📚┃suno-school

[torpid thicket]

But this song has to be fast :L

[sleek stump]

Maybe ill try 4.5 again with Metal after its out of beta..

[torpid thicket]

What's your SUNO?

[tawdry scarab]

https://suno.com/song/355b87bb-9ab3-4e52-abc9-1ddc0c369f58

[sleek stump]

@gingerm but I dont have a lot public on Suno.. just some discord challenge stuff mostly.. but a couple are metal

[tawdry scarab]

relevant to the mission of the channel

[polar spade]

That's considered "normal"
Before clicking Create , we should go through the check list from top to bottom. Check if you have custom or instrumental on. Check lyrics for spelling, tags for proper syntax and brackets, click the version wanted 2, 3 ,3.5, 4.5, check extension is OFF (maybe still on from a previous) and so on ...... get to the bottom, click Create.... then remember you forgot to go back and past in the correct prompts when you added a persona. 😄

[wise trellis]

A system error occurred. Please tell me how to solve it.

[tawdry scarab]

what are the things you were doing when this happened, and if there is a prompt associated, are you comfortable sharing it

[sleek stump]

You wait

[wintry nacelle]

hey all, not sure if there's some tutorial out there but i'm jumping back into suno after a few months and the editor has been amazing

i see that i can replace sections but is it possible to add a new one?

or if i crop out just the chrous, can i use it to generate sections around it?

[sleek stump]

You are limited to replace, crop or extend (from a set point). People have mixed these to try and get what they want, some with success, others frustration.

[wintry nacelle]

ohh so if i have say a first verse and chorus i like, i can try to extend from that?

[half dragon]

Yes, you can.
Either use the Edit Tool > Extend... or 3 dots menu Create > Extend.

[stray lynx]

You probably have to stick to 3.5 or 4 unfortunately. Right now all 4.5 knows how to make is one specific pop song that it plays in every genre. It's just dressed a little differently each time.

I'm trying right now to make something interesting by replacing small sections and cropping but it's ridiculously labor intensive.

[silver rampart]

Hi! I have like 4000+ song in my library. My question is, is there any storage limit per account? Is there fast way how to delete them all at once? I want to try if after deleting junk my account will generate better results 😉

[wooden trail]

If I want to use music created by someone else in Suno in my YouTube videos, how do I contact that user for permission?

[stray lynx]

How it started
https://suno.com/s/j2cA11bRi0WAqZtG

How it's going (replacements starting at about 0:18 up to about 1:32)
https://suno.com/s/l5FZ64Kha3aFSO8l

I'm tempted to slash that entire first chorus off and try an extension, and repeat the process until the whole song is in a pile of scraps.

[half dragon]

There's no storage limit as far as I know.

[half dragon]

Currently there's no messaging system in Suno, so you can't contact them. You just have to know the person.
Actually, you can comment to the song you want to use... ask them that way.

[silver rampart]

Hm but I’m worried if all those gigabytes of data can affect final music output on my side. There’s no unlimited hosting 🙂

[stray lynx]

That would only matter if the music that is stored (probably in s3) is actively part of the training model.

I doubt it is.

But ask again in 3 months! Maybe they're feeding all our generated songs back into the model.

[half dragon]

I haven't heard of that so I wouldn't worry about it.
Also, I think YouTube has unlimited storage hosting ...so not impossible 🙂

[stray lynx]

You know, if the 4.5 model wants to do weird pop song things as it approaches [Chorus] then I wonder what happens if you just don't have one, but instead pretend the song is just verse after verse

[torpid thicket]

Go see what most of the staff members make for songs - 90% of it is Pop/Rap

[stray lynx]

Then I guess abandon all hope 🐦‍⬛

[polar spade]

If you have section A B. C D.... and you like A and C , hypothetically, you could remove from 6 second right of A all the way to 6 seconds left of C. Then Bracket that new formed 12 seconds created between A and C , and replace with something new. Removing section D, you can Replace it with something or just extend after it.

[pallid leaf]

https://suno.com/song/1cc40ddb-c4d3-4105-a188-b956ddf485c4

Why does Suno so frequently mispronounce the first word of a song? 🤔

[half dragon]

Maybe because it senses a French mood..

[wooden trail]

The question is whether they will answer or even see the comment...

[half dragon]

Oh they will see the comments!
Now whether or not they will answer is another story.

[sharp olive]

Thanks for your answer. I uploaded a track and I make a cover of that track. I'd like to make a persona from a cover generated

[fiery crag]

Is it just me or does Suno have a hard time with pronunciation when more than one language is used in a song.

[neon oracle]

Hey guys, I'm having an issue with Suno. When I try to generate a song, the process starts loading, but then nothing happens. Has anyone else experienced this?

[daring lava]

Hey there!

Is this via web or mobile?

[neon oracle]

mobile !

[daring lava]

Okay thanks, iphone or android?

[neon oracle]

Android

4.5 version

[daring lava]

Okay thank you! When you get a chance (if you'd like to) could you make a mention of this in the #📱┃bugs-android channel, so the appropriate team member can see what's going on. Any non account-specific details would be helpful (please do not share your private info). May want to try deleting the application and re-downloading it too if the problem persists.

[neon oracle]

Okay, thank you

[bronze vortex]

Yeah, but even the walkthrough from that link tells me that it creates a "whole new song", which isn't what I want. I just want to be able to change the lyrics and keep the song the same.

[hollow hare]

What that means is it creates a new song track rather than permanently changing the clip you originally selected.

Though I hear Replace can be very fickle to use. Can't really say for myself since I rarely use the edit tools due to using mobile

[polar spade]

I misunderstood your post when you said "adjust" the lyrics. That usually infers replacing out a word here and there. Replace is what's often used for that.

Regardless, the walk through was the process to change out words, phrases or verses.

Are you online at the moment? Since a better question would be for you to walk through what you're trying to accomplish. Change the lyrics and keep the song the same, can mean different things with Suno. You want to use a Suno song you created, then change the whole lyrics? You want to upload a song and replace the lyrics? 🤷

One means is by separating out the "vocals" , which may be what you mean, not just the words/lyrics that's sung but the actual singer's voice as well?

You first then, mate. What is your end result youre trying to get. And we can better answer.

One thing to remember, generative AI means with each change, Suno takes all changes and re-assesses to retro-fit the square peg (new lyrics) into the round hole (instrumental /song) so there will ALWAYS be changes just as if we did in real life songwriting. Syllables change, inflections of the new words, etc. But we can get close to the intended melody and nuances and we can switch them out (all of it) and often get something nice that works.

[polar spade]

Quick note: if its been process spinning for at least 1 minute , we can often just refresh , or even click back to Library, and the gens then show up completed.

To say "process starts loading" and then "nothing happens" at the same time leads me to wonder if something did happen next: so 1) as of now the process is still loading and has never stopped and you're stuck and cant use Suno? 2) you now eventually see two generated outputs and it just seemed to be grossly delayed ? 3.a.) the loading processes of two outputs just disappeared evetually as if nothing happed at all (and you lost 10 credits)? 3.b.) the same but you were refunded or never lost credits? 4) you got a notice of "system error" or "can't process" or any other which ended the processes? Any of those?

When reporting to the bug channel being more detailed-specific will help. If possible, update your report to include more details. Thanks.

[quick wedge]

Try the phonetically method and see if that helps to get the song to pronounce the words correctly. I’ve seen comments about it before. You can try it too.

[burnt zephyr]

Hey Guys, i'm having a issue with the refund of credits. When i want to extend a song it costs 10 credit. But when SUNO fails to extend it says credits refunded, The only problem is that is only refunds 5 credits instead of 10. How can i fix this?

[polar spade]

Refunds are give back 5 credits per the output that didn't get extended more than 8 seconds. If you had both outputs under 8 seconds you should get 5 credits per. So only one "failed" output is only 5 credits.

Did you get one output or both under 8 seconds?

[burnt zephyr]

No they where 3 and 4 seconds..

So both under 8 seconds

[polar spade]

"THEY" means both outputs of one click? So 2 "songs" extensions were under 8 seconds ?

[burnt zephyr]

Both outputs of one click to extend.

[sleek stump]

If you refresh your browser you will likely see that you were refunded for both, theres a visual reporting bug with the web UI reporting updated credits when they are refunded.

[burnt zephyr]

Thank you. That helped, it took some time tho 😄

[polar spade]

Got it... you added text.... sorry, I see now. Yes. You should get a total of 5 credits per failed output. They could show up at different times within about 30 seconds of each other. But if this is repeating, then make note of it. The only refund by Suno would come from billing/accounts, for the most part.

And as Ginger says, do a refresh to see if they all accumulate and finally appear.

[pallid condor]

When I try to replace a section recently, it's acting odd. I have a segment selected, when I listen to the segment, it starts playing from the the start, but then continues to play past the selection. Never used to do that, but ok. ...but then I go to create the replacement, and it replaces from the start point to the end of the song.

[burnt zephyr]

Thanks!

[polar spade]

Thanks

[pallid condor]

Then, I am trying to cover a song that is 3:44... and all the covers are coming out at 2:40s and 3:04... missing a large chunk of the song it tries to cover.

If I download the wav and re-upload it... will it be able to be editable?

I'm in love with my song that only needs some minor corrections, and I'm going to have to scrap it if I can't get this working again.

[polar spade]

It's hard to test through replacement issues now , for me at least, as there is no indication in the suggested replacement samples of how many seconds it will replace.

Before if I were replacing 10 seconds from 3:00-3:10 , it would say 3:00-3:10 on the suggested replacement sample. If longer, it might say "3:00-3:22" and i know that sample will go 22 seconds rather than 10 seconds.

Since there is no indication, we assume it will be exact as requested, and yet, as you say, continues to play past the selections.

I'm reporting that to bugs first so I personally don't have a good response for your question until that's resolved.

[pallid condor]

The title of the song replacement includes the range just after... in this case it says it's 10sec, but it actually goes from 03:05-05:31 ...which is outright

I am thinking it may be related to the song selection preview not being contained in the selection area. When I am trying to select the area to replace, it used to stop playing when it reached the end of your selection... to let you know what you're replacing... but now it just plays beyond the replacement selection.

[polar spade]

Preaching to the choir , mate. As I said, the samples are not showing the proper timemarks of what the segment will actually replace. #📱┃bugs-android message

Certainly , I play the sample, see it's not what I need and do another one. But it's not that helpful to NOT know the exact second the sample will end. It's a guessing game. And the timemarks are therefore incorrect in the samples. Yep. We're saying the same thing.

I'm not helping anyone further with Replace until the glitch is fixed.

Of course, to help, I'd at least say to make sure your "keep time duration same" box is ticked. You know where that is? The cog settings button down by the timemarks in the playback section.

[ebon sand]

there are just some spots where it bugs out, sometimes this can also be useful, but duration timers on the sections would be neat again

[polar spade]

Honestly, I listen to all samples as sometimes the 2 minute sample is better than what I have anyway and so is better to use.

I can work with samples that "push" the rest of the song out a bit and may get a slight repetition of the song rather than a full overlap. I just go in a splice out the double up bit.

But , right, if it completely overlaps a few lines, it's very difficult [possible, but very difficult] to insert those lines back in and retain the same melodic integrity. So , whether the timemarks are glitchy, we just listen to and reject those samples at cost.

[polar spade]

Download wav and re-upload is editable to the limit that you can only upload 2 minutes worth. Keeping in mind that any vocals on the upload will render the output un-publishable, if thats any concern.

[neon oracle]

For more details: When I click the "Create" button after writing my lyrics and choosing the style for my song, it takes me to my library as usual. But then, instead of seeing two songs loading, nothing appears in the menu. I never lose any credits and I don’t get an error message — it just seems like the app glitches.
I’ve tried refreshing and restarting the app, but the issue keeps happening even after deleting and redowloading the app.

[daring lava]

Are you able to test how your account functions via the web?

[smoky crow]

Question: Why is nobody talking about the fact when you create instrumental music with v4.5 that after the break near the end the song becomes mono sounding ?

[daring lava]

Do you happen to have a link to an example that I can check out?

[smoky crow]

let me get it

[polar spade]

I see David is responding. Ive rescinded my response. He's got it. 😄 carry on.

[sleek stump]

HAve you provided feedback to Suno about it? None of us here can do anything about the issues you are having with 4.5...

[smoky crow]

https://suno.com/s/O4IQrw2zzOTg1MfV

this happens to me the whole time and i can reproduce it within pro eq 4., its like the sides are turned down

[daring lava]

Thank you!

[sleek stump]

4.5 is still in a Beta state, so getting all the things you find to them through their official feedback form is important if you want them to at least be considered.

[smoky crow]

This is how i can reproduce it manualy with pro eq 4. When i turn down the stereo placement -> sides downwards in the high frequency range. I push it to mono then

[daring lava]

First off, dope beat!

I 100% agree with Ginger! It helps tremendously to share whatever findings you come across where things aren't working as you'd like them to in this form #📝┃user-research-feedback message

I hear that shift you are describing. Curious about your workflow with this. Was this a one shot that you shared with no editing?

[split gale]

I can second the Mono issue, I just noticed it, it went from stereo down to muffled mono then back to stereo over the course of maybe 30 seconds

In my latest song

[daring lava]

Could you explain your workflow for that song. Interested to know if this is a response to editing functions or straight gens. If you have a link, I'll begin compiling examples

[split gale]

Also to confirm, the mono issue arose during a cover

[pallid condor]

I just had luck remastering my 4.5 as a 4... was damn near close the the original... then I could edit it just fine again.

[smoky crow]

Thanks. Yes, this was a one-shot without any editing. Also when I try to remaster it with v4.5, the overall sound ends up the same — that character spreads across the entire arrangement. If I try editing just the final part by cutting and extending it, the opposite happens: the sides get boosted, especially in the high frequencies. Also i wanted to use that part of the forum but i am not allowed to post there. I have more bugs i can reproduce. I have 30 years of experience in music design and i produce and master dance music as wel.. that's why i notice these things

[split gale]

https://suno.com/s/DuQ4JRBJ3eNIVvpO <--- I think it originates bang on 4:00, you'll notice the muffled tone which appears

And by the time it hits the final chorus, it goes back to stereo

V4.5 is starting to act ALOT like V4 used to

[polar spade]

Great to hear. Yes, one recent "trick" is if 4.5 Remasters (or originals) sound off, do a 4.0 remaster of the 4.5 Remaster to kind-of get the best of both models.

[daring lava]

Thanks for flagging,( if you want to) please fill out the suggested form with as much detail about what you are bumping into about this and anything else where you see improvement can be made. Making a note of this and other shared examples and will pass it forward

[daring lava]

Thank you for sharing this!

[split gale]

And adding to this convo, the awful CLICK which appears at the start of a remaster, which points towards DC Offset

Yeah I can remove it myself, but meh

[daring lava]

Thanks for sharing what you are bumping into, all feedback is welcomed and encouraged! If you have time, please consider sharing these pain points and others you experience in the above mentioned form 🙂

[smoky crow]

Ok will do that. I also submitted myself as a tester as since i can really address issues which i can replicate and as a professional musician with 30 years of experience with producing and mastering. I hope can be added to this team. Suno is so close to perfection and i stumble everytime on the same issues. I even programmed an ai bot to help and guide me into this. Even spend more then 2000 credits just trying to find the bugs or workarounds.

[daring lava]

Oh wow 🤯 glad to hear you've reached out about that! Admission for testers is on a rolling basis, but sounds like you'd be a great fit! In the mean time, if you see/hear anything that could be improved, never hesitate to reach out here, the bugs/suggestion channels or that form. 🙏

[smoky crow]

Will do.

Here is the same issue playing :

https://suno.com/s/UbDdYjLtYaCNSUHj

[livid trout]

lol, I wanted to make an "i like to unwind from time to time" type of song with my "big room EDM" persona, and she tends to always want to go back to that same adlib: i wonder how you prevent the system from doing that 🙂 around 1 minute

Lyrics are far from completed but it was an initial thing, the focus here is how personas revert to the same adlib motif
https://suno.com/s/3RhQp69DyAmh77SL

also , i've made a rock ballad song, and had to change a word in the lyrics, and it pops between the replaced segment. I know the easy solution is to remaster, but this version has a specific sound to it that I love... is there a chance that the replace system can be improved to avoid pops/crackles ?

[neon oracle]

I tested it online and it looks like the issue was due to an artist's name in the lyrics. Too bad the error message doesn't show up on the mobile app

[neon oracle]

.

[daring lava]

avoiding adlib re: try extending from the point before the adlib, including the necessary guidance in the Lyrics/SoM fields. Furthermore, you could create a persona from the cropped section that you'd like it to mimic as well.

replace improvement: definitely expecting there to be a lot of growth and further development with this for sure 💪 Anything you bump into that doesn't align with your interests please feel free to share in the suggestions/bugs channels as you see fit 🙂

[daring lava]

Great job figuring that out! Noting that you didn't receive an err message

[stray lynx]

One nice thing in 4.5 is that it doesn't seem to have the midi guitar and tin can drum problems that they introduced into v4 in April.

I haven't tried yet, but if you generate a v.4 song where the prompt properly suppresses reverb and muddiness but still has midis and cans, coes 4.5 Remaster fix the midis and cans and turn them into actual instruments?

[neat nexus]

Has anyone got any tips or tricks they use to create short melodies? I mean short, no more than twenty seconds, beginning and end.

[stray lynx]

I was able to do this pretty reliably in v4 by prompting enough emphasising I want short loops. specifying a duration, saying it's a loop, etc

4.5 doesn't seem to be as responsive to this

[neat nexus]

That's still a great hint for me. Thanks! I drop t o v4

[stray lynx]

short, loop, [my mood], [my genre], [elements], [desired], 30s, [stuff]

It rarely caps at 30 seconds but usually it's between 20 and 50, and always either loops or feels complete

[daring lava]

Adding to what's been suggested... I've had some pretty interesting smaller responses to commercial jingle, tv theme song + lighthanded lyrical field usage

https://suno.com/s/kO623hoTrQiD6HRK

[neat nexus]

I'll try working with that...

Basically I'm aiming for those short instrumental themes you hear in the movies. Kind of a CODA of the main theme. They're usually really short and inspiring instrumentals.

...

Those helped. I was stuck with usual length and these dropped me down to about 45 seconds which is headed the right way.

[stray lynx]

It might be too short but you might also prompt for things like stingers. I sometimes get refunded when I do that since the stingers can end up under 5 seconds from time to time. But that might not be what you need either

[raw wraith]

Im having a issue, subo is not letting me login in the desktop version of suno on mobile, I even pressed the desktop button, but it leads me to the mobile app, I need to access The desktop app so I can generate videos and download my music

[neat nexus]

This got pretty close: [Removed links]

[neat nexus]

I'll give that a try too. I'm going to need some background music and it needs to be short for what I'm doing with it.

(I have only three minutes to work with so everything needs to be compressed for that.)

[stray lynx]

you could probably crop the last 3 secodns off and fade down

[neat nexus]

I probably will with a DAW.

[pale palm]

Hello. So... since yesterday (or at least that's when I first noticed this), I've been getting these two errors (both at once) every time I try to use the "replace" feature. And it happens no matter what song I do it with.

I'm on PC, and I primarily use Chrome. I tried to use Firefox to see if maybe switching browsers changed anything, but the same thing still happened, regardless.

[torn spoke]

@ashen basalt ait so

[pale palm]

Actually, wait, I think I'm in the wrong channel. I should probably go to feedback/bugs, shouldn't I? Sorry about that.

[torn spoke]

could you be hyper specific with exactly what you wanna accomplish cause I'm on my phone and it's a hastle to browse

[torn spoke]

tried a private window?

[pale palm]

I tried that, too. Didn't seem to change anything.

[torn spoke]

credits?

[pale palm]

I've got more than enough credits.

[torn spoke]

icic

[daring lava]

When you share this in the #🪲┃bugs-web channel, try to be as specific as possible with the workflow of this error (if you want to). Would like to try to replicate it on my end! Thanks for flagging this! Also, be careful to not share anything sensitive to your account please. Want to make sure you're privacy and account are safe.

[torn spoke]

And it's only replace? None of the other editing features?

[pale palm]

I actually haven't tried any other editing features during this time to see if it's "edit" in general doing it, but I can go ahead and test that out right now.

[sharp moth]

Are they offering a tryout for 4.5v at all idk where else to ask this question

[torn spoke]

for the love of every mod/admin in existence, please change your name

[polar spade]

I didn't think to review for that as you didn't get an error at the time. Typically, if the auto-moderator limits it from generating due to some content , it would give a general error. Then we would check title, prompts, and lyrics for anything suspect and run a couple of gens with as much generic text possible or none at all as a control.

But simply "snubbing" the output due to content issues can happen as well.

Glad it worked out.

[ashen basalt]

it is just that I had a song I missed a misspelling on and I really liked the sound and how the singer sounded in it. when I tried to do a cover of it with the persona of the song it made good covers but it didn't sound exactly the same as the original with the sound and all. and the singer sang differently as the original. lower tone or higher tone or completly different how the original sang. would be nice to just recreate the song with correct spelling without the song changing at all.

[torn spoke]

And the replace thing isn't working for you?

[torn spoke]

^

[pale palm]

Okay, went back and I can confirm that crop, remove, and fade-out all work as usual. It's only replace that's giving me the errors.

[ashen basalt]

I managed to generate a replacement that sounded almost the same as the original but replacement is kinda a hit or miss. I generated replacement for around 2-3 hours before I got the perfect one. I just fellt that it was alot of work to fix one misspelling. I was only thinking if there were an esier way to fix a simple misspelling if you hade managed to generate a song that you are really happy and pleased with it would have been nice.

[torn spoke]

Yeah it is a lot of work, and 2-3 hrs isn't that bad when you're looking for an exact output. If you want to be more efficient do it in larger batches

I bet you'll be more accurate with your lyrics from now on huh?

[ashen basalt]

thats what I did. I did large batches for 2-3 hours before the perfect one droped.

[polar spade]

A quick means is by starting with a instrumental that has a button ending, remove everything from 17 seconds to about 5 seconds from the end. Then run a Remaster to clean it up. https://suno.com/s/7lOr83sFHSN2RBsM

To run a few varieties of incidentals of the same motif , you can Cover before doing the above , or sometimes better, Replace the first 25 seconds in the edit UI so you get a varied intro, replace the end as well for a varied ending, Remove the middle and voilá.

[thorny copper]

Hello everyone, how are you?

I've been using the 4.5 model and I've noticed that it's more intentional when it comes to what we want as output. One thing I tested was to be more directional about the chords I'd like to be played and followed, as well as the tone and tempo, but I noticed that I didn't get good results. Does anyone have any prompts to share that can better direct the model to follow these rules?

[gilded shoal]

burnt my monthly credits in 1 day trying to get the perfect tweak on 4.5v

[polar spade]

@neat nexus These samples is to then Cover the incidental as a "jingle", as @daring lava mentioned. I did nothing but Cover it.

https://suno.com/s/8zBRSZ7kBHHqsUmo was a full 22 sec.
https://suno.com/s/v6yfK82rHdhBy9kS This was a refunded 7 seconds [Free!!! ]

Playlist: https://suno.com/playlist/206a22a2-f7a8-463e-8836-1e63fed6f1ea

[daring lava]

Genius move Dasvon

[polar spade]

Sorry. I had wrong links. Replaced.

[sleek stump]

Currently no. 4.5 is still in Beta so they may decide to do that when its fully finished, but we just dont know.

[polar spade]

A starting key or chord progression has never been a thing as far as I've known. Either generative AI is generative or it's going to be a DAW. So chord progression control would simply be a legal liability, no different than control how the vocal will sound.

Tone and tempo is relative and general to how you're expressing them. Using words like adagio, presto, grave, lento, largo, etc. Suno has a concept of those. Certain adjectives such as dark, brooding, slow, bouncy, happy and so on are going to corespond with changes to the tempo as well as to the overall tone.

If you're including lyrics, then the lyrics themselves , what they are saying, expressing, will alter the tone a bit at least.

[lunar nexus]

The quality of melody generation has been very low in recent days. Voice generation also raises questions - melisms sound unnatural. After generating about 20 tracks, about 15 were immediately deleted!

The 90% remaster from version 4 to version 4.5 is disappointing.

[polar spade]

It's so far generally considered "iffy" it seems to be an inbreeding issue. 😄 (tongue in cheek).

[polar spade]

Are the melismas better in 4.0 or 3.5? I simple don't care often for hearing runs as I like more a casual vocal anyway , so they always sound a bit unnatural with anyone I've gotten. I will usually try to use replace to find a smoother sample. So I agree with you.

I enjoy lounge, big band, broadway and musicals, so its that non-verbrato style that i go for often. SUNO is getting better I think, but if any hint of pop, then it goes into Aguilera mode. Hate it.

[lunar nexus]

very emotional review, I read it with pleasure!

[thorny copper]

Thanks for the clarification, Dasvon. I didn't know that Suno still can't follow some technical music standards, but rather something that's in the environment around it. However, when we talk about Generative AI, I understand that there's the creativity of the model itself, but Promtp has a big responsibility in directing it (examples of models like Midjourney for the photo do a lot of good). But I understand that Suno is constantly evolving and I confess that I was very happy with 4.5. It was really cool, but as a producer I really hope that it can follow technical rules like tone, tempo and chords.

[tawdry scarab]

it seems simple but from the ai's perspective you are asking for it to unbake a cake, adjust the recipe only in one piece of batter, and bake it with complete consistency

[tawdry scarab]

i find that presenting the model with a thorough music theory description of the tune helps it make better tunes regardless of adherence to your sequencing

[ashen basalt]

true was only sharing my thoughts tho. don't know anything about AI. I can se the point.

[shadow plaza]

Why 4,5 model vocals sounds so quiet compared to V4 ?

[tawdry scarab]

4.5 often gives you precisely the complexity you ask at the cost of lower rms areas that allow you to turn it up and preserve the dynamic range

which is arguably superior

4 was about bright ear candy that is more universal

but can only contain so much fine articulation

in this way 4.5 is like 4+3.5

4.5 also seems to include more era/style noise

[polar spade]

Though it's certainly possible for Generative AI to follow prompts for chord progressions , tone and tempo and the like, there has to be limiters manually inbedded to keep an AI from recognizing these prompts. Just as Midjourney, for instance, has limiters on producing nudity, even though it can.

I'm all for it, but sad I can not create my own style , as you say, which would include, tone, tempo, key, and chord progressions (manually prompted). And believe me, if that were true, then AI would no longer be considered "generative" and we could copyright those songs immediately as they would be 90% directly controllable and pass the "human hands-on" test.

I dont think it's a "why cant AI do that..."

Of course, thats all my opinion.

[tawdry scarab]

if you upload a chord progression and rhythm in a short sample you will get many extends basically following your sonic proposal so to speak

and i'm not sure but i think you can specify the scale

i can't tell but i put in myxolydian (sp?) and to my dumb ears it gets the emotion of that scale across

[granite mountain]

why so slow

[daring lava]

S/O to @polar spade & @tawdry scarab for sharing their knowledge/expertise here with members! Thank you for all that you do!

[tawdry scarab]

i would, because nothing is at stake, go so far as to say scale, narrative, and emotion comprise a more complete definition of a song than a progression

[polar spade]

I think it is set to reveal the nuances of most things, even if also limited to emulating specific details. So . If I say Dark Blue, and even if it is set to ignore creating Dark Blue things, it will give things in that range -- sky blue, purple, teal, blue polka dots, a blue 🐋 and play the blues and rotate those relatives among the others prompted for.

I'm with you to say, though it would simply be fun to copy my handwritten song completely, and knowing I could then copyright it as well, it's equally cool to come at songs in another direction avoid of the conventional needs.

Certainly, give me every tool available and then some. I won't refuse them, so having those options are always welcomed. For us that pass a poly, criminal record test, psych exam, sign an indemnity agreement and pay $200 a month, we should have those tools as part of the Elite level subscription plan. 😂

[tawdry scarab]

the last thing we ever need is making music reserved for people with more money than soul

not that that you

just saying uber-tier has social implications

[smoky crow]

I’ve reported all the bugs and now I can only hope for a fix, because in its current state, Suno is no longer usable for me. The strange part is that when I try to fix the issue by using the Extend function from the break, the newly generated part introduces the opposite problem: the highs are lifted around 4800 Hz and the stereo field becomes noticeably brighter. So I can’t even correct the original issue.

There is clearly something going wrong with Suno’s audio quality right now. Even when remastering a v4.5 track that initially sounded good, the remaster comes out sounding dull and almost mono, with a significantly narrowed stereo field across the entire frequency range.

Here is the same track with the extend part from around 03:11. You can now hear its more brighter then before that point.

https://suno.com/s/OmImVlwXXh9a4f5q

[tawdry scarab]

sometimes extend and replace create slightly different sound, when getting the whole song this is partly smoothed but remains if the change is severe

it's up to the user to select wisely a replacement or extend that matches

[smoky crow]

This is not a feature but a bug

[polar spade]

Understood. No worries. Right on ✊️ brother.

[tawdry scarab]

define bug in the context of generative ai please

to be frank what i'm asking is, how can it be a bug when there is no procedure to ensure it is a deterministic system

it functions by not being deterministic

[smoky crow]

I understand that generative AI can introduce slight variations when extending or replacing parts of a track, but what I’ve encountered goes beyond creative differences. This is not a subtle shift, it's a recurring, structural audio issue that impacts both the original generation and the extended parts in opposite ways, which is not expected or user-controllable:

The original track flattens in stereo width and dulls the highs towards the end.

When I try to fix this by replacing or extending that section, the new segment becomes unnaturally wide and bright, especially around 4800 Hz, clearly altering the sound balance.
This is not a creative variation. It’s an unpredictable loss of sonic consistency, making it impossible to finish or correct a track reliably.
Even remastering introduces the same dulling and narrowing, further proving the issue lies within the model’s current handling of stereo imaging and frequency detail.
In generative music tools, it’s critical that different parts of the workflow (generate, extend, replace, remaster) work in harmony, not against each other. Otherwise, the user cannot make meaningful or professional use of the tool, especially someone with decades of production experience who can immediately identify these inconsistencies.

So yes, this qualifies as a bug, because it creates repeatable, unintended degradation of the output, not an artistic variation.

[tawdry scarab]

you are invited to use a world class mind bending system that is not programmed but trained, it doesn't do what you want it to. okay? it doesnt give you studio quality? none of what you say constitutes a bug, but rather incongruity with your expectations, though bug reporting for these things is how they get taught to the model or how guardrails get made. you're right to talk about it

but your attitude is myopic

also worth noting, the fact that it does these things when you try to use it does not mean there is any line of code anywhere ever causing that to happen that can be fixed

you may just be digging around in a poor quality region of the model output

[smoky crow]

Thank you for your feedback, I appreciate your perspective on how the AI system operates. I understand that generative AI functions based on patterns it has learned, rather than being strictly programmed with fixed rules. This distinction is important, and I agree that it leads to a more experimental and less predictable output than traditional production tools.

That said, the issue I’m encountering isn't just a matter of creative variation. The specific and repeatable degradation in sound quality that I’ve outlined such as the narrowing of stereo width and dulling of highs, followed by exaggerated width and brightness when trying to correct it goes beyond normal variation. It’s a structural issue that disrupts the integrity of the track’s balance, making it difficult to use the tool reliably for professional-level production.

While I understand that the AI may not always behave in a way that aligns with studio-quality expectations, my concern lies with the tool’s inability to maintain consistency across its own workflow. This creates a situation where feedback loops (like replacing or extending parts of the track) only exacerbate the problem rather than improving the outcome.

I believe this is more than just an incongruity between user expectations and AI limitations. It represents an area of systematic behavior within the model that could be improved. This kind of issue not only hinders creative control but also makes it challenging to create something that’s musically consistent.

While I do recognize the AI’s potential for unique experimentation, I still see value in working towards greater stability and coherence within the system, especially for those looking to integrate it into professional workflows. I’m sharing this feedback not as a criticism of the AI's creative capabilities, but as a step towards making it more reliable and usable in varied contexts.

[tawdry scarab]

i stopped reading when you said it was experimental but somehow your issue is different

this is a punishing amount of saying the same thing over and over

that your issue is special and not related to ai and learning

[smoky crow]

I can repeat the mistakes

[tawdry scarab]

that's vital information to report

and you are encouraged to do so

[smoky crow]

That's what i did

[tawdry scarab]

have you found it to be true in all other user content

and what about is the breadth of your prompts that cause the issue

[sleek stump]

I think a distiction needs to be understood... Suno is NOT a professional level production tool, it's no more than a Tech Toy at this juncture. Can it be used for production, probable, is is capable,my opinion is not at this time. Will it ever? Thats still to be seen, but Suno itself is not billing itself as a professional tool, not in the slightest.

[tawdry scarab]

you can tell because your downloads are 48khz@16 bit

also happy birthday!

to ginger

[sleek stump]

Thanks Sang! ❤️

[polar spade]

In my opinion, We have to think beyond quality sometimes to realize that "adequate", which covers more demographic area, can often sustain the strongest momentum.

[tawdry scarab]

the best art rarely is without significant degridation

"best" "art"

[smoky crow]

Doesn't matter if it's for professional or not, the output and errors are there and repeatable. That's why we are now at version V4.5 and not at v1.

[tawdry scarab]

ok guy

you have fun

[smoky crow]

So because i address bugs in the system you starting to act like you have been stepped on your toes. omg..

[tawdry scarab]

no, not at all

[polar spade]

I always like pointing out the song "Somebody That I used To know" by Goytre. A brilliant song and worldwide streaming sensation , yet the production sounds like it was made on $20 reel to reel tape recorder. 😅 i want to get one of those! (The song, not the recorder).

[tawdry scarab]

i just think you insisting it's worth drama only you imbue it with causes an emotional strain on others

@smoky crow you wouldn't happen to have made spectrographs or fft charts showing the issue across different generations sampled at random from other users

or with a clear diversity of prompting

that would be some next level application of effort

[smoky crow]

I understand your perspective, but I want to clarify that I'm not trying to stop or create drama. The issue I’m raising is important, and it’s something I believe needs to be addressed for the tool to be truly useful in a professional setting. It’s not about emotional strain. It’s about making sure the system delivers consistent and reliable results. I’m bringing this up because I genuinely believe it can be improved, and I’m committed to ensuring that feedback like this helps refine the AI's functionality.

[tawdry scarab]

if extend points were noted in the player we could do some sick automated analysis

can you let me in on your process to discovery so i can examine it directly or is it like an egregious spike

like any casual analysis of extend points would find it happening

[stray tree]

what are you guys talking about?

[tawdry scarab]

they are getting noise and stereo separations correlated with extend points

sorry

loss of separation

[stray tree]

woww

[smoky crow]

Never mind. I am already in contact with the devs. They have acknowledged the situation. Finaly

[tawdry scarab]

sweet

sorry to waste your time

[polar spade]

I can neither boast of a decent IQ or Expertise in music so I'm going to go get some noodles 🍜 . All great debates folks, I get a good perspective from everyone. Nice to have some civil chats about things like this. They are important to all of us involved.

[tawdry scarab]

well this should be able to be visualized

with a couple hours in python maybe even able to be automated

by checking the derivative of the stereo separation and fft profile over time

[polar spade]

Ooooo......#2) Country Christmas Songs for Family Holiday Movie

Needs Christmas songs (public domain or original) for an upcoming movie! Upbeat, driving country or country leaning styles: Country rock, bluegrass, folk, Americana, acoustic pop, etc. Mid- to up-tempo songs are needed most. Male or female vocals are accepted.

Oooo.... now here is where I shine!! I got about 6 of those already in the pipe.

Yes yes.... im off to get my noodles. TTFN.

[smoky crow]

A tip for people who work often in the extend modus. If you want to zoom in and out with the mouse wheel you need to place your cursor on the time locator and a hand icon appears on it and then hold CTRL while scrolling your mouse up and down and you can zoom in and out on that specific position without continuesly moving the slider below.

[daring lava]

Thank you for flagging this!

[keen iris]

I am writing to request assistance regarding an ongoing issue with my Suno account.

Since my paid subscription expired on April 16th, I have been unable to access any free functionalities of the platform. Additionally, I am no longer able to rate other users’ works.

Whenever I attempt to generate music I consistently receive the same system error message:

"An error occurred. A system error occurred."

This issue persists across different devices, including on my phone, where I receive a similar error:

“Create failed due to an internal error. Please wait a few seconds and try again.”

I have carefully followed all available troubleshooting instructions: I have cleared my cache and cookies, tried other browsers, experimented with different prompts, and waited in case the problem was temporary - but unfortunately, nothing has resolved the issue.

I have also attempted to reach out to your support team via email but have not received any response. I sought help in the Discord community as well, but no one seems to have a definitive answer. This problem has now persisted for nearly a month, and I am quite confused and concerned about the situation.

As far as I know, I have not violated any terms of use of your platform. Therefore, I would like to kindly ask:

Has my account been banned or restricted in any way?

If so, is there a way to appeal or resolve the situation?

If this issue is related to policy regarding expired subscriptions, could you please clarify what the limitations are?

I am a great admirer of Suno’s capabilities and am fully willing to renew my subscription if necessary - I just wish to understand what is going on with my account and how I might be able to restore normal access.

Also attaching screenshots with the errors I get

Thank you in advance for your time and support.
I look forward to your response.

[half dragon]

I know you said you have tried everything... and it didn't help... let's see if it even logs anything in the console log on the browser...
View the dev console on browser and then perform the action on the app, then see when the error occurred... anything could be a hint printed on the console related to the error? We might not be able to solve it even if we know the error, but at least it'd give a peace of mind of what's going on.

[daring lava]

Hey there! Thanks for reaching out about this! I see you mentioned that you have reached out to support! Would like to note that if you have sent in multiple emails, it usually takes longer to get a response and often traps you in a loop of waiting while support has more requests to process and check through to ensure there's no crossover between agents and end users, etc. Would you mind verifying when you sent in your initial/last support ticket?

[bronze pecan]

Is there some prompt I can use to force Suno to use more treble at 4.5? :/

Because 4.5 has no treble (at metal)

[keen iris]

It was on 23 of april at 17:05 (in my time zone)

[keen iris]

It was on 23 of april at 17:05 (in my time zone)

[keen iris]

It shows 403 (Forbidden)

[daring lava]

Thank you for the heads up and for your patience with this! When you get a chance, please DM @steady kelp with a brief message about what is going on and the email associated with your account/support ticket 🙂

[half dragon]

You're on free plan right? sounds like it was accessing something it wasn't supposed to.

[bronze pecan]

With people who feel their output is muffled, I’d also point toward Cover first to try to get a bigger variant in output (and hopefully something more interesting to your ears) than the gentle rebuild that a Remaster typically provides.

@steady kelp its not a per song basis issue. Every single (metal) song has very very low treble at 4.5 (v4 was nice treble though). Its probably one of your shimmer fix attempts. But the 4.5 model doesnt need it. you got the shimmer reduced enough that you can run the treble at regular levels. It sounds absolutely terrible to listen to metal like being under water :/

Personas sound different/not working as expected
There was a similar feeling when going from v3.5 to v4, and some feedback around v4.5 has been that previous personas don’t work as well. In my personal opinion, I’d want to have separate personas for each model. Since each model handles styles differently and has a unique sonic profile, it would also make sense that Personas would sound different across each. I’d recommend making some great music in v4.5 then creating new personas from those songs! Also, keep the older personas in case you create in v4, etc.

You are aware its bad when you build an AI Band though and are creating an album currently?
I mean if its not possible to apply them for techical reasons, well ok unfortunate. But otherwise it would be a bit of shame :/
I mean the 4.5 model is very welcomed that it finally can do change persona at extends. thats a godsend if you have 2 singers. Also Creativity is much better, the shimmer is much less, so continue using v4 is not ideal either :/
Maybe you can extend support for personas at v4 model too though??

[keen iris]

I already did DM Mike some time ago, I know you guys hard workers, is it ok to disturb him again?

[daring lava]

For sure, it's never a disturbance. You're right the team stays super busy, but never so busy that you should ever hesitate to reach out!

[keen iris]

Well it show a bit more than just 403 (Forbidden), here is a whole thing that I see in console

[keen iris]

Will do, thank you

[arctic marsh]

I am having an issue, not sure if anyone can help.
When I am on my iPhone or ipad, logged into the Suno APP or Website (on safari) I can see my entire library (100+ songs via paid membership).
However, when I log into suno.com on a desktop browser (Chrome and Edge tested), my library shows completely empty.
I have cleared cache, cookies, tested incognito mode, and confirmed that I am logging in with the same email and credentials. The library only appears on mobile, not desktop.

[daring lava]

Very welcome!

[smoky crow]

Sure np. The only fix i can do is to extend the part which was monotized and then filter that part down with some surgical eqing. Turning the highs down are easier then to add but you have to cross your fingers for how the ai outputs the extension. I also noticed and reported this to the devs as welll that a remaster from a created v4.5 version downgrades also the bitrate.

[daring lava]

Glad you have a work around and thank you for submitting what you are bumping into through that form! That will help get team eyes on what you are seeing!

[tawdry scarab]

what kind of prompting have you tried so far to bring out high end detail

[bronze pecan]

nothing. I'm open for tips. But honestly I would be more glad if the model could do on its own proper treble

[tawdry scarab]

try using treble in the style

https://suno.com/song/44bf6d90-43a6-4631-95d3-c86bc91139a2

[smoky crow]

I'm currently experimenting with a new technique: rendering in reverse using the extended option. I start by extending from the break, find a creative sound extension, note the timecode, merge the files, and then extend the track again. This time choosing a spot just before the previous extension. It gives some surprisingly interesting results while maintaining clarity.

[daring lava]

That's so cool!!

[thorny copper]

Good points. So you think it's unlikely that Suno will add these control fields for copyright reasons? I understand that copyright is more for lyrics in conjunction with melody, which could infer possible plagiarism. But tone, chords and tempo are a set of universal variables and could hardly violate copyright, since they are open variables. That said, for example, if you want to take an existing lyric from another author and paste it into Suno, it will allow you to create the song, right?

Well, I believe that having this would be a giant leap for us customers, because we will go from an AI that generates freely to something more tailored to what we really want to create. We'll see in the next chapters of the product update... I know that there are other tools on the market that already allow some adjustments like these, but when I looked, it's not possible to use them commercially and maybe that's why you brought it up.

[bronze pecan]

wow what a prompting road. Are you musician?^^

[tawdry scarab]

i've just been working to get what i want and figured the ai knew music theory

haven't been disappointed

i have studied music more than i've made songs from scratch anyone wants to listen to

but suno takes care of the "what would other people do here" aspect

[bronze pecan]

having this knowledge like you would certainly help a lot

[tawdry scarab]

what i do now is mostly writing lyrics and then negotiating with gpt or negotiation with gpt then lyrics to fill

so i am not pulling things out of memory, but music theory is more of a reference than something anybody needs to memorize

you tell gpt you want a scale and rhythm grid for random thoughtful phrase x it will instantly tell you the scale and rhythm

[bronze pecan]

interestingly on this song the treble is not bad either

[tawdry scarab]

higher quality prompts seen but i cannot prove to bring out higher than baseline quality

[harsh furnace]

Did Suno break Persona's voices since yesterday?
None of my personas sound right since today.... not happy 😦
regardless of V4 or V4.5

[tawdry scarab]

regardless of adherence

[bronze pecan]

how did you get the treble better on your song though?

[tawdry scarab]

i assume the model can recognize high quality training captioning even if it did not develop key or progression specifics

by what i'm telling you

there is literally no other step

make better prompts get better prizes

study what you want to have

then you can ask for it thirty times by names from every angle

and you aren't sitting in the pool of low quality training captions

https://suno.com/song/d0977be0-60a3-4c0e-b1ea-3a36711e0ead

4.5 feels different in idle but it's an entirely different animal with specificity

as for chord progressions if you're using one of the 60 or so crazy common ones just prompt for their emotional qualities as part of any reference of what they are

gpt can help with that

[bronze pecan]

What is also a bit weird and also happening on your song: It gets louder over time compare beginning with 9th minute

[tawdry scarab]

it does crescendos in sections for powerful rock

if you look at the whole wave form it's just introducing journeys

[bronze pecan]

the volume goes louder though, I'm not comparing calm vs heavy section, heavy vs heavy section is later louder

[tawdry scarab]

yes

that's what a crescendo is

across sections of the song that have any peak it will rise to a peak and then drop and do it again, having sometimes the drop portion rising too

this embues the song with an emotional journey unavailable with flat dynamics

the kind of variance you talked about is like pianissimo vs fortissimo but a crescendo is a steady elongated traversal of intensity up

not that it's safe to use pianissimo

because piano

negative promoting sludge can also help metal

sludge, sludge metal, death metal

and melodic metal in the style can brighten it up more

i wanna say power metal increases the likelihood of major chords which can feel like brightening

[daring lava]

@magic oak Hey there, much easier to keep up with chat here

[magic oak]

hi

i mean it may be me using it wrong. when the new editor came in i didn't extend until now so

[daring lava]

Could you share a link to the song you are attempting to extend and let me know what point in the song you are trying to extend from

If you'd rather DM me to retain privacy of your song we can do that

[magic oak]

lemme find it. but it was going to be from the last few seconds of it

[magic oak]

sure

[daring lava]

That might be what's going on then, if you are attempting to extend when the model is already seeing a song on it's way out, it may continue to just complete (end) the song

[magic oak]

how far back do i need to go

[tawdry scarab]

for change it's best to extend from silence and push through free extended taking them until the model can make the switch from a lack of any longer context, avoiding getting whole song

if you want there to be continuation it's good to be right in the pocket of vamping, in the middle of the sound you want

any closing cadence tends to cause an ending

[magic oak]

i mean in earlier models you could just add lyrics and then it would extend whatever

[tawdry scarab]

looking up closing cadence is invaluable to understand when it's a dead zone for extending

[daring lava]

Perfectly said

[fervent oriole]

Would anyone be able to help me get a refund initiated after accidently doing yearly?

[daring lava]

Hey there! Thanks for reaching out! I think I caught up with what you had brought up in the other channel! Glad to hear that you've reached out to support. Just checking to make sure you reached out to the right team member, can you verify that you reached out to [e-mail]?

Also, saw you mention that you reached out 3 times, that may be adding to the delay. Each time you email support it kind of resets the cycle so to speak, trapping you in a loop of waiting while support has more requests to process and check through to ensure there's no crossover between agents and end users, etc.

[fervent oriole]

It was just [e-mail] they replied that they would get someone on their team to reach out. That was over a week ago.

The last that I emailed was may 3rd

[daring lava]

Okay thanks for the heads up! When you get a chance could you reach out to [e-mail] to make sure this gets directed to the right team member? In that email please include a brief message explaining what's going on, the email associated with the account and any billing statements tied to the erronous purchase. That will expedite the process best

[fervent oriole]

Alright. I'll give that a try. I appreciate the help.

[daring lava]

Also, thank you for your patience with this! Looking forward to getting this resolved

[steady kelp]

DM'd!

[arctic marsh]

Hey David, I don't mean to interrupt your convo, but I ask a question above here about an issue im having. Do you think you may be able to point me in the right direction?

[daring lava]

I can try my best

[arctic marsh]

Okay.. here is what I typed above:
I am having an issue, not sure if anyone can help.
When I am on my iPhone or ipad, logged into the Suno APP or Website (on safari) I can see my entire library (100+ songs via paid membership).
However, when I log into suno.com on a desktop browser (Chrome and Edge tested), my library shows completely empty.
I have cleared cache, cookies, tested incognito mode, and confirmed that I am logging in with the same email and credentials. The library only appears on mobile, not desktop.

[steady kelp]

hmm. i havent really experienced low treble in my heavier music, but i think as a general feeling, i'd rather have a very flat starting point and use a DAW for some EQing, rather than generating something more harsh that is actually more complicated to tame or correct 🤷‍♂️

[magic oak]

extends are still not working on this song. the red text is still showing and i've gone before the outro

[daring lava]

That is odd indeed. Are you sure the mobile account isn't logged in with a phone number? Accounts (both mobile and web) should be tied uniformly.

[bronze pecan]

I can show you soooo many songs. like any metal song

for some reason sangderenard manages to make it sound good, but would be nice if people can have good treble again who did not study music xD

[magic oak]

i think maybe it could be me at this point. would you mind showing me how to use it properly

[chrome junco]

How can I get in touch with the Suno billing team? I sent them an email, but I haven't received a response."

[arctic marsh]

I signed up using the google login.

[tawdry scarab]

this is precisely the high value in 4.5

[bronze pecan]

all my and any others songs have terrible low treble and its frustrating

[arctic marsh]

I signed up with Suno with the.. login w. google. so thats the only way I have ever logged in... I've never had a password... or anything...

[tawdry scarab]

i wonder if suno uses a system prompt or a null prompt to start

[arctic marsh]

DAVID I used a different email when I signed up for PRO>

but it was IN / Under the Gmail account.

[bronze pecan]

I'm really tired of having this low treble in my songs @steady kelp It sounds like being under water :/

v4 didnt have this problem

[tawdry scarab]

woops

[steady kelp]

can you send me this song? (the suno link)

[tawdry scarab]

if people could get accounts without their original email whoever phished me from this server (i assume) would also have the opportunity to trash my suno

[arctic marsh]

I think I said it incorrectly. When I signed up for pro... (I used paypal... which had a different email address)... but the credits went through... so no problem there..

[daring lava]

Okay thanks for clarifying! That explains it. By using two different emails you have inadvertently set up two different accounts. I'm guessing the other account (non-pro) is a free plan account?

[steady kelp]

probably a good reminder that humans can roughly hear 20-20khz, but mostly in ways where certain frequencies interact with nearby things. a sharp cut at 20khz wont do much for most people, but there's plenty of space at 10k where this will just kill a lot of the ultra sharp stuff. v4 was pretty hyped in areas, but this looks maybe a bit steeper than i would have anticipated, even for 4.5

[arctic marsh]

but the credits showed up and were used in the Google login account... and I have done it 2 or 3 months.. (i cancel as soon as I sign up so it doesnt autorenew). it was only after the update to 4.5 that I couldn't see the library anymore on desketp.

[tawdry scarab]

@steady kelp @bronze pecan is it anything like this: https://suno.com/song/be3295bc-955d-4cce-b714-cc1d8625fabe

or harder

[daring lava]

Let's get this figured out! To get you to the best form of support, I would suggest you reach out to an official support team member about this with a brief explanation about what's going on and the emails associated with both accounts. Reaching out to [e-mail] would be the way to go!

[tawdry scarab]

this would be an era/style appropriate frequency response is why i ask

to me

[arctic marsh]

excellent.. I will do that today. Thank you so much for taking time to help me with this. It is appreciated !!

[daring lava]

Oh you're very welcome! Thanks for reaching out!

[tawdry scarab]

i guess i should've just asked to see the song too

[bronze pecan]

yes, give me a moment 😄

but honestly its almost any metal song on suno :/

https://suno.com/song/e60fc2dc-fceb-44bb-84e4-8ee425b7c698

example 1

the suno muffled, under water, edit is fixed

I mean where is the treble at suno's variant? And its almost all of the 4.5 songs

and I only put the "almost" in here because somehow the music studied sangderenard somehow can bypass the issue with his super prompting. But I would love if normal people can have better treble too again

[shadow timber]

So it looks like it’s no longer possible to make a persona from a song made using a persona? Why not? That enabled a lot of tweaking and precision which is gone now.

[bronze pecan]

example 2:
https://suno.com/song/ce94c979-6b61-411e-b0bf-fc3973527506

vs the fixed version:

[steady kelp]

going to sound harsh, but i'm going to say overcorrected

i hear what you're saying- have you tried remaster at all? sometimes those variants are just sort of EQ shifts that you may find beneficial

[bronze pecan]

go for example 2. massive difference too

example 2 is even bigger difference imo

[steady kelp]

sure, agreed. widening, etc, too

[bronze pecan]

All I did was adding treble back 😄

because suno has none

and dont you find example 2 sounds A LOT better than suno???

[steady kelp]

all you did was add treble? i'd like to see that

im using laptop speakers, its different, for sure. its not necessarily that suno doesnt have treble though, which is why i think theres more to it

[bronze pecan]

this was for one of the songs. not sure if it was these examples, some need more, some need less, but they have all in common - way too less treble

[polar spade]

Glad you'll reach out to support Quick question so you might add when talking to them.

You say you have used the account which that one shows no outputs in your Library, if understand correctly. When you generate an output in that account, do they appear to vanish after that point?

[bronze pecan]

hmm why it didnt embed no idea

[steady kelp]

i can see it, no problem

[bronze pecan]

but thats all i did, nothing more. for real

but dont you agree that especially the song by nightgirl sounds a lot better at example 2 my edit version compared to suno??

suno is so muffled 😦

[steady kelp]

bluntly, it still sounds overcorrected to me. its not bad, and i'd bet that most would agree that it feels more alive (or better) because of the brightness. that said, more harsh exaggeration will give you faster and longer ear fatigue as well.

[bronze pecan]

the edit sounds fantastic to me

maybe its your laptop speakers idk

the author nightgirl loved my version too WAY MORE

[steady kelp]

yes, it sounds good. but i mean if you listen to things hyped that way for 30 minutes, your ears will get tired faster than if you didn't have that kind of hype

this isn't an argument in any way, im super glad you got a mix you enjoy

[bronze pecan]

but when I listen to non-ai music, its like this too and not like suno it does currently

[steady kelp]

im not saying the suno mix is better, im saying its less fatiguing, but there's a few reasons for that

[bronze pecan]

my assumption is you did it for shimmer fix attempts. But your 4.5 model has so low shimmer that it can keep regular treble imo

[steady kelp]

not exactly something carried over from v4 - as i mentioned earlier, v4 had some overhype too. just the sound of that particular model

[bronze pecan]

I really cant enjoy it like that. I always enable an EQ before listening to Suno and put external files into my lyrics as recommendation, because the non-existent treble is frustrating me 😦

[steady kelp]

4.5 is generally pretty flat, i made the joke in the office that it sounded like i was listening on NS10s, which are historically flat studio monitors - not a bad thing, but definitely different

[bronze pecan]

When I listen to non-ai music the treble is the same there like with my edit versions of suno. Its not normal that its so underwater like this

[steady kelp]

do you mind if i ask your age?

[bronze pecan]

38, but everyone agrees on my example songs

also young people

[steady kelp]

we're close in age- it wasn't a "your ears are old" haha

the type of thing you're hearing is... relatively new, but by that, i mean 25 years. there's been a huge shift in the production style in more recent decades, which also plays into your feelings here re: EQ and "good" quality

[bronze pecan]

and also did you read regarding personas the edit?
I assume it has techical reason that they dont apply properly to a new model? Its just a bit unfortunate if you did build up an AI Band and make currently an album and then the voices change - but on the same hand I absolutely did need the possibility to use 2 personas in one song which extend finally makes it possible, and the creativity of 4.5 is amazing and the removed/much less shimmer is great too.
so staying at v4 isnt ideal either. Its at least somewhat similar though and I at least like Lucy's new 4.5 voice. but still weird when its suddenly different on the same album lol 😄

btw: would it be possible to add the extend + persona combo to earlier models as well or not?

[tawdry scarab]

4.5 highlights a generational gap between 4 style brightness and most of the history of music

[unique stone]

I agree about 4.5 being flat, at least compared to v4 which went through a phase of having too much high (I saw users saying it hurt their ears), and a later phase where it sounded a lot duller to me. 4.5 seems well-balanced IMO. Probably would be nice to have the option to adjust EQ or do other postprocessing within Suno eventually though.

[bronze pecan]

a phase of having too much high (I saw users saying it hurt their ears)

the problem there was not the treble, it actually was a noise artifact by the ai which you gladly could fix in 4.5

there was a weird high frequency noise in v4's treble when the shimmer appeared

v4 also had often this whistle noise, which you could fix with 4.5 too

and I'm glad you could fix it, this whistle noise was not that nice to hear over time

[steady kelp]

without diving too deep, i'll just say this: if you hear something and it sounds way brighter than something else, chances are the bright stuff will cause some level of ear fatigue, or even potential hearing damage if listening loudly or over long periods of time. source: me, audio engineer 😅

i totally understand the feeling of perceived liveliness in songs with a wider range of frequency bumps, but its probably not very healthy to experience regularly, even if it sounds "better" in the moment.

[bronze pecan]

but this is so exaggerated low treble, you find it nowhere that low on regular non-ai metal music either

[tawdry scarab]

brightess can be hard on equipment to a microscopic tiny degree

without filtering

[steady kelp]

i do have a feeling this will shift a bit as we tune the model, but it likely wont be to the level of v4, and probably somewhere in between

[slender oracle]

what causes suno to chop the last minute or so of a song off?

[steady kelp]

the dropoff is a few different technical things- stopping shimmer is not likely part of the equation. i just dont want to get your hopes up and say "it will sound like v4 minus shimmer next week" because its unlikely to happen

[sleek stump]

At least it will be an improvement

[steady kelp]

too much context for time constraint, typically in the older models

[slender oracle]

this is on 4.5

[steady kelp]

link would be helpful for me to diagnose a little more clearly!

[slender oracle]

https://suno.com/s/FgOmJTlnLAUgJtji
the 2nd one it generated alongside it seems ok

[dense nexus]

I am working on cloning my voice using the Suno audio file upload feature. The only problem that I am running into is that it will not accept any of the recorded (WAV) files that I try to upload. It says that each of them contain copyrighted lyrics. The first 2 I tried were, in fact, copyrighted material, but the last 2 were nursery rhymes, "Twinkle, Twinkle, Little Star" and the "A, B, C Song". Both of these are well known public domain lyrics. Anyone know how to get around this issue?

[tawdry scarab]

vinyl didn't protect extremes in the medium, it depended on eq after the fact, but it has an enormous latitude of available resolution for that eq boosting

[sleek stump]

Generally, you cant. The filters are going to catch what they catch, and they are also catching stuff thats in the PD. Until Suno retunes their filters, you cant get past them.

[dense nexus]

That is most unfortunate. Does corporate monitor this channel?

[sleek stump]

I would suggest singing something , anything, a page frpom the phone book ot something.

[polar spade]

You cant use previously published material, public domain or otherwise. I wondered about the ABC song, but now you've mentioned it.

[dense nexus]

Good suggestion....might even become a HIT! 🤪

[steady kelp]

😎

[tawdry scarab]

while they are public domain songs the performance of them is protected content

[sleek stump]

No, there is no official support in this Discord...

[dense nexus]

What is the optimal audio file upload track length that will insure that Suno has enough vocal material to produce a quality result?

[steady kelp]

here's the scoop on public domain:

Suno's terms (including the pop up when you first use upload audio) say that you should wholly own the material you upload, or have exclusive rights to use the material.

Public domain songs are non-exclusive by nature, so it would be ruled out based on not falling in line with the terms as stated

[sleek stump]

Although you may attract some attention from Mike from time to time

[bronze pecan]

i do have a feeling this will shift a bit as we tune the model, but it likely wont be to the level of v4, and probably somewhere in between

Yeah better than nothing, maybe its even fine then, we will see

[sleek stump]

Keep in mind that Persona are not vocal presets, the model will wander..

[steady kelp]

so a holiday song or traditional song may sneak through the passes for content checks, but in reality, they still fall outside the boundaries that ensure you're only using material you own. the reason we want you to have ownership or exclusive rights to use is because you're then transforming the material, potentially without permission, rather than having expressed exclusive rights. by using the function, you agree to being the rightsholder, which could in turn land you in some hot water

just safer for all parties involved not to get caught in any weird rights issues

[dense nexus]

I totally understand the policy. Keep in mind that I am only using the material to capture my personal, individual vocal characteristics for the purpose of cloning my voice to be used in original material I create...not for cover songs or copyrighted material.

[sleek stump]

Distributors that own copyrights will not hesitate to sue anyone.

[steady kelp]

totally fair. so you could melodically sing something like "my name is __ and i made this song to clearly demonstrate my voice"?

but again, we're not much of a voice clone tool, at least not yet. other tools may be more proficient with that at the moment

[dense nexus]

OK, so I simply sang some material from a medical journal. Here's what Suno told me after uploading that....We've detected vocals in this audio, so you won't be able to create public songs or Personas with it. This audio will be link only.

Obviously it has vocals in it. So, now what? I can't use my personal voice to create a song that can be published in the public arena?

[steady kelp]

also correct. there are some ways to make personas from later iterations, but again, not allowing a direct copy of that voice is an intentional choice by Suno, because we dont know that you're uploading your voice. you could be uploading some other persons voice and using theirs (famous or not), so we play it safe and whenever we detect voices, it just cant be public

reiterating- we're not a voice cloning tool

[dense nexus]

In the past, they allowed it pending my agreement that it was, in fact, my voice and not someone elses.

So, does anyone know if Suno has voice cloning on their radar as a future feature?

[tawdry scarab]

it's not an attractive avenue

but I'm not suno

I dont speak for them

[half dragon]

Guessing not anytime soon from Mike B msg above.

[tawdry scarab]

we are driving on a mountain road and on one side is legal oblivion

it is not presently time to let the kid steer from your lap

[slender oracle]

suno take the wheel

[dense nexus]

For any true vocalist/artist, having the option to use your own voice in producing a finished track is honestly the desired result. While it is fun to hear the lyrics and musical tracks you've created performed by a paid musician, it remains virtually impossible to get any traction with this music on radio or in the professional music world...even though the quality and uniqueness of the music passes muster.

[steady kelp]

is a clone of your voice the real representation you want on the radio as yourself?

this isnt a jab, btw, just sort of a technical question. i make music in a lot of different mediums, so im open to interpretation here

[dense nexus]

Isn't the power of the AI platform the whole point of this discussion. Isn't it supposed to take an audio representation (tones, diction, feeling, range, vocal frequencies, etc) and juxtapose it across the entire lyrics? So, for all intense and purposes, it is you.

[steady kelp]

to me, the power of Suno is giving people the magic I get as a songwriter in this ultra fast, ultra responsive way. I write songs pretty quickly, but in no world did I expect to iterate this fast, nor did I expect I could give even bit of that feeling i get as a writer to someone else.

there are lots of great tools for vocal capture specifically, and I think that's something we'll have a stronger feeling on in the future, but right now its more of capturing a whole experience of the music making journey. in the case of personas specifically, thats about capturing the essence of a song style, so you can make whole bodies of work that are cohesive

[dense nexus]

Not to worry, didn't take your comment as a jab....appreciate the openness. I am a professional vocalist. Over 50 years of performance experience. I am currently performing (I retired from the IT world in 2014) in Independent, Assisted Living and Memory Care facilities throughout the St. Louis, MO area. At the end of 2024, I had completed 800 shows. While I am having fun with this new technology, I am also seeing it as a way to create and publish music that before was not affordable.

So far, I have produced and released 9 albums and over 50 singles to all the major music streaming platforms. My concentration is contemporary Christian, Gospel and Smooth Jazz.

[steady kelp]

love this!! thanks for sharing your use case- i do think we'll get to a place that will be helpful for you, but there's a little bit of tooling to work through before its feasible to be in that space

[dense nexus]

Appreciate the responses. I look forward to the progression and improvements to this platform and the AI technology in general. One thing that I've discovered that helps me manage all the aspects of songwriting and creating the prompts, lyrics, images, etc, is RUBI.AI. It is the only comprehensive, fully integrated AI platform available. I invite you to visit my website to listen to what I've been doing with my music: https://www.scancotech.com/songmaster

[tawdry scarab]

so, when making metal, if the voices seem too quiet or muddy, do an experiment and turn the volume up and see if the vocals ever sound perfectly natural, and ask yourself if they aren't balanced well as a proper listening volume, since it is the case that any mastering will have a most natural listener volume, where it feels balanced to the track's nature

[small bridge]

does suno still not understand if in the tag part for styles you put like a specific word like a name or something? It said before when it first started months ago that it wouldn't recognize if you put in like a specific song or artist. Is that very much still the case?

[sleek stump]

Ya you cant name names, it will get moderated pretty handily

[tawdry scarab]

do you mean a specific word or specifically you're trying to use an artist or song?

[sleek stump]

What you CAN do is get a chat ai to help you get a description of the artists style and use that.

[tawdry scarab]

60s pop rock, psychedelic, baroque pop, verse-chorus, AABA, sitar, Mellotron, vocal harmony, swing groove, tape tricks, modal shifts

Prince
funk rock, synthpop, Linn drums, slap bass, falsetto, dry mix, syncopation, chromatic hits, analog synths, breakdowns, extended chords

Radiohead
alt rock, IDM, ambient, polymeter, glitch, falsetto, drone, jazz voicing, Max/MSP, post-rock, experimental

Daft Punk
French house, vocoder, TR-909, filter sweep, electro funk, sample loops, sidechain pump, analog synths, disco feel, build-drop phrasing

Nirvana
grunge, loud-quiet-loud, power chords, detuned guitars, rasp vocals, raw drums, angst lyrics, garage mix, simple structure, 90s distortion```

for instance

[small bridge]

I wanted to input the style specific to one type of song and I feel my gens when I have put in the styles used for that song are still fairly off.

[small bridge]

i will try this again

thanks guys

[livid anvil]

Why is it so impossible to reach out to the Suno team about billing problems? I've literally tried everything 😑

[tawdry scarab]

did you try emailing billing

[sleek stump]

/billing

[flint haven]

SUNO BILLING HELP

For billing/payment issues, email: [e-mail]

Include:

• Your Suno @handle
• Description of the issue
• Transaction ID(s) or payment receipt(s)

Before mailing:

• Check if your card was charged
• Verify funds were available
• If Stripe rejected your card, try using the iOS or Android app to pay via Apple/Google Pay.

Do not send non-financial issues to this address — it may delay your response.

[sleek stump]

Billing usually is 1-3 days, normal support is 2-3 weeks

[prime prairie]

How in the ever loving HELL do I get certain Voices for certain parts? I'm getting really effing pissed off here!

[tawdry scarab]

the short version is you ask for them and the long version is you carefully strip away and clarify and focus your prompt on the most immediate next section and use extend to only pick the parts that follow the instructions you made, and you confer with others on how clear and actionable your instructions are and whether you have sufficiently provided enough clues in your tagging, styles, and affectations or stereotypes in writing lyrics

this worked in 4

#1365192916687126650 message

[livid anvil]

I've tried emailing both, I've tried sending messages here, I've tried sending messages on social medias. I've been waiting for a week. 😔

[tawdry scarab]

I've done all I can

I gave you an animated react

[livid anvil]

Thanks lol, I appreciate 😁

[sleek stump]

If you keep sending emails before you get a response, you get pushed to the back of the queue every time.. Billing is 1-3 days, support is 2-3 weeks. If you emailed billing with a billing issue, just wait for them to contact you.

[tawdry scarab]

I asked gpt to write a song about how it sees our relationship (in the platonic sense) and it's spooky and emotionally hurts and I don't like it I need a refund

[dire rose]

I'm curious

When will suno finally try and incorporate reasoning into their AI model

In the sense of, just like GPT did with the 4o image generator

Incorporating reasoning in the AI model allows for a much better control on nuance and context, which would catapult Suno to a whole new dimension

[sleek stump]

It may get there sometime in the future, AI tech is massively evolving at a very fast pace.

[dire rose]

I mean obviously it is, but now the possibility is real and should be heavily considered by the staff team

In my opinion the one thing that Suno lacks in is control over the songs.

It's absolutely great at creating music, but it doesn't give the user a ton of control over how to generate that music.

Except if you're a technical mastermind who uses 50 advanced techniques to try and bend Suno to your will. And even then, the results are very often unpredictable and luck-based.

Same thing with COT agents...

When will Suno incorporate a chain of thought to their process? It could create an insane improvement to how the model perceives the prompts and music

Silently waiting for some updates...

[bronze pecan]

hmm cool I got extra credits, but who was that nice to me? 🙂

[tawdry scarab]

@dire rose the reasoning in gpt tends to involve a lot of guided training from huge teams of data annotators as well as the continual recursive escalation of resource demands by practicing recursive tasks or multi step tasks, but at each step of the process the model can manipulate and understand the content being passed around.

In this case you're talking about expanding into parts a system which cannot presently meaningfully read back in the content with the meaning it was embued with because they are different languages

Suno clearly has some theoretical capacity to be broken into many tasks with many models, many subunits

But cannot implement compartmentalized sub-models directly off the back of their existing workflow because it is not text to text

[dire rose]

because it is test to music?

[tawdry scarab]

or music to music

yes

[dire rose]

test*

text* bruh

okay that does make sense

[tawdry scarab]

so it's possible, it's probable someone will do it

but it is going to be a feat, a monumental achievement

[dire rose]

but what about modifying the workflow to work on a text-to-text reasoning of "How the music should be" before transforming text to music?

[tawdry scarab]

whereas gpt just is a snake eating itself until it's the right answer guided by thousands of people working from home

[dire rose]

or even better, are there not any ways of training some kind of model to "recognize text from music"

[tawdry scarab]

which isn't to say, you know, suno probably has data annotation contracts too

but gpt is insane

[dire rose]

such as converting music into a readable format (hz for example or something)

[tawdry scarab]

that is kind of what 4.5 is now

it can do some contextual understanding to guide the generation

[dire rose]

so basically gpt is heavily trained nonstop whereas suno just doesn't have that capability yet?

[tawdry scarab]

the crucial point is that suno can't eat its own produce

in the same way

[dire rose]

okay that makes sense

[tawdry scarab]

so it can't just be chained up in a loop

[dire rose]

but what about focusing on some kind of break down of a user's prompt using a cot agent, in order to structure it in a specific but more context-aware format

and then use perhaps a model trained to specifically generate music based on an input like that

is that realistic at all or no, i am still pretty much a beginner in the domain of ai

[tawdry scarab]

again, 4.5 is what you're asking for

it performs reinterpretation

to some extent in the conceptualizing process

I'm wording this very poorly

[dire rose]

ur name is french right

[tawdry scarab]

yes, though I don't speak it

[dire rose]

ah damn, im french, it wouldve been easier lol

[dire rose]

give me some concrete, vulgarizing example of what happens when you enter a prompt in suno 4.5

"ideally"

like, say I ask suno 4.5 some stuff with an actual context, what does it attempt to do?

to the extent you know atl east

[tawdry scarab]

Suno, en revanche, repose sur un processus de diffusion. Il part d’un bruit aléatoire et affine progressivement une représentation audio, sans séquence causale explicite. La génération est stochastique et non-linéaire, ce qui signifie que les sorties ne peuvent pas simplement être reprises comme entrées. Contrairement à GPT, les résultats de Suno sont le fruit d’un effondrement complexe de probabilité dans un espace latent, ce qui rend leur réutilisation directe impraticable sans transformation préalable.```

@dire rose I am too far from that as a user to even approach a response

[dire rose]

did u generate that using ai

[tawdry scarab]

also my understanding is that suno is kind of both of those types, but it is the diffusion part that makes it so impossible

yes is it terrible

[dire rose]

okay that actually makes much sense

i never really questionned how suno was working but that makes sense

[tawdry scarab]

it is the two forms of magic of ai kind of, one anticipates and one reveals

[dire rose]

so it's kinda just increasing the probabilities of a specific timing being the correct note

but because it does that process on the entire thing at once, there's no injecting it back in the input

thatd make nos ense

[tawdry scarab]

I don't know, they could have all kinds of scaffolding up for music theory

[dire rose]

that is interesting icl

[tawdry scarab]

the things to know I imagine they wouldn't say

[dire rose]

yeah obv XD

[tawdry scarab]

but yes, your prompt in all models is just adjusting the probability

but according to what the words mean to the model

which is hopelessly abstracted

just wildly chaotic in a mathematical sense

[dire rose]

that feels kind of crazy

what about a model that would take a sequence of hz as a input and attempt to predict the next hz out of it?

is that theorically possible?

it goes very far from what suno is doing I understood, but i'm just asking that question outside of suno's context

[tawdry scarab]

theoretically you could do anything, but anything new and different will have unknown optimizations that could take many years to work out to find they are poor in comparison

[dire rose]

yeah that's understandable

i wish i could try lowkey

[tawdry scarab]

to, say, filtering it through diffusion and llm

[dire rose]

the idea doesnt sound bad tbh

[dire rose]

mhm mhm

[tawdry scarab]

if you use python, you have nvidia, you learn pytorch coding

[dire rose]

oh?

[tawdry scarab]

you could make a concept model of a NN that will try to make something from what it learns of your own music

[dire rose]

oh right python is the language used

[tawdry scarab]

and it won't make you songs probably

but you could see how it lifted patterns

[dire rose]

yeah thats my issue

true

very true

[tawdry scarab]

I found making NN in python gave me a lot more insight into AI and clarity on what people said about AI

[dire rose]

what did you make?

what was the purpose of the nn?

[tawdry scarab]

it would reproduce images through ascii by cutting up a set of bytemasks from a ttf font and comparing them to reference images to learn generally how to make an image

[dire rose]

the only thing ik about ai is that we have virtually no knowledge on it, it's too recent

people who give life hacks abt AI

eh

mostly bs imo

you can't even imagine the mathematical complexity about AI, it's as good as attempting to predict how a human will behave, sometimes it works sometimes it doesnst

imo

🤷‍♂️

[tawdry scarab]

that's all kinda fair, the large spooky effects we have no idea about, and there are various ways I anticipate they're going to have deep cultural impacts

well

[dire rose]

that's insane

did it give interesting results?

[tawdry scarab]

I personally could almost articulate most of the math of what an AI is, but NN I could straight up tell you how they work

it did

[dire rose]

nn i know how they work

[tawdry scarab]

let me see if I have any example

[dire rose]

i watched too many videos about neuronal networks

[tawdry scarab]

do you ever watch three blue one brown or something

[dire rose]

the most interesting ones being a series of videos made by a french (💪) youtuber, who made a neuronal network on Trackmania driving

[tawdry scarab]

oh man I think I know that guy

he is awesome

[dire rose]

yeah

it doesnt go TOO far into the nn itself but it showcases so well the effects of it

amazing stuff

[dire rose]

rarely

i used to watch him for maths and cuz his videos are made so well

[tawdry scarab]

they are one of my fave

totally

[dire rose]

but i never watched anything abt him for nn

[tawdry scarab]

I don't remember if he did any I just wanted to see if we saw any of the same stuff

[dire rose]

ohh XD

yeah a bit

i'm not that much of a math guy but his vulgarization is always top notch

ive always been going around saying that maths is half visual+examples, before diving into anything more concrete

sob

ok lowkey tho i might try that "predict the next hz" thing

[tawdry scarab]

I have to step away for a moment and I forgot what I was going to say/do

[dire rose]

LMAO

all good bro dw

i'm just coding and chilling on my side, i got the informations i wanted

ty for the chat :)

[blazing atlas]

hey anyone know where i can go to get help with my account? paid for pro but didnt get the credits.

[half dragon]

Contact [e-mail]

/billing

[flint haven]

SUNO BILLING HELP

For billing/payment issues, email: [e-mail]

Include:

• Your Suno @handle
• Description of the issue
• Transaction ID(s) or payment receipt(s)

Before mailing:

• Check if your card was charged
• Verify funds were available
• If Stripe rejected your card, try using the iOS or Android app to pay via Apple/Google Pay.

Do not send non-financial issues to this address — it may delay your response.

[blazing atlas]

thank you, I did do that so i guess just wait?

[half dragon]

Oh yeah... it might take sometimes, how long ago was it that you sent?
I hear it usually takes 10-14 days for support, but billing it might take sooner.

[blazing atlas]

7 days was my first one

i guess 8 days ago actually

[half dragon]

Give it a few more days... I guess.

[blazing atlas]

thanks that does make me feel a bit better if i know its typical. i was feeling ignored which is very frustrating.

i was able to create a song based on this issue https://suno.com/s/IpUkCTbX3I8JNrt0

[polar spade]

Now that it's been a week, did you check your bank /credit statement to see if had recentky gotten credited back for some reason?

.... and are you getting 50 credits a day AND can used 4.5 and all tools that go with Pro privilege?

[blazing atlas]

I was charged. I did check my bank statment. I did not get any credits today and cannot use 4.5 or the other pro tools.

however i did get 50 credits per day until today as far as i can tell. i have made several songs, including the song i posted, but it was using the 3.0 model

[blazing atlas]

correction - Actually i did get 50 credits today. and I have to use 3.5 model. no pro features and didnt get the pro credits

[delicate bloom]

hi yall

how's everyone

[blazing atlas]

I cant speak for anyone else here, but I am doing fine. I found some credits to do a song i been thinking about.

[fair tapir]

Hey all. Anyone else having this problem? I'm subscribed but I don't seem to have the option to make a persona out of my upload. Please advise! Thank you

[sleek stump]

Weird, its not available

Are you on Mobile? Are you trying to create a persona from an uploaded audio with vocals in it?

Suno doesnt allow Persona from uploaded Vocal audio

[near flume]

If there are vocals in the upload you can't make a Persona from it. You also can't make a Persona from a track that has a Persona in it. When you upload audio with a vocal you should see a popup stating the limitations.

[upper heron]

how should i be wording my style descriptions to trigger beat switches

[daring lava]

Hey there!

Here's an article from help.suno that gives a good introductory example: https://help.suno.com/en/articles/5782977 Additionally, browsing through other people's songs has been incredibly helpful for me personally. You can learn a lot by looking through what other people do when they create!

Hope this helps! If you have any questions please don't hesitate to ask!

[torpid thicket]

So the vocals on 4.5 sound a bit better; But the instruments for Rock/Metal songs still sound muffled as hell...

[daring lava]

There was an update that Mike recently shared here that touches on the muffled-ness you're speaking of here: #🚀┃updates message
Also, if you have time and would like to, the team would love to hear your thoughts (pain points/areas of improvement) in this form shared here: #📝┃user-research-feedback message

[fair tapir]

thanks and @near flume as well! I'm also looking for suggestions on the best way to cover uploaded audio that's longer than two minutes with some continuity. Thanks again!

[polar spade]

As you said you sent an email 8 days ago regarding this, then it being 8 days is a bit long to have fund post in Suno account. Typical for many countries outside of the USA is 3-4 day (international banking system transfers, reviewing transactions for fraud for first time purchases, etc, is all typical bank activities causing delays).

Seems you've not receive any receipt from STRIPE/suno, it's been 8 days, your accounts don't show any pending or returns and youre still in Basic mode as if Suno hasn't received funds.

Certainly, check your profile info for this account to verify your email that you think was assigned when you made payment. If any chance you have another email which was set up on another Suno account and the email happened to be attached to your payment method (maybe you chose a different payment, such as Paypal) then it may have activated another account? Check all accounts if you happen to have more than one, or easier, just check any other emails for a receipt from Suno or Stripe.

No need to respond here unless you found the issue. Just make sure they know all of these details when you speak with support. Good luck then, mate.

[daring lava]

Hey @polar spade 👋 Truly appreciate all that you do here! Always looking for opportunity to help people! Big 🧠 + Big ❤️

[tawdry scarab]

i assume it's known the create menu has 3000 char limit when the site is in profile mode on a phone but it's 5000 in landscape mode

or is that reportable

[daring lava]

Good find! Remember that being brought up, but wouldn't suggest you hold back from flaaging it! Squeaky wheel gets the grease and all that jazz

[sage rapids]

4.5 is a wonder

sorry tottally out of context but I am hearing a song and Im like wow it finally follows the prompt Its complex and minimal and liminal and syntheic 80's and freaking glorious

like wow.

As a real musician I finally feel like I can use this as an instrument.

[tawdry scarab]

real musician™️©️

[sage rapids]

I was makig music long before AI

My ai music making skills are not with instruments they are with words, so I woudnt call myself an ai musician....more like an ai writer

music writer

I can decode music and label it words that an LLM can understand..i can put words to feelings and sounds but I am not playing an instrument, I am writing a sound

[tawdry scarab]

i didn't mean to call into question anything

i just was riffing

[sage rapids]

right on...i didnt mean to sound superior. I just mean i play instruments

[tawdry scarab]

i write notes sometimes lyrics and prompts too often but i can barely play the clarinet, like at a child's proficiency

[sage rapids]

hahaha just keep playing...play to your favoite suno songs

I like finding on piano songs suno made...i try to find them so i can play them...Im not like a concert pianist..im more drummer than piano but its fun.

[tawdry scarab]

on suno i can pretend to have a contrabass or lower clarinet, also in a daw, but i can't make dubstep on a tenor or alto clarinet whichever is "normal"

[vale arrow]

Why are personas speaking gibberish

[sage rapids]

maybe thats not your true instrument. Sometimes you might like the sound but if it doesn't flow out of you, that's not it...play with other instruments. The ones that come easy --thats your instrument.

[sage rapids]

i cant even get them to work Says I dont own them and that the person that owns them doesnt allow my use Im like OKAY THEN NVM lol

[little ginkgo]

Rooted in Afro House and melodic deep house, the track glides along a steady pulse, driven by a warm, rolling sub-bass and a crisp four-on-the-floor kick, Shimmering synth stabs and syncopated hi-hats dance atop congas and rim clicks that nod to West African rhythms, while spacious pads and lush chords create a sunlit, open-air vibe, Soul-tinged guitar licks and jazz-inflected chords weave through the mix, adding depth and warmth, A raspy vocal hook floats above, treated with gentle reverb and delay to enhance its call-and-response feel, The result balances hypnotic groove and emotive resonance, perfect for both festival main stages and underground dancefloors

[tawdry scarab]

that's a cool prompt

[prime prairie]

Yeah. Nothing is working for me. I have two men, one older, one a bit younger (not a child). I specify [Male Voice 1] and [Male Voice 2]. I even put [Older] for the first one. Every once in awhile I get a female voice in the second one, or just the same Male Voice all the way through.

[tawdry scarab]

by every once in a while you mean you're trying over and over and never once doing it by extends

or that's what you get each extend failing to find the right voice

[fair tapir]

Beautiful prompt there! i'm still wondering about this if anyone can help please.

[prime prairie]

Correct. Not extending because it's not giving me anywhere NEAR what I want.

Spending more credits as if I have them to just waste!

[stray lynx]

I've enlisted chatgpt to help me test the theory that v4.5 more or less only knows how to write one song.

I create a series of prompts that by all rights should break it out of its programming and emit something that doesn't fit its bias if there is any way at all to break that bias. Nothing actually does. A few come close, but they tend to all do the same thing by the end.

• standard progression through to verse 1
• some kind of exciting or "up" chorus compared to the verse
• a rising energy through verse 2
• a tension break by the bridge into something either uplifting or sweet
• a victorious verse or chorus afterwards

It's kind of hilarious how it fights so hard to write this same song against all efforts to prompt against it.

style
A brutal fusion of Scandinavian black metal and industrial darkwave EDM. Guttural, dry-throated female vocals whisper and snarl over mechanical blastbeats and glitching synth bass. Frozen tremolo-picked guitars layer with distorted analog leads. The tempo is unstable and twitching, like a corrupted clock. The mix breathes with cold air, yet pulses with synthetic dread. Every rise collapses into a mechanical breakdown. There is no euphoria, only the endless surge of corrupted data and the hiss of static.

exclude
Wide Stereo Arena Space, Bright Mix, Full Spectrum Pop Balance, Warm Organic Room Acoustics, Symphonic Layers, Soaring Lead Vocals, Clean Belting, Operatic Delivery, Smooth Glide Vocal Lines, Pop AutoTune, Lush Chorus Hooks, Euphoric Build Ups, Cinematic Drama, Warm Bass Bloom, Dancefloor Rhythm, Soft Rock Metal Edges, Delicate Female Tones, Emotional Rise, Major Key Resolution, Ascending Tonal Release, Heroic Cadence

result - a variation of the above formula that mostly obeys the prompt except for a sweet and delicate little bridge breaking through like a ray of sunshine

[tawdry scarab]

you can't get one line right ever?

in isolation

also you've cracked the code on music with verses and choruses

[prime prairie]

It will start out with a Male Voice and either switch to female on the next one, (even though I specify [Male Voice 2]), or just simply use Male Voice 1 all the way through.

[tawdry scarab]

and so you took the first male voice that sounded right and extended from there correct

[prime prairie]

Did not use Extend.

[tawdry scarab]

you didn't try what i said works then

[prime prairie]

Ok. I'll try it.

[tawdry scarab]

[man] and [other man] works, everything on top of that to help specify, more or less, works, in the sense of pushing probabilities, if it isn't working, you have to reduce to step by step and clear out all the confounding information

this is true for most anything that is tricky

the more other content around the less likely it is to be noticed and followed over the noise of all else going on

so if you must do something tricky and you must do it right, genning a thousand times to get the one perfect round is madness

every time it does something right get whole song

and it will fold the entire previous context into the next gen

[prime prairie]

🖕🏼🖕🏼🖕🏼

I'm done with this!

[tawdry scarab]

did you just try to extend a thousand times without cutting it down to only the next section

or you couldn't get a second voice?

anyhow I'm not making people happy rn so I'm gonna step back

[prime prairie]

It's not you.

It's the stupid site!

I tried extending twice, and it keeps giving me a dayum female voice in Verse 2!

[torn spoke]

bro

twice

is nothing

thank you woodhouse

[prime prairie]

🖕🏼🖕🏼🖕🏼🖕🏼 @flint haven

[tawdry scarab]

it's not a robot taking your commands

it's ai

you have to learn to work with it

[prime prairie]

@torn spoke I have been working on this dayum song for hours!

[tawdry scarab]

maybe don't do a duet yet, I do crazy length complex songs and I don't bother with that

because it only works easily if it's simple

[prime prairie]

It was written as a duet. It wouldn't make any sense to have him talking to himself.

[torn spoke]

that's not a lot :/

the more specific you want something, you should expect the time required to increase exponentially

[prime prairie]

Some of us weren't born into money.

[torn spoke]

I guess I've just accepted that suno is an investment

and have the 10k credit subscripton

[prime prairie]

👏🏼 Good for being wealthy!

[torn spoke]

I work for the money lmfao

[foggy oar]

Just skip food for 1-3 days per month.. its good to fast and then you will afford it 🙂

[torn spoke]

HEH

[foggy oar]

Less excuse and more happy creating 😄

[torn spoke]

also @prime prairie
Did you make a persona with the part of the song that has a male voice? and then extend using that persona?

As in, crop the original song down to before the female voice shows up. Create a persona from that. Extend the cropped version with the persona.

[prime prairie]

I need a second male voice for verse 2.

[torn spoke]

Right, but if you don't crop the female out of the persona, I don't think you'll ever get a male voice for the second verse

[prime prairie]

So make a full song using just the first verse, save the persona, then extend it?

How will it know to add a new voice since in trying to do that anyway?

If I use the persona, won't it just use that voice for the full song when I extend?

[torn spoke]

in the lyrics, attempt to establish clearly using tags that the first verse is sung by "voice 1" or "male 1", etc.

In your cropped song, go to align lyrics in the "editor" UI, and put in your updated lyrics.

[prime prairie]

Yeah, that's what I'm doing. [Male Voice 1], and [Male Voice 2].

[torn spoke]

In the description of the song, you can also specify that there are two vocalists.

You can even specify that they alternate between verses

the good thing about 4.5 is how much you can specify things in the description

obviously it's not going to do exactly what you write, but it let's you try a bunch of different ways to influence it

So I HIGHLY recommended that you use a google doc or something to organise your songs and prompts

if you're going to change the prompts a lot, so you don't have to search through your songs for specific prompts

[prime prairie]

I will have to attempt this again later. Thank you. Both of you! Sorry for being a...$#!t.

[torn spoke]

Here is an example with google docs, the tab feature is really nice for this kind of thing

[torn spoke]

don't worry, suno can be frustrating ;)

trust me i know

[daring lava]

S/O to @torn spoke for being super cool and helping @prime prairie, and S/O to Dan for sticking with it! Definitely a learning curve to things for sure 😅 When you get a chance please check out this list of helpful articles (if you want to) #1294418290348921007 lots of good stuff there!

[prime prairie]

I didn't realize one had to have a Masters in Rocket Science to use Suno! 😂 I have been decently lucky enough to generate good songs and not spend a lot. But duos are a whole other level, it seems.

[torn spoke]

yeah, and sometimes you get duos when you don't even want them lol

[prime prairie]

Yes. I've had a few of those.

And one time I've had a song turn from studio to live near the end. And I never prompted that. But I can't get a live song now for the life of me.

[hearty moat]

Will we ever be able to download stems of all instruments and vocals?

like, bass, guitar, main vocal, back up vocals, drums etc - All the instruments that are making up the song?

[prime prairie]

As individual stems?

[hearty moat]

Right now its just vocal as a stem and then everything else in one stem..

[hearty moat]

yes, whatever A.I used to build my song, will we be able to create stems of everything?

[daring lava]

That would be so cool!

[hearty moat]

Then I could add them to my DAW and copy paste, double them, tweak them etc cut it up

[prime prairie]

That would be awesome! You should go put that in the ideas channel.

[daring lava]

#💡┃suggestions-web

[prime prairie]

Yeah, that one. 👆🏼😂

[daring lava]

Always feel free to share any ideas! The Suno team actively looks through those channels to better understand member interests and possible next steps!

[hearty moat]

Done

[prime prairie]

Well, thanks to@tawdry scarab, @torn spoke, and @daring lava for help and ideas and not telling me to stick it where the sun don't shine!

[tawdry scarab]

hey good luck making music that resonates with you

it sucks to see people frustrated

[prime prairie]

Thanks!

[daring lava]

Thank you but they did all the heavy lifting 😅 I just came in and saw cool people figuring things out!

[prime prairie]

You shared the link to a much-needed resource.

I appreciate it!

[torn spoke]

Anyone else having issues where the replace feature will completelly ignore the selection it is told to replace? As in it will over ride the lyrics coming after the selected section.

[torn spoke]

It'll respect the start timestamp, but not the end one.

It'll just keep going, and redo the entire rest of the song as well.

[plush rain]

Why am i unable to create songs with russian lyrics?

Thanks guys

[torn spoke]

Could you specify exactly where the process is going wrong?

[torn spoke]

First image showing the selection, and then the second going "bet" and doing the entire rest of the song

[gritty flax]

Any reason I can't seem to download a video from my song anymore?

[torn spoke]

hmmm, can you circumvent it by making a crop? Like remove 0.001 seconds of the song in the editor, and see if you can download the crop

[inland ivy]

Does anyone know why I can't play any music in my library on my new computer

[scenic sequoia]

so i have tried using — before the word, i have tried having a comma, i have tried with a period... is there another way i can get the ai to stop extending a word. it keeps holding a word in my chorus and it always sounds bad when it does it.. i have generated so many times and it refuses to listen. it has changed everything else about the chorus.... besides the 1 thing i want it to do lol.

anyone got any ideas

[formal cape]

You have to generate the video first now.

[verbal mortar]

is anyone else noticing how the vocals in 4.5 just sound off? like wobbly and hollow?

[crisp zinc]

does suno has an network error? cuz i can't add audio

[formal cape]

I have a song with Socrates in it, is there any way to make it sung as the name is pronounced, it keeps doing the Bill and Ted thing 🤷🏻‍♂️

[livid trout]

i wish added to replace, crop, cut etc, there would be a way to "add" a section. I sometimes find that when I initially make the song, with a distinct bridge and break, that the bridge and break are glued together and I can't replace them easily. So it would be nice to "add" a section, which you can then sanitize with new lyrics or instrumental properties with full adherence to the song instrumentarium, adding or substracting from it.

[inland ivy]

Any idea why the application lost all icons

[gritty flax]

I've been trying to download the video for a few days now

[pallid glacier]

Hi there! I’m looking for a custom credit package (enterprise plan or smth). I’ve emailed support a few times but haven’t heard back. Has anyone here managed to secure special terms or know the best way to reach the team? Thanks!

[inland ivy]

My browser issue was related to the below issue: Access to fetch at 'https://m-zephyrtone.prod.suno.com/agg-receiver-service/v1/events/m' from origin 'https://suno.com' has been blocked by CORS policy: No 'Access-Control-Allow-Origin' header is present on the requested resource.

[slender oracle]

i cant get this to work at all! it keeps generating nonesense babble worlds instead of the lyrics i specified for this replacment section

how do i get it to use the lyrics its told to and not replace them with non-english babble that sounds like singing

[torn spoke]

click on the cogwheel and choose manual lyrics

[torn spoke]

^

#💡┃suggestions-web message
@dapper quail
Tried using the replace feature?

[slender oracle]

yes thats what i was doing

[brittle jungle]

Terrible stems as well. Useless.

[sleek stump]

Are you a free user or a subscriber? Video downloading is a pro+ feature, but you get a limited number when you first make a Suno account (You get up to 10 v4 gens, 5 for Web and 5 for mobile and a few video downloads).

[ebon sand]

If the song has lyrics, align them first, even gibberish has to be aligned
Then use manual replacement, it's better than the automatic one

If the song doesn't have any lyrics yet and you get gibberish only from doing what you're doing, you could use the gibberish replacement first, then align that and replace over it a 2nd time

[scenic sequoia]

yeah, after you give a kidney worth of credits it finally gives you a solid one, its new tho, they will update it

[blazing atlas]

I did receive a receipt from STRIPE, when I paid 9 days ago. It was with the same card and bank I used to pay for my pro subscription every month for the last year.

[boreal harness]

I keep getting this when trying to get my stems. Any ideas?

[wintry sandal]

Bom dia, gostaria de saber, pq não consigo fazer o downloads?

[boreal harness]

Entendo. Não sou só eu.

[wintry sandal]

No meu não aprece o cmpo. Qo clica nos ... aparece a janela edutor de musicas.

Tenho ainda 2480 créditos

Quero uma solução, ou meu reembolso.

[boreal harness]

Parece ser um problema no servidor. Vamos esperar um pouco.

[wintry sandal]

Já faz uma semana..

Mas de uma semana nessa situação... cadê os responsáveis???

Alguém!@@ sabe???

Pelo jeito estou no lugar errado. Agora não aparece ninguém para responder.

[boreal harness]

É isso que vejo, mas não consigo baixar nada.

[mental plank]

Hello everyone None of the cards are being paid. Everything was fine for a year and now. What is it?

[wintry sandal]

Mas o download, não abre?

[boreal harness]

Não. Não há ponte entre o Firefox e o servidor DNS.

Zork, has it been exactly a year? You might need to update your card details. Could be a deadline was reached internally.

[unkempt grail]

I'm using the android app and I've just tried the v4.5 version but can't find a save to device button, can anyone help

[half dragon]

There isn't.
Downloading isn't available on the app yet.
Use the browser to download your song.

[gritty narwhal]

I have a 3.5 song (That I'm redoing) that I cannot turn into a persona (greyed out, telling me to subscribe even though I'm premier). I can create personas on any other song except this one. Any ideas why?

[half dragon]

Was it from an upload? that has vocal in it?

[gritty narwhal]

Song is so old that i don't rightly remember. Its entirely possible. Odd that it would be telling me to subscribe rather than the actual reason. Bug?

[half dragon]

Could very well be, since it's grayed out already and shows that error when clicked on it.

[gritty narwhal]

Yeah, even a remaster wont let me, so im going to assume that it was an upload and the warning is bugged.

[half dragon]

Wait... Remaster doesn't work on that song as well?
Strange. I've never heard of Remaster restriction. It's grayed out?

[gritty narwhal]

Sorry, no, A remaster works, but I cannot make a persona from the remaster, therefore, it was probably an upload

Before covers came out, to get consistent themes (working on a musical) I would download the song, crop it and leave the end silent and upload it so that Suno would read the song and give me a consistent melody with extensions. So the upload has no real voices, but the restriction is triggered anyway.

[shrewd willow]

Hey guys, is it still advisable to create personas using just a small part of a song or did they fix that?

[wintry sandal]

Como traduzir para o português?

[shrewd willow]

Traduzir o que?

[wintry sandal]

As msg aqui

[shrewd willow]

Acho que nao da, o discord nao tem essa funçao. Se voce tem alguma duvida eu posso tentar ajudar

[wintry sandal]

No meu suno não consigo fazer o download

[daring lava]

Olá! Infelizmente, essa não é uma opção disponível no servidor. Temos um canal dedicado aos membros que preferem falar em português aqui: https://discord.com/channels/1069381916492562582/1154779408989687848 or https://discord.com/channels/1069381916492562582/1230630513299947581

Se tiver alguma dúvida sobre as ferramentas do Suno, não hesite em me contatar!

[wintry sandal]

Tenho uns trabalhos para entregar, mas está sem a opção download

[shadow plaza]

Why v4. 5 sound so suppressed in Metal style?

[wintry sandal]

4,5

[shadow plaza]

How can fix this

[shrewd willow]

@wintry sandal Eu nunca tive esse problema, sempre consegui baixar. Talvez vc tenha feito upload de algum audio que o Suno esta identificando como direitos autorais de terceiros?

[wintry sandal]

@shadow plaza

[daring lava]

Hey there! The recent update shared here #🚀┃updates speaks a bit on utilizing the Cover function "to get a bigger variant in output (and hopefully something more interesting to your ears)". Adjusting the style of music could possibly help with this when covering! Wishing you the best of luck working it out! You got this!

[wintry sandal]

Mas, aí iria aparecer nas notificações ne?

[shadow plaza]

But i dont want to do covers but brand new songs

[shrewd willow]

Nao, ele só bloqueia as funçoes mesmo. Por exemplo, se eu faço upload de algum audio eu nao posso criar personas. Talvez o seu nao esteja deixando baixar por conta disso. Vc fez upload de algum audio pra usar como base?

[daring lava]

I see, if you adjust the style of music enough you'll be able to achieve a very unique sound from it too! Might take some fiddling to get iit where you want it, but worth it for sure! Plus that's half the fun!!

[shadow plaza]

Is there any promts that i can include that fix this problem is very anoying.. I like the song like the melody but this effect that music is played from the toilet next room is ridiculius

I notice this more frequent in heavy hitter music like rock metal industrial

[daring lava]

With each song being a bit different, it can require a different approach. Wish I could say there was standard surefire do-it-all prompt. Best we can do is try to start simple and build out. One thing that is super helpful (if you want to), look up songs that match the style you are looking for on the platform and study what other's have used that works well and adopt those approaches. That will save soooo much time and help build on what you already know!

[shrewd willow]

A few of my songs in 4.5 feels like that @shadow plaza . It happens way less often if I use a persona.... Although I dont do heavy metal so my apologies if it doesnt help you, but thats what fixes it for me.

[wintry sandal]

Pelo jeito não, irei conseguir nada aqui.. obrigado

Não tem um canal direto esse suporte???

[stray lynx]

If you do want to keep trying though, you can get better sound quality by trying to prompt positively for "dry, tight, close-mic" and exclude "cathedral acoustics, cavernous, excessive reverb." But you might be better off waiting until June to see if they've improved the model. Metal is an uphill battle right now.

[hollow hare]

There is no official support team on Discord, you can only contact them by email.
This channel was suppose to be talking about different techniques (e.g. prompts) but most people assumed it was for assistance/tech support

[shrewd willow]

Is there any update scheduled for june? I am honestly enjoying 4.5

[stray lynx]

They seem to release a new update at the start of every month. I'm curious but not optimistic that they'll make 4.5 better for my use case.

What are you currently making?

[shrewd willow]

Eu acredito que eles tenham um email, porem esse email é conhecido por demorar bastante para responder. Boa sorte!

[wintry sandal]

Obrigado Eclipse

[flint haven]

SUNO SUPPORT HELP

Heads up: Discord does not have any official support team members.

For all non-financial issues, please send an email to: [e-mail]

Include:

• Your Suno @handle
• A short summary of the issue
• What you've already tried
• Any relevant screenshots or references

You'll get an automated reply when your email is received. Sending multiple emails about the same issue may reset your place in the queue.

[shrewd willow]

O email eh esse que o Woodhouse postou
[e-mail]

Escreve em portugues no chat gpt e pede pra ele traduzir em ingles, vai te ajudar porque eu acredito que os membros do Suno nao falam portugues

[shrewd willow]

Sorry I didnt see the question!
I am doing electro swing! You?

[stray lynx]

Metal

[shrewd willow]

It reeeeeaally improved from 4.0 to 4.5

I really enjoyed describing the song over feature. Since I am a complete illiterate baboon in music, I couldnt fully express what i wanted and hindered the potential of my creations, now its going much better!

[stray lynx]

They did improve a lot of things. I think the instruments sound a lot better now than they did in 4.0's final version. I think the style of music you make determines how well the new model work for you.

A lot of what I make is gothic / romantic thrash metal. 4.5 doesn't know how to make these kinds of songs. It really only seems to know how to write one kind of song and wraps your prompt around that one song, and that one song isn't what I'm writing.

[shadow plaza]

I will try this thanks

[stray lynx]

but it writes that one song extremely well

[livid trout]

meh, i have like a litle bridge that is now expanded into a whole instrumental piece, and i need to correct it, but i can't have the generation to sing the bridge exactly like it was .... would be nice if you could just force it to sepparate the bridge part from the instrumental intermezzo

[polar spade]

Ok. The emailed receipt should mean payment was as good as posted to the Suno account and that you should have a paid account. True. From Stripe but says SUNO. (Not just a confirmation of payment from Stripe)

You mentioned , though, you were recently upgrading from Basic to Pro . So at some point you downgraded to Basic? We all do this, no worries. As Basic then becomes a day-by-day refresh, there shouldn't be a period where paying for upgrading to Pro again would be glitchy.

So I'm out of thoughts other than , again, ask if you have other accounts this payment may have applied to. Nonetheless, no need to answer that, just check on your own time.

I wish you a quick resolve, friend.

[shrewd willow]

Hows 4.0 for your style?

[stray lynx]

They wrecked it in April. It was perfect before then. In April they changed the model to make guitars sound like midis and drum sound like tin cans.

[shrewd willow]

I abuse of my personas 😄 But I always dedicate some credits into persona free musics to try to fish some new gems and transforme them into personas

[stray lynx]

Yeah, Personas are so helpful. Sometimes they can be a little buggy but mostly I think they do a great job of making sure you keep a similar style from song to song

[shrewd willow]

A thing I am really happy in 4.5 is that its doing a better job into doing male or female voices. Half of my credits in 4.0 went to the trash because suno made a male voice where I wanted female. Now its going 100% accurate and I am so glad. It was literally making me quit!

[stray lynx]

If you continue with 4.5 you'll probably really appreciate that the model can help preserve the spirit of the songs you liked best as they add to and remove from the database

[stray lynx]

I think 4.5 is better about this, but I also feel like I might be the only person on the planet who never had problems before. It was very rare that the generator would pick the opposite for me and I always wondered why it was giving other people so many problems

[shrewd willow]

It was mostly due to not using personas, back in 4.0 I was obsessed with finding the perfect voice to create personas so I had to spam generation, it was literally 50/50. If you want, link me a song you made, Id love to hear

[stray lynx]

I usually don't link songs directly to suno but I'll be releasing one literally friday afternoon to YouTube. First one, in fact. Video assembled by hand, not AI created (aside from generating base images)

[shrewd willow]

Thats cool, drop me the youtube channel. I always happy to hear new stuff 😄

I am on the same boat, trying a youtube channel.

[stray lynx]

I'll have to DM it. the automod didn't like me posting the direct link

You're welcome to share your channel too, I'll give it a listen.

[crimson yoke]

I have a question. I have the Pro plan, but I was thinking of switching to the Premium plan. If I switch to the Premium plan, will I receive all 10,000 credits from the other plan, or do I have to wait until my Pro plan ends on June 1st? Can someone please explain this to me? I want to change plans so I can get the 10,000 credits immediately.

[tawdry scarab]

spell out the sounds instead of the name, do it phonetically

proration means you will be charged the difference, the portion left hanging, by not waiting

which means, presumably, you would get either 10k OR 10k times the duration of month remaining over the total length

what I can't tell you is how that part works

hopefully someone will come along who can

[half dragon]

Upgrading? You'll get it immediately. I remember that was what happened to me.
The other way around, on downgrading the tier.. you'd have to wait until the end of the period to make it effective.

[brisk shore]

does anyone know sunos current context window?

[daring lava]

S/O to @tawdry scarab and @half dragon for being thoughtful and helping people! ❤️

[tawdry scarab]

all I know is you can blur it with extra short extends without using get whole song

[brisk shore]

I meant like, when you're using extend, how much of the previous song does it take into account?

[crimson yoke]

Yes, now I get it, I had that doubt. Thank you so much for the support. 2500 credits run out quickly when there's a good model like this one (v4.5). I'm already looking forward to v5 at the end of the year. Hahaha

[brisk shore]

oh and also what is the length limit on a suno song?

[hard beacon]

depense on the version you use

[plain pumice]

Any good ways or tricks to use the remove function in the song editor?
..especially when the two pieces you're trying to in effect combine don't match well due to composition, sound levels or other issues?

[hard beacon]

v4.5 is up to 8 minutes... v4 is up to 4 minutes

[half dragon]

for first generation, max length is 8 min with v4.5, 4 min for v4....
If you need longer, extend it and 2 min max on v4... on v4.5.. maybe longer ..maybe 4 min? I haven't tested that.. yet.. but definitely longer..
if you need longer song... keep extending it... sky is the limit using that technique!

[brisk shore]

gotchu, but there's no offical word what the time limit is using extend?

[half dragon]

I haven't seen any official thing about extensions...

[brisk shore]

okk, much appreciated

[hard beacon]

btw.. was there an update i miss ?? i have the feeling the vocals and sounds are slightly duller than before...

[half dragon]

Check #🚀┃updates channel... you might not have missed anything if you already read those.

[hard beacon]

nope.. not really.. 🙂

[hard beacon]

mabye @steady kelp can give us an answer 😉 I've noitced the last message in the updates channel.. but i am sure yesterday i didn't have this duller sound...

Regardless of this, I have the feeling that the guitars have been improved... but the end result is still worse than it used to be 🙂 Everything is still very mechanical... it used to be better

[sleek stump]

in 4.5 I have only been able to get up to 2 minute extensions sofar

But I regularly see 7:50-8:00 full gens.

[plain pumice]

Anyone know if it's the inner or outer part of the remove/crop box where it cuts from?
..like wise, is it possible to make multiple section selections at a time?
Not sure why, but when I use the feature, it seems that it cuts more, so for example, if I for example have a song structure that looks like this:
Intro -> Verse 1 -> Pre-Chorus -> Chorus 1 -> Verse 1 -> Pre-Chorus -> Chorus 1 -> Verse 2 -> Pre-Chorus -> Chorus 2 -> Bridge .
I mark this:
Intro -> Verse 1 -> Pre-Chorus -> Chorus 1 -> **Verse 1 -> Pre-Chorus -> Chorus 1 -> **Verse 2 -> Pre-Chorus -> Chorus 2 -> Bridge .
Expected Result: Intro -> Verse 1 -> Pre-Chorus -> Chorus 1 -> Verse 2 -> Pre-Chorus -> Chorus 2 -> Bridge .
Actual Result:
Intro -> Verse 1 -> Pre-Chorus -> Chorus 1 -> Verse 1 -> Pre-Chorus -> Chorus 2 -> Bridge .

Any one experience issues like this?

[smoky lichen]

How do you get close to 8 minutes on the main gen?

[sleek stump]

Its the amount of content in your lyrics.. if you give suno enough to work with you can get up to 8 minutes.. I also get 8 min instrumentals so... Style also plays a factor in track length, if you are making music that generally only has 3-3.5 minute lengths, its going to be more difficult to get Suno to make really long songs in those genre.

[smoky lichen]

Thanks! Helpful, though I guess there's nothing more concrete.

[polar spade]

If you have 10 days left to renewal date and upgrade to have it prorate immediately, do you therefore get about 3300 credits for 10 days and charged $10. Is that's the idea? I have never chanced it. 😊

Or you get 10,000 and charged $30 immediately, but they prorated and give back $3 and 1/3 of remaining credits from the Pro balance you had paid for previously.

It's never been explained what prorated and immediately means.

[sleek stump]

Theres a bit of RNG that goes along with this... so roll the dice.

[half dragon]

To be honest, I've done only once 🙂 hardly remembering now 🙂 but I could have sworn I got 10K credits immediately, porating or not... I didn't pay attention ..lol
No, there's no documentation explaining how that works anywhere.

[prisma marsh]

Apologies if this is answered somewhere else, I couldn’t find it. Is it possible to specify key, time signature, and tempo in Suno? And can you also specify bar counts (like if you want a specific length instrumental between verses)?

[daring lava]

With the new 4.5 model there is lots of improvement shown on things such as BPM, Key, etc comprehension. Regarding bar counts, you can guide it with bracketed imfo but still has room to grow for sure! Some cool 4.5 articles at the btom of this channel worth checking out if you want to: https://discord.com/channels/1069381916492562582/1294418290348921007

[ornate nimbus]

⚠️Hello everyone, I have an important question, namely I want to make one of my songs also in English, is it possible to achieve a melodic line the same as in the original, or am I unfortunately doomed to a completely different sound of the song?⚠️

[daring lava]

Hey there! That could prove to be difficult if trying to generate a perfect translation over. The model will get creative and things will change. You coul stem the music from the original and upload that and use the newly-translated lyrics with a cover, but I would expect significant changes in the response

Also, you could try the Replace function as well, but that may burn through more credits than you want to. IAnd again, wouldn't expect it to be a perfect translation while keeping the melody intact from the original

[sinful juniper]

Can Suno V4.5:
C3 E3 G#3 -.𝅘𝅥𝅰𝅘𝅥𝅰𝆡𝆢. | D3 F3 A3 -𝇒𝅘𝅥𝅰.𝆺𝅥𝅯𝇝𝆡- | D#3 F#3 A3 .𝆝..𝆡𝅘𝅥𝅲𝆠𝆟-𝅘𝅥𝅲. | F3 G#3 B3 .𝅘𝅥𝅲𝆝𝅘𝅥𝅲.-𝅘𝅥𝅲 |
F#3 A3 C4 .𝅘𝅥𝅰𝅘𝅥𝅰-𝆡.𝅘𝅥𝅰| G#3 B3 D4 -𝇒𝅘𝅥𝅰..𝆡. | A3 C4 D#4 .𝅘𝅥𝅲.𝆡𝅘𝅥𝅲𝆠𝆟-𝅘𝅥𝅲-.𝆢 | B3 D4 F4 -𝆢𝅘𝅥𝅲..𝆡𝅘𝅥𝅲𝆠𝆟-.𝅘𝅥𝅲𝅘𝅥𝅲
𝄕𝄙𝇑𝄙𝇎𝄙𝆡𝇓𝇖𝄙𝆥𝇖𝆺𝅥𝅯𝅬𝇙𝇒𝇝𝄋𝆥𝆳𝇒𝄙𝇛𝆺𝅥𝅯𝇓𝄱𝆲𝄳𝆟𝇅𝄙♮𝇕𝆹𝅥𝆖𝇦𝇝𝆳𝇢𝄙𝇛𝇧𝆱𝇦𝄙𝇨𝇝𝇧𝇘𝅬𝇝𝆟𝄏𝄙𝄉𝇨𝆺𝅥𝅯𝇖𝇧𝆳𝇒𝆖𝇐𝄙𝇛𝆹𝅥𝆺𝅥𝅯𝆹𝅥𝆖𝇕𝇓𝇢𝄙𝆡𝇨𝇤𝆺𝅥𝅯𝅫𝇖𝆳𝇗𝇦𝄙𝇓𝇚𝆖𝇝𝄙𝇙𝇒𝆺𝅥𝅮𝇒𝄙𝇝𝇜𝆹𝅥𝄏𝄂

Or still not recognized

[tiny kraken]

I'm looking for some assistance with accessing my account, and I have yet to get a reply from the support e-mail.

Is there anyone who can point me in the right direction?

[daring lava]

Probably not quite there yet 😅 But haven't tried that either. Interested to know how it turns out if you decide to test it out!

[daring lava]

Sure thing, what seems to be the issue with your account access? What are you bumping into?

[tiny kraken]

If I'm honest, my password isn't working and I'm not sure what the problem is.
I've sent an email in to support with plenty of private details of the account, but no reply yet..

[daring lava]

Have you tried resetting the password? Are you getting a visible error message that you can screenshot? How long ago did you reach out to support?

[tiny kraken]

I sent the email to support 3 days ago.
I am unable to even reset the password as it is also the Gmail account (which I cannot access for the same reason, because they are tied together).

The error is just that I am using the wrong password.

[tiny kraken]

I don't want to tie up this channel with an issue like this, I'm mostly just hoping I'll be able to hear from support and actually get this resolved.

[sleek stump]

This sounds like a bigger issue with GMail than Suno..

[daring lava]

I agree with Ginger, wouldn't hurt to check all fronts on this. When did you send in the suppport ticket?

[sleek stump]

In my brains logic, if you get your gmail issue resolved the Suno issue will also resolve

🤷‍♀️

[daring lava]

Also, hi @sleek stump hope you're having a great day!

[sleek stump]

Hey there! Its not bad at all, getting ready for out of town work weekend so today is kinda lite! Hope you are doing well!

[plain pumice]

No one knows?

[daring lava]

Awesome! Hope you have safe travels!! All is well here 🙂

[daring lava]

When removing/cropping each instance is independent --eg one section at a time. It should be functioning so the inner most edge of the identified area on either side is where the action takes place.

[sleek stump]

Crop only works from one end or the other, you can not remove or crop a section from the middle.

That has been my experience with the feature

[plain pumice]

That's good to know, because it's not clear where it actually remove/crops from.
How about when you zoom in, it seems that you can zoom in pretty accurate, but I wonder, is it enough to just select the section by default with a single tap, or do you actually have to zoom in to get the entire section with no overlap? Thanks for responding by the way.

[sleek stump]

With Cropping you are selecteing the part you want to keep.

[daring lava]

Certainly! Happy to help 🙂 would you mind sharing a screen shot of what you're seeing so I can point to things?

[plain pumice]

That makes sense, and you'd need to use remove if you need to remove something in the middle - thank you for responding.

[polar spade]

Crop means to copy and retain.
Remove means to cut out and the remaining splices together.

If you have a 3 minute song you can Crop out from 1:30-2:30 and keep that segment.

Or you can Remove 1:30-2:30 and what you keep are the remaining ends which splice together as a "Full Song"

So, no, you do either , one at a time. Cannot do multiple sections simultaneously.

[smoky lichen]

Does "high quality" work as a style or in lyrics description?

[plain pumice]

Sure. This is what I'm working on.

[smoky lichen]

I often forget to try, but less than crystal clear gives character.

[daring lava]

I haven't tested it, but interested to see. If you decide to try it out, please let me know if you recognize a significant difference 🙂

[daring lava]

Okay great! Thanks for sharing! Now do me a favor and set your tool to crop and select (click) one of the sections --eg verse 1

[lunar nexus]

wow! I apply the replacement and nothing happens!

[lunar nexus]

The final version does not appear in the list.

[plain pumice]

Notice it doesn't only select one section in it's entirety, it selects more than that.

[lunar nexus]

wow! I apply the replacement and nothing happens!

and now everything has appeared, after 5 times

[daring lava]

I see what you're saying, you can adjust it manually too, but it may be splicing it to the best target area for blending purposes. Best practice is to give it a go and adjust as needed if you aren't happy with the result.

[formal parrot]

HI, I'm new here, so forgive me if I say anything newby-ish:) I'm a PRO level user of Suno. I'm trying to create a song using custom lyrics that is going to end up being at least 8 minutes long. When I input my lyrics I max out at 5,000 characters, I have at least another 2,000 or so to complete the song. I've tried using the extend feature but it doesn't work and just refunds my credits. Any "work arounds"? I'm willing to pay for more characters if that's possible. Any advise? Thanks! Marty

[daring lava]

What a small section, how long is that area you are attempting to replace?

[plain pumice]

Of course it makes sense to target a specific point for blending purposes, but the question is, if it matters and the AI can intelligently blend them?

[daring lava]

When attempting to extend you need to make sure that you are extending from a part of the song that has significant subtance so the model doesn't get confused and think it's on the way out. When extending, if it's not quite taking hold, try extending from an earlier place in the song! Also welcome to the server! Here's a link to the server layout so you know your way around #suno-server 🙂

[daring lava]

That is indeed the question, and with each case it can be different. I expect there to be lots of further refinement in this down the road! If you bump into occurances where it seems off/not working right feel free to share your experiences here #🪲┃bugs-web for the team to see!

[formal parrot]

hmm, I see. I've been trying to extend from the very end of the initial clip. You say this is unadvisable? If I do it the way you suggest it should allow me an additional 5,000 characters if required?

[plain pumice]

There is also a level of ambiguity, because when you zoom in, it seems sometimes it overlaps the sections, and then it doesn't - this may an entirely UI issue, and not function issue.

[polar spade]

"Intelligently" ? When using Remove it will literally splice the two sides together with no blend. You would use Replace to insert or replace and Suno will try to blend. So when using Remove you need to add a bit of skill and use your ear to find a good matching beat, rhythm, phrase, to be as close possible on the splice. Then by using Remaster, that's when it can try finding off beats and odd sounds and blend them.

[daring lava]

Thanks for flagging!

[polar spade]

It does that on zoom, same as if zooming into Google maps, the dot is discovered to show up hovering a bit off from where you think you placed it.

I listen to the song and micro-manually type in the timemarks. I never slide the bracket as an end move. And always check playback after each micro-time entry.

[plain pumice]

Yes, that's what I suspected, that there would be no blend.
So when there is a problem of it has produced a dublicate entire section of Verse+Pre-Chorus+Chorus and the next verse doesn't align with the previous Chorus end point, as in my case, would it be better to remove those section and then extend from those points?

[daring lava]

As long as you are extending from where there is substance you can continue to extend on and on as much as is needed. And yes, 5000 character limit for the 4.5 model 🙂

[polar spade]

It, Remaster, might get creative a bit and give something decent, but for the most part, we help it out by planning where , to the micro-swcond we should place the remove splice marks. However you feel best to fit the arrangement. Here the vid, [poor audio] of a remove with a random splice mark ... and the Remaster blend after done. 8 seconds around the 1 minut splice mark.

https://suno.com/s/U8kV7FBDyahkE66e

Sorry... video too large. Just the blend of the spliced section to post.

[plain pumice]

It's good to know that is paramount down to the microsecond.
So when we have loose ends (two connecting sections that blends poorly), the best bet is to try and use remaster?

This is weird.
I've been getting this lately when I use remove, but reloading the page fixes it.

[polar spade]

Not paramount. Just the better we exact the splice, obviously the better /easier for Remaster to blend later. Adjust 5 times , splice once, i prefer over sliding and guessing. Even that crappy splice i did blended uniquely, at least, and might be enough for some songs.

[polar spade]

Yep. I wait 10 second after the error. But 98% no problem. Refresh or just click Library to get the song.

[polar spade]

One day I might write a 1 minute song and remove 2 sec of every 3 seconds and then remaster the remaining 20 seconds and see what I get. Haha.

A recording studio cant do that.

[plain pumice]

I actually tried to zoom as far in as possible, targeting the outer bounds of a specific section, apparently it decided to remove 3 sections, the targetted and the two adjacent ones.

To follow up, when I selected the section it defaulted to, it only remove that section.

[polar spade]

Just stop. 😉

It's not so much glitchy as it's trying to quantize the marks to a reasonable spot that it knows it can blend easier, regardless of what you want. As An AI will blend anyway, it likely won't matter much, but if we manually put in the timemarks, that is obviously more to our perceived liking hypothetically.

It's not going to blend lyrics well of course unless there is enough blank splice between the splice. But it will adjust the instrumental as best possible.

[plain pumice]

My point wasn't really about precision, the fact, that when I try to be super precise, for me it actually removes more than I target.
Like I said, it removed 3 sections, not 1.
But when I only used default one click selection, then it removed just that one, at blended fairly naturally.

[polar spade]

Even manually putting in micro-swcond numbers, Suno can prefer :02 seconds more or less and might change it. So to make them stik, if even needed, at least go from left timemark to right and from micro, seconds, then minutes

[tawdry scarab]

quantizing = happy users, precision = nothing ever working right with the rhythmic flow

[polar spade]

I think 3 is the limit. If i remember. If you're removing 1 second then just use Replace.