols
I haven't heard anything, but people do periodically bring up .srt files. I'd make the suggesion in one of the feedback channels.
i ask only because i doo at my channel english and japanese songs sub edit for english is okayish but the japanese .... iam too bad in it and cant till now read the signs ๐
How do I properly attribute credits to Suno when I am using the free version and uploading outputs on YouTube and TikTok? I am writing "Made in #Suno" and tagging Suno on both YouTube and TikTok every video, but I wanted to clarify the proper credit phrase or process. Also, I am not monetizing the videos since I am only a small content creator.
I only wanted to know the right process and credits to avoid any future conflicts, since I am planning to avail the premium membership once I get the hang of it from posting demos. Thank you.
Cool. I made a few song with your method and it works great! Trying to figure out how to get black female vocal now :).
Awesome! I'm glad that was helpful.
You can actually create your own accents with Suno by phonetically spelling out how you'd like the words to be pronounced. I've gone back and fixed all of the lyrics after this track was generated, but the original prompt used text like:
Deh en nite, in deh dahk.
Oh, I cahnโt konteyn,
Yohr spahk... Et lit my haart!!```
instead of the (now) displayed:
```Oh, I call your name,
Day and night, in the dark.
Oh, I canโt contain,
Your spark... It lit my heart!!```
https://suno.com/song/5254d589-30a7-4851-b396-882d3902a296
Playful, Cabaret Crooner, Velvet Jazz Whisperer, Smoky Room Serenader, Champagne Swing Vocalist, A Cappella, Velvet Lounge Diva, Midnight Melody Charmer song. Listen and make your own with Suno.
you can see the original prompt in the video lyrics, with all of the wonky spelling ๐
Unfortunately, if something is made with the free plan it can't be distributed. But I'd like to know if Remasters/Covers can be?
Do I have to delete it on my social media platforms?
That's probably a question for support@suno.com . You'll surely get some answers here, but this is a community discord - we're all just users like you. We may not be the best source of legal advice. ๐
I tried emailing them, but they told me to chat first on discord, before following up the email, but I resent the email already
Ah alright. A friend asked me to upload my stuff on youtube so I'd like to know.
I'm missing some context here...
@heavy laurel Ahh.. I see your original post - I missed it earlier. I'm really surprised that Suno's redirected you to other Suno users for the answer to that.
What happened is that I uploaded like 4 of my prompts on YouTube and TikTok, then every video, I tagged Suno and such. Then, just now, I read the terms and conditions of the free user on the basic questions, and it says that the distribution under free accounts must attribute credits to Suno, and would only use it for non-commercial purposes and such (I have not monetized my videos since I am only testing it out and posting it to see if people actually like it while tagging Suno on my first four demos). Then, I tried emailing the Suno, but then the bot asked me to addressed the question first here, then I just resent my email
there might be some information for you here https://discord.com/channels/1069381916492562582/1294423389515284552
Thank you, I will also wait for the email
Damn.. That's a good song. Funny. My friend call me Ducky for real.
So the original (first part) was created with:
Oh, I kohl yohr naym,
Deh en nite, in deh dahk.
Oh, I cahnโt konteyn,
Yohr spahk... Et lit my haart!!
and then you extend it with the proper lyrics after right? It sound like it follow the style right after it too
I don't plan to monetize my stuff either, just to upload it.
Here you go, then. lol https://suno.com/song/24d87536-c076-4f44-8cb1-d819b959bd0d
Playful, Cabaret Crooner, Velvet Jazz Whisperer, Smoky Room Serenader, Champagne Swing Vocalist, A Cappella, Velvet Lounge Diva, Midnight Melody Charmer song. Listen and make your own with Suno.
No, I wrote out the entire song using lyrics like that. I did use a couple of extensions, but it was only because I didn't like the direction Suno went with the original track after about a minute, or so. I used the extensions to steer it back in the direction that I wanted.
Once I had the completed track, I used the "Edit Lyrics" feature to put in the properly spelled lyrics.
hai loff yurr sonks
Oh. I see. Every instruments in your song sound crispy clear.
That surprised me a bit because I was mostly having fun with the styles, but Suno ran with them!
I don't think people know it's from suno for sure. lol
i've been trying so hard to make suno detect that the base i made doesn't has any lyrics And it doesn't has any lyrics so at this point it's worth asking here what i can really do to fix this.. how do i fix this?
it's for making it public
cuz
yeah
i need some help
I think the problem is that there are elements in that sample that have a vocal quality to them and Suno is likely detecting them as actual vocals
that's the thing, i asked here so maybe any staff could help me, well trying to solve this
cuz i don't want to make this from scratch again
and because of free fl studio
i don't have the freedom to go around
and do some fixes
can someone help me or guide me?
when i extend a song the vocals suddenly change to male instead being female
even when the prompt is Female vocal or anything with it.
This is a community user forum, tho. We're all just users like you.
Hi everyone, are there any updates on the distortion issue, code name Shimmer! LOL
mhm
well
i know they have a mail
do you know what's it?
i only have ali at suno's mail
thanks
No official word, but users have reported getting less shimmer. I have mixed results. Sometimes I think the shimmer is finally getting tamed... and the very next gen comes out super-shimmer.
I just tried a new remaster, and its much better.
I have noticed close to zero shimmer lately, but I have also been working on rock-style songs instead of EDM. I assumed that was the reason but maybe not
does v4 refuse to do sound effects or what? Trying to do a simple gunshot noise and he just keeps saying "gunshot" lmao
Hi can someone maybe help me. In the app I signed up for the pro plan. Unfortunately when I log in on the website it still says basic.
I would love to download the cover song for one of the songs I made in the app. Unfortunately I can't do that now, since it seems not to be possible in the app. And the website still says we're on a basic plan. Does anyone know what to do? I use the Android app.
@abstract flower that's pretty interesting... I'm even curious to see an example of a song that contains zero shimmering ๐ฅน
You must be really lucky.... Or you must be creating songs in the music style that's not really common in the inside the model's latent space
sadly its still there, just reduced. You can see all the black dots in my spectral display on Adobe Audition. Its improved but not totally gone.
can you give me the song, I wanna see what it looks like in Steinberg Spectralayers
To confirm:
- You have the same app as the redirect from https://suno.com/Android?
- The profile in the Android app is the same profile as the one on the websitr?
- You received an email confirmation about the payment?
If yes to all three, please contact billing@suno.com or support@suno.com so they can investigate the issue.
Do note that Support is currently experiencing high volume so it will take longer than usual to get a response. My email two weeks ago took about 7 business days
Hi, yes to all three indeed. I contacted support@suno.com they suggested asking the question here as well. I'll wait for their response..
Am I on an island when I say how much it would be helpful if in the EDIT option we hade FADE like we have CROP? There have been times that the tempo assigned by Suno makes a song that should be 3:48 be over 4 minutes and it just cuts off. I download it and do the fade out in Audition easily enough but if I put it into the public view basket in that state I feel like I just finished working out and went to lunch without showering. It's just "wrong" to do that. So please, powers that be, Suno gods, "fade"!
Hi, you should suggest that at #๐กโsuggestions-web - it is a very good idea
I would be curious as to what your tool does with the "sunothing_original.mp3" and how it's output compares to "sunothing_shimmerdimmed.mp3"
how is this shimmer thing going? I am trying new creations with v4 and it is still there... BTW, I have noticed that it kind of exists in v3.5 results as well, but sounds like in less annoying way...
It's still shimmering and now sometimes clanking. ๐ It gets worse if you extend songs. Kind of okay until three and a half minutes.
hold on, I'm restarting my PC ๐
Any fixes for laggy playlists not auto playing and being unresponsive? I have abotu 125 songs in one but it lags way too hard
ok, so...
I think something in my settings is off right now
I was rewriting it a lot today
will get it fixed soon enough
but it works extremely well the louder the shimmering is
with this syle of music, I get a lot of songs that have electric guitars that only play in the left stereo channel. Anyone have any idea if this a common pattern in 70s-80s rock or is Suno just being weird?
70s hard rock, 80s glam metal, guitar-driven, heavy riffs,, energetic, stadium anthem, powerful vocals, aggressive drums, electric solos, amplified, intense, headbanging, thunderous
but lately I've also been trying to tackle the high frequency interpolation (aliasing) artifacts that now exist even in v3.5 generations
That's due to mid/side encoding of the older codec they had (taking a wild guess but v3 used to have a pertty bad phase inversion in side channels)
so, in some styles the side channel is off-center
anything that can be done about it?
okay, changed a bit parameters and re-ran it
yes.... unmix stereo into mid (which will sound dual mono) and side channels (a stereo field without the mid component), then balance the side channels and join it back together with the mid channel to get a new, fixed stereo
most audio editors support mid/side <-> stereo conversoin
thanks. I don't understand any of that but you give me a starting point to research, haha
side channel is basically the same thing that mp3 format uses when join stereo mode at lower bitrate is selected... or what's being used by analog dolby surround encoding
this recommendation is suno-specific.... you can't fix all kind of music like that.... but it works extremely well on suno generated stuff
get yourself the free audacity editor... and then you can get youself:
https://www.voxengo.com/product/msed/
basically with this you can do all by just turing a few knobs
I'm starting to run into issues with the program being able to identify styles. I've made over 1000 songs on Suno and as it's involved, I'm seeing it lose its ability to pinpoint styles and sounds and genres. I can provide examples of where I put in house music and come up with heavy metal. Or put in dubstep and come out with R&B. A month ago, the last time I was working on this it was dead on the money with its genre IDs. Now it's way off... I'm curious what has changed in the past 30 days because the past few hours I haven't seen it Nail is single genre. I've tried excluding some styles and that's not helping. please let me know what I can do differently to get this back on track. I've considered maybe starting a brand new account because I feel like maybe the outputs are being affected by previous tracks that I've liked and dislike and it's just getting clogged with input to the point where it can no longer be specific. Any thoughts?
back in the day with a single track audio editing software (like sound forge), you could mix a stereo file with a phase inverted copy of itself... thus effectively cancelling the mid channel out, leaving only side channels...
and then you can mix the resulting with a full stereo track again, and you will end up with middle channel...
Can you share an example song that it affected? And does the song artwork by any chance have the word "SUNO" over it?
there was a huge discrepency when v4 first came out... I just couldn't get it to produce same styles with the same prompts as in v3 and v3.5
which one (I have a mac)... "AU, AAX for Mac" or "VST, VST3 for Mac"?
took me more than 10000 credits to finally kind of get a feel for it
"for Mac" obviously - I mean... Audacity works with vst fine
but there are 2 mac options
I absolutely can, is there someplace specific I can send these for review? I will I'll put a couple and add the prompts to them and you guys can have a chance to review them and see what I mean. Happy to share, and by the way this isn't a huge bummer so much I find it fascinating that the way this program evolves there are a lot of things that are affecting what comes out and the final product. Obviously Suno is a work in progress and so is the music we make here so it's all interesting regardless of whether the track is a hit or loser, there's always always something to be learned from each new track
not use about AU and AAX format, although they should work.. one is used by logic pro and the other by ableton, if i'm not mistaken
well, we don't work for suno... you'd be better off writing support ๐
go with VST
Just post a link to one of them here in this channel is fine. If anyone has any ideas about prompt changes that might be needed etc. then they'll make suggestioins.
I got myself that VST plugin now too... in other DAWs and Editors I usually use built-in stuff
or iZotope/Waves stuff but it's always nice to have VST installed which works also in OBS studio and Video Editors and whatnot
Another problem I'm seeing here is that when you try to make either a cover version or you use the extend feature, the tracks are coming out are absolutely 100% identical to the original no change whatsoever. I've listen to them over and over there's zero evolution from one version to the next
Interesting.
Now I am intrigued what your tool does.
my "sunothing_shimmerdimmed.mp3" is the result of running the "sunothing_original.mp3" through Descript Audio Codec which seems to eliminate most of the shimmer.
Your result sound a bit different I think so I guess your tool does something else?
well... my journey started with trying to understand and improve upon Descript Audoi Codec... I'm working towards higher fidelity and ability to build a lightweight GPT-2-based audio processing AI... so I could use it to turn all kinds of garbage into CD-quality, not to mention eliminating artifacts...
So yeah... I was in the middle of screwing around with DAC, as I was looking for a way to fix the center channel balance issue it slightly has before I start modifying codebooks or changing the decoding part of the vocoder
could Audacity be used to clean/master a suno track? (total noob to this entire world)
my main point is to create a audio "remastering" toool that more like audio enhancer and cleaner and upsampler and whatnot... I'm not trying to build audio production helper here. I have more ideas from generative AI perspective
as good as any other...
what I'm trying to achieve is to neural edit and uncompress / denoize, and fix AI-generated audio (not only from suno but from other stuff too)
TRIBAL HOUSE, DARK, SLOW , SUB BASS, MALE VOCAL EFFECTS
https://suno.com/song/618a36d6-a9e4-4944-8164-368d4637f3cf
not even close to the prompt. You'll see the prompt is pretty minimal which is where I've been trying instead of these very brief/minimal prompt, in the past I've had a great luck with that but I'm obviously seeing diminishing returns on this
TRIBAL HOUSE, DARK, SLOW , SUB BASS, MALE VOCAL EFFECTS song. Listen and make your own with Suno.
so, in other words... while I made modifications and was able to get DAC to produce higher fidelity output for CD-quality inputs (compared to the defailt behaviour), I accidentally discovered that I had somehow made it to unshimmer suno stuff... especially the songs that are shimmering or very noizy... the files you sent me were really light compared to what it can actually do
slow techhouse, warehouse aesthetics rolling sub-bass with a detuned chorus effect, evil 4th tritone chords, deep pitched down male vocals repetitious tribal percussion, vocal snippets
slow techhouse, warehouse aesthetics rolling sub-bass with a detuned chorus effect. evil 4th tritone chords, deep pitched down male vocals repetitious tribal percussion, vocal snippets song. Listen and make your own with Suno.
your tags usage is really inconsistent...
sections should all be in square brackets
Are all your songs ending around this same length (under 3 minutes)?
I get that, but I could also show you tracks where I use the same formatting and they came out absolutely perfect. Like I've said I've made over 1000 songs on Suno I've put out two full length albums using these tracks that I can point you too and I've had absolute no problems. so brackets are no brackets definitely something and what's happening now and what was happening over the past few months with these outputs
I have actually no idea what style you're trying to get out of it, or how it should sound
Then why why are you commenting?
If you have no input to give then what's the point? I'm looking for insights here
It should be pretty cut and dry, those prompts are extremely self-explanatory. House music
I'm sorry, how exactly are we supposed to help you if your problem ain't 100% clear to us? that's the reason why we're asking for clarifying quesitons.
Bro, come on. These prompts are extremely self-explanatory and it's so obvious that the music coming out of the system is nothing like the prompts. I'm asking for house music and I'm getting rock 'n' roll. How much easier can that be to understand?
Simplify it dude. You're over complicating the question
maybe take 5 mins and chill first lol
well, style prompts have show to be case sensitive...
does v3 give you a different result with the same prompts?
also, does v3.5 give you wrong styles all the time? sometimes all you have to do is roll with the same prompt agian and the next pair might work
Here's a full length album that I put out, I use just as simplified prompts for this rock and grunge album as I did for this electronic/house music. And the tracks came out flawless
Or change the prompt when you're fed up and then Suno finally gives you what you aked before ten times. ๐ Just now it seems very tricky to get a decent outcome.
Ah yea even just running it through DAC alone without modifications is quite useful as seen here.
This person is 100% correct. Something has changed in the program in the past month or two that is definitely not outputting with the same precision that it was say this past summer.
also, you have a literal contradiction in your style prompt... SLOW vs TRIBAL HOUSE
these two styles will be clashing
Start with House in style, capital letter. Try putting [Steady Beat] Next to each section, Try putting [Tempo Lock] [Steady Beat] or [Tempo Change] [Steady Beat] if you want to change, try [Syncopated Rhythm] since you are sticking with house.
This is what Happens when you have somebody who's not well-versed in a specific genre giving advice. Tribal house especially the Afro version of tribal House is often below 120 BPM's. You're thinking of progressive tribal which is a completely different genre. And well establish I might add.
It's still house...lol. If it is house then it literally means bom bom bom bom bom...house music.
Respectfully there is no changing tempo in house music, you have to be able to mix the tracks using the beat, and if there was a shift in tempo it would create a "train wreck" if somebody was trying to mix it.
yes... by default DAC runs at 8kbps and in that mode it sometimes would clean things up... I discovered that later...
in 16kbps mode it won't have much effect on these artifacts
Good morning, Iโve been trying for a long time to separate the tracks of the songs Iโve created or at least remove the vocals using another website because I want to sing the created song myself since Iโm a singer. Unfortunately, despite trying various dedicated websites, the instrumental track, after removing the vocals, ends up being of very poor quality, which doesnโt happen with songs not produced by Suno. It seems the vocal track also removes some of the music's frequencies. I specifically purchased the premium version hoping for a better result. Could you kindly help me solve this issue? Thank you so much for your support.
Respectfully, I know house music and I am not telling you to change the Tempo. I'm telling you ways to get suno to do what you want. So you put it at the start.
remaster the instrumental stem with v4
Yea I tried in 16kbps, it reduces it a bit only a tiny amount. 8kbps DAC seems like a quick and "easy" way to clean it up outside of suno
But you got this so....
It's coming and going. ๐ I'm currently fighting with Suno because it can't differentiate between certain Latin genres. It gave me a few good results but mostly it mixes the genres as if tossing the dice.
unfortunately 8kbps degrades the rest of the audio quality too
I'm struggling with it too, I can't find any kind of consistency from track to track. One will come out dubstep, the next one will be rock, I can never guess what I'm gonna get when I press that button again with the exact same identical prompt. It wasn't like this before honestly, this program used to be consistent. I'm getting the sense that somebody messed with a good thing, and now it's not quite as good
But so far i think less than doing a V3 or V3.5 extension inside of suno.
How long ago did it start doing this? And do you always get shorter songs than you used to?
but.... the fact that it at lower bitrate cleans it is the reason why my mind went straight towards building a tiny music transformer... I only need a 2-3 second long context window for doing all kinds of audio manipulation....
it won't even be that hard to train.... maybe a couple of H100-s in the cloud for 2-3 days
but I believe it will be easier... as if I make it a clean-up and enhancement model alone, wihtout any text prompting support, it will be small and fast
in less than 24h after Karpathy made his GPT building youtube tutorial, I already had a 90M small LLM up and training on 1GB of wiki ๐
so, I already have GPT building experience
So, try reading. See the first said [Tempo Lock] [Steady Beat] is there a reason you needed to explain house to me for that?
I hear you. The last days have been very bad. Genre inconsistency is not the only problem. The instrumentation gets distorted and the voices sound tinny at the end of 4 minutes. And extensions are atrocious. ๐ฌ
It's a pretty great idea.
keeping it pretty simple for now i guess?
One could always try and do something more complex later on.
yep...
all I want is a sliding context window such that audio comes in one end and modified stream out from the other
it would be able to learn to clean up all kinds of unwanted stuff from audio ๐
It is amazing when people look for help and then talk to the ones they ask as if they are idiots...lol
Can anyone give me a hand
Not digitally
I have no idea how you would create the dataset or how you would actually train it to learn how to clean things up, I'm not smart enough for that. I can only use A.I. stuff and slightly modify it (like the code around it not how the models work themself) to fit my needs ๐
So are you in the programming area or training area with all this stuff. I mean at the moment.
Are you referring to me?
I'm more like a regular programmer who as a lot of interest in A.I stuff
not someone with actual A.I science experience
No..sorry
can you give me some clarification?
Same
Good morning, Iโve been trying for a long time to separate the tracks of the songs Iโve created or at least remove the vocals using another website because I want to sing the created song myself since Iโm a singer. Unfortunately, despite trying various dedicated websites, the instrumental track, after removing the vocals, ends up being of very poor quality, which doesnโt happen with songs not produced by Suno. It seems the vocal track also removes some of the music's frequencies. I specifically purchased the premium version hoping for a better result. Could you kindly help me solve this issue? Thank you so much for your support.
it's basically the same as training an LLM, for proof reading and automatic puncutation, for example.
most of what goes in goes back out.. the only change woule be grammar corrections, capitalisation, punctuation fixes, and so on.
so, from that perspective, it wouldn't be any different... it's just instead of text tokens you train it with audio tokens
Suno creates full tracks. I'll explain
But you would need examples right?
Like audio tokens that have artifacts and a clean reference?
Or am I thinking wrong
Since it creates full tracks you get the vocals and instruments all together. You can separate stems but the vocals hit most of those frequencies. That means that with out some AI to fill in the gaps you are going to have ghosting. Leftovers. The vocals will be pulled decently but the instruments will have some echos left
yep... and I've got plenty already...
see, I've burned tens of thousands of credits lately... and since most of the music I make is prompted with my own uploads and then extended and/or turned into covers, I will would get back parts of my own stuff with SUNO "effects" added to my CD-quality inputs.
from there comes the diff.
for this particular restoration task I woudln't need a lot of audio data, as we'er not actually trying to learn long sequencer nor try to map audio features to text tokens
So, if you just want tracks to sing over you can create instrumentals.
right forgot about uploads+cover
There will probably still be other differences in the 2 versions but i guess it could still work.. makes sense now yes
And isn't it solvable for you even with other paid websites? I also tried to make the instrumental piece, but the logic of Suno is that the instrumental piece is fully instrumental, so it's not accompaniment for a singer because it creates the melodic track played by an instrument.
Right, maybe soon, but since the frequencies are the same you can't separate them any more than you can separate them with your own ears. They exist together. I'll explain the Instrumental part next
There is no perfect stem splitter as far as I am aware
well.. not for extended songs... I've made tracks that start with 90 seconds of my own upload and only extended at the end. as soon as you do "get full song", it will go through the whole pipe... so I get back degraded versions of my uploads..
besides... I've been kind of able to isolate and/or replicate suno-like degradation of audio myself during my GPT-experiments already. I was able to reproduce all the distorted generations + also the shimmering myself. so I woudln't even need to use that much
There actually is. Steiberg Spectralayers 11
I moved over from RipX
well... I mean as perfect as it can be
You can make custom instrumentals. Unless you mean you want backup vocals.
a few weeks back I started using spectralayers to clean up suno generated audio crap... and then put songs back together... you wouldn't even know that something was there
What's even more amazing is when Suno puts people into the mod position to answer questions that don't have any clue about certain specific kinds of music. Bottom line if you don't know about specific genres. You shouldn't be commenting. End of story. it's not about them being idiots, but it is kinda lame when people are condescending regarding certain genres like "house music what's there to know"? Obviously there's a lot to know or they'd have it more figured out. The bottom line is Suno is behaving strangely, and I think a lot of people here are feeling that same sentiment. Something has indeed changed in the programming here and is not behaving as consistently as it was a few months back. I think it would be smart to add specific mods instead of having one person answering questions about every kind of music. That just doesn't work and it shows a kind of over simplification of this complex process.
It can splitt into more than just vocals + 1instrumental_stem right?
Gotta have to look into it.
So far I only really needed vocals+instrumental and have been using bs-roformer or mel-roformer localls
We aren't mods
I also do a lot of denoising like this... that suno noise in the higher bands is really annoying... and it usually sits in the "Drums" layer's mid channel, while shimmering is on the same layer's side channel. So easy to separate and clean
yep... a lot
Symantics. If you're answering questions you're a mod
And if you don't know the answer to the question you shouldn't be commenting, commenting simply for the sake of having your voice heard on something you know nothing about is dumb
yeah... and when I tell flat-earthers that the earth is not flat, I'm a paid nasa shill
same logic
What do you mean by custom instrumentals? If you only make an instrumental, itโs like creating music that is difficult to sing over. There are some pieces that are not singable because they are originally instrumental.
Oh just checked the price lul.. gotta wait until I am a bit more serious about putting the music up for people to listen to ๐
Had an answer, you said it was wrong. Don't know what to tell you
Ahh...yes you are correct