📚┃suno-school

[polar spade]

We're having a Glitter Rock Opera convention?

[jovial bough]

LOL

[hidden yarrow]

I tried to eleminate it with replace here and there, but it is always messy ....

[jovial bough]

Me, too. A lot of work

[hidden yarrow]

From my experience it happens often, if the lyrics line is ending early and there is some time left for the section and it stretches the last syllable out....

[polar spade]

Since you said trills: suno knows trills, so add trill to the exclusion box?

[jovial bough]

Ha! I'll give that one a shot. Wish that box had more sway with Suno.

[polar spade]

Dude @summer solstice . Are you posting and walking away before you see you're being blocked? What are you trying to post?

[polar spade]

Sounds like a plan. Maybe even ad "mmm" here and there after a long word to shut their mouth for a more pleasant sound. (Not joking)

[hollow hare]

They are one of those spam bot promoting crypto via images which is why you see a single image at most before Woodhouse deletes/bans them

[polar spade]

Thought ive seen the name before. Wondered if he wasnt seeing the block.

[ripe galleon]

what are certain instruments that v3.5 suno cannot do well?

i once tried doing something with a glockenspiel type instrument

and it sounded really,,,,

screechy

idk how to describe it

[ebon sand]

dobros... reverbed desert guitars... mix with the vocals often

[ripe galleon]

reverbed desert guitars?

[ebon sand]

yeah that canyon sound, sounds similar to a dobro, it has a hard time discerning vocals from that so the dobro might form into a word or vice versa

but that's basically all it 'cannot' do

[ripe galleon]

could you send me an example of this?

[ebon sand]

i can try to generate one of those again haha i usually delete them but when in doubt you could edit those spots to be clean separations

[ripe galleon]

i mean if you are able to for the sake of showing me

then go for it

im curious

[ebon sand]

sure, i'll give it a few tries and if it doesn't obey with failing i'll try to search up one of the fails, i have plenty of them somewhere because i like the sound (when it works)

[ripe galleon]

alright

thank you

[ebon sand]

https://suno.com/s/dLdB87vhFD7KNqyy good start but not quite there, if you get a 'bad take' the sounds merge more but i think it gives an impression of what happens

[ripe galleon]

0:11 "sss- guitar string"

something like that?

[ebon sand]

yes, if it fails gloriously then the sound of the dobro or twangy guitar and the words are the same, i stopped using those two instruments because i didn't want to have this problem all the time

[ripe galleon]

does it just think human voices sound similiar to this specific type of guitar

[ebon sand]

yes

it cant discern them well

[ripe galleon]

weird as hell

[ebon sand]

i think v3.5 is weird enough without these

[ripe galleon]

dude i love it when my v3.5 songs have musical noise/ ambient instruments(???) that i like and cant determine what they actually are

maybe its reverb????

idk

[ebon sand]

you can try to prompt "atmospheric [something]", v3.5 responds well to mood prompting, like instead going in with professional terms tell it what the song feels like, i had one song where i just wrote "cthulhu", it will figure

[distant basin]

I'm trying to get a specific instrument sound but no matter how I phrase it, the model doesn't seem to understand 😓 Namely, I want to hear the mysterious drone-like sound of the 笙 (Sho), a free reed instrument similar to the Sheng or Khaen

Even when I name those instruments specifically, it routinely produces koto or shamisen sounds instead

[ripe galleon]

LOL

[ebon sand]

for example "dark canyon guitar" also works better than western spaghetti guitar with [insert exact playstyle and reverb style], it knows 'the thing'

[ripe galleon]

what the hell is a spaghetti guitar

sorry

[ebon sand]

i think google would be faster than me explaining it

[ripe galleon]

this sounds more like a dish than anything lol

[ripe galleon]

i know but like "western spaghetti guitar" sounds funny, especially to someone who never heard of them before

lol

[ebon sand]

one day i will figure out how to make it use heavy drums and a canyon guitar... it generates mostly either or

maybe i just have to write 'and'

[humble verge]

SUNO is generally pretty poor with drums.

[ebon sand]

oh i like them a lot but i want my bwwuiiiboiiing

[ripe galleon]

even v4 and such?

[humble verge]

I find them to be pretty flat most of the time. Especially when you do heavier rock and you want them to be powerful and instead you get something that sounds very machine-made and shallow. I've had some luck telling it to do taiko drums for more power.

[humble verge]

v4 may have been better? I don't actually recall since I've been using 4.5 for a good while now.

[ripe galleon]

i mean im saying this from the perspective of a free tier doofus who only was able to use limited v4

[slow magnet]

total control should mean that you can do everything, otherwise they should have written extended control or more control

[humble verge]

Ah, well it all depends on the style of music you do, really. The louder and heavier you want it, the weaker and flatter they sound compared to other instruments. But for other genres where you're not looking for a raw sound, they're alright I suppose. Like if you're doing types of music that's generally built in DAWs from the get go - which is basically all music nowadays. I work with genres that SUNO isn't super great at, so it can be a bit of an uphill battle to get good results.

[ripe galleon]

damn

i see

[ebon sand]

i like the drums of v3.5

when it whips out a grand opening and then gives you a lame singer

[safe sentinel]

I refreshed the page and suddenly hundreds of my credits were gone. I’m sending emails but I’m not getting any response.

[ebon sand]

did maybe your subscription refresh? mails can sometimes take a while until they receive an answer

[safe sentinel]

No, this is my first subscription. I specifically recorded it to keep track. My credits decrease by the production cost when making a song. After a while, I refresh the page and go to my library, and nearly 300 credits are gone.

[umbral shuttle]

Hello everyone, been using Suno now for a few days and Subscribed to Pro.. I have two issue one, I Finished my profile but when I close Suno and go back it it still says for me to complete my profile.

Does anyone know why?.... Now when I I do click on create or any other stuff in like library its fine.

Second issue is I lost about three songs in my play list is there anyway to get it back or is gone and none retrievable?.

[sturdy dove]

Alright, I'm back.

@polar spade @young geode and anybody else: thank you for your kind words of encouragement.

I'll be back soon with the full story. For now, a quick recap: I'm have a result that is 90% acceptable for my personal project, and getting the final 5% (dunno if I can get the full ten?) has proven soul crushing. I've invested hours over days over weeks trying to get just a little further.

The editor consistently produces gibberish.
The editor is not properly synched with the lyrics in the song.
I've tried the editor, I've tried extend. I have personae, I have samples.

It's a duet, and it's fantastic up until the last bar and verse, when the male and female vocalists switch lines, making the song such that it just doesn't make sense for the project. All I'm trying to do is fix that, such that the male and female vocalists switch back. Well, that and shorten it - I think it stands at 3 minutes with a lot of instrumental that will be easily cut out. I need it to be 2 minutes in length.

I understand this is s difficult and complicated thing for AI to accomplish right now... but the song is SOOOO CLOSE to being exactly what I want.

I'm open to all options at this point.

I did about ... 100, maybe 120 generations to get where I am. I'm now almost out of credits because after spending a couple of days and a couple of hours each in the editor, nothing useful came.

I'll share my full prompts and workflow if anybody is available and interested in helping me get this across the finish line I would be ever so grateful.

What details can I provide?
I've retained all of the prompts I've used and how I've tried to adjust them.

It also occurred to me that I've remastered the song a couple of times and that may have thrown off the lyrics sync. So I could go back to an earlier version to work on.

[ripe galleon]

hello i have a very weird request

i wanna extend this song i made (both voices are from me) and give both of these voices lyrics and stuff, HOWEVER:
does suno v3.5 even view these voices as human voices or would it see them as synths

if its the former, please tell me how i can prompt this properly

[sleek stump]

Check again after a while or a couple hours.. its likely a web display error.

[sleek stump]

Suno's web services arent immune to general internet issues. It's quite possible the back end was under maintenance or doing a reset or something, or just could have been lag and packet loss. Try again later, it will likely work (or it may have already have set your profile and will display correctly later). As far as songs in your library, make sure your library filters arent filtering out things like Disliked, its easy to misclick and dislike a song. Songs generally dont get hard deleted until you do it manually.

[safe sentinel]

It happened a few days ago. I recorded it when it happened and sent an email, but it wasn’t addressed.

[sleek stump]

Support is about 3 weeks behind, you wont get a response in the short term.

[rose lake]

when will the Cover function being fixed, all Personas are being ignored and the original voice sings????

[sleek stump]

What are you setting the audio influence slider to? If you set it 70+ it will likely override any persona.

[torpid haven]

Are there any prompting techniques to make longer tracks by default and instrumental tracks like in electronic music which have a clean long intro and outro for mixing in and out ? Those kind of tracks that haven (Extended Mix) in the name some times.

[neat nexus]

Broken Covers?

Original (60s style song):
[Removed links]

Glam metal cover:
[Removed links]

The trick is you have to change the style prompt and it helps to play with the sliders when doing the cover. Using a Persona can also help.

[cyan jewel]

what exactly do the weirdness and style influence sliders do exactly

[half dragon]

Weirdness: I guess like creativity

[cyan jewel]

the hover over dont really explain what they're actually doing under the hood

what does suno consider "Creative" or "wild" or "unexpected" what is suno doing when its "matching style description" exactly

how does it handle multi-genre songs?

does it emphasize specific genres?

CAN you emphasize/deemphasize?

[half dragon]

Not specific but it's only influence as close as possible to the style description, whether it's a single genre or multi genres...etc..
That'd be my guess.

[deep nova]

Any way to get American rap / freestyle that isn't African American?

[shrewd sage]

Hello everyone!!. Please explain what regenerate function does in the get streams page, and what is the diference betwen Ver 1 and Ver 2 on the stream extracted.

[tawdry scarab]

you could think of it like the style is the direction and the style adherence is how sure or forcefully you want to go there as opposed to anywhere, weirdness is how far you're willing to stray when you get there, and the adherence to previous music is like how closely you want to generally stay to where you're coming from or what influence you're asking for

i bet if you ask for college rap or comedy rap it will skew whiter sounding

[sour dock]

Works for me

It is the correct use of a double negative because the resulting positive is what we want not what we want to exclude. If we write "do not enhance' in the exclude box we are avoiding unenhanced audio which is what we want.

[young geode]

Post your song and the extension.

What you're running into is a pathway issue. The direction of the generation is making that switch a high probability.

A few things that may be causing this:

1. You're reusing the same prompt you originally generated the song with.
2. The lyrics near the end strongly resembles a previous lyrics already sung by one/both singer(s) or the pattern is to distinctly different than it switches.
3. You're extending from a bad point.

The trouble shooting will take credits, and you might lose some of what you already like, however when I have this kind of difficulty I do the following:

Start a generation from an earlier point in the song. this can fix the issue of play patterns and length.

If that doesn't resolve it, switch back to the original extention point and change the generation prompt. sometimes your original prompt is forcing a hallucination within the extension/regen.

And if that doesn't work, change your lyrics (word choice, structure, punctuation, ect). Often the Vocal shifts are related to structure and punctuation.

And from there it's a combination of two or all three. Effectively what you need to do is try and pull SUNO from that pathway. If you already have a pattern that exists within the song, it's likely that you've introduced something that's broken that pattern.

[sturdy dove]

Interesting. I believe you... I'm trying to think back to the failed generations (this was over a month ago at this point) and remember. Indeed, it seems like it insisted on making the female sing a male part no matter what.

1. When you say "I'm reusing the same prompt"... well,... yes, of course I am. I don't understand how I could not?.. When I've tried to regenerate the incorrect portions, I'm not using the whole prompt, I'm reusing the musical prompt for continuity but with the corrected lyrics.
   ... And I take it from your comment I shouldn't be doing that.

Cool... So what should I do?

2. This is entirely possible.

3. This is also entirely possible. I was attempting to extend from start of the problem area. I could indeed extend or regenerate earlier, from the non-problematic portion, but that seems incredibly counter-intuitive.

How should I modify my prompt? Whenever I let it try it's own thing, it ends up far worse than what I have. Is it a matter of repositioning portions of the prompt such they have slightly different weights, but still generate the same overall feel?

I can certainly change the punctuation of the lyrics. That's no issue at all!

I think I get what you're saying.

You'd like me to post my song and extension. Sure thing! How do I do that? If there's a tutorial on here, just point me in the right direction.

Thank you!

[half dragon]

Share your song and extensions: just post their URL here or PM for privacy.

[sturdy dove]

Oh my, is it that simple!?!? Okay. I don't think I've published anything yet, I'll figure that out first...

[half dragon]

You don't need to publish to share. Publish only makes it Public.
You can share private link as well; only the people you share with will see your private songs.

[sturdy dove]

Oh!

https://suno.com/s/0GIlz9ARqm3DxkHU

[sturdy dove]

I haven't figured out publishing the entensions yet... But thank you! I feel kinda foolish. : )

[half dragon]

Extensions? same way - use the URL.

[sturdy dove]

But... I'm so sorry, I again feel foolish. How do I navigate to an extension? Isn't that part of the song?

[sturdy dove]

All I've got is complete songs...

[half dragon]

Extensions are the gens that say Part 2 or Part 3... before you merge by Get Whole Song

[sturdy dove]

... Thank you! Checking

[sturdy dove]

VR, I'd love to keep this going, but I have to knock off for the night. Thank you for your offers of help! I realized I posted a crap version with a bunch of editing stuff in it. It makes sense in the editor. I'll find a clean version and post it up.

Are the prompt and lyrics helpful to post?

[half dragon]

No worries - remove the post if it wasn't the correct one.
Just come back when you have more time, @young geode would be more than happy to check out the outputs you got for further discussions.
I'd let the master ask for whatever he needs 🙂
Now go to sleep! 🙂

[neon ferry]

Can someone remove my automatic payment td

Tf

[spring mountain]

chirp-bad-hallucination

chirp-bad

[lyric kestrel]

Hi looking for help with a prompt,

What would you put for this?

https://youtu.be/1I9T8Fni9qU?si=JgLrx-xJ6Rp2yBXR

[hidden yarrow]

try german rap 😉

[azure crane]

Why doesn't every song start from the beginning. Almost every song I've created lately has started a few bars into the song. Is there any tip to avoid this?

[young geode]

I can't really tell you how to modify the prompt, because I don't know how exactly you wrote your prompt.

Sometimes it's too much guidance, other times it's not enough. You just need to make incremental changes to it and the scales.

I can say one thing however: SUNO has a 1000 character limit. If you're not even using a quarter of that, you need to flesh out more of what you want.

[polar spade]

Subscriptions are ONLY set up as auto-pay. If we cancel your subscription a week ahead of the next renewal , then it should not auto-pay the next month. Then we can re-up at any time.

It doesnt cancel simply by not having funds in a bank account at time of payment, for instance, but will continue to ping for payment as a courtesy, assuming we have not wished to cancel but simply ran low on funds in the account.

[polar spade]

As in mid-sentence? I know Suno often doesnt have an intro, but ive not had it start as if something was left out of the beginning parts. What's it sound like?

[azure crane]

Yeah, like jumping 1 bar or 2 bars into the song.

[polar spade]

Thats odd. Lyric-wise too , or just the music part of , say, the intro?

[polar spade]

Is that a new Create or a Cover of an audio upload so you know what we should have heard? )

[dry ingot]

Try this:

Lyrics:
[Intro – dreamy echo, delayed male spoken word over soft electric piano chords and airy effects]
(spoken, half-whispered)
Yeah… it’s late…
I’m still here…
Chasing something I can’t hold…

[Verse 1]
City lights blur when I’m movin’ slow
Heart’s still cold but I let it show
All these highs just feel so low
And I can’t tell where I should go
Pour my trust in a paper cup
Every word that you say, I hold up
Feels like I can’t get enough
Every hit just pulls me up

[Chorus – melodic rap-singing]
You got me caught in a wave I can’t break
Pull me under, I still want the taste
Feels like you’re in my blood
Like a drug, like a drug
Even when it cuts, I just stay
Chasin’ shadows in the mess we made
You’re in my blood
Like a drug, like a drug

[Verse 2]
Faded thoughts run through my head
Half alive but I’m full of dread
Every move just leaves me spent
But I’m hooked on the way you bend
Toxic touch but it feels so right
Pull me in when it’s late at night
Can’t step back when I’ve lost my sight
You’re the wrong I won’t rewrite

Style:
Dreamy ambient intro with delayed, echoing male spoken word in a modern hip-hop tone over soft, reverb-soaked electric piano chords and subtle atmospheric effects. Gradual build into a deep 808 trap beat with crisp hi-hats, tight snare, and punchy kick. Smooth transition between melodic rap-singing and rhythmic flow, emotional delivery with raw, vulnerable edge. Arrangement blends moody, cinematic electronic textures with contemporary trap energy. Warm analog sub-bass, wide stereo ambience, and airy reverb tails for depth.

[azure crane]

Well, obviously it's not starting from scratch. I've been playing music for 40 years so I know what it sounds like when a song starts. This applies to both Cover and Create New.

[dry ingot]

What genre are you making? I have this often with electronic music, my theory is it might have been trained from data that is from dj mixes that don't start correctly.
Are you uploading audio and getting the ai to cover it?

If you are uploading, perhaps you could try adding like 5 seconds silence at the start of your audio before the song starts? 🤷‍♂️

[azure crane]

It's 80s funky soul stuff and it's no audio uploading.

[dry ingot]

Yea I dunno, I have had it happen too, I just typically do the thumbs down (in the hope the model learns from it) and sometimes if it is bad I click the report button and then I try again

Other struggles I have include sometimes trying to extend and only getting like 2 seconds, or trying to finish a song (when i use the extend feature) and it just keeps going and going....

[azure crane]

I added [Start here] and it worked with one of the generated tracks now...

[dry ingot]

ahhh nice!

someone made a suno wiki here with some good tricks:
https://sunoaiwiki.com/

I noticed suno themselves use [outro][end] at the end of all their examples in
https://suno.com/explore

so I do that now to end mine. start here makes sense too!

[polar spade]

If Start Here worked, then good.

If its ignoring [intro], perhaps be more literal and spell it out [introduction], that was a thing back with 3.0 I think. Always worth a try.

[nimble pelican]

Hey there guys! Can someone help me out with some info about how I could influence longer songs with my paid plan?
I'm particularly keen to push the song lengths beyond 6 minutes in all cases, since I'm designing a game theme album and the runtimes are usually 5-6 minutes long.
Thanks in advance!

[dry ingot]

Are you unhappy with the extend feature?

Personally I have found after 2 or 3 its a whole new song

[nimble pelican]

I can't crack it... Tried it about 4 times in different ways and it's just adding me 10-15 seconds at-a-time.

[dry ingot]

oh

weird.

[nimble pelican]

10 credits for 10 seconds (sometimes only 2 or 5) is not a really good deal...

[dry ingot]

are you doing instrumentals?

[nimble pelican]

Yes!

[dry ingot]

yea I have had that sometimes

ahh perhaps do prompts inside [] in the lyrics section.

this sort of thing maybe?

[Intro – soft ambient pads, distant melodic motif, slow fade-in percussion]
[Verse A – main theme on plucked synth or soft piano, light bass pulse, gentle hi-hats]
[Verse B – same theme with added strings and layered arps for motion]
[Bridge – breakdown with atmospheric textures, low drones, filtered percussion]
[Verse A return – main theme restated with subtle variations, light rhythmic build]
[Chorus / Peak – full instrumentation, driving bass, wide stereo arps, layered pads]
[Outro – gradual fade, motif returns in soft piano, reverb tail]

[nimble pelican]

Will try, thank you!

[dry ingot]

no worries good luck

[twin walrus]

Hey! I've a question regarding duets or different singers in a song on suno. Does anyone know if there's a good way in v4.5+ to tag two different singers switching in-line, especially if they've the same gender? I know it's still pretty unreliable anyway.
Would I do this with leading metatags like f.e.

[Vocals:] gravely mezzo-soprano female, supporting young alto female, etc.;

And then reuse those tags [gravely mezzo-soprano female vocal] and [young alto female vocal] in-line?

Oh also, does l/r panning have tags? Although I can imagine that would be better done outside of suno.

[unborn star]

If you ever find a way to make it possible, just tell us. Looking for it as well.

Suno should put some Tags in, which AI recognize like [person 1- description], [person 2 - description] and then you can mention them. Would be awesome

But

I have never tried to give instruction to Suno, maybe it can understood them.

[ebon sand]

i'm not working with 4.5 but i usually just assign roles to them, my main singer is the hierophant and the other singers are the cultists
if i have a duet then they have roles in the song as well

[unborn star]

oh this works?

you have an example of this?

[ebon sand]

sure i can search some up

[unborn star]

1-2 is enough thank you =)

[ebon sand]

might take a few minutes, the duets are quite a bit in the back haha

[unborn star]

Don't worry take your time, you can take me anytime even if you need until tomorrow

[hard beacon]

i had good results with "male and female polyphonic harmonized vocals" in the past

[lyric kestrel]

i will try, thank you

[twin walrus]

Oh that sounds interesting! This would get rid of the need to write out the voice attributes all the time

[ebon sand]

I couldn't find my old duets so I just generated a few new ones, here an example, lyrics generated by gpt because it's just a demonstration haha
https://suno.com/s/XkjVQqf8E0GudtEm

it could be easier with the newer models because you can tell them the purpose in the styles box better, 200 symbols is quite limited but it works

[unborn star]

I was free to create the exact thing, which you created right now as remix

lets listen how it turns out now

[ebon sand]

it's not a real remix, i just never change the song names

[twin walrus]

So wait, is Hierophant just a tag for a male name which suno will assume is a male singer?

Also the "-banjo" made me laugh for some reason

[unborn star]

I made one

but it looks like it doesnt work this way

[ebon sand]

you can tell it its male when in doubt, i have used odysseus/siren also at some point

it's never 100%

but a good 60%

[twin walrus]

I see, well time to put it to test myself :D

[unborn star]

so you need to handle it different

[ebon sand]

oh you need to prompt 4.5 a looooooot differently

[unborn star]

I said, I just used exact your thing, to listen how it turns out. For experiment excatly because you used 3.5

[ebon sand]

yeah but i used 3.5

[unborn star]

=)

[ebon sand]

for 4.5 i'd write full sentences or have gpt convert it into a paragraph

[unborn star]

Wanted to see the difference if you use the same instruction as in 3.5

paragraph 🤔

[hard beacon]

for 4.5 and 4.5+ you can still words and use the "creative boost" function to write it down in sentences 🙂

[unborn star]

never heard of this before "creative boost"?

ah this beta button

I guess

[hard beacon]

for example: "Celtic folk, dubstep, electronic" is transformed into "The anthemic track opens with Celtic flutes and lively plucked strings weaving folk melodies across syncopated half-time dubstep drums and a deep, gritty modulated bass, The drop surges with dense distorted wubs, fierce synth stacks, and rapid triplet grooves, Vocal and instrumental chops punctuate sections for extra rhythmic punch"

[ebon sand]

and bluegrass is transformed into banjo with banjos and banjo elements

[hard beacon]

it's available on 4.5 and 4.5+

it's that little button in the styles window 🙂

[unborn star]

ye that's what I meant with beta button

[hard beacon]

ah..ok 🙂

btw.. there are a lot of beta buttons out there 😄

[ebon sand]

v4.5 responds alright to duets, v4.5+ is mainly just screaming for me

[hard beacon]

yapp 😄

4.5+ = Scream at me model

that's why i am creating in 4.5 and cover it into 4.5+

[unborn star]

only screaming?

xD

in duets?

[ebon sand]

i had 2 different people screaming at each other (men)

[hard beacon]

i don't think that he is screaming at you if you try to create a ballad... but.. well..

4.5+ is a very aggressive model

perfect for the metalheads out there..

[unborn star]

I dont feel like this

you mean just in duets right?

[hard beacon]

are you talking to me? 😄

[unborn star]

Im talking to you guys, but you said 4.5 screaming model

[hard beacon]

no no

not 4.5

4.5+

[unborn star]

sry

I meant 4.5+

[ebon sand]

i had very few results from 4.5+ i ever liked, it like metal, but i only like good metal and not like... well i cant name bands without offending someone

[hard beacon]

4.5+ has the "issue" that male vocals are way aggressive (depends on the prompts=

[unborn star]

🤔 but it happens in duets right?` Not in normals songs

[ebon sand]

and notes are held longer than any human ever would

[unborn star]

but since you said it, you can control it with prompts

[hard beacon]

@ebon sand : yapp.. ultra long 😄

[unborn star]

there is something xD
But I need to say I like it, but there are singers out, which can really hold that notes that long xD

[hard beacon]

4.5+ has an amazing stereo sound.. also clean vocals... but i hate when the man is screaming at me

😄

[unborn star]

dimash for example haha

[hard beacon]

but i think it's a prompt "problem".. if you want to create a song with "High-Energy" it looks like, that suno makes the vocal High-Energy (Screaming)

[unborn star]

Can't you say soft scream?

[hard beacon]

even with exclude prompts he is screaming at me 😄 but i think it's a genre problem

[unborn star]

so which genre especially, want to test it now

High-Energy?

Metal?

[hard beacon]

if you want to create a soft rock ballad or something like that, i don't think there are screaming vocals..

[unborn star]

I just want to test a lot, so you got a point, which I want to investigate

[hard beacon]

let me try to create an anthemic dubstep song with the prompts "emotional vocals"

[unborn star]

And I should remix it?

[ebon sand]

they also scream in the genres I create and there there shouldn't be any screaming, it's model bias

[unborn star]

to make it less screamy

🤔 Are you creating stuff with 4.5+?

Didn't see any song Lunacy

But maybe its not just public

[hard beacon]

I am not Lunacy, but my songs are all private 🙂

[ebon sand]

not anymore, i have 1 with 4.5+ public i think but it was just for the contest and no one will care about my contribution but there is screaming as well

[unborn star]

So should I try a song like this?

[unborn star]

ah ok

[hard beacon]

let me do it... 🙂

[unborn star]

ok

[unborn star]

You usually do it for yourself, don't you?

I use Suno just for creating piece for my pen&Paper game or contribution to birthday parties with a custome song

[ebon sand]

yes but in contests it's obviously to partake in a contest

[unborn star]

ye ok

If you want to compete that's something else

It's imiliar to politic anyway, if you have a lot of friends which care, you might win.

[ebon sand]

not seriously, i think there is no real competition but when they sometimes do contests i do things i wouldn't normally do so that's enough for me

[unborn star]

alright

Robin it looks like you made something =)?

[hard beacon]

working on it 🙂

[ebon sand]

i use suno only to create songs for myself basically, i don't share them much, but that's what suno is for, to create stuff I like

[unborn star]

That's the spirit =)

[unborn star]

I would say don't take to much effort in there. =)
I guess it was an example song, wasn't it?

[hard beacon]

well.. i try to create a remix of a track i've create in 4.5... so there a no credits only for you 😉

btw: the track wasn't with creaming vocals in 4.5

[unborn star]

I´m not saying that I want to take the credits for the song you created, I just want to investigate. After I finished it, I will erase the song

That's all for me.

Don't erase the lyrics, otherwise it would be pointless.

[hard beacon]

if i have a good example i will send you an private message

[unborn star]

ok

[sturdy dove]

Ah, interesting. Understood!

In my experience, the longer the prompt, the more confused AI tends to get. I've done a LOT of AI prompting across all sorts of genres, but not much with music. Since I was getting pretty close to what I wanted (aside from who sang what lyrics) I figured my prompt was okay. Again, I'm happy to share it all...

I hope to have more time this evening to jump back at this. Thanks again for the tips!

@half dragon @polar spade

[rocky zenith]

It was recommended that I ask my question here, although it isn't about prompts. I'm trying to edit a song and I can't get the edits to fall exactly where I want them. I can't fine-tune where the "cut" point goes. Is there a way to be precise?

[tawdry scarab]

I think of this as though there is a finite conceptual space, and the more you add to that space the more shallow each piece gets

[sturdy dove]

Vicki, I'll tell ya... In my (admittedly limited) experience, there's NOTHING precise about Suno or in general AI. Hopefully some other people here will chime in and help, but my experience with the editor is awful so far. And I've put hours into it...

[rocky zenith]

Ugh! Not what I was hoping to hear, but thank you for taking the time to answer. If I don't hear anything more helpful, I'll try downloading the stems and fiddling with them in Reaper.

[sturdy dove]

Yeah, if you have that technical capability, you'll be much, much better off doing that.

I'm not trying to be a pessimist, I'm trying to help you save some time and heartache. Suno is great with general concepts but precision is not its strength that I've been able to find.

[half dragon]

Even when you selected a section but it generated on a different section?
Or it generated more than what you selected?

[unborn star]

Can you erase your songs completly from bin into non existing?

I saw if you erase one song, you are able to access it still.

[half dragon]

Permanently removed from trash bin?

[unborn star]

y, I don't even find bin

[half dragon]

On your Library page, there's a little icon on top right?
or https://suno.com/me/trash

[unborn star]

you're right

never have been in library

[young geode]

Labeling, Ordering and punctuation help with this.

If you write a prompt that's something: Guitars, Rapping vocals, Metal drum, fast drums, vibes and moods

Then it might confuse it, but something like:

The track opens with heavy, thumping distorted 808 drums and a boomy wet bass, establishing a deep, sexy groove, Sparse, pulsing Minimal House hi-hats layer in, with subtle Darkwave synth pads weaving atmosphere, A mild, hypnotic 808 pulse floats in and out, while raw textures and gentle filters maintain an enticing, andante pace, Through out the song, the beat drops into deep, reverberating 808 drums, which drives the song forward with the energy of a race, The Vocals of the song are desynched from the main song, on their own time signature, Verses are sung in drift phonk, while Pre-chorus and Chorus are done in Phonk R&B.```

Won't.

What really messes up Suno is if you're trying to blend genres that want to fight.

[devout sinew]

Hello. I’ve spent hours crafting multiple song versions in Suno with the goal of generating modern pop-soul or reggae-soul male vocals with emotional delivery (rasp, vocal cry, fry, etc.), but I keep getting strong country or twang vocal tones, even when the genre and vocal style I’m specifying are far from country.

[unborn star]

@half dragon thanks for the advice =)

[young geode]

What's the prompt and negative prompt you're using.

[sturdy dove]

That's awesome to know!

I always thought the auto generated prompts it made from mine were... interesting. And so complex with frankly a bit of jargon I was a bit uncomfortable with. So I always deleted them and put in mine.

Let me find mine.

[elder dune]

so I have a random question - it seems we can't remaster with 4.5+ .. it can only remaster to v4.5 (no plus) and below, which is annoying, and here's why: I got a mostly decent song put together using 4.5+ (which is REALLY nice with metal, btw), but I had to edit / replace a few lyrics due to typos on my part resulting in bad pronunciation. So I fixed them up, but the volume level of the replaced section was 3-6 db noticeably louder than the surrounding stuff .. not blended very well at all. Previously, I've been able to resolve this after the fact by remastering the whole track - the remaster picks up the new section, but blends / mixes it nicely. However, since I can't remaster to 4.5+ now, when I try that, I get a result that's good for 4.5 (no plus), but sounds muffled and hollow compared to the rich, full sound I get from 4.5+ ... I worked around it and eventually got a good song, but that's kinda annoying anyway. Anyone know what's up with that?

[sturdy dove]

Oh yeah! Mine were SO minimalist. I was just kinda letting it do its thing.

"duet, synthwave, japanese electronica, sad but upbeat, 160bpm"

I can't find the prompts it came up with, but they were a little long and nuts and they didn't make sense to me.

My lyrics on the other hand, that's where I played a lot with direction. Again, I tried to keep it minimal to keep from confusing it, but that's where I did a lot of tinkering and moving stuff around. When the man would sing where the woman should, I'd put in brackets right next to where he was supposed to sing and explicitly bracket that line rather than just start the verse... stuff like that. And I even tried counterintuitive prompting in the lyrics.

[half dragon]

Yeah, there is no 4.5+ Remaster. Not sure why but it's definitely annoying when we needed to run it through a Remaster.
Report it #🪲┃bugs-web - if more reported - it might come sooner.
It should always come with a new model it seems but it didn't this time.

[half dragon]

Have you tried the Magic Wan on the style box? after you put in your minimalist prompt, click it and it'd enhance your prompt

[elder dune]

^^ that's what I do, FWIW. type my minimal prompt in and then hit the wand for AI to expand it. The results are usually pretty awesome. ChatGPT can sometimes be a great help, too, if you ask it for a suno prompt and flesh out the style you want.

[sturdy dove]

1/3
Here are examples of my lyrics prompting... Here's where I started:

[a duet between a male american baritone and a female japanese soprano]

[Verse 1 - Male Vocals]
Twenty years we've been together
From the showroom to today
Every mountain road, every corner
You respond the perfect way
Since 2006, every apex kissed
You're the one who never lets me down
But I can't take you where I'm going
And it breaks my heart

[Verse 2 - Female Vocals]
I'm always waiting, always ready
I love to do just what you say
I obey in every way
Twenty years of girls that come and go
I've outlasted them all
You kept me cozy every night
I perfectly serve you
And you've kept me [emphasize] so nice

[Chorus - Male Vocals]
Geisha, my sexy JDM Geisha
I respond when you call to action
Across the ocean, I'll remember
This perfect dance we've shared
Geisha, elegant hawkeye
I'm faithful, strong, and true
You keep my wild on line
[emphasize] Now I'll give my all to someone new

[Bridge - Male Vocals]
Built when they cared about the drive
Perfect ones are hard to find
From that first day at the dealer
I knew you'd be forever mine
And life is pulling me away
[emphasize] So forever's got an ending

[Chorus - Female Vocals]
Geisha, my sexy JDM Geisha
I respond when you call to action
Across the ocean, I'll remember
This perfect dance we've shared
Geisha, elegant hawkeye
I'm faithful, strong, and true
You keep my wild on line
[emphasize] Now I'll give my all to someone new

[Bridge - Male Vocals]
Built when they cared about the drive
Perfect ones are hard to find
From that first day at the dealer
I knew you'd be forever mine
And life is pulling me away
[emphasize] So forever's got an ending

bah... ran out of characters...

[rocky zenith]

No. What I'm doing is uploading my original songs and having Suno generate covers so I can use the stems for my own production in Reaper. Suno adds a lot of instrumental stuff and I wanted to edit some of it out. So I put a split in, but I can't get the split exactly where I want it. I end up leaving too much in or it will cut off too much at one end of the cut.

[sturdy dove]

2/3
Attempt to fix:

[Chorus - Alternating]
[Male Vocals] Geisha, my sexy JDM Geisha
[Female Vocals] I respond when you call to action
[Male Vocals] Across the ocean, I'll remember
[Female Vocals] This perfect dance we've shared
[Male Vocals] Geisha, elegant hawkeye
[Female Vocals] I'm faithful, strong, and true
[Male Vocals] You keep my wild on line
[Female Vocals] Now I'll give my all to someone new

[Bridge - Vocals 1]
[Male Vocals] Built when they cared about the drive
Perfect ones are hard to find
From that first day at the dealer
I knew you'd be forever mine
But forever's got an ending
And life is pulling me away

[Short Outro]

[Final Chorus - Duet]
[Male Vocals] Geisha, my sophisticated lady
[Male Vocals] I must say goodbye now
[Female Vocals] Good ones like me don't hang around [Emotional, drawn out]

Anyway, it ignored all of my brackets every time, so I thought I was doing too much.

To answer your question, yes, the Magic Wand. Every time I tried it, the result was far from what I wanted.

And I didn't understand a lot of what it was saying, so I preferred to just keep it simple.

This was all 4.5+ btw, I had Pro then.

[sturdy dove]

I'm no pro, but that definitely sounds like a job for Reaper to me.

[devout sinew]

Emotional reggae-soul fusion ballad with slightly upbeat groove. Male vocals in a modern, soulful tone — husky, youthful, raw, with rasp, vocal fry, and emotional phrasing. No country or folk tone. Guitar uses mellow reggae-style skank rhythm (offbeat chords) with deep pocket bass, subtle organ, and rim snares. [Intro – syncopated electric guitar skank, soft bass groove, 80 BPM, 4/4, key of E major] [Verse – intimate vocal phrasing, breathy rasp, storytelling style, minimal rhythm] [Pre-Chorus – slight pad swell, offbeat tension builds, vocal cry okay] [Chorus – soul phrasing with reggae sway, vocal fry optional on “send,” falsetto cry on “oooh”] [Bridge – stripped rhythm, vocal break, reverb echo] [Final Chorus – harmonized background vocals, mellow swell, no belting] Feel: grounded, emotional, human — a groove that holds the heart, not the feet.

Exclude: country

[half dragon]

When you say "put a split in" ... so you loaded one of the instrument Stem in the editor and split it? but doesn't work?
Are you trying to get Suno to regen it without a specific instrument?

[sturdy dove]

I had a song I was really happy with months ago - one of my first AI projects, and I did something similar and it came out perfect.

[rocky zenith]

Yeah, I'll have to try that next.

[rocky zenith]

No...I'm sorry; I'm not providing enough information, I think. I uploaded my original song. I told Suno to "cover" it. I selected one of the resulting covers and clicked "edit". The editing screen shows sections for the verses, choruses, and what it calls "outro"--which are instrumental bits between vocal sections. I tried to select and delete one of the "outros" in the middle of the song, but it cuts off part of the vocal at one end and leaves too much of the section at the other end. When I try to move the split mark, it jumps too far. I can't make tiny adjustments.

I don't know if that helps. It still sounds a little confusing to me, but I can't think how else to describe it.

[sturdy dove]

You can zoom in substantially on the timeline....

[half dragon]

Ah... yeah... use Zoom tool at the bottom of the editor to zoom it all the way to the waveform to make a precise selection

[sturdy dove]

It looks like it deleted a couple of messages. That last one was preceded with

OOOOOOOOOH, YEAHHHH Actually...

So try that first before diving into Reaper, @rocky zenith - I thought you meant a particular instrument. gtg, good luck!

[young geode]

First, remove the country and folk from the prompt (that can confuse the AI), as well as a lot of the Intro/Verse/ect commands. Prompts like 'storytelling style' are likely moving it more towards country as well, given the nature of modern country.

Second, what kind of Reggae-soul sound are you looking for? A lot of users have this issue, oddly enough, where they get the country sound instead of whatever else they're looking for. Typically the solution is to first zero in the vocal sounds they're looking for, by taking the artist/genre they want to emulate and having an AI print out that vocal profile. Then, condense that profile into Vocals: Sex, Type, Range, Style, Qualities.

Third, expand out your negative prompt, Including the Mother Genres Country and Folk: Country: Bakersfield, Nashville, Contemporary, ect. Folk: North-American, ect and the Eras you don't want, Eras: 19xx, 20xx

[storm pawn]

Maestros, when I create songs with my own lyrics (custom tab) - is there any way to specify within the lyrics which should be sung by male and which by female? Suno sometimes does it to choruses by itself, but I would like to have more control over it

[rough reef]

I can't access the website from the Canary Islands, Spain. Neither through Firefox, Microsoft Edge nor Chrome. Any solution?

[young geode]

Not consistently, no.

I have found that 'sandwiching' lyrics between the metatags works, but I haven't fully tested it.

So

Say this, then that < Female
[Vocals: Female]
Say that, then this < Male
[Vocals: Male]
Say what, who, when < Female
[Vocals: Female]```

And for whatever reason that seems to work. However, this was for one song I did and I noticed that it was responding in that way, after having difficulties getting the two vocal patterns to stop hallucinating.

[graceful condor]

Honestly I am really loosing my patience. For hours I am sitting onto one freakin song. Suno is so annoying when it comes to create Rock, Progressive Metal songs. All I want is a nice brutal riff ending for the song. All is sang. All vocals are done. And I just wanna extend the already existing brutal riff with nice pentatonic solos and then END.

All that is left is this in the Lyric section:

[Instrumental Outro: Heavy, brutal aggressive guitar riff with soaring solos, transitioning into a massive, aggressive Metalcore breakdown with chugging guitars, thunderous double-bass drums, and deep bass. Use layered heavy guitars, orchestral strings, deep bass, tom-heavy drums.
[End: Hard stop after breakdown]

Und Style:

{Tempo: 70-80 BPM} {Genre: Progressive Rock, Post-Rock, Metalcore} {Instruments: Layered heavy guitars with brutal riff and chugging breakdown, orchestral strings, deep bass, tom-heavy drums, double-bass drums for breakdown} } {Mood: Apocalyptic, sinister, aggressive, cinematic} {Mix: Crystal-clear, saturated, vibrant mix, crisp, no distortion or bleed} {Production: Polished, professional, no artifacts, natural resonance, consistent volume and EQ matching the original track, no dumpf sounds, full extension with hard stop, no cut-offs} {Form: Outro} {Performance: Dynamic instrumental swells, aggressive breakdown}

Nothing says anything about sining. All lyrics are done but in each freakin extension for 2 hours already he screams "the night" all the freakin time and it makes me nuts. I even downloaded to Song I had so far. Edited it in Software cutting off the none essential part at the end and uploaded it fresh on Suno. I refreshed the page and extended the upload file just for suno not to remember some weird stuff and mix it in again. AND YET AGAIN "the night!!"

What is wrong with this AI? I wasted hundreds of credits for nothing.

[storm pawn]

Superb, thank you! I will testing it today 🙂

[rocky zenith]

Oh! I was looking for a way to zoom in and I couldn't find it. Until you said that. Thanks!

[unborn star]

between Metatags, what do you mean? Do you have a better example?

Don't understand, what Say that, then this < Male should do

[young geode]

Exactly as I wrote it

Lyric
Tag
Lyric
Tag
Lyric
Tag```

The <Sex is just signifying whose singing that line to you, the reader, so you know.

[unborn star]

Hmm? But isnt it sung?

[young geode]

The tags won't be sung if they're tagged properly.

[unborn star]

I just mean < Male =)

[near flume]

what were you using in the Styles box for the voice declarations?

[young geode]

Ya, thats not actually in the lyrics. I put <Male so that you know who is singing that line.

[unborn star]

ok

never worked for me, when it song is much longer

[near flume]

You're putting too much into the Outro tag. Break it up into sections as yobu want them to occur. You can also get rid of the the content in the Styles box since your extending from the songs styles. Check out MasterofSFL in #workflows there's a spreadsheet that has tags you can use for sections and events.

[half dragon]

I think that was just a "note" to indicate - don't add it .. use tags [..] only
Keep in mind, making duet is not reliable - it will take lots of time, lots of trial and error and lots of credits

[unborn star]

Ik tried a lot with it

but never got it perfect

[hybrid nexus]

/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

[graceful condor]

When can you leave the Style section out? At the end?

[young geode]

There is a way, but the sad fact is you have to do it with the AI in mind, and heavily texture the lyrics so that the lines you want sung and switched have distinct qualities. But unless you are aware of that, the duets are going to come off as random.

And most users are already at a point in their generations that it'd require a complete rewrite, and a loss of what they already achieved.

[hexed oriole]

Why not leave the lyrics box empty and hit the Instrumentals switch?

[graceful condor]

Style influence slider has nothing to do what is written in lyrics correct?

[dense moat]

Try this in the Lyrics Prompt....

[Key: D major, Tempo: 92 BPM, Time Signature: 4/4, Genre: J-Pop Ballad, Style: Romantic, Melodic, Backing: Soft harmonies, Lead Vocals: Male & Female alternating lines, Male Voice: Warm, baritone, Female Voice: Soft, airy soprano]

[Chorus]
"Geisha, my sexy JDM Geisha" (male)
"I respond when you call to action" (female)
"Across the ocean, I'll remember" (male)
"This perfect dance we've shared" (female)
"Geisha, elegant hawkeye" (male)
"I'm faithful, strong, and true" (female)
"You keep my wild on line" (male)
"Now I'll give my all to someone new" (female)

Try this is the Style Prompt...

Key: D major, Tempo: 92 BPM, Time Signature: 4/4, Genre: J-Pop Ballad, Style: Romantic, Melodic, Backing: Soft harmonies, Lead Vocals: Male & Female alternating lines, Male Voice: Warm, baritone, Female Voice: Soft, airy soprano

https://suno.com/s/D2wdmJgeWQhjV0ti

[sinful marsh]

Does Suno really notice anything in () style brackets? I find it doesn't all the time... usually [] ones work better but then my knowledge is very limited at this stage. Is there any good reference material explicitly for Suno AI when it comes to prompting SFX and dueting vocals? I tried making a song with a couple of characters (Duke Nukem vs Commander Keen) are there such a thing as child vocals on this platform? (to keep the sound authentic).

Example of the styles I added in:

Begins with a cheesy, heroic 90s synth intro, launching into aggressive, chugging metal—distorted thrash riffs, powerful drums, and frantic fills under an over-the-top macho vocal. Sudden switches to bouncy, off-key 8-bit chiptune dominate during Keen’s high-pitched, innocent phrases. The sections jarringly intercut; in the finale, Duke’s metal devolves into noisy chaos and collapses, leaving only Keen’s cheerful synth melody, which playfully drifts away on fading 8-bit arpeggios. Commander Keen = male boy vocals, duke nukem = A deep, gravelly, and exaggerated baritone. His voice is rich and masculine, a perfect fit for his action hero archetype. The voice is a direct caricature of 80s action movie stars, specifically drawing heavy inspiration from actors like Arnold Schwarzenegger and Bruce Campbell. It's a stereotypical "macho American" accent, often punctuated by a deep, throaty chuckle after a one-liner.

[little cargo]

So i have a persona i have been making a album of songs for.. I just went to download them... and the download option is gone .. i can download all my other music... but for some reason .. this one persona will not let me download anything made out of them. the options to download are grayed out.... please help

[half dragon]

(..) in lyric usuallly makes Suno sings as backup singers Metatags are [...]
Children vocals? maybe try adding "children music" in the style?
This has been shared https://sunoaiwiki.com/ - not maintained by Suno but it might help?

[half dragon]

No Download option? Gone? you got a screenshot?

[sinful marsh]

Thanks for the tips VR!

[tawdry scarab]

yes it does and yes not all the time

it's maybe the least acceptable way

[sinful marsh]

Sometimes it does the vocals purely by accident and really well. when I try to prompt for different vocals how ever.. very hit and miss.

[tawdry scarab]

any character consistently and intuitively used as syntax has a better chance than () but anything not evidently syntax is worse than () but still can work

[midnight belfry]

Is there a structure that is recomended to create the tracks?

[little cargo]

sure.. its my only music that is grayed out.

[half dragon]

Interesting.
I noticed that on couple of my songs too - but they were "contest" songs
I don't have Persona songs so can't test to see.
You might want to report it? #🪲┃bugs-web

[honest cargo]

Any good tags for doing national anthems

[half dragon]

Check these out:
https://suno.com/search?type=public_song&q=national anthem

[sturdy dove]

Dude, this is a craaazy prompt! THANK YOU for doing this! My goodness. Let me check your piece!

[sturdy dove]

So, interesting! This is not at all what I'm going for in any way, but the lyricist switches at the same point. "You keep my wild on line" is supposed to be the male's line and it switches in yours, too. Very interesting!

[boreal zinc]

Why does it insist on repeating the 4th verse instead of continuing? The song was perfect up until that point! I tried to extend from there, but it still repeated the 4th verse even though it was not in the lyrics I gave it for the extension. EVERY TIME

[sturdy dove]

Alright all y'all punks.

I've resubscribed and I'm going to try again. I went back to an earlier version of the song and the lyrics are not remotely synced, which makes editing impossible. Any tips on that one?

@half dragon @dense moat @polar spade @young geode

[sturdy dove]

Bah. I'm starting by trying to extend, but it won't let me add new lyrics in extend...

[sturdy dove]

So here's what's in the lyrics in Editor:
[Duet featuring Vocals 1 as American Male and Vocals 2 as Japanese Female]

[Short Intro]

[Verse 1 - Vocals 1]
Twenty years we've been together
From the showroom to today
Every mountain road, every corner
You respond the perfect way
Since 2006, every apex kissed
You're the one who never lets me down
But I can't take you where I'm going
And it breaks my heart

[Verse 2 - Female Vocals]
I'm always waiting, always ready
I love to do just what you say
I obey in every way
Twenty years of girls that come and go
I've outlasted them all
You kept me cozy every night
I perfectly serve you
And you've kept me so nice (draw 2+ beats)

[Verse 2 - Female Vocals]
I'm always waiting, always ready
I love to do just what you say
I obey in every way
Twenty years of girls that come and go
I've outlasted them all
You kept me cozy every night
I perfectly serve you
And you've kept me so nice (draw 2[Chorus - Alternating]

[Male Vocals] [Female Vocals] [Male Vocals] [Female Vocals] [Male Vocals] [Male Vocals] [Female Vocals] [Male Vocals] [Female Vocals] [Bridge - Vocals 1]

[Male Vocals] [Short Outro]

[Final Chorus - Alternating]

[Male Vocals] [Female Vocals] [Emotional, drawn out]

So at Chorus - Alternating, it goes off the rails and there's no way to fix it. Let's try something...

@half dragon - check this out. This is what's in the window, not what I entered. And I can't find a way to edit it...

Hmmm... I thought I'd figured out how to fix/align the lyrics, but it's not taking my edits.

I had it do a replace with the correct lyrics, but it sounded the exact same and didn't fix anything. The lyrics are still incorrect in the main song lyric box. : / This isn't going well. Again.

As you can see here, the highlighted section is called "VERSE 2 - FEMALE VOCALS", but what's highlighted is a copy of the second verse. What I just "replaced" this whole section with was the alternating chorus lyrics, which is what the sound is but is not what's showing in the box. The main lyrics box remains all incorrect with no apparent way to fix.

Thanks again for all your help and encouragement! I'm hanging out working on something else...

[half dragon]

I'd first and foremost : make sure the display lyric matches what's the audio sound first.
Not sure if the lyric you posted above before or after the edit but I see this part
(draw 2[Chorus - Alternating]

It seems missing a ')' ??
Make sure all tags are closed and all parenthesises are properly closed

This line might be problematic too:
[Male Vocals] [Female Vocals] [Male Vocals] [Female Vocals] [Male Vocals] [Male Vocals] [Female Vocals] [Male Vocals] [Female Vocals] [Bridge - Vocals 1]

Editor is trying to tokenize the lyric into sections and there's no way it knows what those are and what section.
That might need to be clean up before loading into editor.

Edit your display lyric to match the audio as best as you could into basic sections for it to work.

[Intro]
....
[Verse 1]
...

[Verse 2]
...

[Chorus]
...

[Outro]
[End]

Give that a try and see if it helps better.?

[sturdy dove]

That's what was generated by Suno, not me.

The is the end of the second Verse 2 that Suno threw into my lyrics. My lyrics aren't like this.

I can't edit the lyrics in the song I'm working on... Or, I haven't figured out how to do that yet.

[half dragon]

You mean after your attempt to edit?
Yeah, my edit also ended up like that - it's a bug it seems - so I usually had to clean up after each edit

[half dragon]

The display lyric? 3 dot menu on the song > Song Details > Edit display lyric

[sturdy dove]

Ok, checking

[sturdy dove]

Oh, so I'm in the editor.

I'll go to the song

OH
In the Editor, there's a selection under the 3-dot there "Match Lyrics to Audio"!!! Trying that.

@ebon sand

[half dragon]

no from your workspace orl ibrary

[sturdy dove]

Okay, one sec, - thank you!!!

[half dragon]

This is the last resort.

[sturdy dove]

OH. Oops. That's okay, this is a saved edit

Honestly, I kinda think this thing needs the last resort at this point! As soon as it's done, I'll go do what you suggest

[untold briar]

I'm just skimming quick so this may not apply to your problem, but it's possible to fix lyrics using Covers and other features instead of Replace as well, which in some cases works better.

[sturdy dove]

Oh, it didn't work remotely anyway. It just left all the garbage as is. OKAY back to your suggestion1

[untold briar]

Or a Cover can smooth over a rough replace patch or first draft.

[sturdy dove]

I was originally (tonight, anyway, this is a month-long project, trying to fix this) going to just extend from where the problems start, but I can't do that, either.

[sturdy dove]

Okay, I'm in the song now...

I see it! Edit displayed lyrics!

Okay, so change them here to be correct???

Woo hoo

Done and saved on a new edit!

Okay, editing

LOL In Edit it reverts back to the wrong lyrics

Back > Displayed lyrics show the correct lyrics. But when I go into Editor they're wrong again.

What step did I miss @half dragon ?

(AND THANK YOU AGAIN for your help.)

Here's the song I'm working on btw
https://suno.com/song/65403cde-cb53-41a7-a944-219d8cc8edbf

[ebon sand]

It's better to use the old editor and you need to align the lyrics in the editor, both legacy and the new one have a feature where you can align

[sturdy dove]

Doesn't work

[sturdy dove]

I'll try the legacy though...

[ebon sand]

The lyrics alignment in the editor doesn't work? At what point does it fail?

[sturdy dove]

The lyrics are aligned in legacy!!! That's cool!

In the new editor, I can't select the lyrics. And they're completely wrong in the first place in the Editor

[ebon sand]

You have to select the alignment option in the new editor and correct them there but the new editor doesn't work anyway even if you do that

So just give it your best shot in legacy

[sturdy dove]

I know, but the lyrics are picked up wrong in the new editor, so you can't edit them. See the screen shot above?

Yeah, I'll try legacy... I'm watching / listening now to see if they're truly aligned

Thanks for the tip!

[ebon sand]

That's why you need to align them in the new editor, there's an extra mode to do that

[sturdy dove]

Okay, in Legacy when it gets to the chorus it loses track of the lyrics. They're there, but they're not highlighted

I know, but it doesn't work

I can't align what's not there...

Well, okay, talk me through it - maybe I'm doing it wrong!! I've been trying for a while

I'll go back to modern Editor

[ebon sand]

this is lyrics alignment

[ebon sand]

if you never saw this you were not there

[sturdy dove]

See how it's picked up Verse 2 twice?

I can't edit those lyrics

[sturdy dove]

from here. That I've been able to figure otu

Right, tried that a little bit ago

didn't work

[ebon sand]

then try this one in the old editor

[sturdy dove]

Trying it again

Saving changes

Oh wait, crap, old editor! Bah, I'm in the new. I'll catch up in a sec!!!

Didn't work in new, going back to legacy

[ebon sand]

what does it do when it doesn't work, usually it needs around 1-2 minutes

you reported back quite quickly that's why im asking

[sturdy dove]

It just didn't pick up the right lyrics. When I play the song, it's highlighting the wrong lyrics for what's being played

[ebon sand]

and when you change the lyrics in the alignment? e.g. removing the tag

[sturdy dove]

Sorry, I'm trying to figure out legacy now. Should I be looking at Legacy or modern editor - want to be sure I'm with you here...

[ebon sand]

i'd not use the new editor for anything, the replacements work poorly and when you save the song the quality gets worse

[sturdy dove]

Ok. So, I'm in Legacy. It seems to break down at the Chorus. The sections are split incorrectly

I don't see how to create a new split...

[ebon sand]

[1] for example

[sturdy dove]

I'm so sorry, I don't get it

See the break between verses? In VERSE 2 there's a clear break between sections. But I don't know how split that section there. Also, I see that the lyrics are repeated here for some reason. There's only one Verse 2 when the song plays, but there are two in the lyrics section. So I could delete one of those

[ebon sand]

yes only type whats there and to create a split you can add any tag

[sturdy dove]

Then I hit "Confirm and Align Lyrics"? ... Add a tag....

I deleted the excess lyrics. So I hit Confirm & Align now?

[ebon sand]

yes

[sturdy dove]

And it'll figure out the break?

alright

[ebon sand]

lyrics
[some tag]
lyrics

"some tag" is a split

it just splits wherever you have a tag

[sturdy dove]

So the Chorus tag is there, but there's no split

[ebon sand]

reload once and you could remove the inline tags [male vocals]

[sturdy dove]

ok

ahhh.... it's coming in!

[ebon sand]

yeah it's not more than this, you can type anything there

some words are just color coded like verse, chorus, intro

[sturdy dove]

I see. I wonder if I have too many tags because of my alternating chorus, that's what might be throwing it off

Still waiting for the song to load

Hm.

The break is still not there.

[ebon sand]

you'll figure it out, I'd remove the tags and just structure it again into whatever you want

[sturdy dove]

Dude I've been working on this for a month. I'm not likely to figure it out.

[ebon sand]

why don't you remove all the tags and set new ones

[sturdy dove]

So you're saying ... just have a Chorus tag....
I can do that!

[ebon sand]

aaalll the tags

[sturdy dove]

Just have a chorus tag and then leave the vocalist tags out altogether?

[ebon sand]

why do you need them now?

the song is created, do you still need them for something

i'm very sure the issue is not the chorus tag but other things around it

[sturdy dove]

Because the problem with the song is that the vocalists switched in the middle of the alternation, thereby making the song not make sense.

Yeah, I'm trying to fix the problem and make the vocalists sing the correct lines

[ebon sand]

yeah you don't need the tags in the lyrics for that, you can safely remove them and just assign a new instruction when you edit the line

[sturdy dove]

But this convo is giving me insight for how Suno works and perhaps why it's breaking

Okau

trying that

[ebon sand]

it's not broken, currently it's just a formatting thing, where the visual separations are doesn't matter, i usually have none, if the lyrics are aligned you can edit and if they are not aligned you have to align them and sometimes that means to spell a word phonetically

[sturdy dove]

Okay, so I've removed the vocalists tags, now I'll hit confirm and align

[ebon sand]

what about the verse 1 tag and verse 2

[sturdy dove]

Oh, you want me to take that out too, I forgot! Yep, on it

[ebon sand]

I'd try it and see that the song has nothing and then align everything and then structure, if even

[sturdy dove]

Alright, all tags removed - just lyrics remain

So this is weird

I can't get the orange to go away even though there's no tag in that last line. That tells me it's seeing something that I don't want to be there.

[ebon sand]

is it taking a while to align or is it instantly done?

[sturdy dove]

I haven't ... AH
I didn't hit the button, but yes, now it's all green

Okay, so now I Confirm and Align?

I never would have gotten this on my own.

[ebon sand]

yes, when it's completely green its at least a good sign that its picking up the changes, now you could structure it again or make sure first that the words are all aligned

[sturdy dove]

Okay, checking

[ebon sand]

and when everything is aligned then just add [verse] [chorus] [1] [2] or whatever you need to navigate the song, you can also leave it blank, it's really just visuals

[sturdy dove]

so far, loosely correct

Okay. Then.. COOL! I get it now. So once that's done I can go back to extending or whatever because the lyrics will be right.

OKAY. I see the problem.

It breaks down about halfways through the second verse. Loses alignment

so.... How do I fix that?

[ebon sand]

depends on what's there, you could split the syllables at the position it doesn't understand

[sturdy dove]

It's in good alignment I SEEE

Okay

So the problem is that it's not understanding the words. I need to help it interpret somehow

okay

I don't see how to fix it

It breaks at
"I obey in every way" - it's matched through in, but then it stops matching.
I've tried changing it to "evry" instead of "every"... Is that the right track?

sorry, it's matched through "in" in the line "I obey in every way"

but then it stops

[ebon sand]

you could try fun like "ev ree way" "e ry way"
but it could be that the next line is the issue and not this one

[sturdy dove]

Oh, I see

[ebon sand]

when there's an issue afterwards it often will drop the alignment a line earlier

[sturdy dove]

Okay, so when I make changes, how do I test it to see if the adjustments have fixed it? Is it automatic, or do I need to do something?

[ebon sand]

you need to hit save and basically just watch it again, sometimes its a bit jumpy and if half a sentence is missing in between its also not that bad if the rest aligns again

[sturdy dove]

Okay!

[ebon sand]

you only need a syllable match if you're replacing

[sturdy dove]

So confirm

Trying that

[ebon sand]

the new lyrics alignment is less clunky with this but the new editor just isn't good

[sturdy dove]

lol, I'm it matched up

Checking the whole thing now

I've learned so much tonight'

Okay, I'm a keep going

Okay.

[ebon sand]

i hope it even brings you towards what you want to achieve haha, i only align lyrics when i really want to replace, but at least you'll be a master in it after this

[sturdy dove]

So then I retag it. And then it'll have the right breaks. And then I can bring it in to the new editor and try to fix it with regen or extend or whatever

[ebon sand]

haha no, if you want to edit in the new editor you'll have to align in the new editor, they are not interacting with each other

[sturdy dove]

Okay, I see what's happening. Some other fixes I need to make. I'ma git to work and report back later!!!

OH I see.

Hm.

Okay, welll... I guess wish me luck then! I'll just stay in Legacy.

And you're correct, the new editor is not good.

[ebon sand]

but you can copy what you have now and try in the new editor, maybe it aligns now

[sturdy dove]

Yeah, that's what I was going to do. But anyway, I have a lot of work to do now. I've taken way too much of your time, Lunacy. THANK YOU

[ebon sand]

you are welcome, good luck!

[sturdy dove]

Thanks! I'm going to need it!

Are you still there?

If you're still there, and I'm so terribly sorry to bother you again @ebon sand but the next problem I think is that there are two words spoken over each other. I don't know how to fix that. Should I just remove the lyric that's spoken over and causing the alignment issue?

[ebon sand]

you can try either, whatever aligns them, it doesn't have to be truthful, i can show you how my aligned lyrics often look haha

[sturdy dove]

Okay, I'll just keep messing with it. Thank you!

[ebon sand]

here an example, broken down to vague phonetics, it didn't align in any other way

[x]
Ai no sere na no skon se,
o pho re exu see ya
En the os legi toi he mon; o te mon so fee ya
os cross exu see ya
an hura ti ex u see a
a hu de du da ee vos en do hi a nu at a fo os
Ina ri us, Ha gi os Kee ee ee

[sturdy dove]

Holy crap. That's Celtic, right?

[ebon sand]

no, it used to be greek

[sturdy dove]

OH, cool!

I was joking, but cool

[ebon sand]

or here, had to split "ruin", "spe ris" is "spirits", sometimes its like that...

Scars carved into the land, lines of ru in.
Spe ris silenced by hatred.
Where Astaros flames died, a path way o pens

[sturdy dove]

I see

Interesting. I'll keep messing with it. Thanks again for all your time and help. I'll report back when I've made some progress and I'll leave you alone until then

[ebon sand]

no problem, it's funny sometimes to align them, with english you'll have less issues though

[devout sinew]

Hello. I’m experimenting with prompts and I want to make sure I’m steering Suno correctly. How can I prompt if I want:
Only guitar and piano as the instruments.
No drums, no bass, no strings, no extra layers — just pure vocals with guitar and piano accompaniment.
I’ve tried stripped/acoustic prompts, but sometimes Suno still adds percussion or other textures. Is there a best practice for phrasing the prompt so it strictly keeps only the instruments I want?

Thanks in advance

[neat field]

My current process is not pretty for this. Make the song in the style you want most. Do not include acoustic or piano in styles.

Extract the Stem for the vocal manually or with Suno stem feature.

Reupload the vocals

Cover the vocals with 94% audio influence, 100% style influence, 0% weirdness with style "acoustic session, minimalist isolated piano"

[half dragon]

Like this?
https://suno.com/s/IzR67iWoc5sEEVWw

"The song opens with acoustic piano chords and delicate nylon string guitar fingerpicking, The interplay between piano and guitar establishes an intimate, organic texture, Subtle dynamic swells highlight transitions; the arrangement remains sparse, keeping focus on the picking style"

Weirdness: 33%
Style Strength: 90%

[devout sinew]

is this at prompt box?"acoustic session, minimalist isolated piano"

[devout sinew]

thanks I will try this

[sturdy dove]

I hate Woodhouse

If you're following, CraftyWolf, VRKing, Lunacy, Dasvon, Jonathon Fly, MasterofSFL...

I've got my lyrics aligned in Legacy Editor. Extended there to create a new version that I could bring into the modern editor. Using Crafty's lyric prompting to see if I can fix there.

The reason I tried to @ everyone is to say thank you and keep them updated if they're interested.

[young geode]

Unfortunately, I have zero knowledge of how to mess with the editors, except that you can do some free generations with them for minor fixes.

Are you looking for a very specific sound or the vocals for this?

[sturdy dove]

Yeah, the whole song is perfect but for a couple of lines.

It may have been your advice that I regen the entire section rather than just the broken parts. I think it was either you or VR. I'm on it. Thanks again for the help and advice!

[hollow veldt]

My apologies I am new to this Discord and also new to Suno 4.5 but loving it. I have a style on the song I've already completed and simply says Rock. I would like to be able to change taht information and save that song or even save as a new song with that updated style info. I edit the song and replace the style with what I want and Save As New Song and the new song still shows as rock in the style box. Can you help with instructions on how I can accomplish this? Thanks for your time

[half dragon]

So you want to edit the style display after the song created?
That's one thing you can't change after the fact.

[devout sinew]

Thank you VRKing, I used your prompt and W33% and S90% it worked thanks

[hollow veldt]

Thanks for the fast reply, I was afraid of that though not sure why that is locked. Only wanted to change it because how it displays in my Library but not really a big deal. I will know next time. Thanks again!

[sturdy dove]

Welp, even with the aligned lyrics, it's not responding to the prompts again. It just chooses female or male willy nilly regardless of the prompting. Same as before. Trying extend now, but it doesn't match the rest of the song well.

The voices don't match the rest of the song.

Same problem in Regen in Editor.

[half dragon]

Unfortunately, duet is the most challenging thing to create.
It won't be a smooth ride for sure with all of the quirks that the current system has.

[sturdy dove]

Yeah. So what I'm doing right now is generating three choruses at a time with each extend.
What I could do is pull that into editor and pull the lines that are sung by the correct vocalist and delete the others, and then try to heal the whole thing?

Sorry if that's not clear. Feeling defeated again.

I'd ah... I'd pull the whole extend into the editor, make a full song. Then cherrypick the good lines and delete everything else so I have one full chorus sung by the correct vocalist (or at least as close as I can get) and then heal the whole thing that's all cut up and see if it brings it back into some consistent musical expression?

[half dragon]

You can try to delete the wrong vocal Chorus and then copy and paste the correct vocal Chorus into its place, then heal it - hopefully it'd stitch well.

[sturdy dove]

It's so frustrating. It's like two lines in the chorus that need to be fixed and one line in the finale. Everything else is fine.

[sleek stump]

With all the attempts you have been trying over days (a week?) now, I might suggest taking all of it out and into a DAW to fix the final part and then uploading the complete qnd fixed track back into Suno.

[sturdy dove]

Yeah, okay... That's what I was thinking.

[sturdy dove]

Ginger, it's been a month. If I knew how to use a DAW I would.

I had fully given up, but then this place came alive with helpful people again, so I wanted to give 'er a go.

[untold briar]

This can work, basically if you make a rough draft of what you want, you can sometimes "heal" it with combinations of Replace/Cover/etc.

[sturdy dove]

Cool man! I'll see if I can get a full song that's at least kinda close. The thing is the extends are completely changing the style of the song even though I'm extending from the whole song as I like it leading up to it. It's got over a minute of music to mimic and it's just kinda doing its own thing, ignoring the persona, ignoring the audio, ignoring the prompts.
I've got weirdness at 5, Style and Audio at 100. I could turn Style down to 0 or five...

Turn style down to 0?

[sleek stump]

Myself, I just cant use the current editor, If I edit anything in the Suno ecosystem I use the legacy or OG editors. But there are other things I just have to do in FL or Audicity. By all means keep tryign to find a solution inside of Suno, but I would suggest at some point in the near future, go get FL Studio and use the free month of pro and the many tutorials to learn how to use it, even if its just basics. It can make your life much easier in the future!

[untold briar]

At this point, it's almost certainly best to try and accomplish whatever you are doing using a different method or combination of methods, than Extend. But I have just skimmed this conversation and haven't looked at your problem in detail, so I can't say for sure there is a better way to solve it. But for Example, a Replace is interchangeable with an Extend, in that they can both fill in a gap in a song. And you even insert silence into a song, a gap, and Cover over it, and fill in a gap. And so on and so on.

[sturdy dove]

Yeah. Last time I had great success (in my own mind - me being happy with it) I just pulled the stems and used some manual tools. In this case what I want is a little too complicated for my limited technical skills, unfortunately, and the project was due like three weeks ago. I just have to make whatever I get tonight work. Really frustrating to be so close and yet so far away.

[untold briar]

It really helps if you are okay with the music and voices changing a little bit, so you can use tools like Cover, and just get a very close copy.

[sleek stump]

Yes, as Jonathan noted, getting it very close and then using Cover can solve a lot of problems.

[sturdy dove]

Yeah. Unfortunately it's the story of a boy and a girl, and when they change positions and read each others' lines, the story just breaks down and doesn't make sense.

I've used all kinds of other AI tools with massive success. At least in my own mind. This one has been the most frustrating project I've ever come up against.

[sturdy dove]

Yeah? Talk to me about this...

Let me give you both a quick synopsis of the problem real quick, since you're so kind as to offer your support and advice.

[untold briar]

You may be able to get Extend to work if you really dig into the details, and create a custom song just to customize the context window "history" of the audio, and force it. I am just able to chat at the moment, but if you post the details, I will take a look this weekend at them.

[sturdy dove]

I've got a duet I've been working on. I ran like 100, maybe 120 generations to get something close, which is what I've been working with since.

It's REALLY COOL. It's almost exactly what I wanted. But, halfway through the chorus, the singers start singing each others' lines in a way that ruins the story. So I've spent the last month using Editor, Legacy Editor, Regen, Extend, whatever to try to fix these couple of lines.

[sturdy dove]

This sounds complicated and over my head. But I'm willing to try and work for anything.

[untold briar]

Also just FYI, the male/female selector in Suno is not super reliable, and you still want to manually prompt "male, male, male" etc if it's not working.

[sturdy dove]

I've tried every method of prompting. It ignores them all mostly, unless it decides to just do the opposite of what I want - which is what it's doing tonight. lol

[sleek stump]

Ok, question: Are all your chorus stanza the same or are they different?

[sturdy dove]

I'm not using the selector because duet is not a choice

[untold briar]

But for a duet, imagine you create two versions of the same song, one female, one male. Then blend those clips and Cover to create the duet at the end. That kind of alternative workflow may end being easier than trying to get the voices right on the first try

[sturdy dove]

There's really only one chorus.

Yeah, that's what I was thinking of doing before Jonathan. I might try that again - the issue now is that the musical back is coming out quite different every time.

[sleek stump]

SO the entire song only has 1 chorus in it? This might be part of the issue.

[sturdy dove]

No no, sorry

It only has one chorus

Here

Bah, it's over 2k

one sec

[sleek stump]

So it has a chorus, that appears after each verse, its the same? When did it go off teh rails, at the end?

[sturdy dove]

You know what, I see what you're saying. The pattern is inconsistent. I'll fix that first.

Halfway through the chorus. I'm asking too much of the AI. I need to simplify

[sleek stump]

So the chorus fails every time it is sung?

[sturdy dove]

Uh... yeah. Yes, that's correct. I have about 30 gens in front of me and it's doing it wrong at least a little every time.

[sleek stump]

How many times does the chorus repeat in the whole set of lyrics?

[sturdy dove]

I think part of the problem too is that I can't specify which parts I'm happy with and which I'm unhappy with. So it's seeing I'm continuing to regen and so it just keeps giving me variation - which I don't necessarily wnat

It's only there once

I keep regening it to get pieces to pull together to make one chorus that holds together.

[sleek stump]

you onlu have a single chorus in the entire song, one time? Like I said that might be one of the issues

[sturdy dove]

At this point, it's varying the vocalists and backing track a lot, too

so I'm trying to pull something that's close to the original and that's tough at this point

[sleek stump]

Can you post me a link to a complet-ish version of the song so I can see what you are doing?

YOu can DM me if you like

[fresh wave]

Wild idea, make an instrumental by putting all of the lyrics in the style description instead.

[sturdy dove]

This is the closest one:
https://suno.com/song/651daba5-d7b9-47ff-81b1-802c6c8cd807

[sleek stump]

Let me take a look...

[sturdy dove]

Again, I think you've identified the problem: the Chorus pattern is inconsistent. That's on me. I'm asking too much of the AI.

So I just need to rewrite the Chorus and simplify it.

Shenanigans, thanks for the idea!

I'm honestly pretty burnt at this point. It's been a really long day. I need to crash and start again tomorrow.

[sleek stump]

Ya, you have a symmetry issue with that back and forth chorus, From my experience, if you want to do alternatring lines, it needs to truly alternate.

[sturdy dove]

Yep

One half is 4 lines, the other is 5

I'll fix that, that'll help a lot.

I think they're similar syllable counts, but again, that kind of inconsistency is too much for a musical AI to figure out, it would seem

[sleek stump]

M F M F MM (This is where it breaks) F M F

[sturdy dove]

yep

[sleek stump]

Remember, Suno wants to give you what it thinks is correct from itys training, and doing weird stuff makes it very difficult to get the results you want.

[sturdy dove]

yepo

[sleek stump]

Genre also plays a big part on if Duets work at all.

[sturdy dove]

Yeah, I thought going Synthwave would give me a little more freedom

All good, I'll rewrite and try again tomorrow. : )

[sleek stump]

Ask yourself this... When you listen to Synthwave in the real world by real artists, do you hear a lot of Duets?

[sturdy dove]

I hear some

[sleek stump]

If the answer is no, then you know you are in for a huge battle

[sturdy dove]

Ginger, I'm really, really tired. I really appreciate the help and support! But I've got to go. Thank you so much!!!

[sleek stump]

Good luck!

[sturdy dove]

Thank you! I'm going to need it!

[crimson junco]

Hi all,
How to instruct Suno that the cold start of the song is mandatory?
I tried some variations by instructing in the style, or using [vocal intro] tag for some reason, but this was ignored very often, even I have cases that last a minute with no intro instructions.

[half dragon]

Start with no musical intro?
"Opens with a stark acapella line, layered in tight harmonies, gradually introducing subtle finger snaps, The first verse adds muted synth pads and light percussion, The chorus bursts in with bright pop keys, punchy drums, and melodic bass, keeping vocals front and center"
https://suno.com/s/dC0V8hPqdFUfoyy8

[crimson junco]

Thanks will try.

@half dragon By the way, I just tried the new editor. You can just delete the intro, which is also a variant.

[half dragon]

That's another option.

[polar spade]

Have you used the edit UI enough to get used to 1) duplicating things like Choruses and pasting? (That would be my direction these days)....and 2) being ready to jump into blending and splicing and remastering when splices might not match properly (and thats why I dont suggest it at this moment) ? If not, stick with trying different prompting styles as you are. Thats what we did before edit came along, so its good to learn basics first, but not give up learning the new stuff as well.

[polar spade]

Every one thing to alter , changes three other areas, such as learning to align lyrics. So its a lot of fires that start, to put out all at once. Some of these things, I feel, one should take three of four songs at the same time and just practice one skill for awhile among them to see a pattern or realize why it works or doesn't work. That goes for any of these suggestions , really.

At some point, even altering the perception of "perfection" may need to be done so you dont burn out again trying to complete one song "perfectly", whatever that may mean to you. There is beauty of finding perfection in the imperfect. Most all my songs have "issues" and they feel more organic from them. My opinion, i guess.

Find a nice point to stop and finalize a decent song and forget about "perfect" for the moment. Start another or six or thirty more songs and try your new skills. Eventually, come back to these "inperfect" songs , two months later maybe, and try other learned (and hopefully then, better understood) skills to perfect on them a bit more if you want.

[polar spade]

By the way, do you use a DAW in any capacity? I think you mentioned you dont? That will eventually be your best friend, if not yet. Try Audacity as a free source and practice some basic things such as rearranging and splicing segments together using a less "random" program than Suno. These will then apply to being able to use Suno with a better skill mindset.

[ebon sand]

@sturdy dove I'm awake again, I'll backread a bit later, only scrolled through once now but it sounds like things were made more complicated than they have to be

You can definitely create whatever song with whatever pattern you want, I don't like the newer models in edit much either but they will bend

[polar spade]

Kudos to @young geode , @sleek stump , and . . .

[polar spade]

@half dragon , @untold briar , and....

[polar spade]

@fresh wave , @ebon sand and anyone else misssed for being the village of inspiration in all this. Ive ordered 12 pizzas for the Master Class party. 🥳 🍕 🍕 🍕

[ebon sand]

you select only the lines with the wrong gender and replace those again with a command to use the other gender e.g "male vocals", i cannot see which model you used, on v3.5 it's as simple as "use male vocals" in the styles, on 4.5 upwards it might need more elaboration or emphasis but it's the same principle, if only a few spots are wrong only regenerate the spots

[pastel plaza]

hi guys does anyone know why the remix is only half time of the source ?

[polar spade]

When you use Cover, Personas, Extend, suno reserves some of the new output to allocate for , let's call it, "code" which represents the charcteristics and vibe of the original output to seed or feed off of to make the new output.

If your original song was a 30 second jingle, then you have more room in that 8 minute limit, to prompt for a longer remix. If you have a 7 minute song, your remix will likely be shorter and not exactly 7 minutes.

That how its worked since we were introduced extensions and covers.

[rugged field]

I've had a lot of variation on that

Sometimes i gve a 14 second thing and it'll produce from like 1:30- 5:50 ish range

[topaz pulsar]

How does one get fast lyrics? I've been trying for over an hour

[half dragon]

Maybe by having more words into one verse? in one sentence no line break?

[topaz pulsar]

Tried that already

Maybe I need to do a combo of merged lines and non-merged...

[half dragon]

Is it Rap? Maybe also add in the style rapid-fire delivery, double-time flow, high syllables per bar?

[topaz pulsar]

Not rap, but I'll add them to the long list if this generation doesn't work out

[half dragon]

Some examples on rap though - maybe helpful
https://suno.com/song/2de5348c-7b6b-48b4-b08f-62b61b61f953?sh=32oINlXNOuyf9gev
https://suno.com/song/16196222-18ea-4300-984d-5124ba5b4cc6?sh=zO8GkU8iM4xPiU6O

[topaz pulsar]

This combined with line merging worked

Has a CVS receipt of a Style Description but I can't be bothered to keep going on my phone, I'll probably try more when I get home to my PC

[raven turtle]

This is the philosophy of Suno I’m working with. I have enough library songs now that are decent enough to leave until Suno reaches CD fidelity and/or we get Suno Studio to finish them properly 👍

[neat field]

yup

[sturdy dove]

Yeah, I've been trying to duplicate and regen, but the results aren't very good. A big lesson through this was to not just regen what's broken, but to regen the entire section.

Given the disjointed nature of the regens (no matter how much song I select to extend), I may go back to full renders rather than partial.

The heals sources for the heals I've tried so far were so far apart that they never worked, but maybe I'll try healing again once things are closer.

Thanks for the tips!

[sturdy dove]

I'm so far from perfection. I'm just trying to get to a point where the song actually makes sense as sung. That's pretty far from perfection for me! But yeah, I hear you.

[sturdy dove]

Not a real one, no. I did a song on Udio that I was able to break apart and use some really fundamental tools to perfect. That was a really fun, simple process. Since this is a one-off side project, it's hard to invest so much into it. But I do enjoy the ability to do this, so when I have more time, maybe it'll become more of a hobby.

This current project is focused around a real need rather than a want.

[sturdy dove]

I don't know if I'd put it that way... I rewrote my chorus (or am in the process) because Suno can't give me what I want. Which is disappointing, but I have to accept the limitations of the tools.

Yeah, that's what I thought, too - I tried to create an interesting pattern that I liked the flow. But that in combination with the duet is too much to ask right now.

[sturdy dove]

I'm buying!

[sturdy dove]

Right, when I tried to regen only the lines that were broken before, it produced either no vocals or gibberish.

BUT NOW THAT MY LYRICS ARE ALIGNED... .hhhmmmmmm... Might be worth another try!

[ebon sand]

yes that happens when the lyrics are not aligned

[hearty laurel]

Hey guys, Quick question. I created a track in Suno with my uploaded audio. When I download the whole track, there is a nice, long reverbed tail at the end of the song. When I download the stems, the tail is gone. How can I get the tail back in the stems? Thanx

[sturdy dove]

Makes perfect sense

[sturdy dove]

That's interesting. Can you download the song and get it and maybe add it back in another way?

I didn't mean to hijack - I just wanted to respond to everyone who was kind enough to jump in and help. I've learned a lot, I'm learning a lot, I'll try to get back to this project tonight! Thank you all!

[polar spade]

Did you stem out with vocal/instrumental? Or the multi-stem?
Did the tail not show up as the tail of or in ANY of the stems?

[hearty laurel]

Multi-stem, and it was on none of them

[polar spade]

One test is to download it and use another stemming source for "normal" frequency separation stemming. Just to see if it still exists if done that way. I dont see why it wouldn't but helps to omit the obvious sometimes.

[polar spade]

As its noted some third party track separators will produce a dryer vocal stem by removing reverb, I guess we should ask if this of yours was on the vocal or instrumentsl side. There could be some recent tweaking. I haven't "stemmed" in a few days.

[hearty laurel]

It was on the instrumental side. Guitars I think.

[polar spade]

Thanks. Any details help. 🙏 I dont know the answer right off as Ive not experienced it. But curious enough for me to know as well.

It may be a waste of 10 credits, then 50 creds, but you can test the 2-part stem separator to see if it specifically shows up in the whole instrumental first. (If it does, then multi-split the instrumental (without the vocals) so it might recognize it as an instrument than a voice.

And asking again, did both version 1 and 2 of guitar stems do the same? Making sure you scrubbed through and listened to all 12 stems towards the end.

Did you by chance have a chance while the muti-stem panel showed all stems and listen through all the coloured tracks at once and notice the truncated bit on the end, or later after you went to the library?
A) Did hear in the preview. [ ].
B) Did not listen to the end in Stem Previewer. Only heard in separate track in Library [ ]

[hearty laurel]

Both 1 and 2 did the same thing. B

There is a track called systh, that just background noise and seems to be the tail, but it just sounds like distortion.

[polar spade]

Dang it! 🤔 OK. Thanks again.

Yea, thats what I was wondering, if it eventually got labeled as something else as it reverbed off into the tail. It thought that part was maybe a synth instead of a guitar.

Each time we separate, it may hear it differently and finally get a guitar. But that lots of credits to see if that happens..

I can do some testing like that on my account. Would you mind linking the original and set remix open and today ill play with it to copy and multi stem. See if I can replicate. Ill delete all sources when done testing. Or download the mp3 and attached to my DM and ill use that.

[hearty laurel]

Before I do that, do you know of a way that I can edit the lyrics of one chorus and have it apply to all the choruses? I am trying to change 1 line in the chorus, and I would like it to apply to all the choruses. For some reason, after I changed the line in 1 chorus, I have to go back into another edit session to replace the line in another chorus, but when I do, I get a completely different-sounding line, and I would like it to be the same. Thanks again for all your help.

[polar spade]

One alternative: in edit UI, I select the chorus that I edited and which sounds perfect, make a duplicate, then insert where needed and remove the older versions.

The caveate is there may be moments where the splices dont line up as well as wished, then some manual tweaking is needed. Or we can also save the file and run through a remaster and the remaster blends the splices nuch better.

For thr most part, this works good and often I do nothing more as the splices blend fine. Depends on the genre, pre-chorus build, the drum patterns etc.

.

[polar spade]

Ill be delayed in most responses now. Im driving for at least 5 hours.

[hearty laurel]

That sounds like it might work except that the choruses are sung a little differently each time. In my DAW, I even tried to copy the one I changed and paste it into the other choruses, but for some reason, the timing is not the same. I guess it was sung a little differently on each one, and I liked it.
Dasvon, thank you for all your help. Drive safely. I will keep messing around with it and see what happens. I will get back with you. Thanx again.

[polar spade]

Im with you. I like the change ups in choruses. Might need to edit each. Its good practice though in aligning, editing, saving and repeat with the next chorus.

[hidden yarrow]

I often do a cover version on songs that I created with copy/paste sections to get the desired structure. With the right sliders and prompts the new song is very close to the old song, sometimes better than a remaster (I hate those and use them only if the first 8 tries do not mess up lyrics, notes/chords or instruments/timbres/effects).

[hearty laurel]

@hidden yarrow I haven't tried the remastering yet. I will wait until I get the song just right, and then I will see what it can do.

[keen iris]

Hello everyone!
So after my Pro plan ends I download some tracks to my library and get this window on one of tracks I made with Pro plan.
Is this mean that I'm not owner of this track and for owning it I need to subscribe to Pro plan?
If yes, than why do I getting this message if I'm created it with Pro plan?
Also I have some nice old tracks (with v2 and v3) can I became owner of them too or it's only for tracks with my lyrics?

[neat nexus]

Do you get the same message in an incognito tab?

[keen iris]

I never tried in incognito

[granite shale]

I'm in the editor, and every generation is like super quite in volume compared to the rest of the song, how do I fix this?

[hollow hare]

AFAIK, that message simply looks at your current subscription status, not the status when your music was created. Cheaper in resources and faster by checking just if(member.plan.is_free).

You still have rights over the ones you genned while subbed. If asked for proof of the rights, you can prove that you had a membership in the month the song was generated to the distributor or other agent asking for proof of your rights (as
there is no other way at the moment to prove it otherwise) so be sure to document and backup your receipts, creation dates, and the TOS which states about the rights.

[hollow hare]

AFAIK, at the moment, there is no way to adjust the volume through Suno's editor. You will have to do it using external DAWs or other tools/programs (which is something you would have to do for mastering if you are planning to distribute).

I would not be surprised if audio control is planned to be implemented at some point in the future.

Theoretically, you can replace that segment to try toncorrect it, but it would likely be very costly in credits and time compared using an external tool.

[still charm]

Does someone know what style influence does

??

[hidden yarrow]

You can adjust volume for each section on the "new" editor by selecting it, then clicking on the three dots for "more options" and then use the horizontal slider above the color selection. If needed, you can define new sections by splitting.

[granite shale]

Your incrredible, Can't believe I didn't see this. Thank you!

[dry ingot]

Look further down on the left at the DOCS, GUIDES & RESOURCES> #1294423389515284552

"If you make songs while subscribed to the Pro or Premier plan, you own the songs. Further, you are granted a commercial use license to monetize those songs."

[polar spade]

OK. The CBS snippet report of Suno today on AI was just fluff, a non-event piece. But if nothing else , it said nothing relevant on either side, I suppose. Kudos to Rick Beato for staying neutral in the interview. Respect for that.

[polar spade]

That's a relative statement. If I make song using most of the Beatles song, Yellow Submarine, I own that song I make as well, and can monetize and use it commercially. 🤖

[slow edge]

I've a problem with the video/lyrics syncing, it is just showing the last few lyrics (Outro) and highlighting them on the video, but they are the wrong words.

So tried to update the lyrics, re-gen video. Still wrong lyrics showing.

Took it to the editor, to correct lyrics, noticed it had multiple repeats of the 'outro' lyrics. So opted to correct them, align them, saved as new song once it said it was done. Re-gen video, again wait for it to finish, download video. Played it, and it's still showing the uncorrected version.

Any help please? As to how to get the lyrics that appear on the Video, show all lyrics and sync up.
Thanks.

[polar spade]

In simplest terms the style slider adds weight or priority to how much your style prompts and lyrics interpretation, as well as the vibe and character of any persona added, adds to influence the song.

[devout sinew]

Question lyrics box--- I want vibrato on “life”, I would write:

This is the best day of my ~life~ will suno recognize this?

[pulsar willow]

not trying to post invite, just a message

Just waiting for the Creator version beta. on the list. in the meantime here's how I am using Suno just as an assistant. mostly my melodies and all my lyrics. https://youtu.be/14cL3ldmFuI?si=MGyo3JtDsRp33EEl

Almost like just taking my original stuff to a producer

Amazing what Suno does with uploaded live performances of my new songs.

According to Youtube guidelines my channel should stick.

Luddites it's over. Welcome to the 21st century. Music and creativity rolls on.

Prompts: a mixture of agent produced prompts based on my prompts, some suno prompts directly and some I allowed Suno to modify. Not searching the internet for prompts anymore

[sturdy dove]

Crystal, I'm no expert like @Lunacy is, but it sounds to me like you might be trying to do too many steps at once. First align the lyrics, then save a new version, make sure they're lined up. Then proceed...?

@sleek stump , depressing news. I fixed up those lyrics to make them flow better and be a more predictable pattern. It's still ignoring the prompts most of the time. LOL Oh well... I think there's a higher probability of getting closer to what I want with this, but it's still noisy I guess... I'll keep after it!

[jagged narwhal]

Hello I got premium again I heard we can use same voice for other projects too is it true?

[strange pasture]

Really?I am happy to hear that

[jagged narwhal]

I guess I seen on shorts but not sure thats why asked 😄

I worry about fake news

[strange pasture]

Maybe it's a real news,XD。If correct,I can't wait to creat many new and funny music

[jagged narwhal]

I hope a team member can answer

[strange pasture]

+1

[ebon sand]

the videos use some sort of auto lyrics alignment, i have never attempted to get along with that, because it usually deletes all of my lyrics, my videos look like this no matter if i align anything or not

i want suno to fix stuff on that end hahaha

[polar spade]

With the use of Personas, we can attempt to transfer similar vocal characteristics found in the persona while generating a new output.

That is not the same as swapping out one voice for another to have all other aspects of a preexisting output to stay exactly same but the vocals.

[polar spade]

I have not run a video in a year, so I tried one. It synced with the right lyrics, but removed one or two key words on a few lines.

For instance, instead of "Find Your Dream", it just have "Dream" as text , and yet stayed in sync , waiting while the audio would sing "Find Your....."

Ive never fully understood the video/lyric aspects of Suno to fairly give feedback.

[pale condor]

Guys, using the Editor, is it possible to take 2 tracks, and attach the 2nd half of track 2 with the 1st half of track 1

[slow edge]

Mine usually end up with the first word/s missing off each verse.

I had to extend the song twice, to get the lyrics sung by the right voice in the duet.

As when I tried to sort the lyric alignment in original version to get it to regenerate verse 4 and bridge which were wrong singer on each line,

it just either created incoherent mush or start off ok, before just singing the Outro, even though Outro was several more verses and a chorus still away.

The Outro in the lyrics when checked editor version are shown repeated multiple times and trying to correct it with the correct lyrics it didn't work, it said it aligned them, but they were still in the video just showing the Outro repeated.

https://youtu.be/MUjCm-MueqU?si=0LhJulfn2K0B6XKm
The song is on YT now I just put the correct lyrics in the description.

I do find the editor a bit clunky and overly messy/difficult to work with and have it work as want it to.

Say I want to replace say 2:52-3:26 with X, Y n Z lyrics, I don't want it grabbing lyrics from 4:24....

Also if I could type in the times, I'd find that easier than using sliders with Lyme Arthritis damaged hands n wrists.

I don't remember having quite so much issue with earlier/earliest version.

[polar spade]

Simply, no, not at this time.

[pale condor]

Damn really wish this was possible, absolutely love suno

always wanted to craft music but i had no patience using fruity loops and dozens of plugins lol

[polar spade]

Many take to a DAW , then re-upload (now with the 8 minuye limter) and complete in Suno. Thats about all the options for the moment.

[pale condor]

Know a lightweight noob friendly DAW I can use for simple stuff like cutting and pasting?

[polar spade]

People still like Audacity. Not sure. And Bandlab may be a step up.

[winged peak]

Even though Audacity is more of an audio editor than a DAW, it is probably the easiest to use. Burial made many of his tracks with it – without a DAW. It would be worth a try.

[winged peak]

No 😉 - Sorry 2 say.

[sullen wren]

Bandlab is top notch, especially their mastering feature

[ebon sand]

I'd also throw in tenacity, it's brother, more modern

[half dragon]

Something like these?

"A rhythmic rap track driven by sharp, plucked instrumentation—think percussive guitar and pizzicato strings weaving through crisp beats. The production remains sparse yet punchy, supporting tight vocal delivery with syncopated hits and short, clipped melodic phrases throughout."
https://suno.com/s/6zJz0xVszzixErSS

"A rhythmic rap track with sharp, percussive guitar and bold pizzicato strings interlocking over crisp, punchy beats. Sparse arrangement spotlights tight vocal delivery, while syncopated rhythmic hits and clipped melodic riffs accent each section, maintaining propulsive forward motion."
https://suno.com/s/7e7zW8JU3NRXayk9

Just examples ofcourse 🙂

I did - but maybe I intepreted it wrong ..haha... anyways I don't pretend that I know what I was looking for ... thought of string pluck lol it still sounds kinda cool - and put synth pluck in your style didn't work?

[graceful condor]

Ahoi everyone - hope you are doing well ^^

2 things I would like to know:

1. Is any slider (weirdness, style or Audio) relevant for the Lyric Prompt section or is this part handled on its own? So if I would turn down style to 0 this does not effect the prompts in Lyric right?

2. If I go into detail about the type of voice and all its feature should I only do that either in Lyrics prompts or Style or is it okay and more safe if I place this Information in both sections? And would this support each other or interfere with another? And I wonder where has it more weight in Lyrics or Style?

[half dragon]

I don't but I've seen others that do both - I would say Style has more weight but it was my guess.

[graceful condor]

I just noticed if I only add lyrics into the lyric section without a lot of extra detail the songs turns out completely different.

[little jetty]

Hey, all. First time on Discord, so hope I'm in the right place for help and/or encouragement. I'm struggling to get Suno to correctly pronounce a couple of words. I've tried different spellings, I've tried spelling them out phonetically, and still can't get them right. Granted, these are tough words (allocable and improvisational), but I am about at my wit's end. Any suggestions?

[hollow pine]

which words? which language? which musical style?

[modest crow]

Not sure if this was luck for me or not, but I had a similar problem the other day and tried the same things as well as trying to literally tell it to pronounce words correctly and then finally tossed a command to enunciate the specific word and it started generating tracks that pronounced it correctly. Maybe worth a try if you haven't already given it a go.

[wispy laurel]

Allo cabb bull
Imp rove eyes ay shun ull
?
Might have to play around with hyphens and ...s and upper/lower case also, that affects both emphasis and pronunciation

Also what Jenna said. It does understand prompting related to pronunciation and elocution, you might need to experiment to find the right wording though

[junior terrace]

hello everyone, I'm really trying to get a song in the Daft Punk style and I know that they use a Talk Box for the vocals but I can't figure out how to make Suno understand it

[little jetty]

Thanks for the tip. Where did you type the command to enunciate? Did you tell it how to enunciate? For example, no matter how hard I try, I can't get it to put the stress on the first syllable of Allocable.

[little jetty]

Thanks for the suggestion. Tried them both, and Suno still can't figure out what I'm trying to get it to say.🤣

[little jetty]

Allocable and improvisational. English. Upbeat pop shuffle with light swing .

[polar spade]

Spamapaloosa @flint haven

[polar spade]

AL-uh-kuh-bull.

You can TRY spelling phonetically and capitalize the first two letters, but you should also look around at the previous lines and see what parts of the lines are emphasized that Suno may be trying to copy to keep a rhyming scheme going.

So you might get:

We're such a WOKE-able couple
To be the al-LOC-able pair

It may be as simply as putting a dash or comma between any two nearby words to give Suno an additional character that it would then re-calculate the lines. This can force it to change its intonations.

[polar spade]

Reminder: Any of these how-to questions placed in need-help channel are better suited for this channel as well.

Such as how to get the pluck and reverb sound. #📚┃suno-school message

I tried several attempts but see that it will like be several tries before getting just 10 seconds of something close as a starting point.

By the way, you may get some close results from "pizzacato" which means plucking. Most results will sound as a violin plucking so "pizzacato guitar" can help.

Even that, it was always too fast originally. So you may then need to edit or adjust speed down to 25% or so. From there, having maybe a good 1 minute to play with to extend from that, then stem back and add vocals or Cover. Its many combinations and steps that might need be done to get an "exact" sound.

I got my best results from "pizzacato surf" , or pizzacato guitar.

Not what you're going for, but each step could get closer:
https://suno.com/s/8YF6PTByRMieLVuu

[little jetty]

Thanks! I finally got a reasonable facsimile using AL-uh-kub-bull. It's not perfect, but close enough.

Thanks, everyone, for all your helpful tips and suggestions. Much appreciated.:)

[uncut tulip]

Is anyone able to help me make a male / female vocal of a file i have?

im trying to do duet its not worknig so i just need to make a female / male versino

[polar spade]

I can often get some similar sounds, but it could also take 4 or 5 pages of tries and several steps. Nothing really teachable to pass along thst it would work the same if another tried the same prompts. Suno is not geared that way to "emulate" specific hand-picked vibes.

[polar spade]

It sounds to me like an epiphone hollow body guitar with a water drop effect, clipped and add reverb. But yea, could be emulated through a synth. Like I say, its not easy to emulate a specific sound in Suno when its not a specific instrument.

Ok. So Im done and suggest to @half dragon to do the same with this one, given your response. Just cant quite get it there for ya. Sorry.

Good luck.

[half dragon]

What do you mean make male/female vocal of a file you have?
Did you mean, creating a duet song? but you also mentioned it wasn't working.
And you're trying to make just a female and a male of the same song?
Elaborate more....

What's the sucess rate for this approach? I haven't found a reliable way yet but haven't tried this.

Nice! I was working on a song and actually wanted to do this layer vocal thing but haven't been able to.
I'm going to try it tonight.

Did you put anything after the [Outro] - how does it know which to sing for both? is that the instructiion thing you set in the Style box?

Oh so both of the hook lyric and rap verse?

[wanton canopy]

I just started playing around with Suno yesterday. I've tried most of the online prompts to input a chord progression, but nothing seems to really work in that regard. I imagine it's only a matter of time before more options become available that allow for greater micro-control over input. The song brackets make sense, as do the parentheses for lyric input ... but Im a little lost on what the asterisks are used for? I thought I saw someon online using them.

[scenic coyote]

anyone that wants a specific sound... go to beepbox and create a sample. and use that. its so roundabout to get a language model to make a noise that you can create in literally 30 seconds

i know its kinda the "fun" of it but these questions aren't really about tech talk. they are about getting certain sounds which you can do fairly easily

[young geode]

Chord Progressions don't work 100%, and they're likely tied to genres more than anything else. Scales, however, do work.

Asterisk wrapped lyrics have a chance to produce softer, more distant vocals, when used correctly, while curved bracket wrapped lyrics produce an ad-lib/whispered vocals.

If you're new to using SUNO, read these guides and get an idea of how to manipulate the AI: https://discord.com/channels/1069381916492562582/1382019029266600067

[sturdy dove]

Alright, I'm working on this again. I'm trying to use the Editor (current, not legacy) to Replace to clean up some ugly cuts and style inconsistencies. The vocalists are completely different with all the replacements. They style is kinda off, too.

What's my best way to move forward?

Should I just regen the whole song now that the vocalists are singing the correct lines?

[sleek stump]

I would, I would also give your vocalists names, like Mike and Sally, that will help Suno keep them consistent.

Desciribe your vocalists in Styles with their names and then call them by name in the tags

[sturdy dove]

AH. Yeah? Okay!

I found some crazy prompt online that said to use "vocalist one" and "vocalist two" and explain that vocalist one was male and vocalist two was female. I guess the extra steps are helpful. I'll give them names, for sure.

Okay, cool! Will do!

[sturdy dove]

Thanks Ginger!

[sleek stump]

There are many met hods to get from point A to B, thats one I have used and heard other use with success.

[sturdy dove]

You know, I really enjoy creating characters. I find it's really helpful during the imagining process, and I'll bet it helps AI figure out better what I want, too. Makes total sense.

[sturdy dove]

Oh hey, if you're still there, how do you prompt your duets in the lyric window?

is it:

[vocalistOneName] lyrics lyrics lyrics
[vocalistTwoName] lyrics lyrics lyrics

Or something different?

Somebody else here told me to use parentheses rather than brackets ahead of the line

[polar spade]

Might need to avoid the unisex names then, like Sam, Lindsey, Kim, Candy, Charlie, Casey, Morgan, Skylar, Lee, Kelly and River, to name a few. 😃

[sturdy dove]

I'm being explicit: John is a male baritone...

[sturdy dove]

How are you doing today??? You good?

I'm a little sad Sunday is almost over. I was really hoping to finish this project today, but it's been dragging. Well, and I've been dragging it.

[polar spade]

I had a girlfriend named Johnna once, but I digress.

What happened to masterofsfl suggestion of referencing a gender and using quotes. Seemed to work just fine that way. I may have missed a post read along the way.

[sturdy dove]

! I don't remember that. I'll see if I can find that

[polar spade]

I think simply, it might have been thst it doubled a line to be two male vocals so the remaining lines alternated in reverse and you said it didnt match continuity then.

Maybe you went in a new direction then. I dont know.

Anyway, I thought Masters way seems pretty solid and the least complicated. Thats where my mind goes.

[polar spade]

... and im well. Thats for asking. A long day.

Well . . . calculate a little rest in your Suno work.

As said, if you're not enjoying what you're doing, you're doing it wrong. (It means slow down until youre at a more comfortable pace)

[sturdy dove]

Naw man, gotta get it done. I'm way behind on this.

I can't find what you're talking about from Master. But this is what I went with for now, does it look right to you?

[Chorus alternating, sung first by Matthew than by Namie]
[Matthew sings] Geisha, my sexy JDM Geisha
[Namie sings] I respond when you call to action

[sleek stump]

I would tackle that with:
[Chorus - Alternating Vocals]
[Matthew] lyrics
[Namie]Lyrics

I would suggest keeping tags as simplified as possible.

[sturdy dove]

Got it. Okay, I'll try that next.

The last one I generated was actually pretty close. Well the first one. The second one was kinda off

Okay, simple tags, got it

Okay, both of those were completely wrong. Redoing tags

[polar spade]

My bad. It was @dense moat . #📚┃suno-school message

It was a simple and seemed plausible. OK. I see though, you said it wasnt the dueting vibe you were going for.

OK. I guess Im caught up amd stepping back out to sit on my stool in the corner. 🤓

Carry on with what you were doing.

[sturdy dove]

No way man! I appreciate your suggestions! I appreciate you. You're hanging by me through this, it's really cool.

This is frustrating. Like back to square one. The first generation was really close to perfect, everything subsequent it's completely ignoring the prompts.

Here's my current lyrical prompt:

[Short Intro]

[Verse 1 - Matthew only]
Twenty years we've been together
From the showroom to today
Every mountain road, every corner
You respond the perfect way
Since 2006, every apex kissed
You're the one who never lets me down
But I can't take you where I'm going
And it breaks my heart

[Verse 2 - Namie only]
I'm always waiting, always ready
I love to do just what you say
I obey in evree wayyyy
Twenty years of girls that come and go
I've outlasted them all
You kept me cozy every night
I perfectly serve you
And you've kept me so nice

[Chorus alternating Matthew then Namie]
[Matthew] Geisha, my sexy JDM Geisha
[Namie] I respond when you call to action
[Matthew] Across the ocean, I'll remember
[Namie] This perfect dance we've shared
[Matthew] You always kept my wild on line
[Namie] I'm Geisha, your one and only Geisha
[Matthew] You're such an elegant Hawkeye
[Namie] I'm faithful strong and true
[Matthew] I hate it but I have to let you go
[Namie] So I'll give my all to someone new

It completely ignores and has the female vocalist go first 7 / 8 times. And now it's intertwining them again even though it's only supposed to be the male vocalist first.

This isn't the alternating section. The whole verse is supposed to be the male vocalist, then the female vocalist.

I think I have to give up on this and go back to fixing what it broke from before. It was a lot closer to what I wanted.

Tried this, didn't work either:

[Short Intro]

[Verse 1 - Matthew]
[Matthew] Twenty years we've been together
[Matthew] From the showroom to today
[Matthew] Every mountain road, every corner
[Matthew] You respond the perfect way
[Matthew] Since 2006, every apex kissed
[Matthew] You're the one who never lets me down
[Matthew] But I can't take you where I'm going
[Matthew] And it breaks my heart

[Verse 2 - Namie only]
[Namie] I'm always waiting, always ready
[Namie]I love to do just what you say

Keeps having them alternate

Maybe I'll try genning in 4.5 or 4 instead of 4.5+

LOL the 4.0s completely ignored everything and did a blues riff.

Trying 4.5. If this doesn't get close, I'm outtee.

Nope. The female goes first against the prompt or they alrternate against the prompts. Enough.

[half dragon]

As mentioned, a duet is an ambious project to take on.
Suno isn't reliable with duet - so we can't specifically instruct a specific vocal to sing a specific verse or line.
You won't get it by the normal way - there'd be workarounds, different techniques... lots of time and credits to get the way you wanted.

[sturdy dove]

I'm back to editing in editor. It's closer...

19/20 gens completely wrong. Ambitious or not, the hype about 4.5+ was duets.

It doesn't matter. Back to editing what was close. It's frustrating I can't get a contiguous song with the same vocalists and style throughout, but it's better than being completely wrong in almost every way.

[half dragon]

Sorry I didn't mean to discourage you ..not at all.
The oposite in fact, keep on trying... you can do it!

[sturdy dove]

Man, I'm ... what... 50-60 hours into this project over... I think two months now?
And what I've got as a result is a broken disjointed piece of "music". I'm not sure. Pretty frustrating.

I'm not dumb. I use a lot of AI for a lot of things - like hours every day. I've put in way more effort than should be needed. I think Suno just isn't it. A lot of hype from a lot of influencers. I dunno. Frustrated.

Okay, aligning lyrics is Legacy, let's see if we can fix this up again.

[young geode]

Have you seen that consistently work?

[sleek stump]

I have abotu a 75% success rate... which is high for Suno.

[wispy laurel]

I'd kill to have anything in Suno do what I want 75% of the time 😂

[sturdy dove]

I'd kill to have it do it 10%

Alright, so let's talk about this. I tried a couple of remasters, but they're still garbage. Unprofessional, disparate styles and vocals.

Is there anything I can do to save this? The transitions are abrupt - that's on me, I could add a couple of beats here and there, but that's not nearly enough to make this be cohesive.

Here, let me share:
https://suno.com/song/2768de6d-3120-47a9-bb9b-836203374d3c

This is NOT something I would ever put out or put my name on. But I have to get something usable.

What do you suggest? Cover? More remasters? I can burn credits...

There's only about 2:30 in here that I actually want. The first breaks are around 1:30. The new male vocalist shows up at 1:45 ish.

Another uncomfortable break around 2. Another at 2:20. I tried to replace the last stanza but it wasn't getting it so I give up. I some lyrics I like better. At this point, I have to publish and I just don't want to be this embarrassed.

[young geode]

In the prompt are you writing it as Vocals: name, characteristics or as Name: vocals, characteristics

[sleek stump]

In styles I am detailing the vocalists with a name, voice type and register range. 4.5 understands normal descriptions for vocalists...
*The song is a duet sung by a male tenor named George and a female Soprano named Lisa. *
This is one of the ways I describe the vocalists in the Styles, then call them by name in the lyrics with tags.

[young geode]

I'll keep that in mind.
If you're getting it to proc, then I might be able to use something like Vocalist 'Name': {Sex, Type, Range, Qualities.} and the metatag [Vocalist: Name], or some variation.

I'll get back to you if I do, probably use an older song and test it out after I'm done with my current project.

[sturdy dove]

I keep remastering and it it's smoothing out a little...

[crude pawn]

It definitely can if you structure your tags "correctly" and guide it

Proof (this was done free, on model 3.5): https://suno.com/s/TvB9oQfTQe1D76XC

[half dragon]

I never meant to say it's impossible.

[crude pawn]

Try separating the persona/voice with a colon : instead of dash -, and only on the section... Too many tags are probably confusing the AI more than helping... You have to think of prompting as a fuzzy probability thing based on tags, not an actually sentient intelligence.

Find what the proper tags and structure that were used or had the most data for training are, set your tags and structure right at the beginning of the section/line, let the model do its thing until you need a change (which you will just tag again then), and you will get better results.

[Short Intro]

[Verse 1: Mathew]
Lyrics here
Lyrics here

[Verse 2: Namie]
Lyrics here
Lyrics here

[sturdy dove]

"Find what the proper tags and structure that were used or had the most data for training are"

Cool, how do I do this?

[crude pawn]

I had no such impression. I was just saying and demonstrating that it can be done and, while yes a bit fiddly, it's not that hard

[sturdy dove]

I've tried pretty much everything at this point, and it pretty much ignores all of it most of the time... I think I'm just asking too much of whatever compute tokens they have

[crude pawn]

Suno had a wiki or faq page with the tags and how to structure them... I don't know if I had it saved, but you should be able to find it

[half dragon]

I know what you mean. Yes.
There are countless of duets made on Suno https://suno.com/search?type=public_song&q=duet - so it can definitely be done.

[sturdy dove]

I started there a couple of months ago...

I've kept evolving my tags structure based on what seems to work and what doesn't. I'm with a couple of others here, having too complex a tag structure definutely can make things worse.

Again, I think it comes down to whatever the server load is at the moment... If it's quiet, it's able to figure it out. If it's overloaded, it just skips it. And we have no control over that...

But hey, I appreciate you trying to help!!!

[crude pawn]

? What happened to my comment?

[sturdy dove]

Woodhouse doesn't like us.

lol, it's pretty annoying

Woodhouse is a bot who enforces certain rules and deletes comments without giving you the opportunity to edit your infraction. The other day I got some great advice on here so I responded with excitement and a short sentence in all caps. And it was deleted.

[crude pawn]

Well I'm not typing that again.

@sturdy dove I'm a software engineer. The TLDR of what I said and got deleted by a bot is: It is unlikely server load affects output. More likely it's tags, structure, and RNG (literally).

As for you "evolving" the tags you use, don't. The AI will have some tolerance for non-optimal tags and syntax, and can give you good results, but the official tags & structure will be best. Be careful with Skinner's Pigeon effect when experimenting with any changes in prompting syntax. Don't confuse changes giving good results because of flukes/RNG with them (the "ritual") being actually better. Unless you're trying to jäïlbrëäk, the best will be what the AI was officially trained to use, in cases like Suno. 🫡

[sturdy dove]

I hear what you're saying. My tagging strategy isn't random. I'll try 10 gens on a set of tags and if they consistently don't produce the result I want, I'll try something a little different.

There's a strange effect that early gens seem to be closest to the tags, and the more you regen, I think we're signalling that we're rejecting that so it keeps trying to give us variations that get further from the original.

When the Suno tags didn't produce consistent results, I looked to what others were doing with apparent success. And I've taken all that on and tried to adjust.

In the end, what I'm trying to do is apparently very difficult. Things that are easily accomplished with a rudimentary DAW are beyond... or extremely difficult to accomplish with Suno.

Of course it'll evolve and get better. But for now, what I'm trying to do seems to exceed the tool's current capacity for whatever reason.

[wispy laurel]

Suno can't do what a DAW can easily do, because that isn't what it's designed to do.
Regardless of who's using it now, Suno like any generative AI was originally aimed at people who don't know how to make music
So it can write you a song, compose everything, somewhat mix and produce it for you because it assumes that you can't do that yourself.
It doesn't allow for fine, precise control or very specific details because it was designed to work for people who don't have the knowledge or experience to properly make use of that anyway

[crude pawn]

Ok, I've looked further up the conversation and gave this a listen.

Yeah, structure cab definitely improve. Some work on lyrics might help a bunch too (consistent and/or structured stressed syllable counts, as well as rhymes and rhyme patterns, help the AI a LOT).

And I can definitely tell there are some edits/remaster issues there; the sudden cuts; it's been degraded too much.

You have a good basis to use to give it for reference, and to guide it, though. It's actually not that bad. 👍

I honestly don't have the patience to track the conversations all the way up for full context/explanation, but if you tell me here in one comment, or shoot me a DM, explaining exactly what you want to achieve, and obviously if you allow me to, I can give this a crack at and, if I can do it, I can teach you how 👍

[sturdy dove]

No, don't do that

This conversation goes back a month. I'm good. What I've got, while still extremely rough, will have to just be good enough. I'll report back when the finished product is published.

And I don't know if I'll be able to tag you because Woodhouse doesn't like too many tags. I've tried to tag a bunch of people in the past who were helping me and it just deletes it all.

But I'll try. 😁

And thanks for the offer!

[crude pawn]

From a fellow AI enthusiast & creator, I hope my advice is of good help, and I wish you good fortune on the endeavor. Definitely looking forward to listening to the final result. Feel free to shoot me a DM with the link when it's ready! 😁

[crude pawn]

I'm intrigued by the Suno Studio they've been announcing. I think there have been generative AI DAWs and even plug-ins for the traditional DAWs popping up in the past maybe year or so?

Either way, I think it's a matter of relatively short time before we have the best of both worlds; generative AI with fine control... And I'm excited!

[wispy laurel]

What the hell Woodhouse 🤣

[sturdy dove]

It's pretty irritating

[crude pawn]

Yeah the moderation bot deeeeefinitely needs some tuning 😅

[sturdy dove]

If they could set up the bot to allow us to edit rather than just delete the tome we carefully crafted, it would be so nice.

[coarse jetty]

Hi~ i am making Persona to copy the exact music in modified lyrics. However, in 50+ attempts it failed. I would like to have same music and singing style for 85% revised lyrics. Thanks in advance!

[unborn holly]

ping me when it’s not dead

[lyric kestrel]

https://youtu.be/rA9bZYGDi1A?si=X1kBf7WSpqaZZBzR

Could someone help with a prompt for this?

[mild egret]

Is anyone familiar enough with MP3tag to know how to get tags from Suno songs directly?

[gritty glacier]

Hello community, I would like to kindly ask for help with a spoken word prompt. I am trying to create a track with a speaker saying something and a group/crow of people replying. Let’s say a teacher would say "Roses are red" and then a crowd would reply "Yes, roses are red". I tried many different prompts and [tags] but I can’t figure it out. Would someone more experienced please advice. Thanks a lot! 🙏

My example prompts:

[spoken word]
Roses are red
[spoken word] [crowd speaking]
Yes, roses are red.

[half dragon]

There are some free tools online where you can upload the mp3 and edit the tags right on the browser.

[unborn star]

Got a weird question, but AI works with Seednumbers right? How is it normally structured in other AI's. What would be the code to say he should seed this with 123456

[polar spade]

@covert tiger

Regarding your question in need-help billing and account support channel.

Welcome to thr tech-prompt-talk channel:

All song are not created equal. If we prompt a few styles, throw in a smattering of lyrics and click Create, we often now will get long songs. I've done a few instrumentals and got a varied 3 to 7/8 minutes. Fine. "Up to" 8 minutes is is "up to" 8 minutes. Expecting 8nminutes is not reasonable. While we dont agree getting 20 seconds outputs is a good thing, they are also not frequent. Somewhere between 1:30 and 7:50 seems to be the most scattered range of times with an average about 3:30.

As when we REMIX a previous output by using Covers, Extensions, and Personas, the creation of an edit song takes a new form. We have the "code" or "seed" so to speak, of the original , in some form, helping to Remix a new output. You're no longer Creating an original, but remixing a previous. Some room and time is needed and allocated to include that code on the tail end, figuratively. The result is less space to recreate, extend, or form a new output on the front end. Suno can take the original output as far as it can go and so may simply cut it off without finishing. The result, for what its worth, only shows the time of the output, not including its allocated "code/seed" time, which is assumed the "rest of the seed information" up to 8 minutes. The presumption is to put some time in re-extending the song, or creating smaller personas or using less lyrics, or whatever it takes to keep the "seed" information to a minimum to make space for a complete song, if thats our priority over the influence of having the seed information and vibe.

Add it the randomness of prompts and their relating to other factors, some outputs just get screwy and unreasonably short. So an output is 10 credits, no matter. A completed song, on the other hand can be a dozen, or hundreds used to get what we, individually, want.

[unborn birch]

Put what you want it to do inside the brackets and end it with spoken word.
So, [Talking male/female "call and response" with crowd backing vocals spoken word]
Also, put the crowd response inside ().
Try that.

[polar spade]

We use personas, I guess, instead of seed numbers? 🤷 That as far as my ignorance or understanding takes me to answer that. We have more control In a persona and sliders, than a seed number such that art AI may use, I would think.

[unborn star]

That's a fair point, you are absolute right lol
Well I don't used personas that often. But you are right.

I just wanted to make a specific kind of music, so I thought i could give seednumbers to the bot, so it sounds pretty similiar

but

You are right, I can use personas for that.

[polar spade]

Dont know if Im even close. It just makes sense to me.

[unborn star]

What will happen if you remix a song, and use the same persona created for this song 🤔 how much will the song change.

Another test I can do later

Thx for the input. Sometimes it's in front of you, but you ignore it, haha.

[polar spade]

Test and find out. Lol.

But, actually, I do that when making lyric changes, so that i can change lyrics using Cover instead of Edit UI but also emphasize the audio and style sliders to 98% to make sure the song doesnt change.

[unborn star]

I will =)

[covert tiger]

Hi there, I just want to clarify that I’m not using any remix or extension features I’m only trying to generate original house tracks directly up to the 8-minute limit. The issue is I’m consistently getting outputs that are far shorter, sometimes ending in silence after about 3 or 4 minutes. I’m losing credits on these incomplete tracks and really need to understand if there’s a way to ensure I get closer to the full length ideally at least 7 minutes each time. Could you please advise on how to achieve that without these shorter, incomplete results?

[mild egret]

Not what I'm looking for, thanks though. I already use MP3Tag.

[polar spade]

@covert tiger You may have posted before. Or someone else doing EDM with a similar issue. With testing there may be some controled prompts which get around 8 minute songs, but then not in the flavor and vibe you personally want. So , its my opinion that, though ,physically its likely, it may not work within each personal prompts and parameters.

For what its worth, often when I use a few prompts for a dreamy, ambient , new age sourt of piece ill get length over 7:30 frequently. EDMs by their nature, as you might think, should be close to that too.

OK. Im just going to prompt EDM and see what happens. Brb.

[polar spade]

Here, for example, is just using EDM. 2-3 minutes adding ambient French house 3-4 minutes, , and then just ambient/dreamy/nature 4-7 minutes,, as the minutes increase with each one. Coincidence or correlation, I dont know. But it appears Suno treats prompts differently. It comes down to adding and subtracting prompts to get a good recipe that also matches what youre going for.

[ebon sand]

Depends on the model, if you use a persona and just a snippet of the same lyrics on v3.5 you'll get the same song as in the persona again and it largely ignores your new styles and lyrics; the models all behave differently (or not at all) with persona

[unborn star]

I would use the latest

[ebon sand]

my opinion with them is that persona doesn't matter to them at all, but others say they hear a difference

i can plug in my cultist hierophant preacher guy and will get female pop vocals often

[wispy laurel]

I find that personas only really work if their style is consistent with what the AI thinks it's supposed to be doing based on other style information and other cues
If you have a persona that's a female jazz singer with piano and harp, and you use that with the prompt "slow, soulful ballad performed by a female singer with minimal instrumentation", in my experience the output will incorporate a lot of the persona, mostly.
But if you use that same persona with the prompt "epic power-metal anthem", neither harps, pianos nor female jazz singers normally are associated with that so it gets ignored. The best you might hope for in that case is that the singer sounds a bit like the one from the persona
Genre conventions and norms are a huge factor, it's incredibly hard to get the AI to generate anything that it thinks doesn't belong in whatever genre it thinks it's working in. It'll straight up ignore both personas and style/lyric instructions that conflict with its understanding of what that style is supposed to sound like.
So personas generally only work if they're consistent with and naturally fit within the genre of whatever you're trying to make

[ebon sand]

my persona is what i exactly want from the model and it does zero replication of anything in the persona

the biggest part missing are of course the vocals, but the music or style isn't replicated either

[unborn star]

Wait that's what I wanted to use the persona for, creating similiar pieces of music

[ebon sand]

if your taste aligns with what v4.5+ does then you might get the impression something works, I don't like the outputs, thus I wanted to use persona plus prompting guidance but it's a zero success party and doesn't fetch any of that

the playlist inspo thing worked to 20% which is at least more than 0% but it only created usable parts, like 20 seconds at a time or so

[unborn star]

Are you using ChatGPT? for prompting

[covert tiger]

Ok thanks for the info. Not exactly the style of house I’m going for but I will mess around and see what comes out of
It. Do you know of Suno intends to increase song lengths in the next update? Failing that I’d pay good money for API access and pay the generation fees as long as I can get what I want. Either that or go back to using a DAW and learning a few instruments.

[unborn star]

you might get better results if you use it correctly together

[ebon sand]

not really because it has zero idea what music i want

[unborn star]

That's the reason you tell him and he helps you to find it

🤔 But maybe it just works for me

[covert tiger]

I find Claude created much better prompts for music production then ChatGPT but each to their own

[unborn star]

Well I pay for it, so I trained it to write prompts down

[ebon sand]

it just hallucinates a prose from keywords and i can do that myself as well

[covert tiger]

I have a paid version of ChatGPT also. But I still find Claude does much better for me especially for house tracks prompts

[unborn star]

I see

well ok =)

[unborn star]

I would like to try to improve one of your songs just for reference

are you fine with it?

I want to know if my prompting might be bad

[ebon sand]

I also have a gpt sub but i don't use it for anything music related

[unborn star]

I see, I got really good results for me at least, so I was wondering

[ebon sand]

might be very dependent on what you create

[unborn star]

I´m confident, that I can improve any song. I worked a lot now create a workflow to create or improve songs

as long they are good at least, no random songs or something liek this

[ebon sand]

good is a big word

[unborn star]

Ik =)
Music as a matter of taste, but that's not what I refering to.

I said I improve the song, make it more full. Most songs I listen to were on layer 1 if you know what I mean with it.

[ebon sand]

without a comparison hard to say

[unborn star]

Take any pop song

[ebon sand]

oh i have no idea about pop

[unborn star]

What kind of music you do?

[ebon sand]

I'd say sonoran desert ritual chants with bluegrass elements

it takes some of this and some of that, sometimes it's more ritualistic percussive and sometimes its more bluegrass

[unborn star]

You have something to share?

Cant imagine it right now

[ebon sand]

nothing completed at the moment but bases i want to work on
this for example was a parable
https://suno.com/s/P1oftzMtP6EWM8Nu

And this was peak performance of my singer who cant sing well, as it should be
https://suno.com/s/hYzVMG7VKW1vObV9

[unborn star]

Hmm I see can't yet see what you want to reach, but listening to the first song, right now.

[ebon sand]

reach?

[unborn star]

what's your goal, what you want to have in the end. 🤔

[ebon sand]

the glitches gone and some of the responses spoken differently, but i didn't work on them yet to remove the glitches

[unborn star]

I could try to create it in my style, but I guess I will never create something what you would like 🤔

Should I try?

we are speaking about techique to create these kind of samples

I will give it a go, wait

[ebon sand]

if that's of interest to you, but you're lacking the main ingredient, and that's the singer

he's my hierophant

[unborn star]

That's a fair point, are you insteresting what I get out of it?

[ebon sand]

i would give it a listen, but I'm unsure about the purpose of this experiment

[unborn star]

I´m not sure neither, but for me it's like testing if I can reach similiar results and if my prompts make the song in the correct direction. I never tried these kind of songs, so I´m curios

[ebon sand]

haha, then go for it, good luck

[unborn star]

hopefully I get a response after this work. I research a lot of things, how about Suno works etc. etc.
Back 3 Min

ops mixed both version

I guess it would be a bit easier, if I could make remix out of your song xD

So I get the voice at least, you worked on that voice, I see

I get really chanting music with gituar should remove them

but funny

[ebon sand]

it's a persona, but v4.5 and + don't replicate it anyway, they both have a standard voice for some greek though which is pleasent as well, ive gotten it across multiple songs without persona on v4.5 and heard it from others as well

[unborn star]

I have another idea

🤔 ok

[ebon sand]

https://suno.com/s/yFzBqUuL29OVuL7Y i think that's the one, the pronunciation is incorrect but i didn't make an effort with it really

[graceful condor]

I tried this a couple of times now but I am still not sure if this aspect really helped. But does SUNO react to "clean up Audios" you use as reference. For example if I upload an Audio of mine, a piano recording and there are some wrong notes in it if I command Suno like this "Enhance the uploaded melody into a polished instrumental piano piece, with a full, authentic orchestral accompaniment. Enhance the melody, fix any errors like wrong notes or interruptions, create smooth transitions." Is this in the end working? Or if he didn't play wrong notes and create it clear is it just because of the Sliders used?

[unborn star]

But thanks for the experience, never tried something to do with this topic. I might be using it for Pen&Paper Adventure, when I write one

[unborn star]

Well how he suppose to be what clean up means?

[ebon sand]

haha you'd have to exclude arabic structures, it's supposed to be american

[unborn star]

OH thats the poINT

I just was wondering

[ebon sand]

which is why i include bluegrass to not get the arabic desert vibe

[graceful condor]

Don't understand?

[unborn star]

ok so exclude what?

[ebon sand]

i don't know, i don't work with v4.5, on v3.5 it's enough to just tell it bluegrass and canyon and it won't give me arabic maqam or wharever

sonoran puts a location, maybe it can work with that

[unborn star]

I mean, how should the bot need what is wrong in the song. You need maybe to give better instructions, to make it more clean in specific situaiton, what version you use?

[unborn star]

So i need to clean arabic right?

or something else?

[ebon sand]

i don't know, you don't have arabic in there but it gave you something that sounds arabic, maybe its enough to exclude it

[unborn star]

ok

so it shouldnt sound arabic ok

[ebon sand]

it also seems to have ignored the guitar work

[unborn star]

guitar on which part?

[ebon sand]

i don't hear any here

[unborn star]

looks like i need to listen again

ok my bad xD

ops

[graceful condor]

As I said I give him this instructions for example ""Enhance the uploaded melody into a polished instrumental piano piece, with a full, authentic orchestral accompaniment. Enhance the melody, fix any errors like wrong notes or interruptions, create smooth transitions." and in the end he does not play the wrong notes from my Piano audio even if the Audio Slider is at 80% or so. So I was wondering does Suno automatically recognize wrong notes if you give it more room with weirdness or maybe because it detects wrong notes. Or is what I added also part of why it creates it clean? I was just wondering if such instructions have an actuall effect.

[unborn star]

ok wait i will add it

[ebon sand]

https://suno.com/s/Et1r636YOxanYIMf here's another example, here you hear the upright bass (cruicial to have some bluegrassy vibe, it's no true bluegrass) and that a guitar is strummed without playing notes but muted to get the percussive element, not my idea but i liked it

[unborn star]

Hmm

[unborn star]

I guess this would be a bit easier, but it is interesting to work with chants music

It's difficult to make music like that. I will give it a last go

But ye you won with that kind of music. I need to learn the theory about the music, to create really good one. xD

[ebon sand]

haha you are welcome, i'm happy if it was a challenge

[unborn star]

I have problems with chanting and arabic, If I use exclude I wont get the guitar somehow

That's the reason I gave up

you create with?

4.5?

[ebon sand]

v3.5

[unborn star]

OHHH

[ebon sand]

only one of the songs i sent was 4.5+

[unborn star]

This might be explaining a lot, but can't use it with my style of prompting to less characters

[ebon sand]

It's a very different approach, my prompting and especially the model creates very varied results, I don't have a clear plan for the song I want, thus I just give a few core directions, plug in my persona and let the model figure out what the hell occult Sonoran desert ritual bluegrass without a banjo is, it's consistent in itself but the outputs are most of the time very creative, very cool, it has a wide range to hallucinate in and attempt different approaches
I like this about the older model, I don't want to dictate a specific formula
I rarely change my prompt, its the same for months

[wild egret]

hey, i want to recreate this, but i cannot give it the sensual feeling id like it to have.

i've tried simpliest things like :
sensual music for before sex, lyrics talks about love and sex, its low tempo, similar to songs from the weekend
and gave me this
https://suno.com/s/wQV1cvi9BMVvFrlP

and more complex things like
seductive alternative R&B, Lush cinematic strings swell with passion over deep bass and sparse, heartbeat-like percussion, Minimal piano chords add warmth, while falsetto-style vocals float with reverb, alternating between intimate whispers and soaring emotional peaks, The pacing is languid and hypnotic, around 60 BPM, with gradual dynamic builds that feel both romantic and dramatic, The mood is sensual, cinematic, and passionate, perfect for a late-night atmosphere
https://suno.com/s/h18SC4XsTqR46CLk

i feel its like those "the weeknd" songs

[winged peak]

Are you using a Persona and/or Cover? What happens if you add an [interlude] in the Lyrics-Sections or even something like
[interlude]
(8 bars)

Known issue. Regardless of the length of the persona, many tracks stop after the playing time you mentioned. Call it a bug or a glitch. 😉

[half dragon]

if it consistenly cut off then only way to finish is using Extend.

[winged peak]

I think I read on Reddit in the Suno sub that some people had success using the ‘extend’ function to generate parts 2 and 3.
Others seem to have had quite a bit of success adding to it with the new editor.
But I only know this from hearsay.
If you leave out the persona, the song won't have the timbre, structure, instruments, ‘soul’ and vibe of the persona, because in my experience, even low values in the advanced settings (‘audio influence’) have a significant impact on the new track.

Well, there are now a few different approaches you still can try: 1. Extend, 2. Legacy Editor, 3. New Editor.

[near flume]

Anyone else having problems with the output today. Everything has a lot of pops and distorted vocals, even the stems suck. Songs sounded good last night when they were created but now.. ugh!

[unborn star]

Hmmm, I heard that you got that sometimes. If a lot people using the bot, it will get into trouble. We in Germany it was at around 12 pm had the same problems

[polar spade]

Oh. I was just showing that just changing or adding genres or prompts can change song length. Not suggesting that as a viable vibe.

[polar spade]

Its not considered a bug ot glitch, its a constraint placed when allocating space to include seed information from the original.

If you create a new song, your constraints may be just your lyrics , your prompts ,exclusions , ect. themselves which some may require more data, more data means less minutes, but for 8 minute limits, usually doesnt affect much.

Remixing or editing would need more allocated space when making covers, extension and Personas for the seed "code" and so less room for the output /song itself.

[half dragon]

So it's kinda like "context window" for LLM, right?

[robust anchor]

Hello, I am a paid user and wanting to create some tracks but the "Style of music" field is very limited. If I pay $99 I can get 1000 words, but with $9 I get what 25? How can I describe an instrumental with so few words?

[unborn star]

???

[unborn star]

What are you even talking?

why you need more than 1000 Words?

You can create one with limit after this you can extend the song

and get a longer song with your lyrics

But I don't see why you want to make a song over 5 min long

[robust anchor]

That is what is offered if I pay $99 but I am on $9 and I can only fit 25. I am working on some instrumentals that need longer prompts.