📚┃suno-school

[hallow moon]

supposing i have 20 calls going on, in those 20 calls theres an audio generating each 2-5 seconds in avg? then i would need to generate in parallel in order to fulfill multiple calls with tts, but at same time i need speed or there would be too much latency on calls and people would get bored

ye right

[untold briar]

Bark is not well suited for real time streaming becaus of the stacked models. It is not impossible to stream it, but it's not something I have seen in open source public versions

[hallow moon]

honestly was looking for a way to make audios

then i send them

not realtime but fast generation

[untold briar]

Even on a 4090 it's hard to guarantee always less than 2 seconds, unless you force your text prompt to always be short

[hallow moon]

ye right, might have to find something else, but i was hoenstly hooked by the voice timbre quality and [laughs] [break] ecc

[untold briar]

Oh the naturaleness, the humanness, I think Bark is still better than anything else. Even compared to Eleven V2, that's way more reliable and less likely to screw up, but I wouldn't say it sounds more human than the best of Bark

[hallow moon]

yeah

[static moss]

What's the architecture of bark btw?

For sure there's a way to optimize it

[untold briar]

There are some optimizations in the Huggingface Bark and https://github.com/PABannier/bark.cpp for example

[static moss]

I'll check. But I guess further optimization is possible anyway

[wanton pike]

Has anyone found a way in multiparts to effectively instrument switch? I have found success double/over prompting, but one thing that seems impossible to change easily is the underlying beat of the song. I've got a few hard cuts to work, but here's a multi-part example of how double tagging things has worked well for instrumentation switching. If anyone has any advice or better method of doing this, I would be very interested.
https://suno.com/song/1d6fad2a-bd4e-4b4f-913d-c7c2b8e803e7

[untold briar]

What instrument are you trying to switch to?

Oh I see, in the video

[wanton pike]

Just any in any track.. I'm most interested in figuring out how to change the percussion/beat

[untold briar]

Looks like you are getting some switches, yeah, but switching the beat is probably more stubborn

[wanton pike]

I wanted to post this for an example of how it's worked well for me to switch

It was easy switching most instruments/styles with over prompting, double prompting seems to work best.

That took maybe 100 generations in total, which isn't too bad, but I really want to know if anyone has figured out how to change the beat without using 5k credits lol

[untold briar]

So, this can cause problems, but double prompting is not near the ceiling. Not only can you 10x or 20x prompt in the style prompt, but you can add in even more in the [lyrics] like I just switched to Polka in this one by going way over the top: https://suno.com/song/8df23499-64a9-43e0-ad69-9e0cf26f014e

A very important factor in switching music styles or instruments is the length of the clip you are Extending From. The longer it is, the more stubborn the audio will be to any change.

In this case, the the audio clip was over a 1 minute, so I had to spam the word many times.

But there is another way of switching, using a very short clip instead. If you were Extend From a clip at time 00:01 for example, that 1 second of audio, will have very little effect compared to the style words you use.

[wanton pike]

Most of those were 16 seconds, that's good to know, good example here. So does prompting in the lyrics window in an instrumental actually have effects?

If the seed is instrumental I have found the lyrics prompts do nothing most of the time.

[untold briar]

Up to 5 or 10 seconds, the style prompt can have a reasonably big effect. But if you want to completely switch to an entirely new genre than even less than 5 seconds may be required.

In the Polka clip, I actually Get Whole Clip on each section, so it was never extending from just a short clip, it was longer than 1:00

If instead I had extended from the clip direclty without using Get Whole Song, then it would not require going so over the top with repeating 'Polka' a million times

[wanton pike]

Okay, good to know. I will have to try going over the top with the prompts and see what happens.

Another tactic I've tried (Mainly to retain vocals) is merging the song THEN extending, but that has mixed results.

That's exactly what you're saying XD

[untold briar]

So basically, the nuclear option is to use a short clip. The tradeoff is you lose musical memory, and the song is less consistent, less coherent.

However, you can mitigate the loss by letting the music change to add a new instrument or style, and then Get Whole Clip (to put the previous song back in memory) but then Extending from a time after the instrument is introduced. Basically, by putting everything back in memory, Suno will at least have some sense of the previous music.

[naive quail]

Depending upon what you're going for with the song (and how many sections you're making), it can be good to merge like that periodically so that Suno can access some of the earlier sections.

[wanton pike]

Yes and then you are stuck in the perpetual it's a long clip and difficult mode... anything over 7 minutes starts to get a bit nuts.

Retention of Verse/Chorus in those situations would be nice, I haven't had much luck with that once merged.

[untold briar]

Arguably a better default for Suno would be to "Get Whole Song" by default, I think most people don't want to change the music genre. However, in the current system, if Suno made that the default, there is no way to not do that, so it would make it almost impossible for everyone else to every change instruments or genres without extreme difficulty

[naive quail]

This is one that I did recently - you can see each time that I used **Get Whole Song **

[wanton pike]

How did you retain the vocals doing that, it recognized the tags? After I merge it tends to be a mixed bag of nuts if I can even call the lyrics correctly again in any extension.

[naive quail]

When I do sections like that, I don't include the entire song's worth of lyrics or tags (for instrumentals). I do those a section at a time, too.

[wanton pike]

That would argueable like you said, make it very difficult to breakdown, or have a genre change. At least this way it's open to attempt lol

[untold briar]

The new upload feature can also be used indireclty as a workflow to do style or instrument changes, using the Uplaod feaure. You can download your song, make a manual edit, then Upload it and extend. In fact you can get real silly and upload a song made up of two parts that aren't even the same song just to see Suno try it's best to make some kind of musical sense out of the nonsense you gave it.

[wanton pike]

And you have the same lyricist?

[wanton pike]

I've ran some of my stuff through it, it was kinda cool. If we could inject MID-mix it would be EXTRA cool.

[untold briar]

Are the vocals more than 2 minutes in the audio before the time you extend? Maybe if you change genre, the time should be shorter before call back to the voice, or try to keep the voice while donig the genre change.

[wanton pike]

It would be the ultimate remix maker lol

[untold briar]

I should note in case anyone wants to use Suno to make music hybrids, I found it works best to both overlay the audio and use sections of each song that aren't overlayed in the upload. It can work with just one or the other, but using both works best.

[wanton pike]

That would make beat switching WAY easier too... if you could insert audio at any given point in a part-series. You can only use uploaded audio as the seed right? Or does it inject it into the extend?

[untold briar]

If it's not working, the most fiddly variable is the length of the clips especially the very last one before the overlayed audio

[untold briar]

Seed isn't the right term, but yeah the the idea. The upload becomes a clip and you can Extend it.

[wanton pike]

Right, I mean the root of the song, it can only be "Part 1"

[untold briar]

There's a lot of stuff you do, if you aren't worried about the song be in one single piece on Suno website. Because yes, if you use the Upload like this, then if you ever "Get Whole Song" the final song with have to include your uploaded audio. Which in this case is a bunch of weird audio used only to get Suno to generate in a certain way. It's like, Audio File as Audio Prompt, so if file is comletely chaotic nonsense it doesn't matter, we only care about what Suno generates with that as input.

However, Suno is hopefully (fingers crossed) working on some cropping function...

[wanton pike]

That would be wonderful, and sounds incredibly useful.

[untold briar]

And since we're talking about it, cropping would also be useful for style shifting, since it would let you remove very selectively from the music history. Basically now we can set Extend From time, but not a range of time. A crop is basically a range of time, that you treat as new clip. So cropping also adds some advanced propmting features for users who want to dig in more.

[naive quail]

that'd be awesome

[wanton pike]

Yes it would, and let you extract say like a break.

And cropping would also let you stop the credit dump to end. I've just been considering dumping to DAW to end, but then, you don't have the song on Suno, so it's a tradeoff.

💵

Thank you for the advice and info I appreciate the insights. 🙂

[true night]

after alot back and forth my massive 2h mix tape got rendered.. i feel some of the devs might done so manualy if so thank you <3 then that project is done over to the next
@charred merlin just would like you to know so well you know 🙂

[soft bronze]

Just wanted to share that UVR 5 is the best vocal remover.

[unique stone]

It's great

[full lava]

I reported this once in Gen chat and refreshing the page fixed it. but I just got into the website and hit create.

No idea what causes it. idk who to report this too. guess it would eb nice to have a bug report section. This hasnt happened in any version before the Cutom Audio thing was implemented

also this appeared in my create tab, no signs of any of my past works

[marble parrot]

Suno stop stealing credits from me, when I type [End] I want you to end, not make another 2 minutes repeating the middle verses. Thank you. #RantOver

[sage drift]

Oof suno getting sued

[hollow hare]

Extend where you want to end with the new extension set to v3.0 when using ending tags (e.g. [end] in lyric box and "end, short" as the style tag) as 3.0 is much more reliable for ending than 3.5.

(Short to help end it quickly in hope for credit refund)

[marble parrot]

Interesting, I'll remember that. Thanks.

[fresh cave]

how can I make songs ??

[hollow hare]

/help

[flint haven]

How to use Suno

Hi there!
If you’re ready to dive into creating music, head over to Suno.com.

For more detailed information, check out docs-guides-resources.

If you have any more questions, just ask in #general-suno where you can always find our community, moderators and guides to help you with most questions. 😊

[visual nimbus]

Can I leave a question as an issue in the repository?

[untold briar]

Or you can DM me

[graceful condor]

Anyone knows a android app for changing mp3 quality

I cannot upload my song because it's not 320 kbps

And 44100khz

This is the song from suno ai

[analog sonnet]

What nonsense is this. What's with Suno? Why does she start SINGING the CHORUS AGAIN after singing the chorus as a verse! Why is she trying to invent something that is not there, there is just a text with all the settings, why is she trying to invent something from the whole text that is not in the text? I am very disappointed with the latest results. I spent more than 500 credits and did not get any normal result at all, then the words just chew, then new notes appear in the middle of the line and Suno tries to embed the text there, then there is just a discrepancy between genres.

[gray lance]

there is something very strange with the 4.00 generation song : the very last 0.15 sec, it give a real nice and clear composition little bit similar to the song, but like if it's recorded live in studio, funny and strange

[gray mica]

Quick question, are you limited to ten public playlists?

Just wondering since when checking my profile I can not see all my lists

[urban wing]

@ #📚┃suno-school my suno credits have been stuck at 50 for a couple of days now

[unique stone]

Do you mean they don't decrease even if you use them?
Or they don't go above 50? They refill to 50 daily on free plan.

[urban wing]

they dont give above 50, i havent tried using the 50 yet

[unique stone]

If you are on the free plan, that is normal. The daily credits reset/refill to 50. They don't stack/add up.

[urban wing]

no its not normal, if a daily plan gives me 50 credits why dont i have 100 credits or more by now?

they dont stack up?

wow

[unique stone]

Correct. They don't stack up.

[urban wing]

so,,, where do i see pricing for actions?

[unique stone]

The subscription page had information about pricing. There is some extra information at the bottom about how the credits work. Subscriptions have monthly credits - those also do not stack up (so you get e.g. 2500 credits for Pro which can be used during a month and they reset to 2500 at the end of the month)

[urban wing]

yeh ,, thank you! i'll take a peek

so 2 songs a day, and 2 running jobs, that seems toatly fair, but what counts as a running job?

like if i expand one song is that 2 actions already?

so 2 songs / or 1 expanded song? a day ? seems fine

[unique stone]

When you press the "Create" button, that creates a "job" for the AI to process. That job ends when the song has finished generating. It means you can press the button two times and have to wait until those songs are ready before you can press it again.

[urban wing]

okay but i still havent realy found what 1 credit is worth in the documentation

can you link it ?

[unique stone]

Ah... Each "Create" is 10 credits which gives you 2 songs (always 2, this can't be changed.) So with 50 credits you can press the button 5 times each day.

Creating and extending is 10 credits each time.

[urban wing]

thank you for your answer!

one last question when i expand a song how do i get my songtext to be used in a new 'version' ? and i asume that would mean an extra 10 credits for generating it...

[unique stone]

Each time you create or extend a song you can type in new lyrics. Sometimes the AI can get confused and might repeat or skip lyrics which is a bit frustrating (just mentioning so you know this is a thing that can happen!)

Yes, extending is 10 credits too.

With v3.5 you get up to 4 minutes for the first part, then up to 2 minutes for extensions (the times can be shorter, sometimes very short, it's not guaranteed... Hard to explain why but the AI isn't able to know the length before it has actually generated the song)

[urban wing]

yes but i havent found the button to use my 'extented lyrics'

[unique stone]

There's a community wiki which you might also find useful: https://www.suno.wiki/

[urban wing]

i'll just take a look at the suno www, and the wiki then..

[unique stone]

Ah there'll be a button next to the song, called "extend" - or it'll be on the "..." menu jf you are using a phone

[trail badger]

Has there been a quality problem at 3.5 v for 1 or 2 days? Vocal sounds started to sound very poor quality.

[urban wing]

hey another question i get an error when i click the generate lyrics button, is this because i cannot extend an instrumental song with my own lyrics or is it because i have to start a creation from scratch?

i still havent figgured out how to make the prompt use my own lyrics

the error i get when i click the generate lyrics buton is, 'Error please use a different prompt"

[rustic wyvern]

I have troubles using voice input, just registered my and all the iteration I have but the first one are not like my voice at all (I'm a woman and all the iterration are male voice)

[urban wing]

yeah i get this when typing lyrics in the custom lyrics paine "Error generating lyrics , please use a diferent prompt..."

[pearl trout]

How I write him my lyrics and he creates a melody for me

/chirp

[flint haven]

The /chirp command is no longer working

Chirp is retired, and no longer on discord. You can make music now on our website: https://suno.com

[rustic wyvern]

This is getting very frustrating, it doesn't even add the whole thing when I put "get the whole song", I've used more than 100 credits trying to get this right

[charred merlin]

"Get Whole Song" doesn't cost any credits, and assembles all of the previously used segments that lead up to the extension you're attempting to "get whole song" from.

Ex. Clip 1 (02:00), clip 2 extending from the end of it (01:00), and clip 3 extending from the middle of clip 2 (01:00).

If you "get whole song" on clip 2, you'll get a 3 minute song composing of clips 1 and 2. If you "get whole song" on clip 3, you'll get a 3m 30s song based on when the extension points were used.

If desired clips are being left out from the "Full Song", you may need to trace back what songs are extended from what.

[charred merlin]

https://www.suno.com , you can no longer create music through discord.

[charred merlin]

Sounds like the text that you have in the Lyrics field before hitting "generate lyrics" is causing the AI to write something that it considers inappropriate, so it's stopping itself from continuing.

Try to write the lyrics yourself, use another chatbot like ChatGPT or Copilot to write the lyrics, and that way you can examine the lyrics yourself to figure out if anything is inappropriate

[charred merlin]

How long of a voice clip have you uploaded?

[rustic wyvern]

14 seconds, I tried cutting up earlier to see if it would change something but no only the first iterration had my voice on it, reste were male voices, as a woman i have a low voice but the male voices weren't even close to mine. I know get whole song doesn,'t use credits but after trying many times to get the input voice coming without success just to end up using the only itteration with my voice and having the whole song cropped at the end is pretty frustrating.
I redo the "get whole song" a second time and this time I had the whole song coming through

first extend changed the voice, so i extended anoter time at a moment where the voice was singing with a higher pitch it seemed to do the trick and not change again to a male voice

[charred merlin]

The more context, the stronger the "theme" is for suno to hook onto, so I'd suggest trying a longer segment -- spend the full 60 seconds repeating the verses you want to be sung so that it can pick up the rough structure as well.

I know it might make it not as clean for sharing within the social media side of Suno, but if you sing for 55 seconds, then do a hard stop, then start singing again so that you're not extending from silence (which is a guaranteed fail), so that way you get the benefit of "training" the initial generation on your voice properly, but also giving yourself an easy segment to go into a free editor and crop out the singing at the beginning so that it starts "cleanly"

https://www.bandlab.com/studio

[rustic wyvern]

Thank you for the advices about using voice input

[urban wing]

I see so it does censor or cut inapropriate words , out.... i still think its a great song tune generator when using the instrumental slider..
thank you for the the hint

[charred merlin]

I want to encourage responsible use of Suno, but just between us, as long as the mods aren't listening, you can spell words out phonetically if you run into troubles. (ex. "A zombie ate me" might get flagged, but "a zombie eight me" would probably get through)

[unique stone]

This is the correct workaround when lyrics are blocked (so long as they comply with terms of service, FOR EXAMPLE don't make a song for the purpose of harassing or threatening someone.)

[charred merlin]

Yeah, I don't want to party with someone if they use Suno for hatred. 😤

[tribal rover]

Is there a way to change the details of a song, such as the picture, so that when the video is downloaded with lyrics, it shows the updated image? It seems like even if I change the image, it still downloads with the original image, regardless of refreshing. Any solutions?

[charred merlin]

#1254083946279469167 message

[empty anvil]

Can you fix Suno video output because some text get cropped due to its width🙏

[jovial scarab]

hi can someone help? most of the songs that I have created are now not visible on the platform. what can i do?

[charred merlin]

#📚┃suno-school message Thanks to @hollow hare for compiling this helpful checklist

[thin flint]

Has anyone heard of suno-ai.me? It was shared in my FB timeline and I didn't know there were multiple Suno sites. Curious why it was created over Suno.com?

[charred merlin]

Not legit, 100% scam. You're better off using the official Suno site -- they're using the Suno backend without permission, and charging money for it. Plus they'll just send you here if you ask for support, and the only advice will be to use the official site/app 😛

[hollow hare]

There are a lot of scam sites (and apps) that target people that do not know about the official Suno addresses.

Heck, the pricing tells you that right away with no free plans available. Basically acts like resellers (and I would not be surprised if their ToS says that the site owner will have ownership over the songs)

And, sadly, it works because people don't put much effort into investigation, likely heading there directly through social sites like how you had seen (while a simple lookup on Wikipedia would have linked them to the official site).

The only official addresses are:
https://suno.com
https://suno.ai (the original address, redirects to .com)

Original app:
https://suno.com/iOS

[charred merlin]

lmao, in order to use the scam site, you need to agree to not misrepresent yourself as another company.

https://tenor.com/view/a-little-bit-awkward-love-is-blind-uncomfortable-embarrassed-disconcerted-gif-7087180328501280567

[tidal nebula]

Hello, i have a problem with Suno AI on the website, the AI dont recognize the lyric i have added everytime. I have testin different lyric, and the problem is the same, it work yesterday. Someone can help me please ? Thank you !

[hollow hare]

Are you using custom mode?

[tidal nebula]

yep

oh, nvm, it work now

idk why

[hollow hare]

Hm, was it repeating verses?

I know that is one issue people have at times with 3.5

[tidal nebula]

Yes, some verses was repeating

probably the issue

[hollow hare]

Ugh, hate it when that happens but I didn't look into how to solve that problem.

I do not know if [Break] worked in that scenario.

IIRC, someone mentioned they simply extended before the repeated Verse and continued on with the second verse but I do not remember how effective that was during the rest of the song

[onyx glade]

Is there a way to create intro's ?

[bright wasp]

Does anyone know why Suno sometimes can't complete a song and instead gives credit refunded? It's been happening so much to me, for every 10 songs I make at least 7 of them come back with "credits refunded"

[blissful dirge]

Tips on how to procedurally slow the tempo of the song

[charred merlin]

Sounds like you’re extending from silence, or a part of the song that starts to “resolve itself”.

[austere dragon]

Is there something going on? I'm trying to do an intimate piano song about progression and it just comes out as Pop/Rock.

[charred merlin]

Use v3 to start (which has a more broad range), and try to consider how your style prompts could be misinterpreted.

[austere dragon]

the prompt in question: Haunting Piano, Emotive Smooth Clear Soulful Male Vocals, Sad, Minor Key, Melodic, Melancholic

[austere dragon]

the song, wait until the second verse https://suno.com/song/9c8bbfd9-0937-4355-855b-8a6c50c8d3c4

[charred merlin]

Strip the vocal stuff out, take out all lyrics (except for metatags intended to push a style into a specific direction). Work on vocals afterward (stopped typing up this advice after reading your last comment and added the second part).

——-

If you enjoy what it sounds like in the first half, just place an extension point before you stop liking it and work on correcting the second part

[charred merlin]

Are you dissatisfied with how it starts off? Or is it just the second half that you wish was different?

[austere dragon]

It could start of better, but I put nothing in the prompt to make it a pop song. The constant percussion undermines what I'm trying to convey.

[austere dragon]

I had "downtempo" in there at one point, and it was worse.

[charred merlin]

Try listening to other Suno creations and seeing what people used as style prompts for inspiration.

If it’s misunderstanding what you’ve said, the solution is rarely to add more to your prompts to try to explain everything— it’s usually a sign that you need to reapproach how you’re attempting the song.

[austere dragon]

Thanks for that. Make sure it's always the user's fault is a good way to go. I mean I started with 'Sad Piano, Male Voice' but it came out sounding like dua lipa's latest drop, but yeah, it's on me. Somehow I have failed my song. Here's a link to my profile: https://suno.com/@jack... tell me where I went wrong with all those as well. Also you're recommendation for starting in V3 is nothing but a naked attempt to get people to use more credits.

[charred merlin]

This advice is being provided for free, so apologies if it’s a language barrier, but that response came off disrespectful of the time and effort spent helping you figure out how best to use the platform.

As I spend my free time doing this to contribute to a friendly and supportive community, you’ll excuse me if I don’t respond to you any further.

[austere dragon]

Thank god for that.

[uncut pumice]

LOL If a song is removed by staff....will I get an e-maill telling me that it was removed...and why?

[charred merlin]

lol, pretty sure they’d only remove it if it broke ToS, though I’ve never actually heard anyone report back if that happens

[charred merlin]

Sorry I wasn’t able to be of more immediate assistance (or simply do it for you), perhaps one of the other support volunteers will want to help you 🙂

Have a wonderful rest of your day!

[lucid valve]

hey guys

[mystic flume]

You're so much nicer and more understanding than me.

[lucid valve]

so lets say i have a instrumental with me

and i have a lyrics writeen

i have suno pro

can i make suno sing it in my instrumental?

if so, how?

[charred merlin]

Argh, soon!

From how I understand your request, you want to keep the instrumental exactly as is, and only add lyrics, correct?

[lucid valve]

correct

sorry for my bad english

thank you tho!

[mystic flume]

Currently you can upload the instrumental but Suno will just use it as inspiration rather than reproducing it.

[charred merlin]

Your English likely has more consideration and careful thought than a large number of natively-English speakers. ❤️

[lucid valve]

oh, how can i actually do that? cause i tried but suno literally just toke my instrumental and didnt sing anything on top of it

[lucid valve]

ty <3

[charred merlin]

Like Tinda Lee says, the current way we can approach this is to upload 6-60s of the song, then extend from it.

Uploading a song places it in your library like a regular clip, from which you can extend and specify the lyrics to be included in the extensions.

[mystic flume]

So, upload it, then use the "extend" feature, which will add new music to the end of it, and your lyrics. So it may or may not sound just like your uploaded bit, but if you generate a bunch of times you could end up with something close.

[lucid valve]

thanks both of you!

[mystic flume]

yw bb

[lucid valve]

sorry to bother you one more time, but i tried uploading me singing a song so suno would copy the melody and it didnt work

is that even possible?

[charred merlin]

We're still learning about features that might come in a new version -- "inpainting" is something that's been mentioned that could allow us to highlight a section of a song and allow the AI to overwrite it while keeping the rest of the song in tact. That could allow you to approach the first issue that you asked about, but we need to cross our fingers and wait for now 🙂

[charred merlin]

How long was your clip? The more context you give, and the stronger the "musical structure" is, the easier Suno will be able to continue from it cohesively.

[lucid valve]

ill be waiting lol, that would be perfect

[lucid valve]

like 30s

also the clip i gave was the chorus

[mystic flume]

It's possible, but it'll likely take a lot of tries. Give it as much as you can.

[charred merlin]

I would imagine that should have been enough. 🤔

3Daizy shared an example of her using the feature where she sings a chorus, https://youtu.be/6rzt1WNGUSc

Not meaning "see? it works!", just meaning that this is a tricky new tool that we're all still learning how best to use.

[lucid valve]

ill give it more tries

[lucid valve]

thank you for the example

[charred merlin]

Just realized that the community-made wiki hasn't been updated in a bit -- I'm sure there are people out there who are making amazing use of the upload feature and have better ideas how to squeeze exactly what they want out of it, possibly by using the uploaded audio to "prime" specific metatags for try to gain some degree of agreement between them and the AI.

❤️ Best of luck!

[steady steeple]

Why isn’t the app available in Israel :/

[unique stone]

It's not available outside the US at all

[steady steeple]

Oh wow that sucks:(

[unique stone]

For now at least.
I'm on Android and there's no app for that at all yet 😆

[austere monolith]

Thoughts about this project I found 😎

[unique stone]

Using that would be against the Suno terms of service. Please don't share things like this here.

[hallow moon]

hi, on suno, can i generate voices normally without noise or sing phrases?

consistently

[hallow moon]

https://prnt.sc/hVckxXle3JUd
Also what are these things appearing on my terminal, why everytime i run its loading these generation models (3 of them -.-)

[untold briar]

Bark has three models, those are 3 "progress bars" for each model. (Well there is a fourth model, but it's so quick it doesn't have a progress bar)

[hallow moon]

do i need all 3?

desperate try to generate faster lol

Also, whats this, should i solve or ignore? ..

D:\appsD\bark\bark-env\Lib\site-packages\torch_utils.py:776: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly. To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage()
return self.fget.get(instance, owner)()
D:\appsD\bark\bark-env\Lib\site-packages\transformers\models\encodec\modeling_encodec.py:120: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
self.register_buffer("padding_total", torch.tensor(kernel_size - stride, dtype=torch.int64), persistent=False)

[untold briar]

I'll reply in off-topic, this channel is meant for technical Suno music discussion.

[hallow moon]

ops :C

[soft bronze]

Am I allowed to upload on Spotify the songs that are created using free credits after Suno gave me approval to use it for general commercial?

[hollow hare]

If Suno had explicitly given you written permission that granted the song(s) generated during the free membership now has the same terms as a song generated during a subscription, then yes, you can distribute it to commercialized platforms like Spotify.

Before you sign up for a distributor, which you need in order to get your song on to Spotify, investigate and read the terms of service and check the chat here for the history if they will accept your music.

Generally, what I had seen was:

✅ DistroKid and Ditto are known to accept and distribute AI songs and
❌ TuneCore and CD Baby are known to reject #💡┃suggestions-web message

I recommend taking a look at Top Music Attorney coverage regarding different distributors ( #🗣️┃chill-chat message )
It will give you a good understanding of what to look for in their Terms of Service.

You especially want to find out how they distribute royalty and for any clauses that will make them retain it instead of giving it to you if they terminate your account (see TMA video about TuneCore and Symphonic as they had those very shady clauses).

In general, I see most people here using DistroKid and a few mentions of Ditto and LANDR. Check out all services, see what they provide and how much it will cost, etc.

[soft bronze]

Thank you so much. I’ll be using Distrokid.

[charred merlin]

Distrokid gives 100% royalty -- can't say enough great things about them for getting started

LANDR charges a lot more, and offers services that people might not necessarily need. Distrokid lets you pick and choose (and since theres a lot of free tools and community advice available, I prefer having a piecemeal approach to everything)

[hollow hare]

DistroKid is pretty great from what I heard, two items you will want to be aware of:

• Your songs may be removed once you stop your subscription unless you opt-in with Leave a Legacy (if you do this, be sure to do it as an album rather than single tracks) https://support.distrokid.com/hc/en-us/articles/360013535494-What-Happens-If-I-Stop-Paying-the-Annual-Fee
• You will need to pay extra for opting the song to YouTube's ContentID (which a lot of services also charge for iirc) https://support.distrokid.com/hc/en-us/articles/360013535314-What-is-YouTube-Content-ID

[charred merlin]

A note about "Leave a Legacy" -- it's a one-time fee of ~$49, but it applies per album, and they allow you to make albums of ~32 or 36 songs.

When I plan to stop using Distrokid, I'm going to place all of my songs into a multi-album collection and pay the couple fees to protect them with that way.

I'd rather handle things this way than have a "leave a legacy" fee baked into the cost of every upload that I make when I'm primarily focused on getting my content out there, and can take time to see which songs resonate with me on a deeper lever later.

A tiny bit of work to achieve exactly what I want for the absolute cheapest price.

https://www.bandlab.com/studio for in-browser DAW and effects
https://www.bandlab.com/mastering for 4 variations of professional mastering

[hollow hare]

See above (Just in case you did not see out follow up advices)

[rain shuttle]

hey guys how do i make a duet

i am making a musical i need help with it

someone help

[charred merlin]

Guaranteeing specific voices isn’t something that can be done with 100% certainty, and it can take a bit of experimentation if you have a specific vocal tone you’re looking for.

There’s some helpful tips on the Suno.wiki, but there’s a #1252298754787377225 about reusable voices that you could upvote for visibility ❤️

[olive grotto]

how long does it take for an account to be deleted? i requested deletion of my account over 3 weeks ago, and it's still up. if my account is not deleted within five days, this service will be in violation of EU law.

https://gdpr-info.eu/art-17-gdpr/

[mystic flume]

When does your subscription end?

[olive grotto]

i never signed up for a subscription. i got a free account

[daring lava]

Have you reached out to [e-mail]?

[olive grotto]

i did, they linked me to this form, which i filled out. yet my account remains up

[daring lava]

Ok thanks. Have you reached out to them again recently explaining your concern? If not, either reply to the support member that contacted you or begin another support ticket. If you’d like to, you can dm me the email that is/will be tied to ticket and I will pass it forward to the team in attempts to expedite it. @olive grotto

[olive grotto]

thank you, i'll try to contact them again

[charred merlin]

For future reference, if you add a username to a post in an edit, it highlights the post for them but doesn't ping them.

[daring lava]

Very welcome 👍

[naive quail]

@olive grotto Did you fill out the form three weeks ago, or more recently? I'm wondering if that's what starts the countdown in this case.

[olive grotto]

three weeks ago

[daring lava]

Thanks, always so helpful! 🫡 How’s your day going? Scrolling through this chat, I see you’ve been helping a ton of people today! So awesome! Thank you for what you do ❤️

[charred merlin]

Thank you, dude! Between the devs, mods, and fellow Sunoers (Sunoites? Sunors? Sunorians? Sunonauts? Sunauts? Sunoists? Sunosians? Sunovators? Actually, I think I like Sunovators.), I feel like I'm in great company 🙂

My day has been going well -- I've been playing with Websim.ai to see what's possible in the system (as well as taking advantage of unlimited Claude 3.5 Sonnet in it), and today I was able to get some 1~2-shot generations from Suno that sound & feel like I've grown up listening to them.

Lindsey (my wife) and I found a couple shows that really click with us, so I can stop making her rewatch ITYSL and Corner Gas 😛

Hope that all is well on your end! 🫶

[daring lava]

You definitely are in good company! Good people doing good things!

Sounds like it’s been a very productive day, always a good thing!

Everything good here. Busy, busy, busy 😉

[formal tree]

i'm trying to re discover how to make my music, i have alot of different songs i wanna get made but no way to make 'em because i forgot how to use this discord page.

[charred merlin]

If you mean the chirp command on discord, it's been discontinued -- everything has moved over to https://www.suno.com. There's the #1252267444773326848 forum and https://www.suno.wiki if you're looking for advice on how to go about creating a song. 🍻

[graceful condor]

Forgot why?

I remember now

It's routenote

However it's still in review

How long does it usually take??

[charred merlin]

This is the Suno discord, you'd have better luck asking Routenote how long their services take.

[graceful condor]

Unless there is a better distributor that is free and faster

Oh ok

[charred merlin]

I've read that it can take 14 days for them to publish the music -- I haven't worked with Routenote before, but it might be a good idea to ask them if they allow the distribution of AI generated music.

[blissful dirge]

So today about 10 of my songs got their image replaced with this, their previous images were generated by suno i dont know why did they got replaced

Any ideas?

[teal raft]

This is the default image now

Everyone gets it, it's a static thing

You don't generally get it now

But older gens sometimes got replaced with that image

You get it when you create a new playlist, that will be the image for it

[blissful dirge]

Oh

[teal raft]

It isn't anything you did

[charred merlin]

If you go to edit the song's photo, does it still show the old one to swap between?

[teal raft]

Suno used to have a different image default, was bird-related (chrip) but they've since changed it. It was around then that this image started showing up

[blissful dirge]

I was getting sone sick artworks on the final end cuts so its a bit sad

[teal raft]

The Suno logo is in the center

[teal raft]

Are you a free user?

[blissful dirge]

It says it's not available anymore

[blissful dirge]

Yes

[charred merlin]

Always a free user, or did you recently end/cancel your sub?

[teal raft]

Ah, then yeah.
If you are a paid user you can upload your own art and also prompt for Suno to make you new art

[blissful dirge]

I know, thats why i said the previous artworks were also generated by suno

[teal raft]

Any new tracks will get generated art based on lyrics or if its an instrumental, the song style

[blissful dirge]

My country's economy doesn't give me the spare money to get subscriptions for things

I also asked earlier for how should I type for a slow tempo sub 50bpm non-instrumental song

One of my main genres (slam metal) instantly sets the bar to 300 bpm and then slows down a bit

[hollow hare]

You could try down-tempo, maybe even very-slow-tempo? I do not know how effective Italian tempo tags are, like lento or adagio, you could try that as well (see #📚┃suno-school message for the list)

[charred merlin]

Paimon has a brilliant suggestion about using traditional musical terms to describe the "feeling" of the speed, instead of the specific bpm, which suno doesn't quite understand.

Perhaps start the song off with a different style to ensure a slow start, then extend slam metal after it?

[blissful dirge]

Tried [slow tempo] and [50 BPM] today, will try those tomorrow, thanks

Im thinking of type o negative levels of speed

[hollow hare]

To clarify, I meant as style tags. Most times, tempo metatags do not work or not as effectively compared to when used in the style prompt.

Genre has a very large effect on the model. Like harber mentioned, generate a generic slow track then Extend into metal as metal itself would almost always have higher BPMs.

Maybe use FL Studio or other DAWs to make a nice slow tempo intro piece (or even just a metronome or you tapping on something) and use the audio upload?

[charred merlin]

[slow tempo] without hyphens might very well be interpretted the same as [tempo, slow], which might seem like a silly distinguishment to make, but "tempo" by itself doesn't tell it what to do, and "slow" is incredibly vague in an actionable sense. Even hyphenating it doesn't necessarily work as a "magic trick", but can be a helpful way of making Suno tokenize a prompt differently (and therefore understand it differently)

[slow-tempo], [slow], [downbeat], [adagio], [largo], [slow-metal] are some different variations off the top of my head (not suggesting to use all of them all at once, but different ideas to swap out to see how it gives you different results -- after a while you'll develop your own "dictionary" of terms to use)

Also, when you're experimenting like this, check out how v3 (more broad and versatile, as the model selection dropdown says) and v3.5 (excels at song structure) handle the prompts differently

[blissful dirge]

I've been testing since early june but gave it tought only for 2 weeks now

That dictionary has barely a couple pages, i will expand on it

[charred merlin]

I understand for free users it's more difficult to experiment in the same way, as there are only so many generations per day.

I don't mean to dismiss the guidance that people hope for when asking for specific tags to accomplish a particular sound, I know experimenting even once or twice a day means fewer chances to make the music that you actually desired to create.

In the next couple of days I plan to put out an early version of the Suno Simplified workflow -- my brilliant and helpful friends here have helped me learn so much, and I hope I can help turn it into bite-sized tips that anyone can apply to their own specific styles. ❤️ 🤘

[gusty pendant]

Hello, cant find this anywhere after searching trough the discord, can you get Youtube content ID on your songs from suno ?

[marble cloud]

is there a way to exclude instruments (and vocals) from generated tracks? I'm trying to make ambient tracks but all of them include some piano/keyboard, and when trying to generate reggae tracks, even if marking them as instrumental, they automatically include some form of mumbled vocals

[untold briar]

Not a surefire way, but you can add repeated "instrumental" words to the style prompt. This tends to happen with instrumentals in music genres that typically have singing.

[vivid sage]

I saw this and other similar strings in some songs, what does it mean/do ? [>>>4\7/0\6§9/8\3/2<<<]

[untold briar]

Probably nothing, causes a hallucination (random word or scream), or in instrumental, it could do something interesting. Also in general it injects a little 'chaos' in the music, in my personal opinion. So the song might be a little more weird than usual.

[vivid sage]

Thanks

[naive quail]

That one, in particular, is also a required tag for entries into the SSC4 #1258015120697724980 message

[untold briar]

Ahh, a tag for a contest, thanks!

[trail badger]

Friends, have you been having problems for a few days? Sometimes he started producing very irrelevant things. He sings the lyrics in incomprehensible languages, sometimes making the music go too long. I think there is an update to the 3.5v model. It's like a testing phase.

[naive quail]

I had a few mumble/gibberish tracks pop up over the weekend, but it was probably less than 1% of my total weekend renders.

v3.5 has always had an issue with not wanting to end a song - it seems very particular about the song structure and lyrical content for different genres. It'll often seem like it's ending a song... and then pick it right back up. 😄 I really haven't seen much change there recently, although most of the vocal songs that I did over the weekend did actually stick to my song structure and concluded on their own w/o me even using an [End] tag.

[trail badger]

The features you mentioned are what I know, there is no change in that regard. I was getting the same quality results from the songs I produced in the same style. It has nothing to do with these results at this time. I also feel a loss of quality in the vocal sound. To summarize, the v3.5 model is now starting to resemble the v3 model. That's how I feel. Maybe there is a problem with the country language I use. Such a problem may not appear in English. But I clearly see a change. I think that during upgrades to a higher model, testing is done on the current model. To benefit from people's experiences. This is most likely the case. It is a situation that should happen. I just asked out of curiosity.

[mental adder]

admin / moderator here?

[untold briar]

Moderators and Guides are online, yeah

[wheat lava]

is there a way to take the instremental of a suno i already made

[toxic chasm]

i just found out about suno made a song i wanted to make another one but can t do i have to pay for it?

[trail badger]

You have 50 credits every day. You have the right to produce 5 songs with 50 credits. Consumes 10 credits on 1 Production.

[trail badger]

You can find and download the music you have previously produced from the library menu.

[charred merlin]

If you're asking if it's possible to remove the vocals from a song you've made in Suno, it's not currently possible using only Suno. If you search for "stem separation", "OpenVino" or "UVR" in this chat's history you can find some tools to remove the vocals

[earnest ledge]

Can we edit our songs in the Suno app? I was trying to make a song with my voice but it came out with a generated voice. Will I be able to edit it?

[charred merlin]

My response from <t:1720439160:t> should also apply to your question ❤️

You could try to upload another clip of you singing for Suno to continue from (longer is better), or place an extension point before the artificial voice and cross your fingers that the extension might "pick up on" the vocals you're extending from

[earnest ledge]

Hopefully they make it so we can add our voices to the songs that come out with an auto generated voice. I was saying because I wrote my own lyrics and everything and the song turned out really good, for me.

[charred merlin]

That would be so neat!

https://discord.com/channels/1069381916492562582/1252789683176345641 They have this known-suggestion for reusable vocalists, but imagine how could it would be if we could train our specific vocals (perhaps by recording a few minutes of sample dialogue in the same way ElevenLabs does), then use that same vocal model for different songs?

It would be helpful being able to speed up the workflow for normal song creation, but I wouldn't be able to help myself -- I'd have to hear what my voice model would sound like yodelling, rapping, making gibberish noises 😛

[raven turtle]

Does anyone know if Suno recognises vocal "accents" while singing in English. such as "German male voice", "Greek female vocalist", or "British indie female" or am I wasting my time trying?

[silent viper]

Has anyone tried these tags in the lyrics field:

[hyper-modern production with clear vocals, no autotune, Dolby Atmos mix, high-fidelity, high-definition audio and wide stereo]

Did it work?
Share your impressions.

[charred merlin]

I got a few successes with v3, [french] [french-accent], etc. But it's a very specific request, so it can be difficult to achieve with consistency

[raven turtle]

Many thanks. I'll switch to V3 and give it a go there.

[naive quail]

Negatives don't work (yet), so asking for "no autotune" is pretty much the same as asking for "autotune". Most of the modern vocals in the training data are likely auto-tuned/pitch corrected, anyways, since that's become the industry standard these days. It may still be a hard ask even when negatives are supported.

[rustic lintel]

Any advice on stuff like adding slight pauses or extending certain words? I've found that commas, full stops, ellipses etc. often don't really do anything and the AI just rushes through the lyrics as if no punctuation is there?

[naive quail]

I haven't found anything to consistently add pauses, but you can extend words with extra vowels, so "Long" might become "Loooooooooooong". You may have to play with it a bit if the pronunciation gets thrown off.

[rustic lintel]

Yeah I tried that with a "Hiiiiiiim" and it ended up turning into something sounding like "High-em" 😅 Will give it another shot though, ty!

[naive quail]

It can depend a lot on the singer, too. Some may turn that "loooooooooong" into more of a "lune" type vocalization. "Laaaaaaaahhhng" might fix it for them... but then you'll get a different one on the next roll that messes THAT up. It's all a crapshoot.

[solar rose]

What I miss in music styles such as alternative rock, hard rock etc. is the typical electric guitar sound. Unfortunately, almost everything sounds like noisy synthesiser sounds

[naive quail]

I haven't had too much trouble getting electric sounds that are not-horrible, at least. It's the acoustic guitars that drive me nuts. They usually sounds like early 2000's midi guitar Here's some electric. https://suno.com/song/a70f96f3-9c8e-41d0-a016-53808b4b50d2

[silent viper]

OK.
However, everything is a matter of learning and training.
All I.A.s are using us as guinea pigs in a gigantic testing laboratory.
They are taking advantage of our inputs to improve their knowledge and results.
Thus, every type of increment such as Key, tone, chord progression, types of instruments, form of execution that we can add to the prompts will be assimilated and implemented little by little by artificial intelligence platforms.
Remember that at first, artificial intelligence was unable to produce images of hands correctly.

[naive quail]

tho I hope they're not simply feeding everything we do right back into the training data!

[unique stone]

Pretty sure they're not.

[naive quail]

Especially not the emoji songs...

hides

[charred merlin]

Not necessarily true -- a lot of AI products go through a "sanitized" stage of RLHF (reinforced learning from human feedback), but when that data is collected and used without any oversight, or if they allow public content into RLHF, it results in controversies like Microsoft Tay chatbot (4chan got access to it, and they trained it to become a sympathizer of some very controversial and evil political figures)

A more reasonable implementation from what I could imagine would be Suno taking information based on the ... > Report > Bug > ... reports and using that as extra feedback for future refinements.

[unique stone]

Maybe that's why so many vocal hallucinations lately 😅

[charred merlin]

I was wondering why "humma-lumma-nummah, summa-chut London" kept creeping into my songs...

[naive quail]

brace for "crab, crab, crab" randomly being belted out in the middle of opera songs...

[rustic lintel]

Remaking the favourite of my first 5 suno songs using the tips in the pinned messages 📌

[neat nexus]

I'd imagine we'd be deluged with "toasters which were quite insane" and "we choose to go to the moon" if this was a thing so I think we're fine.

[teal raft]

Dang when can I go to the moon Mr. Kennedy?

[frank osprey]

Does someone have a prompt on suno.com to manage the length of the music for the first part ?

[charred merlin]

I'm not sure the length can directly be prompted -- it can be influenced by the style prompt or the metatags, but I don't think you can explicitly request a specific length. At least not that I've seen demonstrated with any consistency

[frank osprey]

Ok, cause a the begenning it was only 1m and after 2m and now 4m. It's getting more difficult to have short music 🙂

[charred merlin]

v3.5 can start off with 4 minutes, and generate 2 minute extensions, v3 can start off with 2 minutes, and generate 1m extensions.

You can try to end a longer song by placing an extension point a little bit before a musical passage ends, so you're asking Suno to reimagine how that section might go afterward, and depending on what style prompts are trying to overpower it to continue a song, you might be able to guide it to a "natural conclusion" by stripping the rest of the style prompts and using "end" for the extension's style prompt

[distant vale]

i can't get suno to extend my songs with an instrumental interlude or solo. it just keeps making vocals. i used to be able to do this. any thoughts?

[foggy phoenix]

Man i wish if suno was able to pronounce some words i can hear it has not been trained on to sound better. Specially on my main language which is Danish. Sometimes even though the setup is on danish it switches to Swedish or Norway language or german with words it cant pronounce which can be annoying in many situations. I suggest for every language you support to get native singers to help you train suno's voice models better. Not to have someone with a slight accent to a language either. Because at times it sounds okay and then you try to do some specific names or cities or something it gets a hard time to actually say it right with the right pressure or way to say it. I guess you do this language switching so it doesnt sound disturbed for words it cant say but i would rather have a wider range of words trained into the voice models.

[neat nexus]

If you're looking for a work around until things can improve I suggest using phonetics to get things sung\said the way you want them to be.

Just be aware this is a universal issue. I still remember pulling my hair out trying to get it to sing lives instead of lives. It kept getting it wrong.

I eventually cheated with leaves, hoping that it would be close enough that when sung it would work. It seems to have done.

[charred merlin]

Phonetic spelling ftw = "she livs a thousand lives"

[neat field]

basic car commercial orchestra experiment is up https://suno.com/song/d66b4e15-2d41-4980-9d1d-9c39d140ce45/

[rustic lintel]

I'm gonna guess that it's nigh impossible to recreate the same sound of a song, but with edited/tweaked AND extended lyrics?

Just a matter of trying multiple times till you get something close to the original sound

[neat nexus]

Are you continuing from something or starting completely over?

[charred merlin]

Not impossible, but often way more work than what it's worth 😛

If you're spending a lot of time working on a specific song, you might start to notice recurring motifs or musical phrases and passages that keep coming up -- it's possible to apply music theory to your song and try to figure out the best place to use extension points right before or after those key moments when those sections end and start anew.

Of course, this isn't an exact methodology and still requires a lot of experimentation and familiarity with the platform. As always, the more specific of a song you want to create, the more time and credits will presumably go into it. Most times it's going to be a matter of finding a happy compromise, imho

[neat nexus]

If you're starting over, your best bet is to consider the original a live performance. Just accept that it was a once in a life time event and the only thing you can do is get a different, also once in a life time performance.

[foggy phoenix]

Ahhh Thank you. Do we think v4 wont require this method?

[charred merlin]

There will always be ways to mispronounce things, even between regional dialects. Expect improvements, but not a simple black and white solution

[foggy phoenix]

Alright going with this in mind going forward Thank you

[neat nexus]

We all cross our fingers and hope. But despite how good it is when everything goes right, it's important to remember we're still very much in the infancy of everything. The good news is that this is probably the worst it's ever going to be and that each version gets better and better.

But given that humans mispronounce words all the time, there's no guarantee.

[neat nexus]

Oh! And you can try your hand at uploads to see if that gets you closer these days.

[foggy phoenix]

Ahhh okay so it depends on the training data for the voice models how good they get?

And even then there can still be flaws?

[charred merlin]

That's my opinion, but AI will always produce better results when used collaboratively along with human intelligence and feedback ❤️

[foggy phoenix]

Alright amazing

[neat nexus]

Training, whether or not Mercury is in retrograde, the position of the sun and whether or not the wind is blowing....

There are all sorts of things that can influence stuff. This is a question that is best asked to a team member when they are around and you can post to them directly. (Not a good idea to ping them.)

[foggy phoenix]

Alright fair 🥰

[devout meteor]

why do my generations not put lyrics even though i have lyrics set and instrumental off???

[foggy phoenix]

I think this got my q's answered Thank you

[neat nexus]

I joke about the other stuff, but the truth is that some of this is random number generation somewhere. So we know that stuff is random at least somewhere. But as I said a team member would know more. We're not all developers here, we're just enthusiastic users of the product.

[charred merlin]

#🪲┃bugs-web message -- UGH, I'm tired, my mistake!

[devout meteor]

do i ask there???

[charred merlin]

No, my post I linked to is an answer to the same question

[devout meteor]

oooh okay sorry i dont use discord very often

thank youuu

[charred merlin]

OH, sorry! I misunderstood!

[charred merlin]

Are you having this issue? https://discord.com/channels/1069381916492562582/1259843144376385576

[devout meteor]

similar except it's a new generation and not an extension

whenever the song generates the lyrics don't appear on the right hand side where they usually do and the song is an instrumental

[charred merlin]

@final hazel seems like it might be an emerging issue based on the amount of recent instances...

[charred merlin]

Please upvote the "missing lyrics and vocals" link that I shared to let the developers know that you're also experiencing this issue -- there's no solution as of yet, as details are still emerging, but every reported instance helps tremendously ❤️

[devout meteor]

okay thank you!

[charred merlin]

I know one of the users who first reported this noted that it happened for a couple hours, then seemed to be fine for a large portion of the day, only to return later. :\ I wish I knew how to advise handling the issue (besides either coming back and testing later or trying to generate different styles of instrumental intros for songs, neither of which are a preferable alternative), but hopefully any information regarding the issue will be shared in the relevant thread so you don't need to keep guessing 😦 ❤️

[hollow steppe]

hi.. is possible to ask ai a female voice on verse and a male voice on chorus or vice versa? i tried everyway but i was unable to make one single song with these parameter

[hollow hare]

Old copy-paste made for v3.0, I haven't tried for 3.5 but the metatag reading is much better:

Getting duets... that is more luck than anything else. Try to use male and female sounding names in tags, that might help, but it is still hard to achieve.

A few people were able to do it but at an expense to their credits:
3Daizy's duet
MrTree's duet
Suno.Wiki entry

Again, getting it working on purpose is hard.

Alternative methods:

1. It is easier to make a duet by doing something like generating instrumental part 1 then Extend it twice (one extension where you specify for male vocals in the styles field, the other for females), or something similar for each alternating part, then edit them together in a DAW.

2. Even easier (and cheaper) is to make a full track then swap the voices gender using other AI tools like kits.ai (see #🗣️┃chill-chat message ) or your own voice with a RVC model (see #💡┃suggestions-web message ).

You can use a DAW to edit which voice track plays when.

[hollow steppe]

tks.. ill look how voice swapper works..

[hollow hare]

Sounds good. Mind you, thst post is also oldish so there may be new services which can do something similar.

Take a look at YouTube, there are a few channels that reviews new AI tools regarding vocal swaps 👍

[modest patrol]

What are the #1 male and female names?

Would like to try them out for voice switching see if it does anything 😛

don't try to paste a wall of text, bro

Yeah just type out the #1 names, I don't need a list of 50

and reformat to not be all-caps

The bot hates caps

gah, I'm getting the missing lyrics bug now :/

Random gibberish not anything in my lyrics

Example: https://suno.com/song/ed6bf10f-8539-45f4-beef-f8fa044e3f61
Lyrics missing from the right side, just random gibberish WAT

guess I'll stop wasting credits for now

maybe it's filtering on suno's end for the prompt? Like it doesn't like proper first names in the [ ] tags?

no... changed to guy and girl and still the same thing

[alpine venture]

pulling is a keyword that idk how it hasn't been used much yet

I just started and already its massively powerful

combined with basically anything. I literally got piano strings.

so plz, help me generalize it lol

bc I want to use the string effect to pull everything like taffy

and make staccato instruments elongaaaaaaate and get rough

but basically its the verb when using stringed instruments

https://suno.com/song/a6179612-5957-4e84-9bcf-6d8ce87c236a

in v2 it actually generalizes better

likely bc it saw no use

but its still IN there

https://suno.com/song/009f66d3-cca3-4b4a-b6f6-4b80925dc85f

inasmuch as it leads to more depth and reverb

which is still a nice effect (and will be further enhanced with community use)

using 3.5 continuations off a v2 open for whatever you want "pulled" is likely to have the best effect

but I really only had a few hundred credits to explore this particular word

(I was too busy in liminal spaces)
https://suno.com/song/761dcd6d-2af7-4adb-bbe3-618e0fb26ad3

[mild rain]

Anyone else noticed that as of late Suno is adding way too much side-chaining on most Synthwave and Synth instrumental creations, it never used to be like this. This just accentuates the currently low quality audio and doubly ruins the creation, song after song is just a pumping mess of FFfffsssshhhhh wishy-washy sound.

[rustic lintel]

Anybody know a way to affect the inflection at the end of a sentence? I.e. upward inflection for question?

[charred merlin]

I heard someone mention this a while back, and I LOVE the idea of using male/female names to help nudge Suno toward an agreement and providing clarification, instead of just repeating "MALE VOCALS, MALE VOCALS, MALE MALE MALE".

Saying "[judy], [dianne], [carol], [jessica], [lindsey]" would be a novel and creative way of re-clarifying what "female vocals" is supposed to mean if there's any degree of uncertainty.

Regardless, some will read this suggestion and instead, they'll get angry at the tool and say "well, it should just understand what I mean by "female vocals" otherwise it's a bad product", which misses the entire point, and would genuinlely make me sad that they're missing out on the "pioneer aspect" of this all.

Thank you for this list, I'm going to put this to good use and share my findings 😄

[charred merlin]

https://discord.com/channels/1069381916492562582/1259843144376385576 Sounds like you're experiencing this emerging issue -- there's no definite details we can provide at the moment, but if you give an upvote for visibility the devs will know it's effecting people.

[modest patrol]

I up voted already, but there are only 5 upvotes. I had the same problem for almost 10 generations. I just stopped trying for the day

[sleek rivet]

Does Suno recognize specified instruments like Yamaha DX7 and TR-808 in "styles of music" section?

[woeful oak]

When is the Suno app going to be available in the Canadian Apple App Store? Using the website from
Mobile is rough.

[charred merlin]

...I'm trying to think of how to say "yes" and "no" at the same time-- Nes? Yo?

You likely won't be able to explicitly tell it what instrument to play, but those keyboard names might be associated with specific styles of music that could be coaxed out by mentioning that specific instrument model.

A vague comparison would be "steelpan" -- there's a chance that it might not explicitly understand it as a request to include a Trinidadian steelpan, but it might suggest that it should make music that is associated with a steelpan (which may or may not include the actual steelpan, as confusing as this feels to type)

Similarly, "moog" might not get you that specific tone of keyboard, but it helps distinguish from a grand piano sound I would imagine.

[warped furnace]

Wouldn't count on it. Sometimes it even confuses piano with violin. Or say, AI favors piano in certain genres and doesn't listen when I wish for other instruments. 😅

[stoic condor]

ok random as hell but was debating on experimental work with the sound bite upload

so say being a gun expert i wanted to use the effect of say a 9mm and dont have an easy way to record a high quality effect so i decide to use a rip from a free to use database

the question here becomes does the stupid, pardon but seriously it kinda is a stupidly worded warning pop up, affect this idea or can i say just proceed as normal

[charred merlin]

#📱┃suggestions-ios message

Very soon, my brother 🇨🇦

[pastel fable]

is it possible to make a duel of instruments? I have seen that it's really hard to put 2 different solos I'm the same track, and sometimes it finishes abruptly the song to continue with the second solo

[hollow hare]

It would depend on the wording of the license from the site you took it from.

If the site had a CC-0 license, you would be good to go iirc.

Otherwise, depending on the license, you may be liable to legal consequences if the original creator decides to if you violated their terms.

(The odds of this actually happening is very low unless the sound taken was unique and could be pinpointed in the original source)

[charred merlin]

The phrasing makes it seem like you can't use open licensed content, and it has to be something that you, and only you, own the rights to.

[radiant cobalt]

This new web design that just went live is actually really bad for my browser creation rn

[hollow hare]

Ah, true, forgot about the exclusive line.

[radiant cobalt]

can we have at least have an option to go back to the old way, this is bad

I can't even click stuff.

[charred merlin]

That being said, "transformative work" is what defines copyright -- so if you make reasonable changes to it... gonna stop myself from speculating too much. I genuinely wonder how much of the safeguards are designed to "childproof" the product to prevent lazy newswriters from saying "this app lets you make songs about hate crimes, or upload stolen content", and where the line between reasonable use is. Maybe reasonable use doesn't apply when it's a closed-source proprietary software.

There I go speculating again. 🤦‍♂️

[radiant cobalt]

Not being able to resize windows that overlap one another is a bad bad thing

[charred merlin]

Does the issue persist when you close & reopen the browser, or try from incognito mode (which acts the same as clearing the cache)?

[radiant cobalt]

YES

Style of music has 'suggestions' that literally overlap the window

[charred merlin]

In #🪲┃bugs-web can you share a screenshot of what you're seeing?

[warped furnace]

You can't resize but you can scroll with the bar at the right. Just found it. 😁

[sleek rivet]

is there a way to tell it to use the full 4 min?

[warped furnace]

you can try to use meta tags like [fade to end] or [repeat]. I fill the lyrics section up to the brim so that I won't get cut too short.

[rustic lintel]

I've found that using the outline from Tinda Lee in the pinned message in this channel, the AI generally will make repeats of some lines in order to fill out a little bit of time before the end of the song. And like Panthera said you can just write more lyrics (adding instrumental interludes and stuff is also nice, otherwise the AI can kinda rush through your lyrics)

[mossy crow]

Also is there a way to tell it to NOT use the full 4 minutes and make something short?

[dusky sphinx]

How do I get drums/percussion only for a section of my track?

Everything I've tried is ignored.

[near ice]

The website is very broken in Firefox; I can't even move the textarea input.

[pastel fable]

have you try

[drums solo]

?

[dusky sphinx]

Yes. Also just [percussion] too

[pastel fable]

I think that Suno needs to offer more control over it

[near ice]

Firefox On mobile

[charred merlin]

Trying to force a drum-only section can be difficult, because it's not simply that you're trying to give it appropriate styles, but you already have an existing muscial theme that it's trying to continue, and you're asking it to disregard all of the other instruments and only use the drums.

It's not a quick solution, but the way I would approach it if I wanted to create this exact section in a song would be to create an extension point before a musical passage ends and attempt to guide it into even a second or two of unique drums, perhaps using [complex-drums] as the only metatag. Once I was able to get a second or two of justdrums, I'd create another extension point to extend from (not getting whole song, because I only want to continue the drum section that's about to begin). Then I'd be more free to play with style prompts without the existing track overpowering the creations, but I'd still probably use v3 to make sure that it was willing to be a bit more experimental and didn't try to "find a groove to fall back into" that contained other instruments.

[charred merlin]

❤️ 🙏 thank you for bringing this to attention

[upper pawn]

Anyone else unable to scroll lyrics in mobile now?

[rigid sedge]

Yeah

[upper pawn]

Lame

[rigid sedge]

Complete useless atm

[upper pawn]

Looks way worse

Font choices ? Please lmk I will email them lol

All love. I like suno and have spent 100$ at this point. I Wish the censorship was better. Premium user. Should not be worried bout phrasing. Then all my edits wasted. I don’t even know what was being censored. Kinda lame

[crisp axle]

Has anyone noticed that since v3.5, the amount of full length (4:00) generation is about 1/10 of the time? I just created 10 songs, all with lyrics maxed at 3000. Style very descriptive, maxed to 120 characters. Out of the 10 generations, none were longer than 2:35 seconds. The same exact Lyrics (prompts) and style previously generated max (3:00) each time.

[rigid sedge]

I have songs between 3:30 and 4:00 🤷🏼‍♂️

[charred merlin]

https://discord.com/channels/1069381916492562582/1260023056013852775 @rigid sedge , I'm sure this is going to be fixed VERY SOON, but at least know that it's on the dev's radar

[crisp axle]

I do as well, but the amount of short generations in v3.5 is staggering

[sterile anchor]

Stupid question, am I blind or is there no way to “re-use prompt” on IOS?

[charred merlin]

Not a stupid question, and you're not blind -- a lot of functionality is only present on the webapp, like resuse prompt and extend.

[sterile anchor]

Thought so. Thank you!

[rigid sedge]

@crisp axle Mostly I make rock songs out of my lyrics and work a lot with Intro, Breaks, Interludes, Solos and stuff

@charred merlin 👍🏻

[earnest ledge]

Ok. I have a question. When I try to costume the song and record my own voice, I get a generated voice. I’m not sure if I’m doing the whole process correctly?

[hard ruin]

The new design of Suno is quite disappointing, particularly the web version. The previous design was much better. The site is now very laggy, and the addition of a [quick Public] option is frustrating, as it often gets pressed accidentally.

remove [quick Public] option is not safe for commercial purpose!

[mossy walrus]

the new UI looks very very bad! not only from a design point of view, but aesthetics as well. with such a big budget, they can't even hire decent designers and engineers? 😕 so disappointing .. some points to consider for improvement perhaps:

1. the fonts and buttons look tacky and ugly
2. the background image used for the placeholder on the right looks disturbing
3. why are the song titles are clipped?! -- a prime example of prioritizing aesthetics over functional design
4. the individual song list items are cluttered with options/buttons packed closely together, making them prone to accidental clicks
5. there's layout shift when the song can't be extended vs. when it can be .. because the "extends" button is completely hidden instead of making it disabled?

[tawdry hazel]

Further, the song image used to be square (a perfect analog for an album cover) and I feel the change to the rectangle is a step backwards from that presentation perspective.

[wanton pike]

Anyone having issues with the time bar/progress bar/tracker at the bottom with the redesign? I can't seem to FF or rewind very well anymore for finding extend points. It like hangs, and I have to click off and on the track again to seek.

[teal raft]

This, I will agree with. I don't like the 'upload a square whoops you only see a vertical rectangle of it! But Playlists are still square' mis-match. I can deal with it, but it is a big ask for more normal people that just want a pretty picture

[bronze helm]

Plenty of thumbnails that depended on the 1:1 ratio are cut off because of this, I don't think it's a well thought-out change
Nothing to lose sleep over, just inconvenient

[languid mauve]

@hollow hare Please lmk if you find a solution

[hollow hare]

Adding more lyrics helps. Or repeating what you had after an [interlude].

Having down-tempo (or slow-tempo, etc.) in the style to slow down the song can help

In my case, I normally pad the end with various instrumental tags, usually so I try to get a good instrumental catch to extend from for a future project.

[languid mauve]

TYm I'll give that a shot. Also The Harber mentioned a convo w/ you about distrokid but I cant find it. Can you help me find that so I don't ask redundant questions por favor

[hollow hare]

Sure, it starts from here and goes on for a few message posts:

#📚┃suno-school message

[languid mauve]

tySM!

[hollow hare]

You're welcome, have fun with your creations 🙂

[languid mauve]

TY. Did you decide to use Mixea services? I paid the 79 fee for my album mastered however Distro offers unlimited mastering for 99.

[hollow hare]

Uhh I really should copy my text before sending it

[languid mauve]

DOH!

Woodhouse strikes again

[charred merlin]

In the message I linked earlier, I show the mixing service I use through Bandlab

[languid mauve]

TY. I'll give them a shot

[charred merlin]

oof, sorry to snipe, Paimon 😅

[hollow hare]

All good, we both are here to help 🙂

[charred merlin]

🤠

[languid mauve]

I'm super grateful to you both

and you patience

[onyx glade]

Hey guys, does anyone know if its possible to create an intro?

Or do i just extend one with 'intro' in the prompt?

And just copy paste occordingly in a daw

[charred merlin]

Different ways of approaching it -- some include lyrics and everything right off the bat and try to discourage Suno from starting the vocals too early by adding metatags like [instrumental-break] [long intro], etc

I prefer to create instrumentals to begin with (still using the lyrics box, but only putting metatags in to "tweak" the style prompts, or try to suggest WHEN to apply certain style prompts), then add the lyrics as an extension later

[cerulean steeple]

Hello. How do i make "i" sound like "i-i-i" not "i i i"?
For example the word "side". If i do "si-i-ide" it sounds like "si i ide"

[graceful condor]

I think u can type it how you say the word like: Sa-a-ayd

IDK

[languid mauve]

I've had decent luck mispelling words as to how they sound. Get creative

[cerulean steeple]

Hm. I'll try that

[hollow hare]

You can stretch out a sound by repeating the letter, like siiiiide

Like jureynn mentioned, you can try phonetically spelling the word such as "eye" for "i"

(A better use example is "they live their lives" where the AI cannot separate how it sounds. Using "they liv their lives" would give the appropriate pronunciation iirc but it may take some testing)

Sometimes adding dashes will cause the AI to try to sound it separately as you likely experienced

[charred merlin]

Figured it could be helpful to expand upon in a section of the workflow, so I thought I'd bug you while I gather my thoughts in case it could be helpful ❤️ (sorry if I was grouchy in the DMs)

Depending how fancy I want to get with a track, I separate the stems in Audacity using the OpenVino plugin, touch up the songs there with basic effects like compression, normalization, high/low pass filters, import each individual stem into the bandlab studio to use advanced pedal effects and their presets, then mixdown as .wav from there & use the bandlab mastering to find the best sound (never used the punchy bass one yet, usually "universal" or "analogue" for my indie tunes, "clarity" can be a bit tinny to my ear).

I'm considering running each stem through the most appropriate mastering filter before taking the knife to them in audacity, but I need to be cognizant of overmixing, because at a certain point it doesn't make the final mix sound any better, and my wife just gets annoyed asking which version sounds better when she stopped caring 50 versions ago 😛

[cerulean steeple]

Tried to stretch but Suno sings it like "siiiaaayd" 🤣
I guess phonetical spelling will be the best way

[hard ruin]

how do we keep the same composition for every verse?

[languid mauve]

This is hugely helpful.. My other half gets incredibly irritated when it's replayed 100x also. Of course the first bandlab i chose was punchy. 😬 I'll work more on making sound more complete

[charred merlin]

Absolutely no right or wrong choice -- if punchy sounds good to you, that's the best one for your music ❤️

I just tend to work on the bass tone before the mastering process, so punchy just overpowered my songs, but I haven't experimented with a huge variety of styles or whatnot

[languid mauve]

It definitely overpowered my track immediately. Wasn't what I wanted. Do you set your release dates for a week ahead or immediately?

[charred merlin]

v3.5 is better at maintaining a consistent song structure, so I'd recommend using that model and trying to keep the same style prompts and metatags for each verse you extend 🍻

[languid mauve]

Consider copy before woodhouse gets you

[charred merlin]

Oh, Woodhouse is a sweetheart. He means well, and there's nothing he can do that a quick ctrl + z into the text field can't undo (relatively speaking)

They recommend 4 weeks if you want everything to release worldwide at the same time, but I just release immediately and let the hyperfollow link they provide direct people as needed. If you set up a free BandCamp page to share your music you can manually edit the hyperfollow page to include that, so people can get access without needing to wait for whatever streaming service to carry it.

Though you just reminded me to look into "waterfall releases" to see if they're what I think they are. edit. not quite what I was hoping for.

[languid mauve]

TY. I set for 1 weekin hopes the majority will release simultaneously. Have you seen your music on itunes / jukeboxes / tiktok yet?

[charred merlin]

iTunes is usually one of the first, but I don't check tiktok.

I think you need to specifically make a facebook or instagram account to link them so you can have your music searchable when people make posts

[languid mauve]

I realistically don't expect any streams or downloads, as I haven't done any promoting. I don't even know where to begin.

[languid mauve]

I'll do this in the AM

[stoic condor]

refer to how i worded my question pretext and why i was asking had go run some errands but exactly my point to why i say it seemed redundant in wording and that being what lead me to even see about the fine lines of it

[charred merlin]

tbh, don't worry too much about it. There's a wild influx of people who are all now able to make the same baseline of quality when it comes to music, so the amount of people advertising their music simply for the sake of being heard is a cacophony.

It might not get wild satisfying numbers, but I'm just casually making my music available and easily accessible for those who are interested -- that way I know the growth is organic and I can track the progression without needing to worry about remembering what promotional stunts got the most inflated viewcounts, etc.

[languid mauve]

This makes sense. I need to remember its for me and I prefer organic reach. I would however like to break the flatline and get a little something to get recommended or shelved

[charred merlin]

distrokid has a little game you can play each day to get your name on their community spotify playlist, as well as a "swipe left or right" song voting thing, a couple ways that you can get your content heard without "gaming" the system.

https://discord.com/channels/1069381916492562582/1260045620538904628 I think this could be a neat change to make songs shared in the #🐣┃suno-showcase stand out a bit more, as a way of improving the visibility

[solid tundra]

I saw @onyx glade ‘s question about an intro earlier, and I was wondering if anyone’s had any luck “pre-extending” a song. I can do some light music editing. Is there a seed available in an API, or anything? I know this is the longest of shots :p

[river nacelle]

Hello everyone, please tell me why the text input field is not active? You can choose the style of the song, but you cannot enter your own lyrics.

[pulsar oxide]

i have a custom beat mp3 file is there a way i can use the beat and use the ai for lyrics?

[dapper fulcrum]

the new text style looks like the default and cheap, and the stretched text and pictures generally lead to a design defect, the logo is also a defect for the reason that it looks bitten and careless.
your previous design was the best solution, as was the old logo on the server here. fire the person who is ruining everything for you.
Sorry for my English, I expressed myself as best I could

[pulsar oxide]

agreed

[outer dust]

Hello! Thank you for the new design, very “interesting and unclear (something has become).” But here’s what I’m talking about - in the description of a song, if the “NOT CUSTOM” mode was used, the “initial requests” on the basis of which this song was made are no longer displayed. Was this done intentionally? Why am I asking this? - For example, I had already accumulated a very large “library” of my songs, and it was those “descriptions” that I used as templates (the styles were indicated there) when creating new songs. And now I don't see them. Will this “option” be returned?

[foggy phoenix]

Why am i getting no sound on generations???

Its all muted

Ohhh it was my bt speaker lmao never mind

[lyric anvil]

Is suno using the wrong major version for V3.5 song requests at the moment?

Noticed what seems like a quality degradation today when using full song prompts on 3.5.

[untold briar]

I'll check it out, hmm

[lyric anvil]

Thanks, burned alotta credits on poor quality songs 😛

[torpid blade]

Is rabbit allowed to use Suno via web ui

[haughty pollen]

Hey wanted to know how to create instrumentals in suno 3.5. Every time I hit custom and instrumental it adds it own lyrics.

[torpid blade]

https://www.rabbit.tech/support/article/rabbit-r1-features

[ashen onyx]

The new interface was completely pointless. You probably have different tasks at the company, but if I were responsible for development at Suno, 100% of my focus would be on improving the sound quality of extended songs. Put all other development aside and fix that first. Everything else is basically just cosmetic right now.

[lyric anvil]

A multi device compatible ui is pretty important for growing their product, and ui devs dont generally do AI learning/data science so they arent wasting resources.

Your opinion is pretty entitled and uninformed 😦

[ashen onyx]

As I said "You probably have different tasks at the company".

[torpid blade]

Why not just do it together

Not saying they are focusing only on UI
But they could be doing both but pushed the UI frist because they are done

Sound quality takes time like tuning the model and what

@ashen onyx

Your idea would be like moving the whole team to that thing

But however not every member of the team would be machine learning specialist

So I would not criticize them releasing new UI first but hope the team gets song quality done quick

[graceful condor]

I didn't get new credits today, I usually get new credits each day without fail

[torpid blade]

It takes 24 hours

So every 24 hours

[ashen onyx]

Since the new interface basically didn't provide any improvement for the user, such an action gives the impression that they are focusing on the wrong things in general. An interface is the outward face. It should not only convey design and function but also the underlying connotations. I have worked with interfaces for 35 years, so I know how important this is. In this case, they just released a worse version of the previous interface, or at best, no improvement, and as a customer, I get the impression that they adopt the same kind of culture in all other operations as well.

[graceful condor]

i have been using Suno every day since April

every day I get new credits at a certain time, each day the credits come in with no problems until today

past three months

no problems for three months

over a thousand songs

[torpid blade]

Well there is actually another bug on top that might affect you

[graceful condor]

every day I do this

[torpid blade]

#📚┃suno-school message

@ashen onyx it's just sent today so

They might be fixing it still

[ashen onyx]

Well, poor quality on extended songs is an old concern.

[elfin robin]

I'm sure I'n not the first one to point it out, but Suno's new interface is a step back, not forward. The lyrics area is way too small, and texts and headers are way too big, that's interfaces design 101. And although you may find it subjective, the fonts are absolutely out of place. And besides that, not much changed as far as I can see.

[tacit aurora]

I came to report problems on the website, don't tell me that's an update

[drifting mauve]

suno chirp bot

[open crown]

is this the right app?

[azure crane]

Why such a large Create button? It takes up a lot of space where you could instead show several tracks in the library, make the player bigger which makes it easier to jump back and forth etc... It's also way too close to the player's slider…

[azure crane]

Nope… This is the one: https://apps.apple.com/us/app/suno-make-music/id6480136315

[patent saffron]

1

[lethal egret]

Is possible to export a fully master wav? Is there any prompts for it?

[charred merlin]

I think you might be asking if there's known prompts to use in Suno, similar to how some people use "best quality", "mastepiece", "8k ultraresolution", etc with image generation.

I haven't seen any indication that this works any differently from any other prompt decoration, like "sleepy", "winter", "dramatic", but you can automaster your downloaded .wav files on Bandlab.com/mastering

[heady parrot]

is it only me or this new UI looks kinda ... squished

[charred merlin]

#💡┃suggestions-web message
Different results with different screen resolutions and zoom levels, I guess.

I'm cool with it, but I know not everyone sees it the same as I do

[heady parrot]

ye but yk i technically should have "default experience" looking that i use the most default thing you can imagine aka -> chrome 1920x1080

[steel crescent]

Hey guys, im having trust issue for something i got in my mail, is it legit?

[languid mauve]

I'm set to renew today but it hasn't processed yet, What time does Suno renew monthly membership at?

[languid mauve]

It's legit, congrats!

[hollow hare]

IIRC, people say it renews at the time you first signed up.
So if you signed up 2:00 PM, it will renew at 2 PM

Remember that credits refreshes, not carry over, so use up all your remaining credits

[charred merlin]

God, I feel so drunk with power when I have 3 days left to spend 1000 credits.

[trim glade]

dont understand what happened with Suno.. I have try to make a song and always suno make it to Rock with the same Voice.. damn When i try to change the Style of Music nothing happens.. always the same damn voice

[trail badger]

I have been saying for a few days that the v3.5 model does not produce quality output. I saw that the site was updated today. The V3 model also gave good results when it was first released. Then it stopped giving good results. They later updated the site. I don't know if this is a coincidence or not. However, it is quite interesting that models do not always produce good results after an innovation. It gives results as if we were using the v3 model under the v3.5 model. The reason why I did not mention this is to support the identification and development of the model, which is in alpha version, by identifying its problems through users. It is for informational purposes.

[charred merlin]

🤷‍♂️ "Does not produce quality output" might be a bit subjective -- it's fair if you're not satisfied with your recent creations, but people are making amazing content literally every second. The v3.5 generations I've made recently have been crisp audio quality, and have had a variety of wild & beautiful results.

[trail badger]

Let me explain it this way. I have never had any bad results since the v3.5 model came out. I'm not talking about not liking the rhythm of music or not liking vocal harmony. It depends on that person's taste. I always produce in the same style in the v3.5 model. And as I said, it always gave the same good quality results. I don't like the music, I don't like the emotion, that's a different thing. The results of the songs I currently make in the same style are of very poor quality. Just like the outputs in the v3 model. I no longer feel like I'm using the v3.5 model. If there are multiple servers, it may be related to the server I am on. Or it may be too busy with the iOS application. I don't know exactly the reasons. But as a result of making a lot of songs and getting good results, I know that I don't get the same results now 🙂

[final hazel]

Can you explain why you had trust issues? What could they have done to make you trust the email better?

[zenith panther]

Have any admin to help me? I need help with my billing

[final hazel]

For billing issues you can mail [e-mail]

[charred merlin]

Admittedly it sets off alarm bells to receive an email claiming that you've won a prize, and you just need to give them information to collect. A lot of employers intentionally try to test their own employees with phishing attempts, so I think having an unprompted email still gives someone pause, even if they were potentially expecting an email.

I think people might have been less uncertain if there was a public announcement that winners would be announced, or what potential emails might look like.

[final hazel]

I personally know all the signs since I've been running a webshosting company for over 20 years. And besides the prize and maybe the google form, I didn't see any red flags.

[charred merlin]

For us tech-savvy folks it's a clear green light, but if I could imagine my grandmother making a reasonable misunderstanding, I'd rather air on the side of overcaution than considering it a failing on their part. Internet safety is horrendously underappreciated, so even if it seems a bit misguided I'm happy to encourage that sort of critical analysis.

Better to get a second opinion about a weird feeling about an email instead of assuming "malwarebytes will probably save me", or mentally letting down your defenses because "you recognize the sender" if you don't have a broad familiarity with tech to look further into it yourself.

I didn't know Daizy was a web-pioneer 😮 -- so cool! 😄

[steel crescent]

Its just that i dont trust any mail so if i have to enter details about me, i double check so i dont lose everything 🫠

[final hazel]

Gotcha!

[ashen onyx]

The fourth or fifth extension is pretty always distorted.

[charred merlin]

Are you using "get whole song" to extend from, or are you just extending from the extension each time?

"Get Whole Song" makes it so the extension uses the whole composition as context for continuations, ensuring that they're more likely to fit the larger "musical narrative".

Extending from extensions that were extended from extensions is like making a photocopy of a photocopy of a photocopy of a--

[mild rain]

I totally get what you are saying and am experiencing the exact same thing. I only create Synthwave or Rock but the last couple of weeks each creation is really low in audio sound quality as in overdriven distorted washy, grainy low resolution sound like every musical element is running super hot, also as of late so many tunes seem to have some sort of side-chaining effect which is terrible. Melodic elements on many tracks don’t sound melodic, a lot of instrumentals just repeat all the way through without changing chords, I’m getting crazy nonsense vocals on what should be instrumental. v3.5 or v3.0 isn’t producing good results for me at the moment. 478 creations and only 4 kind of workable creations isn’t coincidence… listening to tracks from a couple of months ago in quality and composition seems a lot better than that of today. I am also getting a tonne of creations only seconds long but not refunding the credits and many being just over a minute long when I thought v3.5 was supposed to be 4 minutes. Seems to be going backwards as of late. Really hoping v4 sorts all this out.

[warped furnace]

Just now, I'm with @trail badger on that. The quality of my last tries is abysmal at best. Doesn't follow style, doesn't follow lyrics, sings gibberish and so on. Seems no longer to able to mix genres in an interesting way ... Hopefully it's just temporary (and due to the update). 🫣

[torn trout]

Am i the only one who thinks this is kinda annoying?

like i cant Drag down the window on Style Of Music Anymore, But i can do it on the Lyrics Window

[warped furnace]

You can scroll. The bar appears when you put something in the box. But I liked it better before. Don't need an enormous box for the title. 😅

[torn trout]

Yeah i agree the old one was way better at least it would let me Read what i was Looking at correctly, Hope someone Fixes it soon tho

[vivid pivot]

Why did the website change? How do i get to any of the menus? Whenever I refresh I see the menu then it hides into the side, I dont even have any controls to control music... Was this tested before release or is this intentional and Im missing some obvious button?

[charred merlin]

Out of curiosity, are you using "Tablet Mode" on windows?

[vivid pivot]

Unfortunately no

I did get it to work as a workaround lol. If i refresh and very quickly click one of the menu buttons before it goes away I can use it lol

[little cargo]

is there anyhay to go back to the older themes..at first visually the new theme looks nice... but it has design flaws that make it look like crap - text hard to see bad contrast in some locations..

im on a pc im sure it looks great on a phone

poor scaling.. the lyrics now are smaller

[charred merlin]

https://support.microsoft.com/en-us/windows/manage-cookies-in-microsoft-edge-view-allow-block-delete-and-use-168dab11-0753-043d-7c16-ede5947fc64d#bkmk_deleteallcookies I'd try clearing your cache for Suno to see if that quirk is resolved, because I haven't seen anyone else reporting a missing sidebar as of yet.

[charred merlin]

Not as of yet, but soon there will be a #1252298754787377225 for having the ability to go back to the classic view that you can upvote for visibility

[little cargo]

the black and white buttons.. are... they gotta be fixed

maybe blue and white

[vivid pivot]

Hmm still the same. I also just realized I can goto suno.com/create

[little cargo]

ive been making websites for almost 2 decades... im not just some noob spouting 😛

Is there a index of artists yet?

[vivid pivot]

Ill mess with it later tested from a different computer and it does seem to have the menu there

[charred merlin]

https://tenor.com/view/suspicious-confused-worried-urkel-curious-gif-5837124

[nimble nymph]

I've lost a lot of credits trying to force suno to use a female or male voice in the lyrics and it works very randomly and often doesn't work at all. I tried to use e.g.:
[Verse - Male] or / and [Verse 2 - Female] but as I wrote, sometimes it works and sometimes it doesn't and I can't find any way to control it specifically.
Can someone explain how this should be properly formatted to force Suno to choose what voice to use in the song?

[charred merlin]

No guaranteed method, but there's been lots of discussion on the topic in this channel if you search back for "female vocals" as an example ❤️

[warped furnace]

It's almost impossible. Tried everything. It happened to me sometimes when I put the tag "mixed choir" in the style section. 😁 More often than with tags like [male voice] / [female voice] in the lyrics. But no guarantees.

[ashen onyx]

Ok, do you suggest that I use "Get the whole song" from part 2 and continue to extend to part 3 from the Part 2 whole song?

[charred merlin]

It doesn't cost any credits to Get Whole Song, and this is where I'm really looking forward to some of the new organizational stuff they have cooking up, but you should notice a difference between extending from Parts 1 & 2 as a Full Song compared to just extending from part 2, and the differences will just become more pronounced.

That being said, extending from an extension (without getting the full song) can be powerful if you want a section to be completely different from the earlier song.

example - Parts 1 & 2 are a funky groove, but we want to swap into a heavy metal breakdown, but directly making an extension always has too much funk. We could create a transitory part 3, only intending to use a couple seconds of it, perhaps stripping all existing metatags and style prompts and using metal-leaning prompts. Once we have an extension that "breaks away" from the funk, we place another extension point a couple seconds into Part 3, which is a VERY short clip intended solely to aid in blending the existing song to our desired extension (that admittedly falls outside of the cohesive groove that Suno is trained to stay within so rocks songs don't suddenly become opera, but this is how we go about working around that issue if we wanted that mashup)

Now when we make part 4, we extend from Part 3, ignoring Parts 1 & 2 (they'll still be brought back later if we Get Whole Song). Now we essentially have a "seed crystal" of what we could turn into a metal song from Part 3. After Part 4 hammered out the metal the way we want, once we say "get whole song", the full track will have a somewhat cohesive understanding of how to continue the mashup between the two, though any extensions from that new whole song would likely continue to fall into the metal style unless you took steps to push it back toward funk.

[hoary plover]

Guys, the new design is an absolute disaster! 😫 It’s also being discussed in our Facebook group, so make sure to check it out there too. Please fix it ASAP! 🙏🏻

[charred merlin]

https://discord.com/channels/1069381916492562582/1260263379243761776 Let your opinions be heard ❤️

[dusky quartz]

New update changed how vocals are mixed IMO, Wide+Reverby, probally to make up for the instrumentals, Mixes are always WIDE. Dunno about this

[obsidian escarp]

new design is horrific. if it's not fixed soon, I'm canceling. what an awful interface.

[dusky quartz]

Same, It's heavily on the mobile market now.

[warm cipher]

Is it possible to give the user more control over the result through tags? For example. I'm losing a lot of credits because the model is deciding to repeat some verses I think should not be repeated. (Well, I lost 1100 credits because I couldn't spend them before the month went away and that hurts, but this is a different issue). Instructions about instrumental parts duration could help a lot. I also think that it's very important to allow to generate extensions that modify just a time interval on the previous part. I've lost a lot of credits because a very nice fragment failed to pronounce a single word and creating new fragments from the second before the failure solved the problem but ended in worse quality fragments.

Allowing the user to set up the band members, I mean, selecting how many vocals (and their gender) and the instruments to use would also be a killer feature.

Can the model accept context songs? Like in MidJourney you can use character images to make them appear in the new generated images, it would be so good to be able to generate songs using "the band" on your previous songs...

[unique stone]

The only update was to how the website looks. The models haven't been changed.

[dusky quartz]

Dunno my ears got changed then

[dusky quartz]

Let me remix a old song and see the vocals, (didn't try)

[dusky quartz]

No it is definitely the website then because the song audio just sounds different

[unique stone]

I don't know why that would be. It was just a "reskin" of the website. It was available to alpha testers for a while with the same functionality. The audio generation is an entirely separate thing.

[hard ruin]

yo bring old desing 😠

It's a serious shame billion dollar company updated cheapest web flow

[loud copper]

this new font is hideous

[charred merlin]

@hard ruin @loud copper Not much we can assist with in regards to the site design here in tech-discussions, but they've recently added #🚧┃suggestions-layout if you want to share your insights ❤️

[slender trellis]

Does anyone know if you can edit the printed lyrics of the song in the Suno browser interface? Not the actual song, but the print out. I have lots of errors where I merged extensions in the audio resulting in mangled printed lyrics...

[lyric anvil]

There is a bug reported above.

[remote cedar]

Anyone found a good way to get different voices singing each verse? and then maybe a duet for the bridge and chorus. I've tried [voice 1] instead of [verse] and that had zero effect

[loud gazelle]

I used something like this in the Lyrics:
[Storyteller][Female Voice]

[Verse 1][Male Voice]

[Break][Distant Female Voice]

[Pre-Chorus][Female Voice]

[Chorus 1][Male Voice And Female Voice]
as well as "male vocals" and "female vocals" in the Instrument / Tag window. Not working 100% - sometimes AI does AI things, but 1 out of 10 creations had the mixture I want. Some others said to use like male or female sounding names in the lyrics to get better results.

[remote cedar]

awesome, i'll give that a try

[tender schooner]

there’s a problem with pronunciation. I don’t know if to put this in the iOS or browser bug because it’s not really a UI issue. I’ve noticed this problem months ago and it’s persisting.
Some words like “geek” it mispronounces. It often will say “jeek”. Occasional pronounced it correctly but it’s more often and not it won’t. Spelling it “gheek” is still not going to pronounce it correctly.

I have a discord friend who I often make funny songs about and geek is in his username.

[timber adder]

hey guys ...what happened to the layout ? its terrible in my opinion. the to big buttons with "create" and "upload audio" takes so much space and the worst is that the cursor of timeline doesnt work anymore. i cant skip the track on the spot i want... why did they do that ... ? it was just perfect. its like on youtube/google where they try to make things better but it become worse...

PLS change the design ASAP i cant even use the cursor to skip a track ... it doesnt work !!! 😦

[final hazel]

Please leave your feedback here: #🚧┃suggestions-layout

[river furnace]

Hi, i have a quick question - im trying my best to generate the most simple and minimal tracks that preferably are just keys or guitar - what im really trying to achieve is suno not adding in a bunch of crazy leads and solos ove the entire track, is there a way to get it to leave the leads out and keep it chords only or really simplify the output?

[charred merlin]

The only musical term that I can think of at the moment to use as a style prompt or meta tag would be [accompaniment], making sure that you’re not adding too many other style prompts to overpower it before you have a solid “simple” track to extend from and build on

[river furnace]

thanks so much, i will give it a try

[coarse oak]

Idk if this is the right channel but I went on the website and saw all of my stuff was deleted

Not even in the trash section

[merry herald]

Hi all. Is there a mechanism to get Suno to follow the flow of the lyrics as written. What I'm running into is Suno repeating verse and chorus, as well as choosing when it decides to use Bridge and so on versus following the lyrics as written. I'm using tags such as [Verse] [Chorus] [Bridge] [Outro], etc. to control flow. Any thoughts and/or suggestions would be greatly appreciated. Thanks

[charred merlin]

I can't find my song in my library! What do I do? thanks to paimon for this guide

First, reload the page and ensure you are in the library. For cache-bypass reload, see #📚┃suno-school message

For usual causes:

• Could be on the next page (blue circle for pagination)
• A filter is enabled (yellow circle)
• Accidentally trashed (red arrow to access the trash bin)
• You are in the create screen, not library, and the piece was down voted (which causes the entry to be hidden)
• The text filter option can be finicky. Try to use as much of the full title as possible or unique keywords if the first attempt failed.
• Accidentally overlooked the entry (with text filter disabled). May need to reload the page first.
• Logged in to another account (can double check by going to your profile)

https://media.discordapp.net/attachments/1102676217687195679/1246851079182880808/image0.jpg

[neat field]

did some very limited testing of
Gaelic-dialect prompt today

seems less strong dialect than "celtic" for my purposes with v3.5. just leaving a note here for easier Discord searching

[silent viper]

The conception, arrangement and musical execution are cool. Better than competitor Udio. Therefore, I opted for a paid Suno account.
The end result is pleasant.
However, when you separate the voice/instruments for a slightly more serious production, you will notice hums, hisses, distortions and artifacts that discourage any earthling.
Not to mention the voice full of autotune that is almost nauseating.
It's that clean audio we're hoping for.
At least Suno.com had midi output as some platforms already provide.

[charred merlin]

Break the song down into smaller chunks -- the structure tags aren't an explicit instruction and only provide a suggestion of guidance to Suno.

If you have a more specific idea for a song, it may take more than a "one-shot" generation to create it, requiring specific placement of extension points after the sections you want to keep, or before sections you want to discard.

Rule of thumb, the more specific of a song you have envisioned (specific lyrical flow or delivery, particular beat or subgenre, etc), the more time and credits will likely need to go into it.

[merry herald]

Ok. Thanks

[thorny yarrow]

Please roll back to the version before the update (2-3 days earlier).
The singer sings like a robot, and then adds the dialect (not standard accent)..
Please, I want to create more professional stuff..

[charred merlin]

No updates have been made to the backend as far as we’ve been made aware.

[thorny yarrow]

I saw the new version of website, new voices added, and the results when making songs were bad 😦

[languid mauve]

Would Suno ever consider providing an option for artists to make our music style prompts private?

[bronze helm]

How does everyone else go about making sure that vocals don't suddenly change depth and sound more masculine or more feminine? Sometimes they sound like a blend even. Style prompt starts with "female vocals" at the very beginning, and yet they often just slowly degrade to more masculine vocals, as if a second singer sings the next verse (this is v3.5)

[final hazel]

Sorry, but I think you are misunderstanding. Only the website changed. Nothing changed on the model, nor are there voices added,

[charred merlin]

If I get my way, yeah. ⏳🤞

[warped furnace]

I had similar problems and read here somewhere that others had them, too. Changed [bridge] into [interlude]. That helped. Don't know why. 😁 Tried Bridge again and the song started to reiterated the verses above. So I'm sticking with interlude. ^^

[thorny yarrow]

Let wait and see how many unhappy reports from the user you can get from the users. I guess maybe double or triple to 2-3 days ago.
Anyway, i stopped to make the song today. Will back soon.
Best regards,

[merry herald]

Thanks for the suggestion. 🙂

[final hazel]

I am just a volunteer moderator, sharing knowledge. Have a great day!

[snow cairn]

does paid version have better voice quality or its same? From yesterday, the voice quality has gone done.

[charred merlin]

Same

[languid mauve]

This would be amazing. I can't see a single artist that would NOT want the CHOICE

even if Suno discouraged, the default could be on, with option to disable

[mossy crow]

Is there a way for me to use a discord bot or other means to play Suno songs on my discord?

[final hazel]

There is no bot anymore, chirp is removed.

[mossy crow]

Ah, shame. Thanks.

[cobalt rivet]

ok is suno working? cause ive create a song and it doesnt play

[hollow osprey]

All my songs are showing on my profile page as “Untitled” now. When you click into the song it’s showing the correct name.

[bronze tree]

Nice updates; I'll put them through some tests and let you know if bugs were squashed on my end.

[runic bluff]

please come back to the previous layout. new one is ugly.

[charred merlin]

Can you share an example?

[charred merlin]

Working for me 🎶 Maybe close & reopen the browser?

[cobalt rivet]

yeah i dont know why it works now

thanks

[knotty sage]

Hey, is it intended that a song created a while ago can't go to trending even if it got 30 "likes" in an hour ? Does the song has to be newly created AND be liked multiple times in a short time ?

[charred merlin]

To my knowledge, I don't think the algorithm behind the trending page has been made public, unless someone's datamined it.

[austere dragon]

is there a technical way to move the progress bar of a song so I don't have to listen to the whole thing over and over again to figure out where to extend from?

[charred merlin]

The progress bar functions normally after the song has finished processing (when you can see the extend button, and add it to a playlist).

It was functioning as expected during testing, so I imagine that'll be a quick patch for them when they release the first update to the new UI ❤️

[knotty sage]

Ok, it seems to be the case, i have an old song that got a lot of likes today in a short amount of time and didnt get to trending so i guess it has to be a fresh song.

[astral flame]

i have a song it keeps making a rap and i want it singing, any suggestions? the only thing that has changes is my lyrics. the prompts ive used before

[short path]

can you play songs you make with discord bots like musico or maki?

[lament coral]

Are there some musical instruments that SUNO does not recognize?

[final hazel]

yes, mostly niche ones.

[lament coral]

trying to get a Hammond organ as used in rock bands in the 60's. any suggestions?

[fossil reef]

What about comments and daily interaction tasks?

[charred merlin]

https://discord.com/channels/1069381916492562582/1257721892836474930 -- it's written with a different explanation, but I'd upvote this for visibility for the daily tasks thing ❤️

I think comments are unsure because they would require additional moderation, but I love the idea of an iMessages reaction selection!

[fossil reef]

Do a possibility to report comments, let's say for community moderators, let's say they check report they get 1 credit. Or just do like profanity filter. I don't think it would be too wild 😜

[naive quail]

Hammonds have kicked my butt in Suno. I can get them with something like "Tonewheel Organ, B-3, Dave Banging on Stuff" in the style, but adding "Funk" or "Blues" or pretty much anything else and the guitar becomes the dominant instrument (rightly so, but still...)

https://suno.com/song/4de5eb6d-4737-4a58-8285-6e5534d97d0d

[lament coral]

Thanks, SixString

[naive quail]

No prob! You could do something like that clip and experiment with adding styles to tags within the lyrics box.

[graceful condor]

The music I want to make keeps ending at 3 seconds

I wasted alot of credits trying to make it works

Please help asap!

All of my extend of any music ends at 3 seconds !

[hollow hare]

Erm, don't you get refunded if you generated clips that had like < 7 sec of length?

We can't help you much without further details. What prompts are you using? How long are the lyrics? Etc.

[charred merlin]

Well heck -- I thought it was 5 and under, but you're right, it's 7 and under that refunds the credits back.

[hollow hare]

It was a wild guess on my part. I think I had a 6 sec clip refunded once (do not want to dig through my trash to confirm that though lol)

[charred merlin]

I'm looking at a 7 second refunded generation and a 9 second regular generation, so I'm leaning toward Great Paimon.

[unique stone]

It's usually anything under 8

[radiant saddle]

Does anyone know how to generate vocals only?

[rustic lintel]

Gonna put this here as it's quieter

😩 Just made a new song and the AI decided to repeat verses and choruses so much it took up the 4 mins and just cuts off nowhere near the end of my actual lyrics LOL

Actually both songs that generated did the same thing, repeated parts of the lyrics that shouldn't have been 😖 This is the first time it's happened with these lyrics, and I've used em several times already haha

[rough acorn]

Is anyone else having trouble with generating instrumentals now? Before the update it might once in a blue moon generate something just under 2 minutes but usually gives me a full length song. Now literally every instrumental I've tried to generate lasts less than 1:45 with half of them being just over a minute.

[real falcon]

I have that a lot specially today, had a song where a verse was repeated 4 times

[real falcon]

have been specially bad recently, completely ignoring if I want 1-2 instruments sometimes even use others are ignoring the instrument I want to use

[rustic lintel]

Yeah something seems off tonight compared to a day or two ago

[fluid gale]

Why when I do extend the second part comes out for 3 seconds?

[real falcon]

I've never trashed as many songs yesterday and today as ever before 😦

[weary briar]

do you extend from a part with no sound?

[fluid gale]

I do an extend from the end of the song.
Do you think that because the end of the song is quiet this is happening?

[weary briar]

yes, if it's a quiet part it stops the song 99% of the time

[fluid gale]

I'll try to do it a little before the end. Thanks for the help mate 🙂

[weary briar]

np 🙂

[split dagger]

This is a newbie question - but from reading here I think it is not only newbie:
I have gotten a lot of good results, but they are almost never what I ask for. Wrong gender, wrong style, wrong instrumentation. And then, as if just to keep me on my toes, something will be exactly what I ask for.
It does make it embarrassing to post the songs with style sections that call for one thing and something entirely other is on the track.

Is there any place where prompts-that-work are taught?

[hollow sapphire]

🇨🇵 When using prompts, is it possible to include anything and everything, like [Citation] or [Whispered] or [Duet 1] / [Duet 2], or any other directive? Does it actually have an impact on the construction of the soundtrack? Thank you for your answers! Because the ideas are numerous, but if the prompts are well-defined, then imagination is restricted.🙄

[buoyant loom]

hey, I think it's a real shame that since the update of the new design you can no longer add a genre to the song description but can only write the lyrics.

[torpid merlin]

I'm trying things like expanding the lyric prompts just a little. Like [lead melody], [chorus, no beat] or [drop, powerful melody] etc. Seems to work, but since prompts are a bit hit and miss anyway and I'm running on free credits, it's hard to tell. It does NOT work if you type in longer sentences like [bridge, melody with a new orleans sound] - it will speak/sing it as lyrics in the song.

[torpid merlin]

It looks like this might be the right place mate, if you're looking for a conversation at least. I know there are some Youtube tutorials made by Suno staff too.

[shut moat]

are there specific commands or weights to the prompt in the style of music? or is it just comma and order of the words that control what kind of music that comes out.

[torpid merlin]

weight definitely, if you want a style to be featured more prominently you need to put it first, and maybe repeat it or re-word it

[shut moat]

Like the keyword burlesque in the front of a prompt compared to it being in the back end makes a world of difference.

iun the back it makes more slam poetry kinda songs if you mix it with avant garde post <insert music>

[torpid merlin]

Weird thing though - you know how non-custom songs use just a written freeform prompt? You can start your lyrics with a prompt like that, and it will be ignored as far as lyrics go, and probably help with the style. It's helped me set the proper BPM at least.

[shut moat]

more testing is required I want mix music that has no earthly business with each other I am wondering if there is a weighted command, since order gives the 1st more weight than the last words.

i start with generated prompts then translate them into a more simple language then translate them back into a complicated language.

indonesian is a good 1st translation since there are no tenses in it.

I also had chat gpt write a list of 100 words that were not related to each other.

to see how the bot speaks a script of words that doesn't flow.

I also tried tensor calculus and particle physics equations to see how the bot would interpret them as a style of music as well.

it was mostly electronic.

but the lyrics, well it made it a very tinny voice and synth sounding androgenous voices for the lyrics.

[torpid merlin]

I think this is random, I can never get it to stick to a certain voice

[shut moat]

I just started this bot 4 days ago, im just throwing everything i can think of at it.

kinda like my midjourney stuff. I like sending it conflicting information, although my understanding of midjourney is far more technical since I've been using it since it's infancy.

every time i use pure math androgyny is the result.

[torpid merlin]

Hmm, when I wrote lyrics featuring the word London, the vocals switched to a british accent. Might be related haha

[shut moat]

1 r sin ϑ ∂ϕAϕ; ∇ × A~ =  1 r sin ϑ ∂ϑ(sin ϑAϕ) − ∂ϕAϑ
, 1 r sin ϑ ∂ϕAr − 1 r ∂r(rAϕ), 1 r ∂r(rAϑ) − ∂ϑAr  ; ∇A~ =       ∂rAr ∂rAϑ ∂rAϕ 1 r ∂ϑAr − Aϑ
1 r ∂ϑAϑ + Ar
1 r ∂ϑAϕ 1 r sin ϑ ∂ϕAr − Aϕ r 1 r sin ϑ ∂ϕAϑ − Aϕ r tan ϑ 1 r sin ϑ ∂ϕAϕ + Ar r + Aϑ r tan ϑ       ; ∇ · T =  1 r 2 ∂r(r 2Trr) + 1 r sin ϑ ∂ϑ(sin ϑ Tϑr) + 1 r sin ϑ ∂ϕTϕr − Tϑϑ + Tϕϕ r , 1 r ∂r(rTrϑ) + 1 r sin ϑ ∂ϑ(sin ϑ Tϑϑ) + 1 r sin ϑ ∂ϕTϕϑ + Tϑr + Trϑ r − Tϕϕ

this the sort of thing im throwing into the prompt box.

or i make up words or mix them together, im going to try emotion next. then color, how is color intreted into sound

the bot has a bias.

i want to learn it.

when i know how the bot interprets these words then i can sort of control it. I think.

or im just talking out my ass, im no musician.

[torpid merlin]

TEchnically, neither is Suno lol

[shut moat]

it's bult on countless musician's knowldege.

im an artist, and Midjourney knows artists, but it doesn't know art. so maybe that is similar with this bot.

[torpid merlin]

I'm curious if the vocals can be made inhuman, since there's probably no training for it

[shut moat]

what if sounds like grunts and screams and hums etc etc were phonetically spelled.

[torpid merlin]

Doesn't work, tried it

Common vocal sounds like coughs etc work, but not expressive, performative vocal grunts, screams and throat noises

[shut moat]

what if you mixed languages together?

gutteral words in dutch russian and german, with higher pitched languages. i bet you could ask GPT to write a mixed bag of words and languages.

Im not too well versed with GPT. but it is an interesting experiment.

[torpid merlin]

Hmm it might know klingon

[shut moat]

make a song that alternates elvish and klingon.

one ius smooth and serene and the other is abrsive and crass

it's an idea, i think i will work on that.

making a song that is 30 languages at once.

[untold briar]

It can do many languages at once, pretty well. Probably 'hallucinates' with 30 in one prompt, but maybe not if each language has a full line of lyrics.

IPA alphabet does not work well. (A little bit, but I think it's mostly an accident from where the IPA alphabet overlaps with real languages)

Elvish and Klingon work okay, I think.

[torpid merlin]

Oh look the king of non-music, here's your mentor Dem lol

[untold briar]

In the style prompt, creative language use is very useful. Check out Yolkhead's prompt here https://suno.com/song/ec27b5b5-5b2a-4f22-9518-24960ee61546 "cello-cassettefield viola-tapeworm violin-glitchwood insightful, broken composer, post-baroquefield"

"field" in particular comes from "field recording" but field by itself is strong enough you can kind of mash it together wiht other words and it can be pretty cool as a style.

I would say formatting and punctuation are two very cool things to experiment in the lyrics field. And that includes spaces and linebreaks.

[viscid pewter]

hi dear suno devs, i tried your app now for 3 days, and i love it but i have found some limits to the usage :

• greek doesn't sound well
• tried several times duos, mixed or not, never worked as intended
• lyrics instructions slightly too long (like 3 or 4 words) get read by the AI

(and yes, i love to do some original dig through)

[uncut pumice]

This is strange...all of a sudden. I'm getting songs with no lyrics.. and no credits refunded? I mean complete 4 min. songs with not one lyric that was requested. LOL

[viscid pewter]

yea got the same sometimes too

idk if it is possible, but maybe you could add in the custom how many voices are intended, their awaited style, and even say to whom the lyrics are destined

[graceful condor]

hi is there any way i can ask suno to generate a track without lyrics + for a certain duration ? like maybe 30-45s or 1 minute ? i appreciate any help thanks

[ornate dust]

I got that when my page was cached on the old layout
When I refreshed I got the new layout and lyrics returned

[timber adder]

any body here who works for suno ?

[untold briar]

Suno is on EST time, so they're not usually online until for 3 or 4 more hours.

[charred merlin]

The staff chip in occasionally, but usually you're asking stuff to the mods, but the devs do read this stuff eventually.

[untold briar]

For account issues email [e-mail] is the only way to get resolved, in any case.

[timber adder]

no i have suno issues

so many users complaining about the new update... and im one of that ....

[charred merlin]

#🚧┃suggestions-layout They made a channel specifically for feedback -- out of all of the channels, you're most likely to get your comment regarding the new UI seen by a dev by commenting in there, or even better upvoting people who have already made similar comments.

[timber adder]

thank u ur the best

[charred merlin]

We aren't able to "explicitly code" behaviours into Suno, it's all suggestions, but the easiest method would be creating an instrumental track, then creating an extension point a tiny bit before you want the lyrics to start.

[somber badger]

suno is changing your lyrics you put in. sometimes for the good, sometimes for the bad. but if it changed your lyrics and you download the video,it shows the old lyrics and not the adjusted lyrics that suno did.

[reef cliff]

Hi all, I would like to ask a little help here (if is possible of course); at what time the credits renew (I'm Brazilian GMT -3), aanyone here can help me, please?

[flint haven]

Free Credits Refresh

Refreshes at 0300 UTC+0 (<t:1713236400:t> in your local time)

[charred merlin]

If you have a free account, they refresh at the time Woodhouse shared, if you have a paid plan, it refreshes around the time that you made the initial subscription.

[reef cliff]

Hi, thank you very much for you attentio, answer and help

[unique solar]

I have a song that I want to keep the first portion and delete the rest. How do I do that?

[charred merlin]

Easiest way is to open the song in a Digital Audio Workspace (https://www.bandlab.com/studio, or Audacity) and trim off the end of the song.

Since we don't have editing capabilities in Suno (yet! https://discord.com/channels/1069381916492562582/1252773477157699664), the only way to handle that within the platform itself (so you could share the result easily on the site), would be to create an extension point just before you want the song to end, strip all metatags and style prompts and just use "end" "fade" "goodbye" or something similar.

[naive quail]

"Exit Stage Left" 😄

[unique solar]

I extended at the point where I wanted the song to end and in the prompt place only "fade". It worked except the last part of the video was a slide that had the word "fade". Should I have used [fade] instead of "fade" in the prompt?

[charred merlin]

You KNOW I'm trying this.

[charred merlin]

Ideally, don't use any lyrics or metatags, take out all style prompts and just use "fade" as the style prompt.

It's already gathering information and context about the style by simply extending from the track, so doubling up on those prompts in the style section will just nudge it toward thinking that you want it to fade out and come back

[unique solar]

ok, will try this next time. Thanks!

[naive quail]

haha. I've used it as the title in my [End] extensions... usually when I'm just playing around with the extension names with stuff like "A Guitar Walks into a Bar..." for one and "...and Screams!!" for the next. 😄

[wintry abyss]

hii

[merry barn]

Im trying to expand the lyric prompts a little by adding [beat drop], [powerfull melody] but in most of the results it is singing those as the lyrics in the sound.. can anyone help me with this?

[warped furnace]

Weird. Beat Drop should be working. I've collected some tags in a list (I found them on Suno). Those worked, too. At least more or less, they weren't sung.

[Break]
[Instrumental Interlude]
[Melodic Bass]
[Percussion Break]
[Syncopated Bass]
[Fingerstyle Guitar Solo]
[Build]
[Bass Drop]
[Powerpop Chorus]

[merry barn]

but if i make them longer it won't work for example [Light synth pads, a soft kick drum, and subtle hi-hats.]

[warped furnace]

I think, that's too long to work. Although I haven't tried it. 😁 You can't be too specific. Even the shorter ones are ignored if they don't fit into the song structure.

[merry barn]

i see

[warped furnace]

The tags work best for me when I put a complete song with structure into the lyrics box. It's helpful if you learn basic song structures of certain genres. For example pop songs: introduction (intro), verse, pre-chorus, chorus, verse, pre-chorus, chorus, bridge, and chorus. Having that, I can put breaks or other things (e.g. instrumental interludes) in their right place. Most of the time it works out for me. 🙂

[merry barn]

thats exactly what i did.. but ill try again and this time make it a bit short!

[charred merlin]

#📚┃suno-school message
https://platform.openai.com/tokenizer

Using OpenAI’s tokenizer as a means of visualizing how Suno.ai interprets prompts;

’Happy Hardcore’ is 2 tokens, and would be interpreted as 2 entirely separate words.

‘HappyHardcore’ gets broken into 3 tokens, and this introduces a level of ambiguity into the AI’s interpretation of the prompt.

‘Happy-Hardcore’ is 4 tokens, and would likely contribute to the highest level of ambiguity, unless hyphenation makes prompts interpret differently (which would make sense seeing as how common hyphenating genres is)

While it's strange for short metatags inside square brackets to outright be interpreted as lyrics, it absolutely can still happen.

But similarly to the example above, "beat drop" could be interpreted as "play a song with a beat, 'drop' is musically ambiguous in a classical sense, so let's use that as an "prompt decoration" and take inspiration -- ooh, maybe a lemon drop!"

It's technically not wrong, it just didn't quite understand what you meant. Silly little robot tries so hard, but sometimes just doesn't understand some references. So maybe trying [beat-drop], or [complex-crescendo], and adding an extension point at a really intense moment and taking out all other style prompts and saying "explosion" or "glitch-breakdown" or something similar to "get your hands dirty" with experimentation and try to force Suno to act like an instrument -- "hacking" your way into becoming a "suno DJ"

I don't make EDM, if that wasn't painfully obvious -- I hope you ravers and bassheads can forgive me 😅

[neat field]

I believe that I have found a way to make v3.5 end way, way more often

Pad [End] with 3 new lines before and after [End] at the end of the song lyrics

[graceful spoke]

quick question about terms of use: Am i allowed to use a suno logo in my video, if im basically telling people where they can find me?

[charred merlin]

I would imagine if it's being used in the same way that people & businesses advertise their facebook / IG / reddit accounts with the logo, I'd imagine it would be fine? Should only really become an issue if the usage could be mistaken as an official endorsement of your content, or making it seem like the video was released by suno themselves

imo

[graceful spoke]

right, it would basically be my branding, with a blurb saying "Find me at Suno.com" with a logo

[sweet wigeon]

I do it something like [Instrumental intro] or [Instrumental solo]

[hearty raven]

I'm looking to find voice presets for Suno Bark. Could someone please point me in the right direction for where I can locate them? Additionally, I'm interested in learning how to create my own voice presets. Any guidance or resources on how to do this would be greatly appreciated!

[tawdry merlin]

How can I make my track, originally generated with Suno, public? I edited the track in Audacity by cutting some pieces, and now any further extensions from this edited piece can't be made public.

[graceful spoke]

ive edited a few of my songs externally, yeah they cant be uploaded to suno now, but i have them on my personal pages/youtube, as long as proper credit to suno is given and you understand the licence you are under.

[tawdry merlin]

Well this sucks,I did read the licensing info yesterday and it says that if the track was generated while you have membership,you own all rights to it.

[graceful spoke]

you do, but you cant reupload edited songs to suno, because then its not really fully ai generated on an ai music generation site 😛

[hearty raven]

Another question: How do I change voice properties like speaking rate and remove filler words (ahs and ums) in Suno Bark? Any tips on other useful adjustments?

Is this the right channel of these types of questions, or are there channels that are more suited to Bark questions?

[untold briar]

Hi, the Suno Discord has shifted to be focused on the Suno Music service (the web site) and there isn't much Bark talk on here. I'm one person you can ask about Bark though. Probably best in either the channel off-topic (even though it's still Suno) or you can DM me though. The short answer about making good voice presets is you that you can simply use Bark itself and text prompts to generate them. When you don't give Bark a voice, it uses a random voice which has some relation to the text prompt. If you retry a bunch of times, you can get a good random voice. And in Bark you can save any Bark output sample, also as a new voice (typically .npz file)

If you search for .npz you may still fine some Bark voices in old messages, actually. But it's annoying hard to search for file extensions in Discord so it only pulls up some. You can also DM me, as I said, but I may not get back to you quickly.

[silk mulch]

I don't know but many of my songs are stuck on constantly loading and never truly loading and for some of them it's been more than 15 minutes

[hard ruin]

why the hell on loading

[silk mulch]

Maybe they screwed up again

[hard ruin]

oh ghos

[silk mulch]

They were in so much of a rush to perfect their website yesterday that they actually screwed up everything

[brave viper]

your not alone same here

[unique stone]

#💬┃general-chat message

The team are aware of the issue.

Are you uploading something you have made/recorded?

[cold zenith]

So there's a problem for everyone ?

it's fixed

for me there's no loading anymore

i just checked

but it means it on the air

[unique stone]

By "existing work of art" it means the audio you are uploading, not a picture

It's a copyright detector for uploaded audio

[hard ruin]

😩

[unique stone]

What is that

[silk axle]

Hey I wanted to let you know Suno is slower to load since you add a new design to your whole website, I noticed it loads first the Ai generated image + the customized text and after the song. That means we have to wait for the image to load, that's annoying because I think images are in high resolution, I would recommend to make compressed versions of the Ai generated images to load faster for people who have a low connection (for me for example cuz I'm in an airport)

[trail badger]

I hope you can restore the system to its previous state 4-5 days ago. I haven't been able to produce happily for 4.5 days. Sound quality and production quality have dropped a lot.

[silk axle]

Also why my sound can't be public? I used my voice a little bit and I made original lyrics

[unique stone]

I did not say it is user error.
I was explaining that the message didn't mean the picture when it said "art"

It can be a false positive (like Shazam can detect a song as something completely different)

[trail badger]

Production speed doesn't actually matter. Production speed does not matter as long as it is of good quality. I think there is a loss of quality when they increase the speed. v3.5 is the best model on the market. Model developers may not be aware of it. But there has been a problem with the model for a few days.

[naive ore]

i guys any one with acent isues, im want suno making use portuguese of portugal any is very dificult always get brasilian acent, is anoing, any one have any form to get a rigth acent, prompts dont respond

[unique stone]

I'm still not saying it's a user issue, in case that wasn't clear. If you're uploading something you made and the site is detecting incorrectly, that is not your fault.

[mortal wraith]

I used a portion of my Suno song in this track but still It won't set be public?
https://suno.com/song/a4416810-e30f-4974-8585-eeef97dfca63

[bleak shuttle]

Why isn’t my image appearing as the song’s cover?

I set it as an anime and it’s not appearing

https://suno.com/song/9fd0bf96-f4c8-4218-ab6d-5a1d76c66667

Nope not appearing

[naive ore]

another proble music is to low compere to others, wen dawnload nee put 2 db gain in out souce program for listen good in my car. any one fell the same?

[charred merlin]

https://www.suno.wiki/faq/the-music-biz/your-upload-has-been-detected-as-copyright-material/

[bleak shuttle]

I can’t figure this out

I only got a subscription for the custom song cover

😭

[charred merlin]

Uploaded audio with vocals (and works extending from them) can't be made public (for the time being) as a safety guiderail. I would imagine it's an attempt to prevent bad faith actors from uploading someone else's voice

READ the message. Request a human review.

Your hostility isn't making it easier for people to help you.

https://tenor.com/view/thumbs-up-thumbs-up-group-thumbs-up-circle-okay-gif-4307739110186058516

"no one is reading"
doesn't read

[charred merlin]

I know we can't completely regenerate the artwork used for the video, but I know we can set specific album art for the cover -- let me double check how 🙂

[bleak shuttle]

I set a specific album cover already

It’s just not appearing anywhere besides on the website

[charred merlin]

Sorry, I didn't phrase myself properly -- I know it's possible to set custom embedded video cover art

[bleak shuttle]

Oh

Alright

[charred merlin]

My initial observations are that I generated new artwork from the site for a song last night, and that new artwork shows up as the cover when I share the link in discord.

I updated with an uploaded picture, and the link didn't automatically show the updated cover art the same as it does on the website. I marked the time and I'm going to check back periodically. I'll let you know if I figure anything out ❤️

[silk axle]

Ah :( but it's my own voice lol

It would be cool to approve it manually by some moderators idk

[teal raft]

It's not possible, I've asked

And I'm p. convincing

[charred merlin]

I know :/ Luckily it's still able to be fully generated and used within Suno, and can be shared on other platforms like Youtube, or you can look into a distribution partner to get it on Spotify / Apple Music, etc 🙂

[silk axle]

I wouldn't go as far as looking for a partner for Spotify, everything I do is for entertainment, I don't care about the money

[warm cipher]

I think I am somehow into a bucle as the model keeps repeating the same "structure" that is like verse 1, verse 1, verse 2, chorus ... But no indications anywhere telling it to repeat twice verse 1...

[bleak shuttle]

Oh it loaded now

https://suno.com/song/9fd0bf96-f4c8-4218-ab6d-5a1d76c66667

[charred merlin]

Okay, mine changed too! I set it at 1:52 my time, and only noticed it now, 3:27, so at least I have a ballpark estimate when someone asks about that in the future.

Thanks for letting me know 😄 🤘 I just realized the album I was listening to ended like 30 minutes ago, so I'm going to give your track a listen 🙂

[bleak shuttle]

Oh thanks!

[charred merlin]

oooh, that electric violin (?) at 0:48 / 1:14 is dirty 🤘🎸 😍 (I mean that in the in the best possible way) -- definite "Like" from me

[random shale]

I have a 3,5 Pro version of Suno. I have uploaded one of my own songs to extend it with suno, but I am unable to make it public due to content restrictions. How can I verify that I am the owner of the works? Thanks a million

---

[teal raft]

It isn't ownership, it's that your sample has a voice in it

Samples you upload with voice means you can not make it public, it's Suno's policy

[sweet wigeon]

Where to post suggestion regarding this: I would like to create voice only, no music to test if Suno can pronounce the lyrics correctly. Because I found it can have problems to read for example "v panice" reading "v paňice" instead like "v panyce". So when I prepare lyrics I would like to complete them to be ready for production. The lyrics finalization depends on how AI comprehends the pronunciation...

[bleak shuttle]

Tyyy

[sudden spear]

Got a suggestion about Suno normalizing song volume when getting full song. Please have an option to turn this off. I use eMastered to master songs and upload to other sites. Don't use the public option on suno to share songs, so I don't need normalization.

[viral moss]

I was experimenting with Metal infused styles, and noticed if it try "Disco infused thrash metal", I consistently get a singer with a British accent... has anyone else experienced something like this?

[austere minnow]

Does anyone know how to separate the vocals from the instrumental, it would be good if "suno" integrated it because external programs do not separate them well, there is always noise or bad sound

[sage flower]

This is a hugely requested feature. Currently my understanding is that the way Suno generates the audio it doesn’t. It also wouldn’t be any better than AI stem separation services as it generates the audio as a whole rather than compiling the instruments separately. Thankfully there are many ways you can seperate and improve the current audio. My new web based daw (in my name) has just launched in beta and is free to try. It is fully featured with detailed editing and multitracking. We also offer a powerful AI stem separation that definitely can help, which goes live tonight. Feel free to try, it’s web based and no installation / account needed 🫡

[austere minnow]

oh I want to see this

[sage flower]

I try to avoid breaking any rules but the web address is in my name 😃

[shadow sorrel]

did the playback engine change with the UI update? I have a song that sounds way muddier than it did a few days ago and its even missing some drum beats and smearing some guitar parts together. anyone else having similar issues?

[small dirge]

OMG! Hi! Okay, loving Suno! Slight problem though. It doesn't use my custom cover art for the song. Help?

This is the image I'd like to use. I uploaded it but it still uses the default when I download.

[hollow hare]

The album image is only cosmetics when people view the page. It does not affect the video as the generation only occurs during initial generation.

This also explains why, when you combine your extends, it basically concatenated your latest generation to the initial generated video - cheaper on the processing side.

It would be wayyy too resource heavy if they updated the video everything someone generated an image or changed their song title.

Maybe make a suggestion at #💡┃suggestions-web to have an option for regenerating videos with the latest image?

[small dirge]

So, I can't use my custom art for my songs instead of the default?

[hollow hare]

Unless you mean that your default image did not change despite uploading and selecting the picture in edit details?

[small dirge]

Yes, that's what I mean.

[hollow hare]

Ah, I see. I would recommend refreshing the page (in case there was a visual glitch) then repeat edit details > select the uploaded image (which should be present).

If repeating the process does not change the image, it may be a bug and should be reported at #🪲┃bugs-web

[hollow hare]

To reiterate for clarification: if you are talking about the downloaded video, the video is ONLY generated during the initial generation so changing the cover image or changing the song title will have no impact, they must have been done during the initial prompt (well, the appropriate title since you cannot upload the image at that time)

[small dirge]

Oh. I didn't realize that was an option.

[hollow hare]

There are many video editors out there that you can overlay your image by the way. Most people just simply use other AI tools to make videos like Noisee.AI

[small dirge]

Ah, okay. Thanks. I really love the song I created and the image I generated is just perfect for it!

[austere dragon]

Any suggestions for an EDM build up and drop?

[charred merlin]

Suggesting services is one thing, but promoting your own service (which will be paid after the beta) starts to blur the line between solitication and self advertisement.
#📜┃rules message
Rule 4

[charred merlin]

There are many free services that can do this

Ultimate Voice Remover, the OpenVino plugin for Audacity

There are TONS of paid services available online with free demos, but they're free to advertise on their own.

[charred merlin]

We haven't been made aware of any changes to the song generation, only changes to the UI

[shadow sorrel]

it's not really the generation though. its playback. its like everything has more compression on it now or something. all songs I have generated prior to a few days ago sound different now when i listen to them in my library

[charred merlin]

If I understand correctly, they sound differently from playback on the website to downloading the .mp3?

[shadow sorrel]

mp3 and wav files downloaded sound the same as the web player. more muddy, compressed and missing/blending some notes

[charred merlin]

So then it is the generation quality, if it doesn't matter where the playback comes from?

[hybrid vine]

I have a question, not sure if it's the right channel though.
Once you use all your premium credits for a month, do you still get the daily credits that free users get or not at all?

[charred merlin]

You get all of them at once, plus more, when you sign up for a paid account.

In addition to getting a extra chunk of credits, one of the features of the paid plan is not having to wait for your "daily allowance", and are allowed to spend it all at once if you want.

I think @unique stone has mentioned the devs are okay with people using a free account in tandem with a paid account, or multiple paid accounts, or can purchase extra credits that don't expire at months end.

As long as you're not using multiple free accounts.

[shadow sorrel]

unless i misunderstand how suno generates and plays songs back? i assumed what we are presented as a song is just an audio file. i didn't think it was data that is fed into an engine every time its played. i thought generation was done once when you generate and then after that you're just listening to a file.

im confused/concerned because its like opening itunes and all of your songs sound different and are missing some notes. those files are "done" and not being generated. theyre the products of being generated in the past. therefore if generation was affected then all songs I create moving forward would be affected. not songs that have already been generated if that makes sense

[shadow sorrel]

i also have an example of how it used to sound prior to noticing this and then how it sounds now so i kjow im not going completely crazy lol

[charred merlin]

You are correct. After you click "create", the song is generated and processes. After the "extend" button appears, it's fully generated, and this song quality will not change beyond the bitrate difference between the .mp3 and the .wav

[terse loom]

Joined up because after emailing them they suggested joining discord 😄

So I have been having issues (especially this week) where the music I am generated - no matter the Genre - is pretty much hitting the same style of melody (pretty much one that repeats the same type of note structure with no real deviation around why a few notes up and down)

Additionally the chords are extremely limited - like four chords and that is it throughout the whole song.

The thing that changes is the compliments around it, the arrangement, but the chord progression and melodies are almost always the same structure.

Rendered a massive amounts of songs last night just to pretty much get the same structure from Classical, through to Heavy Metal.

The "sound" changes - the tone of the voice, the instruments - but - the structure of the music is the same.

Wondering if other people are having this issue because I didn't used to - when I first use Suno the songs were a lot more expressive.

[charred merlin]

Many things effect audio generation quality from day to day, hour to hour, prompt to prompt.

People have made the same assumptions since December 2023, but the only people who can say for certain are the developers.

[unique stone]

Do you have some previously-downloaded songs that you could download again for comparison? They should be 1:1 identical. The songs just get stored once generated, they don't regenerate them. Only the WAV seems to go through some extra treatment (not sure what.)

[charred merlin]

I'd be curious gets extra treatment, or is just a non compressed version of what you'd get from the .mp3?

[shadow sorrel]

when i extend a song does it re-run the entire thing through the generator or does it just append to the point you tell it to? if not, would it if i bounced between v3.5 and v3?

[charred merlin]

Append, afaik

[unique stone]

I have no idea tbh. It's the audio from before compression, but It seems odd to me that it goes off for a few seconds to do something before presenting the WAV download file.

"Get Whole Song" is where it glues the parts together and I think if you used v3 for some parts it could affect the overall quality when doing that.

[shadow sorrel]

i have two 26 second audio clips i downloaded before and after for comparison. idk if im allowed to upload here but id be happy to send to both of you in DM for your information if youd like?

[sage flower]

Yeah, I’ll be sure to be concise to include all available options of any services that I use personally as a Suno subscriber

[charred merlin]

Hard to give specific feedback unless we know the kinds of style prompts & metatags you're using.

At a glance, overwhelming Suno with prompts and tags can "confuse it", and cause it to lean more toward a generic pop/rock/dance/rap unless it has a strong musical theme to continue from, or the prompt suggestions you've given it aren't being outweighed.

[charred merlin]

Other information is one thing, but specifically hosting the link in your name and telling people to check out your (limited time free) service is questionable.

There's other paid services, but they've usually gone through significant community vetting in order to be recommended, like Spectralayers 11, etc.

What does your service do differently than these existing services?
https://audiostemseparator.maztr.com/
https://voice.ai/tools/stem-splitter
https://cryo-mix.com/separator

[daring lava]

Hey @sullen oar 👋 please tell me you are going to be pumping some awesome Suno tunes on that stage behind you 🥺

[shadow sorrel]

i can PM you the before and after audio clips in the quality im talking about if you like?

[untold briar]

We (mods and guides) don't know how the Suno Sauce actually works, but based on audio models that are open source, when you append two samples together, the result may not be identical to appending the raw audio together. Even in Suno's open source Bark TTS this is can be true (though people rarely append sampled together in a TTS like that, so it rarely comes up). So basically... yeah it could be different.

[unique stone]

Are they from the same song/part? If so yes please I'd be interested to see what's going on.

[sage flower]

Those are all great options for me to include. My service isn’t a stem separation service, but we do offer it. Soniqs is a full fledge web based daw

[terse loom]

When you are calling Prompts are you talking about Lyrics etc - or just the Style?
It seems to be an issue with everything even if I don't use prompts

This is what I did yesterday
https://suno.com/song/e810c34e-3786-4121-b90a-50611b9a238e
https://suno.com/song/c163c0cb-4554-4785-8a36-be7e0a8c12fa
https://suno.com/song/f90fc0a6-3af8-473f-a4b0-02722e0c013e
https://suno.com/song/113f30cf-f4d5-4157-b5e9-13ebb9bd97ce
https://suno.com/song/67350121-7c40-4afd-9924-6020d9dd39fe
https://suno.com/song/b7bca604-2854-4977-b9a3-be942f1182a9
https://suno.com/song/71701e18-abdc-44ff-b232-b59b8fc8a723

So the issue here isn't the instruments, or even the structure of the song.
It is that 1) The vocals keep doing the same thing over and over again where the singing is pretty much the same tune, just slight alterations - between multiple genres.
2) Chord Progression is almost always the same - just changes to the structure - usually 4-5 chords

Not really upset because I know it is AI and I know a bit on how it is built - but - I am not sure if it is something I am doing, but it has become very repetitive since last week (so wasted 400 credits last night to prove this is really bad hit and miss, with a LOT of misses)

Basically the melody structure that keeps happening with "twists"
Verse 1
Line One Melody 1
Line Two Melody 1 (little change)
Line Three Melody 1
Line Four Melody 1 (little change to next phrase)

It isn't treating the verse like a lyrical melody, and instead keeps repeating variations of the base melody.

[mild rain]

totally with you on this, most of my creations as of late sound terrible.

[charred merlin]

Still advertising your paid services in response to user requests. :\ https://www.bandlab.com/studio is an amazing free service, and I trust their reputation.

[sage flower]

Bandlab is also awesome, I would recommend that although I think you need to install their app on mobile devices?

[charred merlin]

Just makes it uncomfortable since we're not familiar with the product. People looking for support are a captive audience of a sort, so it's like shooting fish in a barrel to say "here's a service that will do everything for you that Suno wasn't able to do".

Either way, it would be easier to recommend your site if we knew about it or had used it first, rather than presenting it as an option and using the tech-discussion forum as a testing ground of sorts

Yeah, it seems like BandLab's DAW isn't available in a mobile browser. I know it's relatively complex so it would be awkward to use on mobile, but I'll have to check out their app now that I'm aware of it, since the web app is an occasional part of my workflow

WHAT -- BandLab's app has a splitter built into it? If this is on their website for free, it's just another reason they're the friggen G.O.A.T.

Wtf this is awesome. I’m so glad things happened in a way that I discovered this app.

Pitch editing and speed adjustment (with a bpm counter), looping, live mixing of stem volumes for quick experimentation

[sage flower]

Yeah after hand developing Soniqs for the past 6 months I can appreciate the complexity. Thankfully I had a rich knowledge of web audio before I tackled creating the software

[charred merlin]

I think it'd be neat if we figure out a way to handle https://discord.com/channels/1069381916492562582/1260045620538904628, because that could serve as a place for people to share the software and tools they make with the community in mind -- like a safer, sanctioned place to self-promote paid services so we can become more familiar with then recommend them if they resonate with the community. ❤️

[final hazel]

Okay.. This is becoming more soliciting. Normally we wouldn't allow advertising services. So let's just drop this subject (and thus your app) in full. @sage flower

[teal raft]

So: Have you used the splitting yet with bandlab?

I ask because I have used UVR a few times but the results are p. mid

and that's due to the style of music I'm doing, most likely

[small dirge]

Noisee ai only does 30 second clips for me. Also, how do I upload the cover image that I want for my song to be generated with? I don't see that option.

[hollow hare]

Did you read the manual on their page? Specifying time ranges allows you to generate clips up to 1 minute. (IIRC, you can currently generate just three jobs per day on the free plan)

As I mentioned, Noisee is a (Music) Video AI service, it is not a video editor

[small dirge]

Ah, thanks.

[hollow hare]

Searching "video editor" on your search engine will give you many options. I used KdenLive on my Ubuntu server for example

(Heck, there are video editing apps too)

[charred merlin]

Havent experimented fully yet, I prefer OpenVino + audacity to UVR, to be honest.

[teal raft]

Noted, thanks

[small dirge]

I haven't found any that we're in my price range though. You know...free. lol

[hollow hare]

There are several of video editors for free like Kdenlive

Search for FOSS video editors for example

(FOSS = Free Open Source Software)

[charred merlin]

I saw someone in off-topic recommending to woofi to use 2 separate UVR karaoke filters to cleanly separate vocals, as one is trained on removing background vocals and one removes lead (to address some of the weird bleeding)

It can still happen with OV, but overall i've had better results, since I feel like there's so many considerations to make in regards to UVR that I feel like I'm leaving so much on the table as to not be making the most effective use of it.

OV simply has an advanced option for how many separate shifted passes it should take over the same content to try to correct mistakes (or overthink a simple problem).

I'll let you know if I have strong feedback about the bandlab services -- but seriously, the fact that it's everything that you need from start to finish for free is absolutely mindblowing.

[small dirge]

Ah okay. Thanks is that in the app store for android or are they sites?

Also when will suno be releasing a version for Android? I saw there is one for apple but not android.

[charred merlin]

I recall you having mentioned that your mixes are pretty complex and lush, which I imagine would make specific isolation pretty tricky to nail consistently.

I'm in favour of starting a kickstarter project called "Buy KakerMix Spectralayers 11 For Great Science", btw

[glacial scarab]

keep in mind open vino is just the framework and the audacity splitting is just ht_demucs. (if you where already aware just ignore me)

[hollow hare]

Knew that was going to happen lol

Ah, you did not specify what OS you were using so I assumed Win/Mac/Linux which Kdenlive (https://kdenlive.org/) run on.

For video editing, working on a desktop is much, MUCH easier than on a mobile device.

The one time I used a mobile app to edit the videos... well, it was exhausting to do it IMO if it is not a simple animation transition effect.
I think I used ShotCut (https://play.google.com/store/apps/details?id=video.editor.videomaker.effects.fx) then?

.
As for Suno Android app, it will be announced when the team is ready to release it. We were notified about the iOS release VERY vaguely literally the day before.

It sounded like it would have been for all phones but, as you know, was just for iOS to USA only :/

[charred merlin]

❤️ Have I mentioned that I love you, recently?

I wasn't aware of the underlying model underneath it, so I wouldn't be surprised if there was a degree of placebo effect going on 😛

I know I'm naturally favoured toward it just for the fact that it's built into my DAW, but if I had to wager ANY guesses to the differences I'd experienced, it could be that OV is using standardized/idealized settings for that model, or the fact that running on my CPU due to the Intel architecture requirement means that the process is handled differently than if a GPU were to address it (now I'll admit that's just wild speculation, and it's way more likely the same kind of self-rationalization that people make when audio generations come out different quality from day to day)

[glacial scarab]

I don't think so but have some love as well. You seem very active and helpful to everyone around ❤️

[small dirge]

Okay, thanks.

I just want to change the cover art. sigh

[glacial scarab]

i think the shifting is something only the audacity implementation does. And it could very well be optimized settings yes. UVR is complicated to make it work as best as possible, so if audacity open vino works for you no reason to change i suppose. Never touch a running system and all that 😂

[charred merlin]

I think it's handled by the Segments section of demucs' UVR settings?

[charred merlin]

Give it ~1 hour, afaik 🙂

[glacial scarab]

I think that is something different.. I'll have to check myself tomorrow. I mostly use roformer now and the thing that removes backing vocals always forget the name..

ht_demucs has been some time since i last used it

[small dirge]

No, not what I'd like. Everyone is saying it's not possible because the video is generated at the beginning before there is a chance to add personal art. So when downloading, it uses the video generated image. Thanks though.

[teal raft]

I'm exclusively looking to strip vocals out of hooks that I generate for other songs

I can't have the same lyrics and the begining as another song. That's silly!

Just for the initial like ~15 seconds usually

[glacial scarab]

to feed them back into suno or to use in a DAW or something?

[charred merlin]

Ah, if you're looking to replace the video content that shows when downloaded, that's something that hopefully we'll see before too long 🙂

If you want to change the art that shows on the embedded discord share links, that's possible at the moment, though. 🍻 hopefully not too long of a wait on video regen!

[teal raft]

Back into Suno as a hook for another song is all

[small dirge]

Yeah, the first is what I'm looking for. I hope it's soon. I use TruePerson AI to help me create my custom lyrics because it's an amazing text app that does a lot! MonAI AI is for doing my custom artwork and, of course Suno to put it all together.

[glacial scarab]

then UVR is probably better because it has many models that are specifically for vocal/instrumental splitting. the models that split into more stems are a bit older and there is often bleeding.. which does not matter much for mastering if you merge it all together anyway but degrades the Instrumental. I can share an example of model and settings tomorrow or sometimes this week if you want

[teal raft]

That would be excellent, I already have UVR installed and set for multi-stems because more = better

But I ended up not ever using it and instead just want the vocals out

[glacial scarab]

yea it's a tricky thing.. and always requires some trial and error because every song is different. I am waiting for the day we can split it in quality that is usable out of the box. But feeding it back into suno could work because suno might "fix" it on its own.. might require double upload. Like first the split stem -> get it to repeat in better quality. Download upload again -> use as starter

it seems to work with vocals so why not the other way round

[tall trellis]

is there currently a problem with extending the uploaded samples? somehow the quality has dropped massively since yesterday or the day before yesterday, everything is distorted or sounds like it has been reduced to 9600 baud

[charred merlin]

Not that we've been made aware of, or noticed any emerging trends thereof. It's not an immediate resolution for the issue, but reporting the bad generations through ... > Report > Bugs > Poor Audio can help ensure better results for everyone ❤️

[viral lotus]

I haven't worked much with extending songs. When I want to extend the song, what lyrics do I put in the lyrics box...because the finished product always contains duplicate lyrics in the lyric text. Can I avoid that?

[tired hedge]

did anybody succeed creating a song in reverse from suno? like just a song played in reverse, from the end to the beginning?

[viral lotus]

I imagine that would be a miraculous feat of prompt engineering 🙂