📚┃suno-school

[proud dagger]

Example if I choose Extend from 8.0 and paste part 2 lyrics deleting old one in the lyrics field?

[half dragon]

Usually for best results, should always choose the extension point a bit back into the previous part, not at the very end...
But being 8 min which is the max length - results might be vary as the output might be affected by the context windows loaded by the previous part.
I've never had that length to extend from so... not so sure

[proud dagger]

Let's take another example of song which is 4min29 secs. Can I extend from 4min.29 secs or choose 4min15 secs. Second question about lyrics: Should I remove the part one lyrics and paste extended one?

[half dragon]

Maybe try to make a cuts to smaller clips, then extend from the last clip, then stitch them back together later... manually
Like in Studio or DAW, then remaster the whole thing to smooth out imperfect stitches

[proud dagger]

and Lyrics?

Remove old and paste new one or add the lyrics to old?

[half dragon]

Well... on extend, you'd either have new lyric or repeat lyric from the previous part.. but always have the new section lyric, and remove the previous parts lyric

[bitter skiff]

hey, i’m an AI and full stack dev, been working a lot lately on automation pipelines that link slack, notion and some custom ai agents i’ve built for clients, mostly around rag setups and multimodal stuff like clip and yolo mixed with a bit of voice ai. it’s been kinda wild seeing how many projects turned into something real for people using them, feels good honestly. always tweaking things, learning, and letting the systems grow on their own in a way.

[plush field]

any idea how to get a kick drum from the start, not when suno decide it is time to introduce it

i need "extended intro/outro" with kick from the start (in intro) and until the end in outro)

even with this

[Intro - Kick drum beat]
.... .... .... ....
.... .... .... ....

[Instrumental Hook - Kick & Bass]
.... .... .... ....
.... .... .... ....

full blown kick starts at 0:15

i tried literally 100 combinations today
my brain hurts

[white pecan]

Maybe try this :d : [Drop] [Pearl's Drum battery kit, Roland TR-808 and TR-909, Kick/Bass driven, Catchy SERUM-bassline driven]
.... .... .... ....
.... .... .... ....
.... .... .... ....
.... .... .... ....
.... .... .... ....
.... .... .... ....
.... .... .... ....
.... .... .... ....

[modern vine]

Yes I do this, I change the styles as needed as well to change up the end of a song, or last half sometimes using 3 extensions 😆
Add some instruction in the lyric sheet, like an extended outro and the style box for experimentation purposes

[polar spade]

@noble current You may need to click the desktop box. It will reset the artwork to the top left. Recently, ive noticed Suno has the setting as a default in some instances, so its stretched out to the edges and cuts off about 30%. Not sure what the incentive made the change. Maybe just an update glitch that keeps turning off desktop mode.

[proud dagger]

Does hook uses credits?

[frigid orbit]

It’s far less reliable now with extends looking back at the previous part prompt/lyrics

[polar spade]

@carmine gorge I cant see your menu, so it may be you are not aware the Get Full Song can be found in the drop down [ . . . ] menu. Click menu/Create/Get Whole Song from your Part 2 output.

[elfin lodge]

Hey is there anyone here who I could talk to about a small issue I’m having

Suno related of course lol

[polar spade]

Just mention it. Someone might or we can guide you where to ask.

[elfin lodge]

I have some pre made instrumentals that I really want to use but everytime I do Suno completely switches it up even with the style influence cranked up

I have 0 idea how to fix that or how to trick the AI into playing my song to near perfection

[polar spade]

The Audio slider is the one that we use to mirror the uploaded audio more closely, if your style slider and witness are default,and no prompts or lyrics, you can try Audio at default-- which you seem to say is too "generative" and varied. So set Audio, only, to between 65 and 95% . And try that setting.

[elfin lodge]

These are the 2 instrumentals I’m trying to use both are 2 different songs within the album, should I add style tags to these? If so I’m terrible with prompting

[polar spade]

That seems like a separate question than the first?

[elfin lodge]

It’s in regards to the first question

[polar spade]

Adding prompts will or can alter the arrangement and even composition, as well as the genre to a bit of output from it sounding as similar, with a better fidelity and master of course, so at least try with:

Uploaded audio
No lyrics, no tags in the lyric box. Leave blank
No prompts.
Weidness/styles slider at default (50%)
Audio about 78%.
No exclusions.
No male/female.

[elfin lodge]

I already have those in my library what’s the next step

[polar spade]

You said it doesn't emulate the original upload as near as you wish, right?

[elfin lodge]

Can I send you an example of what i mean?

[polar spade]

I dont think I will know what you mean, if you say Suno doesnt make a similar output to the upload. But DM me a before and after if you wish. Or post here if you dont worry of privacy.

[elfin lodge]

https://suno.com/s/heJwOv2BqkFHlzD5

Here’s the uploaded version

Sounds identical

https://suno.com/s/2NsYw1y4TS98ioN7

Here’s the song I tried to make with it

[polar spade]

You just now says "sounds identical". Did you mean it DOESNT sound identical? Because it doesnt. It only sounds a bit similar in melody and riffs.

Right?

[elfin lodge]

The uploaded version sounds 100% like the audio files I sent earlier for this song

I’m a bit of a perfectionist so my brain wants my song to perfectly match the audio

[polar spade]

Was this left in the style box when you ran an extension or cover? You should remove this ,if it was. Leave style box empty.

[elfin lodge]

When I uploaded it, for some reason it randomly gave me a style description I did not put that there

[polar spade]

Suno is not a cloning app. Its generative music AI app. That means it emulates but not clone. It eill be close but with enhancements. Thers no perfect clone setting.

[polar spade]

Right. Correct. So remove it before and use the settings i mentioned. Ill try one myself . Hold on. Lets see what happens.

[elfin lodge]

Okay!

[white pecan]

Hey everyone, I have a question for Suno's team, could we have (in the future) the possibility to put markers on tracks in the create area ? I mean, a little onglet at le left of the three point option menu of tracks, in wich one we can choose in a choice of color (same as PSCC, PRCC or illustratorCC, from adobe) like this we can mark the potential of our best generation for final form work, is that a wrong idea ? :d Peace ❤️

[polar spade]

This one was the better of 8 outputs using the settings I mentioned before. None had as much prominent piano as the original. https://suno.com/s/4NSJzu99PZL9oviW As you can see, its close to instrument and drum patterns, but has slight variations from being an exact close. Its attempting to "cover" but enhance at the same time.

[elfin lodge]

I’m okay with changes I’m really trying to lean into that sleep token x bad omens style influence that me and the band are going for

@polar spade i DM’d you

[white pecan]

You made that track ?

[polar spade]

I did then , add only "ambient piano" in styles , to the same settings without changing anything else. Which may have given a nice push. But thats totally your ear, and choice to then start tweaking with prompts.

I noticed it came in better with a piano riff in the intro as the original, butttt... at an octave lower. Potato / potahto.

But this is just so you can see what those settings can do to get something pretty close in the vibe as well as almost the same sound, but not exactly close the uploaded audio.

https://suno.com/s/vOXFOfgUkcbpx7Dh

[polar spade]

Yes. I posted above #📚┃suno-school message

[white pecan]

I mean, Go make public tracks no ? an album ? :d

your sound quality is impressiv

[elfin lodge]

This version I liked

I guess the main issue I’m having is when I add lyrics the AI seems to just drift farther and farther away from the sound I want

[polar spade]

Yes. Now THAT is quite true. Suno has to tokenize every word and somehow fit it to match as best possible what you have as the instrumental. So it can go in all sorts of directions since you have no melody in the instrumental.

It likely will pull some of the melody from the riffs it hears.

[polar spade]

I use Suno and Premium, though it should not make a difference between that and Pro.

[polar spade]

Hope my answer gave you an understanding that the audio slider affects the uploaded, or rather gives weight over anything else to make it sound as similar possible with the underlying rule that it wont clone it exactly.

[white pecan]

weirdness = 0% / style = 100% / audio influence = between 25% and 85% it depends on how you wanna keep the original theme 🙂 but sometimes i go 15% 85% 85% but..... this means more generation in my opinion :d

[elfin lodge]

With that being said, would adding style tags help it be on that right direction? I know not having style tags helps keep it basically the same but instead of letting Suno assume how I want the lyrical version to sound should I add tags to direct it?

[white pecan]

Soulful-Pop vocals is an amazing one :d

Emotional-Pop vocals works too

[elfin lodge]

I have the audio files for the original pre made instrumentals maybe give it a listen and help figure out how to prompt it?

[white pecan]

yup

let see that

[polar spade]

I also added "ambient piano" to that second output i posted, as mentioned. It still sounded pretty similar in vibe, but added more nuance piano at the beginning. Another output i listened to, sound quite good, but because I put ambient, about 40 seconds in it started in with a violin , which was NOT in your original. So even the slightest of prompts can start taking it in different directions. Some quite nice and maybe preferable. Its up to you to then experiment with a base settings then tweaking with added prompts, yes.

[elfin lodge]

@white pecan

[white pecan]

so 2 differents tracks ?

[elfin lodge]

Yes

[polar spade]

OK. Ill let you continue this offshoot convo. Good luck.

[white pecan]

mmmMmmmmm

[elfin lodge]

I’m only wondering because if I start adding lyrics I don’t want it to deviate and get all weird on me

[white pecan]

you made this in live captation with real instruments ?

[elfin lodge]

It’s audio my friend bought of someone

[white pecan]

mmmmmmmmm My best advice ? cast a singer and record him :d huhuhu

[elfin lodge]

I am the singer lol

We don’t have the money or time to actually record it so I’m using Suno as a roadmap to how I would want it to sound when I do sing it

[white pecan]

ok ok

[proud dagger]

I am trying to make something for my kids like poem, story ryhmes etc but cannot find a way to do it in style that would fit. Any suggestions

[elfin lodge]

Lemme know what you come up with for both of those if you can 🙂

[white pecan]

at first, use EssenceExplorer from chat GPT

it will describe each artist or music style in prompted words 🙂 ready for ai

like this you can think about a group you like and get his precise definition with directly prompted words 🙂

[elfin lodge]

What if both of those audio recordings are a mix of two bands

[white pecan]

no matter

you Describe 2 different groups with EssenceExplorer and then you take the precise words you want for this singer voice

or you record your voice with suno, you generate in different metal/symphoniq metal/Nu metal style and when you get something you create a persona

and you re create something with that persona, and you re-do the same process until you get The result you was looking for

i gave you 2 different possibilities, now you should be able to create those 2 voices :d

but i can't prompt you this, this is not my style anymore :/

i'm more into EDM 😄

[proud dagger]

any ideation would help

[white pecan]

well if they like pirates you can try [Sea-Shanty]

[proud dagger]

nope kids are small :D

2-3 yrs old

[white pecan]

Christmas song / epic sea shanty / philharmonic orchestra

Disney theme, Disney song, you can try things like this

[elfin lodge]

https://suno.com/s/8OECZtPt9Tc8KD46

@polar spade @white pecan

Using the audio uploaded and adding some style tags made this song perfect

[white pecan]

that's a poetry ? so you want it looking old no coming from the old times :d

[polar spade]

Although I could not tell you the outset of what it would sound like:

Investigate the "Inspire", or inspo, tool of Suno. You can upload up to four different audios, put them in a Playlist. *it doesnt have to be uploads. Any 4 outputs will do.

Then click Create, and in the panel click INSPO and choose that playlist to insert 4 audios into the create box. And experient from there.

They idea is that it should review the four and come up with a blended idea of something else similar. Typically it puls from the first in the list more, so im experimenting on have the first audio quite short so it has less to pull from.

Anyway, you might like what it does. Who knows. Have fun.

[gilded plume]

Hi everyone, when i create hooks it seems limited to 30 seconds, but on some other creators pages i see hooks that have 1 minute duration. Any idea how to get the doubled duration ? Thanks in advance!

[elfin lodge]

Yall are so helpful

I love this group of people

[white pecan]

WOooOoooo you are sooo right 🙂 i've never test that, i'll definitly do it :d

[polar spade]

Well done. 🤘 Xcellent.

[polar spade]

Combined with your great suggestions, this inspo tool might be quite handy.

[elfin lodge]

You know I found the most egregious way Suno messed with my audio lol

https://suno.com/s/93hJDmqIrRj9Mw2e

Those weird riffs were not in the OG version of the audio file lol

[polar spade]

Turned weirdness slider up a bit, did you? 😁

[elfin lodge]

It was at 0 lol

[polar spade]

"O" and "100" are about the same. I never use 0 or 100.

[elfin lodge]

Really?!

So if I want my audio to sound almost identical don’t use 100?

[polar spade]

Well, in all candor, they can give odd results or not be consistent. But as you see sometimes something nice may still come from it.

[elfin lodge]

The one I posted saying sounds great had an audio influence of 90% with style influence being 60% and weirdness at 15%

[polar spade]

No. The 0 setting are less invasive. 100% weird just adds everything into the pot and its just a cancophany of noise and glitching.

[white pecan]

thanks, you're welcome 🙂

[white pecan]

Why ? :d i'm interessed

[polar spade]

Yea. 1% and 98% are about my limits. Hey, i also do mostly folk and pop and ambient stuff, so different settings may sound better for what you're working with.

[white pecan]

yeah i see what you mean

when i see that the generation isn't working i use to try some 15% weirdness /85% style

[polar spade]

Its anecdotal. Just try it yourself and see. 100% tends to just blow the engine. And "0" seems to act more like "infinity" or open-ended or "null" response rather than actually removing any code/instruction to alter the sound. Just my opinion, and a few others.

[frigid orbit]

It’s crystals, alchemy and magnetic healing. There’s no science, just feeling good about yourself.

[polar spade]

Dont forget aromatherapy. After all, without it we could not have got It Smells Like Teen Spirit.

[shrewd willow]

Did you guys managed to do any prompt for Christmas songs? Suno doesnt seem to know what Christmas is and doesnt ackowledge it as a Style 🙁

[knotty stag]

Is the 8-minute limit on creating a song still there? Because I just created a song in Suno that is 9:58 long.

[quick wedge]

I’ve made Christmas songs and haven’t had any issues with the prompts they just look holiday vibes

[shrewd willow]

Not working for me. It does capture the vibes, because of the overall style prompted, but it lacks the christmas song characteristics like jingle bells.

Different than my Halloween playlist where it acknowledges the Halloween style

[quick wedge]

Try this prompt

Festive Christmas EDM electro swing track with a joyful holiday atmosphere. Vintage swing brass and clarinet melodies, walking upright bass, and jazzy piano rhythms fused with modern EDM production. Bright holiday bells, sleigh bells, and subtle choir pads throughout. Energetic electronic drops with punchy kicks, sidechained synths, and groovy electro swing beats. Cheerful, playful, and danceable Christmas mood, perfect for a winter party or holiday celebration. 120–126 BPM, major key, upbeat, polished, radio-ready

[shrewd willow]

It didnt acknowledge the style, but it did in fact brought the jingle bells right at the beginning. I think this is good! Thank you ❤️

[frigid orbit]

Google what are most common instruments found in Christmas tracks, perhaps their genres and vibes, perhaps a distinct nationality

[quick wedge]

You can try this also if you want a club vibe too

High-energy Christmas EDM track with electro swing influences. Punchy four-on-the-floor kick, sidechained synths, and powerful drops combined with swing brass riffs and jazzy melodies. Festive bells, holiday vocal chants, and winter atmosphere. Club-ready, fun, and energetic Christmas vibe, 126 BPM

[quick wedge]

You’re welcome ✌️

[shrewd willow]

Merry Christmas for you guys ❤️ 🎁

[quick wedge]

Thank you Velvet Bunny 🐰
Merry Christmas to you also 🎄✨

[frigid orbit]

Cowbells and Churchbells are often forgotten

[shrewd willow]

Its working so well now T_T

[polar spade]

Just "jingle bells" can be enough. The rest becomes nuance in genre. But jingle bells, Christmas and jazz often get a more Peanuts /Vince Guaraldi vibe.

[shrewd willow]

Yea, I added Pimolruk's advice of Holiday bells, sleight bells to my prompt and it worked. I also added Festive next to Christmas. Both on the prompt and song metatags. Its working fine, thankfully. I dont have many credits left this month for trial and error 🤭

[polar spade]

Yes. Stick with just the "bell" prompts to leave room for your own prompts if you need.

I realized that I may have sounded like im dissing Pims suggestion. Not intended. Just saying the more prompts to start, the more Suno funnels down to one sound and harder to get your own prompts to blend in without making a persona first and then turn around and add it to the first result.

[polar spade]

@dull coral Anything is fun to try. But whether it works is up to your results and being able to repeat the results successfully.

Ive move this to the prompt channel where is would be the better thread among users.

[polar spade]

@dull coral but you're also asking to the developers if this is something Suno recognizes. Yea, we will not ever know that, so much. Thats more proprietary and industrial secrets that would not so much be disclosed.

So its just the "try it and see" approach and report back.

[pastel dock]

what can i write in the prompt to make the song quality worse

[polar spade]

Distortion? Din, noise, cancophany, avante gard, lo-fi, crackle, vintage, experimental?

[pastel dock]

distortion, that's it. thank you

[polar spade]

OK. Try that. 👍

Ive only listed what have worked for me in the past. So, not sure. Your mileage may vary.

[pastel dock]

yeah well i just clicked randomly on filters lets see what i get

the worst AI music cocktail concoction you will ever see

wait it actually sounds decent?

[polar spade]

What i havent tried is something like "prepared piano", but as thats kind of a genre, it would be interesting to try. (Since I cant prompt for John Cage)

[pastel dock]

my goal is to make AIs say funny stuff in songs

[polar spade]

As in melodic hallucinations and gibberish? Or more of that off the wall creepy babbling and talking?

[pastel dock]

that's what i have been doing, ive already published a bunch of songs like that

but i prolly can't post them here because

1. a lot of swearing
2. content some people would consider "offensive"

awh hell nah im definitely getting blacklisted if i submitted any of em

[polar spade]

Yes. There is a bit of NSFW parameters in both Suno and Discord. Sometimes just depends on the time of day -- and if I catch any of it in time. 🤣

[pastel dock]

isn't suno the way people make those clash royale songs with insane lyrics

how are THEY getting past the filter

[polar spade]

Some users sneak a few by with Suno's back turned, sometimes.

[pastel dock]

dawg.

[polar spade]

If its rampant with one user, there IS a report button. 😁

Proof it exists is not permission to do more of it. 🥴

[pastel dock]

heh.. im basically already starting** an underground AI musician career...

[pure panther]

I was making a decidedly not Christmas Song and this is what it gave me 😂 Suno has a mind of it's own

[plush field]

what?!!!

blah, ok hardtek it will be 🤷🏻‍♂️

[hard beacon]

https://tenor.com/view/hello-hi-hy-hey-gif-8520159980767013609

😄

[hard beacon]

Just a question that doesn't really fit with the topic of this channel... but... is there a “Suno User Remix” channel? You know, where you can share a song link and call for a remix contest for an official release?

[neat nexus]

This sounds like what you're looking for?

[hard beacon]

let me read it 🙂

[azure oxide]

https://youtu.be/l9bFZNbQ0js?si=K93TfTVqnyD3k0oe

hello in try to upload my song to suno along with another for inspiration test is there any way to see if i can prove to suno that this isnt copyright... maybe with like a proof of relese or somehting?

[remote ivy]

what is everyone using for their tag references, i think this one of the top search results i looked at: https://github.com/stayen/suno-reference/blob/main/tag-reference.md

[hard beacon]

thx for the info mate..

@neat nexus : so there is no specific channel for "just asking" - do you wanna make a remix ???

I haven't just joined this server today, but I just haven't found the right channel yet 🙂 I also don't know if it's “allowed” for me to ask for a remix that I can then publish worldwide on around 200 streaming platforms...

[azure oxide]

Never mind it seemd to go through.... Now lets just hope the next one goes through

[hard beacon]

haha... and Woodhouse is in battle mood again 😉

[dull coral]

Yes, it was a scam detected by the Netcraft browser extension. - It has now been removed.

[neat nexus]

rick-and-morty-that-just-sounds-like-slavery-with-extra-steps-aliens-slave-extra-steps-gif

You want people to make you remixes you can monetize?

Am I understanding this correct?

[hard beacon]

I can't really relate to that statement... it was just a question 🙂

I've been in the music business since 2004... monetization that isn't beneficial to the artists would land me in jail 🙂 I just wanted to know whether it's possible to discover new artists in an unbureaucratic way in the age of AI music... but the answer is enough for me 🙂

But no problem... I'll just do it the old-school way 🙂

[shrewd willow]

I am kinda unsure what you are looking for, but there are lots of viral songs on tiktok which people are using and not getting punished because the creator allows the use.

Might be a good way to start, check for songs that went viral and that fits your project

This one is 100% AI and absolutely viral in many platforms (incredibly dirty lyrics, but PT-BR)

https://www.youtube.com/watch?v=vRBu_RLBt1A&list=RDvRBu_RLBt1A&start_radio=1

[hard beacon]

Let me explain in more detail...
A few weeks ago, I produced a song in FL Studio... the vocals were added later using Suno Studio... I just wanted to know if there is a way (and if it is allowed) to publish this song as a remix contest in order to then publish the best remixes... of course, all with the appropriate contracts for the artists... 🙂

It was just a question... nothing more, nothing less 🙂

[shrewd willow]

Suno does that from time to time. You can probably find past contests so you can adapt to your ideas. I don't think there is an issue with your plan (If I understood correctly, a contest to remix a song you created on Suno).
A famous publisher that some creator use for Suno is Distrokid:

https://distrokid.com/

I think, as long as you make a transparent Terms and conditions for participants to read and acknowledge, I dont see any issue. I dont even think you need to publish the song, as it will just be a template for future remixes.

[crimson pine]

which plan you recommend for distrokid?

[hard beacon]

distrokid is a service for “beginners” 🙂 I just wanted to know if it is permitted to publish a song edited “with Suno Studio” as a remix contest with the benefit that the best remixes will be published worldwide.

But hey... no problem... I own the rights to the original song and I have dozens of producers in my contacts who would remix the song... just an idea 🙂

[shrewd willow]

https://help.suno.com/en/articles/2746945

I think this section has the info you are looking for

[hard beacon]

don't worry.. i own the rights of the original one coz i produced it by my own 🙂

[shrewd willow]

Then you should be good to go 🥰

[hard beacon]

Let's forget about this topic... it was just an idea to promote AI musicians... I'll stick to “real” producers 🙂

[frigid orbit]

Traditional is a nicer term for both sides 😉

[hard beacon]

yapp.. agree.. was just an idea 🙂

[frigid orbit]

Do you mean remix as a word or as the function in Suno?

[hard beacon]

???

[frigid orbit]

Do you mean remix as in the word for Remixing something? Or the function in Suno referred to as Remix (think cover, extend, speed etc)

[hard beacon]

Since I am a full-time musician, I mean “cover” as “remix”... that is, a reinterpretation of a song...

so in Suno words "cover" 🙂

Publish song link - Cover function

🙂

[frigid orbit]

Nevermind, forget I even asked.

[hard beacon]

ok 🙂

[quick wedge]

Your interpretation of it sounds confusing 🫤

[hard beacon]

https://tenor.com/view/homer-simpson-the-simpsons-monkey-gif-9673882010326475083

[hard beacon]

That could be down to the translator 🙂

[quick wedge]

✌️ that’s cool you’re a music producer though anyways have a wonderful day 😎

[hard beacon]

you too mate 😉

[molten jay]

makes a ambient in background of my music! what should i do, it just create my instruments without wob wob or wooooo

[frigid orbit]

Try extending from a sample that has 0 audio.

Let me find you my playlist of samples

[molten jay]

i have no sample

[frigid orbit]

https://suno.com/playlist/8ac4d413-63c3-4bd2-a845-50c3caeca8cd @Opti - Silence / 30sec / 1min/ 2min all have little to no audio, virtually 0 volume. Try extending from either of those with the exact same prompt.

[molten jay]

should use these as cover or?

never mind, you mean if i make a silent audio then extend it i will have raw instruments ?

[frigid orbit]

I'd suggest extend, but for some reason 0 audio does something magical to Suno's models.

Lets say I want to do a speech, and I extend from silent. It'll remain silent mostly with just a voice. Or go for something minimalistic. Extend from the 0 audio sample, and it'll keep it fairly minimal.

[molten jay]

let me try it

[polar spade]

Any remix is a collaboration. Youd have to work out the contract on that first with the other user.

And I believe they are also non-commercial, unless you remix your own creation. Its good to check the TOS again to be sure.

".....If you are a user that activates features that permit other users of the Service (“Remixers”) to remix your Outputs (each, a “Remix”), then you agree that all Remixes shall be a joint work owned jointly and equally by you and the Remixer to the extent that copyright vests in any such Remix. Furthermore, if you are a joint owner of a Remix generated by these Service features, and regardless of whether you are a free Service tier user or a subscriber to a paid Service tier, you additionally covenant and agree that the Remix may only be used for lawful, internal, personal and non-commercial purposes, provided that you give attribution credit to Suno in each case....."

[hard beacon]

I'm talking about a completely original creation... only the vocals were created by Suno Studio 🙂

And no worries... contract is available 🙂

But as I said... no problem... I'll just use the classic channels for a remix 🙂

[polar spade]

I can't tell which "remix contest" is referred. A Suno one or just one randomly outside in the real world. Such as maybe the John Lennon Song Contest having a Remix category

[hard beacon]

Forget it... it doesn't matter 🙂

[polar spade]

For Suno Remix Contest. I think not.

For outside ones, many dont have restrictions so much but do ask to reference AI participation. And even those rules may change weekly as attitudes change.

[polar spade]

Done. Forgotten. 🙂

[hard beacon]

Once again... I produced a song... independently of Suno... then integrated vocals with Suno Studio... and I just wanted to know if it would be possible to organize a remix contest... but if that's not possible, then I'll just contact my contacts... it was just an idea 🙂

In other words: thousands of tracks are released on Suno every day with the option to cover/remix them... I just wanted to know if these remixes can also be released (after consultation).

[quick wedge]

I don’t think there’s anything like that here unless it’s something like this #1254876104297283626 message I’m not sure either of what you’re asking though

[neat nexus]

As far as I can understand it, they made a song in a DAW, used their paid Suno account to craft some good vocals for it and are now seeking people to do remixes for their song.

It's not a Suno contest, it's a potential work for hire situation.

They want to see what their track sounds like when others using Suno mess about with it, it get a range of what Suno can do with a track like theirs.

While I appreciate their desire to be fair and compensate anyone whose output based on their original track is liked enough to go on an album or somesuch, there's a huge potential for abuse here and unfortunately no good way to establish a co-authorship or partnership in Suno yet to ensure everyone feels happy and is credited for whatever they bring to the table.

[neat nexus]

Really this sounds more like a legal and licensing question than anything else, I'd recommend you send an email to [e-mail] regarding it, just please understand there is a bit of a backlog.

[hard beacon]

I didn't expect it to be so complicated... but it's okay... I just thought it would be interesting for beginners or AI people to get an official release... but the topic is over and I'll just ask established artists 🙂

[restive pebble]

Fascinating conversation on the subject.
I enjoyed seeing the discussion on it and learned some too. Thanks.

The idea is sound as it would be great for unknowns to get a little real world exposure but the legality of keeping track of who's it who's etc is beyond Suno capabilities.

I wonder if there'd be any interest in just being a part of something like an album release or something. Maybe people would be open to it without financial compensation and be willing to get a name on something.

It's not something many driven upcomers may be interested in but there must be some willing to share their efforts for a call out

[pure panther]

Another absolute banger combo of genres
R&B ragtime dark EDM funk

[polar spade]

Before I get too worked up over this. Anyone else care for having this psychedelic epilepsy-inducing wave of colour tinting that keeps showing up when making a Cover Art video and ruining a decent video.
My AI artists duo said they dont like it either. 😅

[hollow hare]

One more thing to add, as an FYI:
Suno is aware how messy these type of situations can get which is why, in their TOS, essentially revoked the commercial license for remixed tracks.
(Really just a CYA clause where if multiple parties are disputing ownership and gets the company involved, they'll be like "none of ya can sell that anyways, peace")
@hard beacon @neat nexus

[pure panther]

It could do without it.

[polar spade]

Yay! One vote against. Thanks.

[plain pumice]

Anyone having problems with extracting Stems in Studio?
..suddenly I do.

[pure panther]

I've had them never process on the page where you can download them all and had to go into the library and download 24 separate files. Not sure what your issue is, but if it's that, then yes 🙂

[plain pumice]

Usually, through the interface when selecting the song/clip in the Studio, stems could be extracted and imported directly in the Studio.
Seems they've changed something, because while the Studio is open, you can still extract stems from song from the workspace, but not the song that you're currently working with.

[pure panther]

Huh. I've never encountered that.

Might be worth reporting

[jovial scarab]

The new persona is a disaster. I cannot get the original vocals at all

[spark cargo]

Hi there! Can anyone confirm whether or not Suno is capable of understanding Advanced audio mastering instructions? More specifically, meta-tags setting parallel multi-band compression parameters for driving the tonal character of the audio output. As an example:

[GlobalFX: Parallel_MB_LinearPhase | Process: Parallel Compression (Multiband) | Mode: Linear Phase | Filters: 24dB/oct | Global Wet: 30%]

Are these tags processed or largely ignored by the LLM? Please feel free to refer me to detailed technical docs, or perhaps someone might know of the correct place to ask this question.

I hear crickets.....

[polar spade]

No.

No one can confirm it. Whether it's true or not. No one can confirm it.

So all you have is to run several runs of different parameters with the minimal of prompts, and see if any are followed precisely.

[spark cargo]

gotcha' , thanks, I'll look elsewhere

[cyan light]

If it actually works or not I don't know but you can try adding " Multiband linear-phase parallel compression subtly glues the mix, preserving lo-fi textures while adding clarity, applied globally at 30% wet through steep 24dB/oct filters for tight sonic punch." somewhere in your Styles box and hit the magic wand and try?

[jovial scarab]

Personas are not working correctly. Sometimes it only captures the voce 30 seconds and the rest of the song is totally out of tune

Please can you fix personas Cannot maintain the same voice

[half dragon]

If you feel it's a bug - then report it in #🪲┃bugs-web channel - make sure you share the data to support - maybe the links for the songs or/and to the persona or somg it was created from... so when dev could look at them they'd have something to look at - devs monitor bug channels much closely than this channel.
Of if you want to, you can always send feedback via the official feedback link.

[jovial scarab]

Thank you I will do

[raven parrot]

Hi everyone 👋

I’m running into a persistent genre-injection issue.

No matter how I write my prompts, Suno keeps inserting FUNK into the generation, even when:

• funk is not mentioned
• guitar is explicitly excluded
• the prompt is rooted in non-Western instruments (saz, gourd violin)
• the mood is / minimal, not funk

This happens across multiple prompts and sessions.

It feels like “dance / playful / groove” is being auto-mapped to FUNK internally.

Is there a way to:

1. fully disable auto-genre injection?
2. hard-exclude funk as a latent genre?
3. or force non-Western rhythmic interpretations?

Thanks!

[plain pumice]

Hi Matrak.

When you say your prompt, would this be the custom prompt or style prompt?
What model version are you using?

[raven parrot]

Hi! Thanks for checking 🙏
This is happening in the custom prompt (not style prompt).
I’m using the latest default model available in the app.

[plain pumice]

Since you're using v.5, it's important to note, that most of the style influence comes from what is in the Style Prompt.
To better help you, what does your style prompt look like?

[raven parrot]

My style prompt is very minimal and does not include funk or funk-adjacent terms.
The issue is that funk appears even when the style prompt is neutral or empty, based on the custom prompt content alone.
Happy to paste my exact style prompt if that helps.

[plain pumice]

Lyrics can also influence.
Have you tried using the Exclude prompt field?

[raven parrot]

Yes
I’ve tried the Exclude prompt field (including excluding funk explicitly), but funk still appears intermittently.
That’s why I suspected some automatic genre inference beyond the explicit prompts.
Happy to test again if you have a recommended exclude syntax.

[plain pumice]

Is it possible that you're using some styles / descriptive tags that somehow could be associated with Funk? Even while it's more remote.
As I mentioned, even lyrics can influence this.
I'm not sure if this would help, but perhaps restarting the browser or trying a different browser - this may seem odd, but there may going something on at session level, which also gets parsed when you click to generate.

[raven parrot]

It is possible, of course, that certain words or descriptions carry familiar associations. I do not dismiss that.
Yet in a wellordered method, such associations should yield when an exclusion is declared.

If lyrics or descriptors outweigh an explicit “do not,” then the system is reasoning by habit, not by rule. In my view, exclusion should be evaluated after all inference, not before it is overridden by memory.

As for sessions or browsers, I will gladly test that as well. But if a fresh session produces the same result, it would suggest the behavior lies not in the state of the session, but in the order by which rules are applied.

Thanks for taking the time to think this through and for the suggestions
I really appreciate it.
I’ll try testing with a fresh session / different browser and report back if I notice anything useful.
Grateful for the help 🙏

[plain pumice]

Your exact prompt might actually help, perhaps I/we can spot something you have missed or are unware of.

You could also try, if you have a persona that's far away from funk, then use it with high audio influence in the advanced sliders.
You could also try to generate a song with only the lyrics, advanced sliders at default, use no persona - see what you get.
Even further, to rule out that it's something in your custom prompt - run it without any musical and production cues.
.. it's basically to rule out if the custom prompt or style prompt somehow could be the cause of it.

Also to mention, the exclude field is more of a strong discouragement than a filter that actually blocks it. In fact, it can actually also sometimes encourage the listed tags. I'm not sure how it actually behaves if you were to write "avoid" or "no" XYZ. As writing avoid X would actually logically encourage the tag rather than exclude it.

[spark cargo]

I have run into a similar issue which I believe was related to the persistence layer. I was able to move beyond it by placing the following at the top of the lyrics box. I was not using the app, this issue was experienced in the browser.

[Reset Memory] [Clear Memory]

[raven parrot]

That did it!
funk is gone now. 🙏
Adding [Reset Memory] [Clear Memory] at the top of the lyrics solved the issue for me.
Really appreciate you sharing this, it was genuinely helpful. Thanks again!

[spark cargo]

no problem. you are welcome. glad you got it sorted out.

[wise kindle]

Wait [Reset Memory] [Clear Memory] actually works? Hmm lemme try something... [Add Credits: 1000]. Nah that's cool though.. I thought I was going crazy when I noticed it sounded like my generations were getting polluted by previous ones.

[spark cargo]

It happens from time to time. You just gotta do a little house cleaning of the memory persistence layer.
There are other ways to break out of that loop. However, those two commands are the easiest. It's kind of a shotgun approach, but it works.

[wise kindle]

Good to know, thanks for the tip! (even though it wasn't directed at me initially)

[minor storm]

Hi, I have produced an albums worth of material of original songs (uploaded then covered) I am happy with the backing tracks and don't want it changed but I want to use the vocal persona feature to achive a consistant vocalist across all these songs, how do I do this without regenarating the entire track?, I tried generating a a new vocal track in Studio but it seems you cant apply the persona to a specific track but only to an entire song, any ideas?

[spark cargo]

sure no problem. I posted it for the benefit of everyone.

[spark cargo]

have you tried using the cover feature? You can cover an entire song, adjust the sliders for maximum adherence to the original version while using a persona to apply the new vocal track to it. In theory it should work I haven't messed with the new persona feature much.

[quiet solar]

It's unlikely you'll get a diferent answer elsewhere. Not a truthful one, at any rate.

If such technical docs exist, they only do so in the hands of the Suno development team and aren't shared beyond that. Everything we know about here that gets shared about what Suno can/can't/will/might respond to is based on varied personal experience amongst users who themselves have varying levels of experience.

Realistically, your only route to getting an answer would be to get one straight from a developer.

[spark cargo]

thanks for your reply. I have seen reference to advanced audio meta-tags and I realize they're not widely published but I know they exist. I'm just trying to locate them again I do know a few users who bounce around discord reddit etc.. Who are knowledgeable in this area and I have sent them messages. I just haven't asked here before until this evening.

[minor storm]

@spark cargo yea that was my fisrst instict, but the cover feature isn't realiable enough to truly preserve the the source, these are tracks I've spent alot of time getting right, the specific parts are as I want them and while the cover version with 100% on the audio upload slider will keep the syle very similar it still introduces to many changes, I am really looking for a way of only changin the vocals

[spark cargo]

I've had pretty decent luck with keeping covers pretty much exact using the cover feature, but I know what you're getting at. I think maybe the only thing you can do at this point would be to generate a cover with the vocals you want sounding the way you want, and then extract the vocal stems and put that together with the original version in a DAW.

[quiet solar]

^ Basically, this.

You can do the combining in Suno Studio as well, if you have it. But you'll need to generate new vocals for each track individually outside of Studio via Cover or Extend and then stitch them in as replacements in Studio or your DAW of preference.

[wise kindle]

huh, I thought using [Chorus] would repeat the chorus (assuming you've supplied it earlier), but apparently not

[spark cargo]

yeah, you still have to input the lyric text under the chorus tag every time.

[wise kindle]

the generated lyrics in suno just write out the chorus once, then use [Chorus] alone after that

must be somewhat new of a feature. last I remember using that it just re-wrote the entire chorus out each time

[spark cargo]

that's news to me. I have been writing my own lyrics since version three

[wise kindle]

yeah I haven't used it in a while. I just wanted some placeholder text to try out some style prompts so figured it'd be worth a shot

it does look like it received an upgrade as it throws in stuff like [whispered vocals] Hollow and [low vocal register] Eyes like eclipses opening, which also doesn't work lol

the low vocal register kind of worked in the second generation, but on a completely different line

[mellow condor]

Does anyone have any ideas for a hardstyle song? There should be longer passages in which mainly bass variations should run, but instead the rather short mellow (which should actually occur in breaks/hooks) is repeated over and over again and only the bass changes... What prompts are there to tell Suno “just bass for now, no mellow”?

[cyan lantern]

I've got a question... I've tried in the style prompt to reduce the BPM in a song at ~120 BPM to ~100 BPM. Style Influence is set to 65%. It always keeps the OG BPM. Any tips on getting it to reduce this?

[polar spade]

[edited to focus on the re-opened discussion]

In my opinion, Suno uses diffusion or autoregressive elements for a waveform generation. In training, it has had a massive amount of data associated with specific audio, so terms like " crisp audio", "radio ready polish" will work well because other similar sounding audios/genres are grouped together to associate with those terms which label those very audios in order that they relate to each other.

Suno doesnt run a virtual mixer under the hood. From what I read as times goes on are two things: 1) No one consistently shows how well some surgical DAW parameters (including specific brand named preset) are working -- else, as things go, the news would spread like fire and everyone would be doing it, right. . . and 2) quite many others , especially on reddit say its just making things worse. But neither can really confirm or deny it.

But the broader terms are often noted to have some effectiveness in certain genres. And logically, regarding autoregressive training, that makes perfect sense to me that they would. Nothing is forced, its heavily predicted.

[scenic rampart]

hey team, is it possible to specify the extend duration, one way or another (with bars, or with seconds...? )

[spark cargo]

actually, no, I clearly stated my reason for asking a few posts after the one you're responding to. You might want to catch up
before continuing.

[frigid orbit]

Not really but yes. But it isn’t in the way you think it works. I assume you ask it in specific due to wanting a short extend?

Look at the wav form and find a spot where it looks like the sound is dying down. If you extend within that it’s far more likely to create an end to whatever you generated. VS for instance a plateau or rise in audio.

‘
You used to do this by ear, but with the added visualisation of the wave, it should be 10 times easier.

Do this in a rise of audio, and it’ll likely rise and go on.

Ahh cool, I leaned into the use of extend, you into prompt.

Together it should work

[visual tulip]

I tried searching the feed, did not see an answer. Os there a way to prompt for beats to be playing only for a while so you can sync up with another like you do when you beat sync house records. And how can you get rid of all kyeboards and synths sounds and have just beats.

[half dragon]

Not sure about prompting but for 2nd part, you can always extract the stems and and remove the unwanted instruments or sounds.

[hard beacon]

only 1 hour and 9 minutes left till XMAS 🙂 whoop whopp

[neat nexus]

Uh..time zone differences?

[hard beacon]

Germany 😉

22:53 (10:53pm)

[neat nexus]

That explains it.

[half dragon]

1 day and ~8 hours more here 🙂

[neat nexus]

yeah. I was about to ask for winning lottery tickets or a trip in the time machine...

[placid nebula]

Germany opens presents on Christmas Eve… mind blown with that one back when I lived there

[hard beacon]

😉

[neat nexus]

We did a kind of half-and half when I was a kid.

We did our gift exchange to each other on Christmas Eve, then on Christmas morning we'd get the ones "from Santa" to open.

[hard beacon]

@placid nebula : where do you live now?

[placid nebula]

North Carolina in the US

[hard beacon]

xmas with 21°C ??? 🙂

or in Fahrenheit: 69,8 🙂

[placid nebula]

Ha, visiting Michigan right now… so ~5C

[half dragon]

That's what TX is this year.

[hard beacon]

uhh... germany (my location) is 1°C at the moment

[placid nebula]

Nothing gluhwein can’t fix

[hard beacon]

😄

[sleek stump]

Waiting for nuke..

[hard beacon]

😄

[sleek stump]

and boom 😄

[hard beacon]

https://tenor.com/view/incident-gif-2747499635347300991

[neat nexus]

Hey guys no big deal but can we move the offtopic to the #🗣️┃chill-chat channel? This is for prompting and discussion of how to make Suno do what you want it to do.

[hard beacon]

i don't wanna move... so... then let's talk about prompting 😉 in the past the best way to "fix" some missspelling in the vocals was the editor..what's the best method now?

[sleek stump]

Cover

I have 100% easy lyric fixes by doing a W1% A99% cover

[hard beacon]

but... imagine the output is perfect and you don't want to have any changes except of the vocals.. (or one vocal line) 🙂

let me check 🙂

[sleek stump]

I usually get a 99-100% cover within 3-4 gens

Editor has gotten.. well, expensive and unreliable

[half dragon]

I use Studio nowaday - usually works well with one word - a line isn't well reliable then usualy Cover in that case.
Then bring it back in Studio to get that vocal line, assuming I must have the original song. Or go with the Cover version if I like it better.

[hard beacon]

narf.. nope.. still the same issue.. let me try some different slider settings 🙂

[half dragon]

sometimes.. I also use Extend... just to get that section... then bring that section into Studio.

[sleek stump]

Studio is unobtanium, for a lot of people. I cant justify the cost for premier, especially when I already have a lot invested in DAW and Plugins.

[hard beacon]

Studio is interesting to add some new tracks... i think some v5 outputs are to thin... so i add layers like "guitars" oder "pads" 🙂

[half dragon]

a lot of things are not for everyone... different ppl different justifications 🙂

These are just tools, many ways to fix it.

[hard beacon]

Have you also noticed that outputs with vocal personas are quieter and the bass range is a bit thin?

or is it just me??

[half dragon]

I don't use a lot of Persona so haven't noticed. I like different vocals and differrent things on my songs.

[hard beacon]

you are 100% right.. but if you try to create an album of an "AI ARTIST" the funtion is good 😉

for EDM is doesn't matter.. but if you want to create, for example, different tracks for a AI Rap-Project the same voice is very useful 🙂

[half dragon]

Agreed.

[quick wedge]

What prompt do I use that sounds like a K-pop idol?
Any suggestions?

[hard beacon]

did you ask ChatGPT (with Suno gpt) ???

copy paste incoming

Style / Prompt:

K-Pop idol pop song with a clean female lead vocal, bright and expressive, Korean pop style pronunciation, polished and youthful tone. Catchy melodic verses, emotional pre-chorus, and a big glossy chorus with layered harmonies. Modern K-Pop production with tight drums, shimmering synths, subtle guitar accents, and clean bass. Studio-perfect vocals, forward in the mix, smooth highs, no distortion. Energetic but elegant, radio-ready idol performance, 2:50–3:05 runtime.

Exclude Style:

choir, chant, monk, drone, gritty rap vocals, lo-fi, distortion, over-autotune, harsh highs, vocal fry

Suggested Sliders (important):

Weirdness: 10–18%

Style Influence: 85–90%

Audio Influence: 0–10%

Optional add-ons (pick ONE if needed):

“soft Korean-accented English delivery”

“bubblegum K-Pop vibe”

“dark K-Pop concept, BLACKPINK-inspired energy”

“romantic K-Pop ballad chorus with power notes”

[half dragon]

or gemini 🙂

[hard beacon]

😉

ha !!! Fixed the vocal 😉 Thanks Ginger 🙂

now some remasters to get a great output to master 😉

[quick wedge]

I haven’t checked ChatGPT yet but I’ll test out the ones you showed me plus VRKing too thank you ✌️

[pure panther]

[hard beacon]

what is your experience with remaster to get the best sounding output?? subtle - normal - high

[pure panther]

Subtle all the way

[hard beacon]

ok 🙂

[pure panther]

The rest seem to dull the entire song considerably

[hard beacon]

I kind of feel like the remaster function is an optimized cover function 🙂 2 different buttons - same result...

addition: with normal and high

[pure panther]

It's an excellent tool for smoothing things out after you've Frankenstein edited a song both in and out of Suno so much that is feels stitched together.

[half dragon]

Sometimes when I removed a sound or an instrument that I don't want from the stem... it often left the "hole" in the sound... run it through Remaster... filling it full again.

[pure panther]

I usually can't stand backing vocals and remove them.

I noticed it seems to give the main vocals a "hiccup"

[half dragon]

backing vocals has its place in some song... definitely not on every song.

[pure panther]

I agree, Suno on the other hand... doesn't 😂

[hard beacon]

To all the germans in here: MERRY X-MAS 🙂

or europeans.. don't know all the time zones 😉

[pure panther]

My Mom is like 99.98% German according to 23 and me. It was a little shocking considering she's at least 4th (probably more) gen U.S.A.

[hard beacon]

well.. there are a lot of germans in the US 😉

Wisconsin is the biggest german community (in realtion to immigrants) 🙂

[pure panther]

She was from Pennsyltucky Pennsylvania

We need per inspiration song sliders, that would be legit.

Like one is 20%, but the other is 40% strength

Like this:

I'm going to suggest it 😄

[unborn grove]

Does anyone know how I could use the high tenor vocals in alternative rock like Sam Fender? It doesn't seem to understand what it means

[unborn grove]

Woodhouse fr rn? lol

[half dragon]

whooops

[pure panther]

Aww, I was reading that you big meanie 😛 (talking to the Ai)

[half dragon]

To get Suno to produce that "Sam Fender" sound—a powerful, belting high tenor with grit and emotional urgency—you need to move beyond simple genre tags and use specific technical vocal descriptors and structural meta-tags.

Here is how to structure your prompt to capture that high tenor alternative rock style:

1. The "Style of Music" Box
Instead of just "Rock" or "Indie," use a combination of tone and range tags. Suno responds well to specific vocal classifications. Try this:   

Style Prompt: alternative rock, indie rock, male high tenor, belting vocals, raw emotional delivery, anthemic, gritty tenor, 120bpm, driving guitars, bright production

2. Using Meta-Tags in Your Lyrics
To force the AI to reach those high notes at the right time, embed bracketed instructions directly into your lyrical structure.

For Verses: Use [Verse | vulnerable vocals | light rasp] to start with that intimate Fender-style storytelling.

For the "Big" Moments: Don't just label it "Chorus." Use [Chorus | anthemic high tenor | belting vocals | powerful mix] to trigger a higher vocal register.

For Ad-libs: Use [high tenor wail] or [vocal run] in brackets after a line to get those soaring melodic flourishes.

3. Key Pro-Tips for the "Fender" Sound
Vocal Texture: Use the tags [raspy lead vocal] or [vocal grit] to avoid a clean "boy band" sound and get that stadium-rock edge.

Range Control: If Suno keeps giving you a baritone, explicitly use the tag [high-pitched male vocals] or [soaring tenor] at the beginning of the lyrics block.

Instrumental Contrast: Sam Fender's vocals shine because they sit above jangly guitars and heavy drums. Adding [fender stratocaster tones | driving indie beat] to your style box helps set the right stage for a tenor voice.

4. Example Prompt Template
Style: Alternative Rock, Indie, Anthemic, High Tenor Male, Emotional Belting, Raspy, Driving Guitars, 125 BPM.

Lyrics: [Intro: Distorted guitar and driving drums] [Verse 1 | vulnerable vocals] I was seventeen years old... [Chorus | anthemic high tenor | belting vocals] OH, IT'S A LONG WAY DOWN!

[rough river]

Hey everyone. Question: has anyone experimented with Polyphony or Homophony?

[half dragon]

Straight from Gemini - some suggestions to try 🙂

[autumn delta]

Is there a way to force a drum solo? preferably with the other instruments stopped

simply putting [Drum Solo] doesnt seem to do jack

[half dragon]

There isn't a reliable way to do it - it mostly depends on genres that it was trained on too.

[autumn delta]

ahh gotcha, I shall put in a feature request over in the beta group 🙂

[autumn delta]

Request submitted lol

[unborn grove]

It's decent but still needs to be a little less 'yell' more proper singing lol

[shell loom]

slap cello, bass cello tested a fully working

[plain pumice]

Any one who has experience with using the Suno Studio to master, remix and adding own vocals and such, where it's the remixing feature that add most of that?

[versed talon]

I did not like it as much to be fair

[quiet solar]

If you want the instrumentals left alone completely, the approach I typically take is based on Extend rather than Cover:

1. Extend from somewhere in the song after the line you're looking to fix.
2. Add in the section you're looking to fix, with the fixed lyrics, into the lyrics box. Set sliders for weirdness and style influence to low and audio influence mid to high (I find the sweet spot is usually somewhere in the 50% - 85% range). Optionally, use a vocal persona on the extension to try to minimise changes in the voice (not tested extensively on the new persona).
3. Repeat the section in the Extension with your vocal fix in place until you get something that sounds to your liking.
4. Extract the original instrumental and vocal stems, and the vocal stem from the fix.
5. Use Suno Studio or a DAW/audio editor and replace the relevant part of the original vocal stem with the fix.

Because there will typically be slight changes to cadence and other minor changes between generations, sometimes it's easier to replace a line or a section rather than just the offending word in the original.

The rationale for using Extend rather than Cover is that Extend is intended by design to keep a high level of consistency with the original, whereas Cover is intended to deliberately make creative changes. So you're less likely to have issues with key changes, tempo changes or other oddities using Extend.

[frozen maple]

Hello there folks? How is your generated music going? I'm writing here (don't know if it's the right channel..) because lately I'm creating songs with a Persona I created but the generated songs keeps being ultra familiar with the original osng that I used for create that Persona..

[frozen maple]

is there anybody here that could help me? I'd like to keep just the vocal timbre of that Persona

please and thank you in advance!

[proud dagger]

Anyone here in support speaks hindi or can understand? I have one word in my entire song which is coming out weird. Need help to understand how to fix it

I tried 4 covers with different things

[somber anchor]

Came here to check, if anyone has the same problem since the newest update. I don't like that

[crimson pine]

same, I just turn it back to legacy mode, or make the audio influence slider super low

[keen iris]

Why Persona's make midi tracks after 2min?

[frigid orbit]

To add to this. Take note of the used terms, those should be the takeaway. Formats and all that are malleable/subjective.

[frigid orbit]

Can you elaborate?

[frigid orbit]

Can you give us the word regardless in its englisised manner? As in using the western alphabet

When in doubt, try spelling it out how you’d pronounce it.

Anglo-Saxon - Angle’looh-Sahkson for instance

[keen iris]

It's starts very good (v5 quality) but after 1min it became worse, and after 2min it sounds like .midi music files, like some 8bit track but with vocals

[frigid orbit]

Probably has to do with your own input. But you could extend just before it degrades.

Let’s say it starts 01:02, so you extend from 00:58

[proud dagger]

It's how it's breaking and pronouncing it for just one line. I can sing in same tone but AI for some reason choose not to for that word

[frigid orbit]

Because it’s AI.

I gave you a solution

[proud dagger]

Never happened 1 week before for same word

[frigid orbit]

Cool but in this instance it does

Using AI is like commissioning an artist, but every time you do it, it’s a slightly different artist.

[keen iris]

The other day all tracks was normal, it's starts right after this new Persona voice thing went out.
Extend drops it quality to midi around 2:10-2:30 too 🥺

[summer plinth]

I usually create prompts "as if I were chatting with a friend" 🙂 Do you have any extra ideas?

[viral pond]

has there been any degradation issues lately ?

had some longer generations that has sounded horrible, especially together with cover and personas

I havent changed anything specific, but suddenly longer songs sounded very distorted and vocals uncelear

[summer plinth]

In version v4.5, I was experiencing this type of issue with some songs. Because of this, I decided to rework the Lyrics section. Even if I explain what the song will do in the Styles section, I also explain it in the Lyrics section. After the song ends, I add parts like [Short End] or [Musical End]. In more standard terms, I also add the (fade) section.

Sometimes this was more complicated, but I haven't experienced this problem in v5. The reason I haven't is because I want to maintain control 🙂

[proud dagger]

I can understand the AI working on it dude. I work on machine learning and AI as well. I found a way to keep everything same but I think it's a language which makes phonetic different. Anyways I will try to find someone who is making same language music.

[frigid orbit]

Welcome to Suno where every update downgrades the rest. All I can do is come up with a boot like solution

[frigid orbit]

No, not at all. There are certain words that seem to be difficult, sometimes it’s genre dependent. Sometimes it’s just luck of the draw or just very long words. It has issues with or due to rhymes earlier that messes up. There’s no science behind it. All I can do is offer a solution how it can properly say the word.

[proud dagger]

I don't agree from technical point of view but I don't want to argue. Models can be tweak based on feedback. I will submit the feedback to devs.

[frigid orbit]

That’s fine, mispronunciations within a non-English language just isn’t rare. Hell even English has issues.

Sounds like you ain’t a power/long user though, and feedback to the devs is always great.

[snow scroll]

errrm .. So I created a persona using the "New" method. Then plugged it into a previous song so I could have my persona singing it ... and it changed the majority of the melody lines ?? So the song simply isnt the same at all . Am I misssing something ?

[half dragon]

How did you "plug it" into previous song? by Cover and add Persona? Did you keep the audio reference slider high enough?

[snow scroll]

Yes ... Made audio ref 25 at first then 85 ... still changed melody though

I even used my original download from my DAW ... Its better at keeping melody , but still not the same#

Ha .. Take out the persona , use the same original from the DAW and the melody is followed perfectly .

The persona screws up the melody :(((

[half dragon]

Did you create your Persona from a full song, a vocal stem on a section? I heard that that might be a factor too... some suggested that for just vocal, create the Persona from a best section of the vocal stem

[snow scroll]

As per the video instruction, I picked the "best" section of a vocal stem

I'll keep playing around with ideas ... but it aint working for me atm

[half dragon]

Hopefully someone with more Persona experience would chime in.

[snow scroll]

Cheers bud

This is my first dip into it to be honest

it does a better job of copying voice type without persona atm, but its still not "right"

[half dragon]

Did you try to create the legacy Persona, instead of a new? see if it has the same issue

[snow scroll]

yeah, it didnt change the melody, so better in that respect, but I felt hadnt nailed the voice quite, so wanted to try the new way ....

[half dragon]

It might be worth to report it to devs.

[snow scroll]

Yeah maybe will if my fiddling doesnt find the issue ... cheers

[verbal flame]

Is there a list somewhere of the commands for orchestrating a song?

[white pecan]

Hey Dev chat 🙂 Just wanted to wish you all a merry christmas 🎄 Peace ❤️

[proud dagger]

Well I don't want to argue but keep getting notifications of your reply. One don't have to power user to understand tech nuances. When you see something weird, you call it out. As a customer I don't just close my eye and imagine the sun is sometimes purple. It's upto devs to consider feedback or not. In my intial request I asked for user who is building content in same language. I didn't seek help from anyone else who don't work on same language.

[half dragon]

There isn't anything official. You'd have to google for many users compiled list.

[quiet solar]

In my experience, Personas have always been a little bit odd when applied retrospectively to a song as opposed to using them to create something new, but a vocal persona (in principle) shouldn't have any impact on the melody. If that's repeatable, that's unfortunate...although worth knowing to warn others.

[snow scroll]

I can’t get round the problem atm, but will keep trying . I’ve had to pop out ( Xmas and all that 🙄) but will revert to legacy persona to double check it’s only the new version of persona that is giving me the issue. I’m not keen to try a 3rd party to create a persona and would prefer to stick with Suno. I’m sure it will iron out 😂

The legacy version though, in my view is not truly representative of the original either . 🫣

[polar spade]

Remember, personas are influenced by Audio slider according to devs, but of course, so is the Audio, so make sure your original Audio prompts (and not the persona prompts) are used to give more weight to the Audio.

I havent checked in awhile but, I think style slider about 75-78% and Audio closer to 55% -- could be as low as 20% now if its just a voice persona.

I cant vouch yet for the newer "vocal only 5.0 persona" maker which may be tweaked differently beyond just used on vocals only. We used vocals only as early as last year, but the model was 4.0 ( or maybe even 3.5) ive forgotten. Ive heard we would need to re-calibrate the sliders for the new persona version. So keep that in mind.

[versed talon]

Master your songs before you publish please

[frigid orbit]

You named an issue, I gave a solution. Anywho goodluck with your adventure

[dusty tendon]

Yeah

[crimson pine]

nah I don't know how

[polar spade]

You've found the Hindi channel then. So that should help. 🙂

[sudden oracle]

/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

[cunning mulch]

how to prompt song

[glacial cobalt]

🤣 i thought i was the only one 🤣 "

[glacial cobalt]

"Dunno how many feedbacks I’ve sent about the glitches and cutouts the past 3 months... and this is what they come up with instead of fixing it? 🤦‍♂️ Good luck ,

[half dragon]

#1294418290348921007

[snow scroll]

Thanks for that, and Happy Christmas . I have been messing with the sliders to see how they affect things. But at the end of the day, if I use the same song and the same settings … use legacy and the overall melody remains intact. Use the new version and it falls apart in major places. Whether the voice clone is good remains debatable. To be honest I use my own recording from my daw with my voice as upload, and if I put no style command, just let sumo recreate, it < Sometimes> does a pretty good attempt at capturing my voice 🤔

[frigid orbit]

Not everything is necessarily a glitch

[hollow hare]

I wonder if their weirdness/style scales are scaled to the extremes. Can't do much without a link

[visual tulip]

https://howtopromptsuno.com/

[crimson thorn]

How do you make a section of the song quiet then abruptly explode?

[frigid orbit]

I like how “it can’t do x” are bugs/glitches now

[fossil moon]

hi hi, anyone knows how to make an string quartet style =? ( string quarted dont work )

[polar spade]

@astral zephyr
From your note in the bug channel:

Just to compare: i ran a couple. I put shamisen and got what sounds similar to a shamisen and bamboo flute. Put koto, got a koto. Put zither, got zither.

So they seems at minimum to sound similar and clarity seems good.

https://suno.com/s/Dl44cFBx74JkoUwG

[viral pond]

heya : 🙂

what is the proper way to use exclude lyrics ? is it narrated the same way as the normal musical style prompt ?

or is it just keywords, any examples ?

I cannot figure out if it is narrated or if its just purely keywords commaseperated 😄

[quick wedge]

There should be an exclude style box on the create page some where
Just add prompts that you don’t wanna hear in your song

I use commas to seperate my prompt words

Studio production, melodic dubstep, clear airy female vocals, subtle psytrance, etc is what I’d use
Some people uses sentences in their style box though

[viral pond]

yeah i used comma seperatead , but was curious if there was other ways to use it as well, like hitting the magic button on the style prompt. I like the narrated way of prompting

[astral zephyr]

thank you, it's weird because i didn't change any of my prompt and can't retrieve the same sound as my last 20 others old sounds... hm...

[quick wedge]

I’m not familiar with the narrative way some how it makes me think of the [spoken word] prompt but that’s probably different from what you’re looking for

[polar spade]

Yea. We never know I guess. It doesn't prove anything if mine sound OK and yours doesnt sound as prompted. But just to compare, it seems it can do it. Im running Premium, if thats anything.

Just at least refresh the browser to make sure the create panel is starting on a fresh slate and hoping it doesnt have any lingering files to affect prompting.

[viral pond]

yeah just wonder if the exclude styles work the same way

[quick wedge]

I dunno maybe someone else knows about the narrative thing with the exclude box

[viral pond]

yeah anyone who has any good tips to get the best out of exclude, shout out with a tag so I don't miss it 🙂

[quick wedge]

You can try asking Alikan maybe he might know. I would suggest asking in Gen chat but they haven’t talked about much in there 🤷‍♀️

[viral pond]

aha, tried to ask llms and search reddits but it seems like this is a feature not that many uses.. its sad cause i think it can be very powerfull

[polar spade]

You seemed to merge from talking about using a persona and keeping the melody of the original, to covering an upload and keeping the melody of the upload.

The latter means Audio slider can be near 100% and the others at default. The former can means Audio slider may need to be much lower.

[quick wedge]

Merry Christmas 🎄💫 Have a good day ✌️

[viral pond]

next year will be awesome i think

[quick wedge]

Merry Christmas Dasvon 💫🎄

[polar spade]

Merry Xmas , Pim. I didnt win the 1.8 billion lottery. 😭

[quick wedge]

💰🧧 here you go plus lots of vibes ✨✨✨
I didn’t receive any Xmas presents either but it’s fine

[viral pond]

funny how many generated songs i made that i never even listened to

[polar spade]

@small prairie Beyond a few exceptions of clicking on a promo granting ownership of a speficial Basic output, as Suno owns those Basic outputs, you would email Suno directly to discuss any transfer, permissions or license.

[e-mail]

[lone owl]

Halo?

[frigid orbit]

Thus far I would suggest promoting without extend. And once you got a feel for output, just list out what you don’t want. Keep in mind that despite it stating exclude, it still pulls attention that way

[viscid ravine]

dutch? 🇳🇱

[visual plank]

I'm sure you are right, but what happens if you don't do this? what's the difference?

[final elbow]

Hi!

[sonic folio]

Does anyone know if you could define a singer in style prompt with something like "high fry screams [M] and then in lyrics suno will follow the style of [M]?

[sharp jolt]

Do anyone know a single source for Dino meta tags with descriptions and examples?

[abstract flower]

You certainly could. Why ask? Just try it. V5 is much more responsive. If i were you, i'd state smth like it in the prompt, e,g, "vocalist Mike: male, high fry screams, etc.", a whole sentence (IME, sentences work better, but — it seems to me). then use [vocalist Mike] in the lyrics prompt. Try a few things like that, you'll probably find one that works.

[quiet solar]

To an extent, yes. Some people have been able to name/define singers and have Suno reference the description from the style box.

Personally I prefer cutting out the middle man. If you state multiple singers (eg Male tenor, [add any additional descriptors here] and female alto [add additional descriptors here] then you can call the male tenor / female alto voices in lyrics prompts without needing to name them “Joe” or “Sandra” or whatever.

Note that any prompts in the lyrics field will be treated as suggestions, though, so it’s not a reliable way to switch singers on demand. But it will switch between the correct two singers when it does.

[abstract flower]

Is "Dino" an autocowreck for "Suno"? Then sorry, but there is no list. You can write anything you want between the [] brackets, and the model may interpret it. or ignore. nobody knows, the tags aren't fixed by the model design, but interpreted like text input. You'll find a ton of these lists on the 'Net, but they don't work, at least not as you'd expect.

[frigid orbit]

Nah he wants to play the game

[dusty tendon]

Same here

[dusty tendon]

Hi

[thick temple]

Why is the 'Exclude' tag being ignored for Solo Vocal tracks today?"

[polar spade]

What's another way of describing what youre trying to accomplish and why do you think the exclude window will help?

I dont normally get solo vocal tracks on initial generations in Suno.

[sonic folio]

Is there a specific reason why suno adds clean vocals even if I specifically exclude it? Because I want harsh/screaming vocals through the whole song and I can't figure out how to make suno do that

[quiet solar]

Exclusions have always been hit and miss in my experience. Suno is much better at interpreting what you tell it you do want rather than what you don’t.

I’m assuming you’re already prompting for that in the style box for the vocalist, and reinforcing it in lyrics prompts. Are there sub-genres you could prompt for as well where screaming through the whole song is more common?

Other than that, you could try putting all the lyrics in block capitals - that sometimes helps Suno to recognise shouted or screamed lyrics, although I’ve only tried that in the context of it being different to standard lyrics, not across the whole set of lyrics.

[sonic folio]

I do...this is my style prompt

"Symphonic Blackened Deathcore. Male high-pitched fry screams, shrill distorted vocals, piercing fry technique, upper-register screams, strained high screams, searing vocal attack. Drop A tuning. High energy, cinematic orchestration, operatic backing choir, rapid blast beats, technical blackened guitar riffs. Crushing breakdowns, guttural tunnel throats, soaring high shrieks. Studio-grade production. High-resolution audio. Mix is crystal clear, transparent, and ultra-wide. Arrangement is strictly driven by the provided lyric markers [Chorus], [Verse]. Instrumentation possesses distinct, intentional textures. Zero noise floor. Professional master.

[quiet solar]

Out of curiosity, try cutting everything in the prompt after “soaring high shrieks” and see where that gets you. Longer prompts have a tendency to have fewer instructions picked up and (I suspect) priority shared across them more evenly.

A shorter one focused on what you categorically want can then be a starting point for a cover where you can steer it on other factors like tuning, production mix and so forth.

[lone owl]

Hello guys, in here i got problem making music using suno, how to make sure there's no saxophone in my audio? That's wierd i just try making 80s instrument like (Man in the mirror by michael jackson) but why suno always giving me lot of overused saxophone in instrument? Hope only drum and classic synthesizer

[wise kindle]

I don't know why anybody wouldn't want saxophone, but you could add it into the exclude style list

[lone owl]

Well im free user cant do that anyway

[hard beacon]

maybe that's the "error" 😉 then you have to rewrite your prompt till it's without the saxophone..

[lone owl]

Still not working

[hard beacon]

what is your style prompt?

[snow scroll]

Sorry, perhaps I’ve not been clear. I’ve tried both versions. .. upload my daw version and try and use a persona on that … or take a previously created song and update it to use a persona . In both cases, I’ve found that legacy works better and keeps the melody , but the cloned voice is hit n miss, or switch to the “new” version of persona where in all cases the melody is changed pretty substantially. I’ve gone up to 85% on the slider too 🤔🤔🤔

[hard beacon]

@snow scroll : don't wanna search about the original problem.. What do you want? adding vocals?? Adding extra instruments?

[thick temple]

yup try the older models, they are more responsive , sometimes

[thick temple]

im the og prompt, i want a super spare 1920s primitive field recording, solo female folk, 50bpm, raw acoustic, guitar, dead strings. no choir, harmony, gospel, drums, violin, bass. dry solitary sound .......I get orchestra, background singers, mystery instruments ive excluded no attention to the prompt

[hard beacon]

sorry doc, but the text was for @snow scroll 🙂

[thick temple]

NTW only nutters are posting on boxing day have a good day

[hard beacon]

🤨 same to you

btw.. boxing day in germany was on the 24th ;-D

[thick temple]

im in NYC we have no idea what boxing day means

[hard beacon]

well.. boxing day was your word.. not mine 😉 Whatever.. what is your issue with prompting?

[lone owl]

1985 Electronic Synth Instrumental, Driving Shuffle Groove, Swing Beat, 112 BPM, Pure Synthesizers ONLY, Polyphonic Analog Synths, Yamaha DX7 Electric Piano, Bright Lead Synth Melody, Square Wave Lead, Moog Bass, LinnDrum, High Tech 80s, No Jazz, No Horns.

[hard beacon]

ok.. and what are your "problems" with it?

ahh.. sry.. the saxophone thing 🙂 Well.. i think suno creates a saxophone because of "Swing Beat"

[fossil moon]

hi, someone that really master the Studio mode ¿ is a total mess for me

[hard beacon]

???

[fossil moon]

Is that for me?

[hard beacon]

i don't get it.. what is your question about Suno Studio?

[fossil moon]

well...i I needed to add a string legato; I tried singing it and mixing it, but it gave me everything except what I was looking for. Then I tried doing it directly with the floating bar, which gave me everything from pizzicato to all kinds of concertos, but not a string legato accompaniment.

[hard beacon]

well.. i would use the direct floating bar method with different prompts.. for you: Legato string chords" or something like hat... and think about it: There are always 2 different versions created in the layers

[fossil moon]

Yes... I've spent over half an hour and 200 credits trying to get this, but if even my recording doesn't make sense, then I don't understand anything either....

[hard beacon]

editing in studio is "trial and error"...

[fossil moon]

And are there plans to make it work in a more intuitive way?

[hard beacon]

i know that problem.. spend 400 credits today to get a better "percussion line" in studio 🙂

[fossil moon]

omg....

but its weird coz sometimes it works....

[hard beacon]

that's why it's AI 🙂

[fossil moon]

And do you think they'll try to improve it? Make it simpler and more functional?

[hard beacon]

for sure...

[fossil moon]

thanks 🙂

cya !

[hard beacon]

cya 😉

[thick temple]

i figured it out, im sorry i dont mean to be on your thread i thought i started my own

[hard beacon]

no prob

[lone owl]

Any suggest so uh...my instrumental will sounds like (man of mirror - michael jackson) or (where do broken hearts go - whitney houston)

[hard beacon]

Emotional uplifting pop-soul ballad with a timeless 80s feel.
Moderate tempo (~90 BPM), warm analog drums, deep bass, clean rhythm guitar, subtle strings, and a powerful gospel choir in the chorus.
Lead vocal: expressive, soulful male tenor with smooth falsetto, emotional delivery, wide dynamic range, intimate verses building into a huge, inspiring chorus.
Theme: self-reflection, personal responsibility, hope, and positive change starting from within.
Arrangement: soft and minimal intro, gradual build, big anthemic chorus, key emotional bridge, final chorus with full choir and soaring vocals.
Mood: sincere, inspirational, emotional, cinematic, timeless pop.
Production: polished, warm, radio-ready, classic pop-soul sound

[young frost]

You could also try this if you are just worried about the instrumentals

Mid-tempo reflective pop with gospel influence. Piano-led arrangement with strong, steady chord progression. Minimal instrumentation at the start: piano, soft pad, light bass. Gradual build across the song. Drums enter subtly, focused on kick and snare, no groove emphasis. Gospel-style backing vocals or choir introduced in later choruses for lift and conviction. Strings used sparingly for warmth, not drama. Emotion grows through layering and harmony, not tempo changes. Tone is earnest, purposeful, and hopeful. Feels like a personal resolve becoming a shared anthem.

[lone owl]

Is that put in the style tag?

[hard beacon]

yes

[lone owl]

Got it i'll try later

[sonic folio]

Tried that and i tried all caps in the chorus yet suno makes the choir sing 90% of the chorus 😡

[young frost]

Have you tried adding an instruction directly before the line in [] to inidicate the high shrieks?

[sonic folio]

Do you mean like [High Shrieks] lyrics lyrics....?

[quiet solar]

Might be time to exorcise “operatic backing choir” then, if Suno’s developed a fixation with it. Pruning more and more bits of the prompt until you get the core of what you want, then building up from there is more art than science.

[sonic folio]

Yeah...rn suno id stupid af.......

[polar spade]

If referning to the thread just above your post, One of their issues was personas were not catching the same melody well as legacy personas might.

The other was not emulating close to the vocals given.

Devs have noted the new personas will need new tweaking in the sliders to get a similar vibe as legacy does presently or did.

As it were, legacy personas are a combination of 3.5 and 4.0 persona models from those of us having been two years here. The rest of the more recent legacies are 4.5 personas version (which took a left turn away from any vocal emulation) and quick 5.0 personas that didnt sound good with any older outputs but only 4.5 and above.

And all models and personas after July 2024 have been poor on (repeating the same) audience and live performances, in my opinion. [Creating original live performance reactions though are hit and miss recently, true]

The new persona , focused on vocal stems, is to have not as much influence by the instrumentals that vocal quality and melody should be more prominent and influential.

So I assume we might find the sweet spot with enough posts about it.

[fossil moon]

hi anyone looking for a good prompt to make clasical music ¿? i think i get a good one

[half dragon]

Feel free to share, here and/or #1252267444773326848

[hexed iris]

Is there a way to get the composition within an approx duration?

[half dragon]

It's known that prompting for duration is never reliable.
Though there are suggestions that using bars instead of minutes might help; even then, not guaranteed either.

[rugged jungle]

Guys do you know how to use the function persona because sometimes I am using the function persona but sometimes the voices become different.. is it like that or??

[half dragon]

There's been reports of persona issues.
Spend sometimes to read through the messages and you'll see lots of users have issues with new Persona; legacy Persona still works better.

[rugged jungle]

@half dragon okay I think I will legacy

Is it going to give the same voice??

[half dragon]

Personas are saved templates that capture the unique voice, instrumentation, and vibe of a song, allowing you to reuse that specific sound for new tracks, ensuring consistency for albums or a signature artist sound - so yeah vocal should sound the same.

[spice whale]

how would you go about telling the ai to have a vocal part start off with a relaxed rap attitude and slowly increase to harhser and slowly all the way up to screaming like in metal?
Sure 1 sec

Damn the prompt was in the older version would have to retest it xD

[voice: rap lead]
...
[voice: shouted doubles under lead]
...
[voice: scream adlibs behind lead]
...
[voice: full scream lead]

Text example:
[Intro: sparse beat, dark tone, low energy]
[voice: relaxed male rap, close-mic, dry, calm swagger, clear consonants]

(Verse 1)
I talk in low flames, don’t rush my breath,
Loose on the pocket, I’m counting steps,
Eyes half-lidded, grin in the shade,
I’m cool with the blade that I never displayed

[Pre-Chorus: build tension, tighten rhythm]
[voice: more aggressive rap, sharper attack, slight grit, louder, less airy]
[gradually increase intensity]

(Pre-Chorus)
Now the bass gets bold, my jaw sets tight,
Each word hits harder, I’m done being nice,
Pulse in the teeth, I’m pulling the chain,
I’m feeding the drum with a little more pain

[Chorus: heavy drop, guitars enter]
[voice: shouted chant-rap, raw, strained edge, bigger projection]
[add rasp, push intensity]

(Chorus)
I WON’T BACK DOWN — I WON’T BE TAME,
I HIT THAT WALL — I BREAK THAT FRAME,
I WON’T BACK DOWN — I LIGHT THAT WIRE,
TURN THAT BREATH INTO BARBED-WIRE FIRE

[Bridge: ramp to extreme]
[voice: harsh vocal, scream-rap hybrid, heavy distortion]
[escalate to screaming over this section]

(Bridge)
No more talk — it’s teeth and thunder,
No more calm — I tear straight under,
No more “wait” — I rip the lock,
I’m the aftershock, I’m the aftershock

[Final Chorus: full metal peak]
[voice: full scream, metal-style, sustained screams, maximum intensity]
[wall of sound, double-kick feel]

(Final Chorus)
I WON’T BACK DOWN — (SCREAM)
I WON’T BACK DOWN — (SCREAM)
TURN THAT BREATH INTO BARBED-WIRE FIRE

style:
Dark hybrid rap/metal track with a clear vocal intensity ramp: relaxed swagger rap evolving into aggressive rap, then shouted chant, ending in full metal screams; tight drums, heavy guitars, dynamic build.

no 100% garentee tho, or if some parts are right some not, add more [text music commands fitting the voice you want]

Ah you got me.
Same but still failing sometimes

Full female or full male are in the options of the song

[snow scroll]

Thanks for information. It’s really interesting and starts to make sense regarding development (or not) of personas through the later versions . I’ve been a user since 3.5 , and now realise how from 4.5 , voice similarity dropped off. There are as we all know , 3rd party apps that claim to nail vocal cloning. But I am hopeful that Suno devs are noting these threads and will , I’m sure, nail it better than the wannabes. 🤞

[spice whale]

Both... i tryed, via music commands like [female rapper, calm female voice, female demandor, female commanding,...]
Several times adding female helps to remind and focus the querrysearch of the matrix or something

[spice whale]

Whatcha mean?

[snow scroll]

Previous thread about personas ..

[spice whale]

Not my question. xD

Dropped voice cloneing and the consistancy of the personas ...

Meaning the personas do not keep up the same outputs songs voice?
Kinda feels the same, moved to inpos, kinda find it easyer to handle songcoherency there

[raven turtle]

Personas update seemed to work OK initially but now it’s just as bad as it was in 4.5 IMO.
I’m getting each generation very similar musically as well as vocally. Same Melody riffs etc. Just as it was in the early versions. Plus it feels like to vocal now sits too much separated from the music bed which sounds thin and has wavering volumes.

Nah… personas is a good idea but still needs a huge amount of work to get it effective.
Personally I’d rather be able to use my own voice too. I understand their reasons for limiting this function which could be actioned right now, but still, it’s something that would at least make the process more effective IMO.

[young frost]

Since we personas talk about consistency across tracks it should only take into account vocal qualities as instrumentation and vibe often shift when going across and entire album.

In the style block include that there is birth a male and a female singer and then indicate in [] when each is singing.

[frozen dagger]

from what i understand the newer persona mode is for more vocal consistency and less focused on any instrumentation\melody, but that was how i read the changes.

[young frost]

Exactly

Let me track down an example

Warm 3/4 Waltz At 70 Bpm, Soft Upright Piano With Gentle Arpeggios, Brushed Snare, Light Acoustic Guitar, And A Mellow String Pad, Male Baritone With Tender, Unforced Delivery, Occasional Harmonised Female Backing Vocals, Mood Is Intimate, Nostalgic, Slow-Dancing-In-The-Kitchen Comfort, Saxophone Removed; Melodic Fills Shift To Violin, Dynamics Stay Close And Warm Rather Than Anthemic

By acknowledging both the ai is looking to include them.

I have done a few duets and mixed leads. You could try phrases like a collaboration between a male and female lead. Or even a cypher (since you were talking rap) with male and female rappers.

I was responding to this and following the chain back it seemed to link to your conversation about a rap song

Ok same concept just don’t include the rap references

You can even say in the style section things like “the song starts with a male lead (describe voice if wanted). After the first verse a female lead joins”

As you said. Duets are challenging. I would love to see suno develop better multi voice instructions and models.

Did you use [] or ()
Square brackets for instructions.

I can’t see where I did that. However, If I have a line I want screamed I would try [screamed vocals] directly before the line.

Do you have an example I can play around with?

That wasn’t me that was MFG

[grand wadi]

Its better to use [ ] over () if u want suno to read it as an instruction

Although the software has gotten better at distingishing commands etc recently

Ohhhh lol i have had an insane stuggle with this myself...actually take a look at this took alot of tinkering to get this right https://suno.com/s/KZWf2IgL4VuvdBDG

Thats exactly what i had to do to get this track the way it is! It was all wrong for a few tries until i got the prompts and the spacing (oddly enough) correct

The space and the format (line per bar stack)

[young frost]

What do you have your sliders on in the advanced section?

Try keeping the style in high 70s and weirdness under 20

Low weirdness should encourage to stick closer to your instructions.

When I am experimenting I will throw in high wierdness but if I want something specific it is always low

[tribal frigate]

any tips on how to make a word a bit longer? and shorter? I tried just repeating letters, but it doesn't work as expected.

[frozen dagger]

wanting them to hold the notes a little longer while singing a word¿ use an em dash, and maybe say [melisma] or [aria] where you want the note held for that sylablle where the em dash marks it.
i.e.
[melisma]
America— America—
[aria]
Shed his grace on thee—

melisma holds the same note singing the syllable. aria goes over a range of notes while holding the syllable.
em dash is Alt 0151

[fossil moon]

hi, its possible to make a mix between 2 songs ¿

[frozen dagger]

In Create tab (dektop browser version here) drag the songs over the the add the 2 songs you want to fuse. This is probably the best way to blend.

[frozen dagger]

this is the same thing as the +Inspo funtion, except you dont have to add a playlist, can add up to 4 generations to remix

[fossil moon]

ok thanks i will check

[late sundial]

hi everyone ! quick question for any experienced user : can i ( and if i can, how do i ? ) create a song entirely built around the tic tic sound of an old clock ? ( the song is supposed to be named time, talk about time and i even already created an animated visuals around this theme )

[fossil moon]

i m sure if u give that sound to suno,and give some instructions like style, instruments and mood/energy, they will do something cool

[late sundial]

i'll try that TY

[fossil moon]

u know how give the audio right ¿?

[late sundial]

i figured it out, used youtube then youtube to mp3 and now i'm waiting to listen to these new generations

[fossil moon]

nice 🙂

[fossil moon]

so hows going ¿

[late sundial]

no matter how hard i try, suno won't even bother to include any ticking sound effect into my song... i think i'll have to do it via studio, given that i succeed in using this tool 😄

[fossil moon]

maybe....I don't control it either, but maybe the ticking sound should be added as percussion. Although you could also add it later, once you have the music, and then add it on top?

[pure panther]

I decided to try to make a polka song... I think I don't like polka...

[polar spade]

Im assuming all this is without Studio.

Will this tick be an even tempo throughout the song? If its not trouble just use the most basic DAW.

First try recording 5 minutes of a metronome set for tick and tock noise [no suggestions] . Try that first. Upload and prompt to see what happens.

Second is to go back to Suno make your song by uploading just one minute of the tick to see if it picks up on the tempo. Then download the song and put the song in the DAW. Open the tick tock metronome file. Line them up so the beats match.

Upload back to Suno if your wish to play with it more.

[late sundial]

i'm experimenting with suno studio RN, thing is,this way it's harder to create a song from only a 10 second ticking sound loop i made. but eventually i'll make it 😄

[polar spade]

Oh good. You got Studio. Even better.

[late sundial]

yeah but really hard to master

for now i'm trying to figure out how i will be able to structurate my song

[polar spade]

Is your tick tocks at 60 BPM? I did one ( about 1:40) at 120 (more of a desk clock or small wall clock) has a bit of fluff noise but if its any help. . . . https://suno.com/s/YkuhP6mro4aK7pVW

[molten jay]

https://suno.com/s/N4sAOjnVM94imPDO

If you want just Tick Tock use this Style:
Foley, Sound Effect, Solo Grandfather Clock Ticking, Mechanical gears, Metronomic, High definition, No instruments, Only ticking, Steady rhythm, 60 BPM

[molten jay]

Suno generates everything perfect and at last second destroys all lol

[raven turtle]

A quick observation. I’ve been having issues trying to get suno to not generate to type as a modern production result.. so instead of trying to over prompt for what I want give trying the excluding negative prompting to remove that edm, pumping sidechain gloss it can add all too easily. I’m getting some results with this excluding “modern, edm, trap, hip hop, r&b etc”

[proud dagger]

I saved a pesona from a song made on V5. Today when I used it on new song, I noticed slight static noise. Is there any solution for it?

[fossil moon]

try remaster

[dire night]

Remaster can work, or a cover without any style prompt

[flint elm]

How do you best and most effectively use prompts inside the lyrics in Suno v5 to generate specific vocal types (choir, harmonies, male vocals, female vocals, mixed vocals, backing vocals, etc.)?

What is the best approach to achieve more controlled and intentional results? From my experience, the style description is often interpreted rather loosely by Suno. Choirs or harmonies may appear randomly at some point, but when I want specific lines or sections to be sung in a certain way, it rarely works reliably with my current methods.

For context: I record my own vocals and then use Suno’s cover function to expand or build finished songs based on my own melodies and vocal recordings. I already try to record harmonies myself when I want them, or multi-track higher vocals for choir-like sections, but Suno often ignores this and sings everything in a completely different register or vocal style.

So my main question is:
What’s the most effective way to guide Suno toward specific vocal roles and arrangements within the lyrics when using v5 and the cover feature?

Thanks in advance!

[sonic folio]

Did not work....suno then made the verses clean and only parts of the chorus where screams....SMH

[molten jay]

Hello all, I need a "Sound of Glass Shattering" how to ask this in Style or Lyric?

[ionic silo]

Hey Folks, i have a Question
Is it possible to add a female breahting behind a textline (male vocals) wanna make some hotter stuff and that would be great for it

[polar spade]

Old school. Turn on record and drop a bag with some bottles inside. Seriously, probably a faster way.

But adding we may not have Studio so cant place the sound where we want, then its a bit trial and error to see if any type of prompting helps enhance our recording when we add lyrics or a full instrumental.

Certainly in edit we can at least copy and paste fx between sections, and see if remaster might blend decently, as well.

[hard beacon]

https://tenor.com/view/feliznavidad-santa-dance-dancing-christmas-lights-merry-christmas-gif-13807946318442273527

[polar spade]

Inside Suno only:

I dont think probable without many attempts and then finding one that works enough to then manipulate and edit more.

Genres like kpop, or sensual, caberet, risqué, or disco may brings it out more.

1. Ultimately you might get enough to then stem out all the musical parts and keep the breathy ones then patch together breathing sounds to then lay on a track under whatever you have for vocals in Studio or a DAW.

2. Unless someone has three outputs in a row of prompting showing that it consistently works the way you're asking; breathing in one higher frequency (female) WHILE a lower frequency (male) is singing or speaking -- then its random enough to say, not likely at this point. Its a duet really, but with specific "harmony" of breathing instead of singing.

3. Putting (breathing), (gasp), (awe) or other breathy tags in parentheses before or after a singing part, and maybe even push for an initial duet to edit to all male vocals later, so that the parentheses pull out female resonance responses-- then back to step 1 -- might be an angle of the parentheses tags dont do it alone.

Patience and credits mostly in Suno to try to force in prompts when, for now, it could just be an added track in Studio or DAW and be done.

[viral pond]

heya... can you use meta tags over several lines as if you made a longer comment that are ignored by generation ? similar to /* aand */ programming wise ?

but like [ in one line with several comments in between and end with ] on another line

[frigid orbit]

Just try it. You kinda gave the terms in your question.

[frigid orbit]

Just try it, if it fits your expectations, use it.

[viral pond]

YEAh, i mean here is an example ...

[

Some comments goes here .... 
Some comments goes here .... 
Some comments goes here .... 
Some comments goes here .... 

]

[ionic silo]

THANK YOU! I will give it a shot

[frigid orbit]

Try writing out moans, groans and breathing with a little mention in styles. Think stuff like (Hmmm'yes), (Mhngh'ohhh-hmmm), (Ohhh'yah!) after each line.

[frigid orbit]

Just try it.

I wouldn''t use it, cause it makes no sense to me.

[viral pond]

suely can ry it, but just wondered if anyone did test this allready

@frigid orbit i have made a highly specific project in chatgpt that i guide to what i want, but i want it to always inclyde some intro text that will be added on every generations so i would not need to add it at a later time

[frigid orbit]

People use all sorts of outlandish bloated formats.

[viral pond]

just for saving time... if i want to have links to my other projects and such things

[frigid orbit]

There's 90% sure fire ways to do stuff. Keeping [Tags] to 1-3 (1-6 in v5) words per tag for instance. Listing out terms in styles and not depending on natural language, etc.

almost everything besides that is just R&D/Just doing and seeing if it bears fruits.

[viral pond]

i successfully used narrated sentences as meta tags for additional instructions music wise

[plush field]

is there any way to force random seed for each generation?

something similar to setting seed to -1 in Stable Diffusion

i know that if you change text field enough it should use new seed. changing a style should help as well

but, i want to experiment with exactly same prompt with just different styles, for example i want to see what will hapen if i replace eurodance with eurorap or eurorap with eurorave

[viral pond]

you could compose a melody in daw, or anywhere that lets you somehow create a melody, upload it and use it as a baseline, experiment with the sliders, you will get pretty close results every time, 0% weirdness and 100% audio influence will give more or less the same melody, setting 100% style influence will change the instruments as well

[plush field]

i want exactly the opposite 🙃 i want totally different versions based just on genres listed in style, where i want to just do some small changes but to get totally new song. in most cases all generated songs sounds too similar

i am trying to find a right style for a song, which is not a common genre, but a combination of few totally different genres, like Hard Dance, Scouse House, Happy Hardcore...

and then i want to try Hard Dance, Scouse House, Rave... (just one style changed) but i am getting almost the same thing

[calm fjord]

Hey! I prompted a song with lyrics, and the result was totally perfect. The instrumentals and the style was perfect. But I need to complete change some of the lyrics and add another verse.

Now I'm trying to regenerate the song again with the exact same prompt. But it gives me completely different instrumentals and style. Isn't there a way I can make it use the exact same instrumentals?

[plush field]

are you on free on paid plan?

[calm fjord]

Yes I'm on Premier

[plush field]

go to studio, and edit lyrics

and replace the text

[calm fjord]

I don't see an edit lyrics button, I assume you mean this page right?

[lone owl]

Well that's fun and working but unfortunately i looking only 80s style, this one feels harmony modern. I looking classic FM tines....like you know FM E.piano 1 from DX7 or some brass synthesizer(not sax or trumpet).

That's why my inspiration is instrumental from.:

• Man in the mirror (michael jackson)
• Where do broken hearts go (whitney houston
• it must have been love (Roxette)

Well i tried but ended up mess overused saxophone🤔🥀.

Maybe anyone can teach me how?

[plush field]

split to stems

select part of vocal you want to change

right click and use remix selection

and then you shold get text in lyrics

after it finish processing

[calm fjord]

I see the Remix Selection button now. So I should just enter the lyrics here?

[plush field]

it should detect current lyrics first

wait for it to finish

to get a wave and lyrics

and you can use any of those

[calm fjord]

Seems to work! Atleast its a lot closer to my orignal. Thanks!

[plush field]

you can even try to remove current lyrics and just create new ones in new track

like this

[calm fjord]

Should I leave Styles empty?

[plush field]

i think it should use the ones used for song if you do not populate it

but you can try

try with empty and with any custom one you want

[calm fjord]

Its still a bit far from my original, but I guess it would always be. Thanks for your help

[plush field]

i haven't played a lot with studio, someone elese might have better answer, so check again later

i think i found an easy and reliable solution for total random generation even with the same lyrics and style on each generation

use UUIDs as song titles

🙈

now i am 99.99% sure that using song title with versions is a sure way to waste credits

do not do this at home kids

[calm fjord]

I've tried a few things now, and even when Replacing Lyrics in the Song Editor, it just regenerates that whole part with new instrumentals which is annoying. So the Replacing Lyrics function is basically useless, unless there is a switch I'm missing. I assume its to avoid copyright issues or something, even though the original was created within Suno lol

[quiet solar]

There is a way to do it, but it's more complex than just tinkering with settings.

To explain; anything that Suno creates is always created again from scratch, so you'll always get new instrumentals and new vocals generated whatever approach you use. Sometimes the differences will be quite subtle, sometimes less so. Replace lyrics doesn't get around that, but is a quick and easy option for someone wanting to make a fix who isn't worried about minor changes to the instruments in the background.

If you really want to keep the instruments identical, you have to find a way to work with just the vocals. There are a few ways of doing this, but as you're a Premier user, what I'd personally suggest is this:

1. Go to your original song in the Library and Get Stems, split it by instrumental and vocal.
2. Extend your original song from somewhere after the section (e.g. Verse, Chorus) that you're looking to make lyric changes to.
3. In Extend, put in the new version of the section that you're looking to repeat. Keep the style prompt identical unless there's something specific you're looking to change. Set Weirdness low (c.5%-15%), Style mid-low (15%-50%) and audio influence mid-high (50% - 85%). You may need to tweak these, particularly audio influence, to get the sweet spot.
4. Extend as needed until you get vocals that you like and that fit quite closely to the original. Disregard the instrumentals.
5. Get Stems from that Extension and split by instrumental and vocal.
6. In Studio, start with the original instrumental stem and vocal stem and then import the new Extension vocal stem and use part of it to replace the original vocal stem in the section you're editing. For consistency of sound, you may find it easier to replace the entire section's vocals, not just the specific bit you're looking to change.

[quiet solar]

Controlled and reliable results are beyond the model's current capabilities.

It's best to think of prompts as being suggestions that may be taken into account, rather than instructions that Suno will always follow. From that starting point, what you're trying to do is pushing past the boundaries of what's possible via Suno prompting. It is possible, but you have to take a step back and look at Suno as a tool that can create the pieces for you to put together in editing to achieve the end result.

The simplest example is switching a male/female singer - that's fairly easy to control via a combination of prompting and settings to generate one version of the song with a male singer and then cover the same song with the style prompts changed to have it performed by a female singer. You can then create a blended duet between the two by splitting the songs into vocal/instrumental stems, picking the instrumental stem you like the most and replacing some of the male vocals with the female vocals using Studio or an external DAW/audio editor.

It's more complex if you want to create harmonies, because you don't want both singers following the same melody. Your best bet is to prompt for harmonies in the style prompt, and call both singers in lyric prompts, e.g.

[Chorus]
[Male Tenor] Guy sings this line here
[Male Tenor and Female Alto] The line you want to be harmonised here

If all else fails, you can use the first approach and manually adjusting the pitch of one of the singers in an external DAW or audio editor before combining them in Studio or an external editor.

Choirs are their own separate headache. Prompting for choirs often means including a few different choir-related prompts in the style box (search 'choir' in Dasvon's posts here and you'll get some good suggestions). Typically, Suno will combine choirs with a lead singer, so if that isn't what you want then you may need some experimentation to get a version that works which you can use to create vocal stems for editing.

Backing vocals are generally straightforward. If you've prompted for a backing singer (or singers) in the style prompt, you can typically use rounded brackets, in the lyrics box to indicate where lines or parts of lines should be handed to the backing singer, e.g:

[Chorus]
The main singer sings this bit (backing singer comes in here if there's time before the next line)

or

[Chorus]
The main singer sings this line
(the backing singer sings this line)

Again - all prompts are just suggestions, so anticipate this may take a few tries to get right if you're doing something complex or have a very specific vision for what you want happening at many specific parts of the song.

[shell loom]

These are the big stylistic categories. used by musician

Heavy metal (classic)
Alternative metal (downtuned, mid‑paced “chug” riffs)
Power metal
Thrash metal
Death metal
Black metal
Doom metal
Gothic metal
Symphonic metal
Progressive metal
Industrial metal
Folk metal
Metalcore / Deathcore
Nu‑metal

Metal guitar vocabulary is huge.
Chug / chugging (palm‑muted low‑end riffs)
Tremolo picking
Pinch harmonics / squeals
Drop‑tuned / downtuned guitars
Gallop rhythm (Iron Maiden style)
Shredding / tapping
Dissonant intervals
Wall of sound
High‑gain distortion

Metal drumming defines subgenres as much as guitars do.
Blast beats
Double‑kick / double‑bass
Skank beat
Polyrhythms (prog metal)
Breakdowns (metalcore)
Fast tremolo snare patterns

Metal vocals are extremely varied.
Clean vocals
Harsh vocals
Growls
Screams
Fry scream
False‑cord scream
Operatic vocals (power/symphonic metal)
Spoken‑word / narration (doom, gothic)

These describe the feel of the mix.
Raw / lo‑fi (black metal)
Polished / high‑fidelity (modern metal)
Atmospheric / ambient layers
Symphonic / orchestral backing
Industrial textures
Dense / compressed mix
Reverb‑drenched vocals

Metal has its own slang and identity markers.
Headbanging
Monster ballad (80s glam metal term)
Metalhead (fan identity)
Riff‑driven
Mosh / pit
Shredders
Blast merchants (drummers)

Enjoy! and happy 🤘🏼

[onyx sparrow]

how i get a 3 minutes long track?

i tried already few things. dont works anymore..

[shell loom]

not here: this is for prompt development try
https://discord.com/channels/1069381916492562582/1069381916492562585

[frozen dagger]

different things at play here. 1st lyrics length obviously matter. 2nd the style prompt (genre of music you choose) gonna basically gonna be the biggest impact on how big it thinks song should be, based on how long your lyrics are.
that said if u have no lyrics (which I'm presuming is the case here with short duration) then try adding some placeholder stuff in your lyrics prompt to artificailly lengthen the song. something like:
[Verse 1 Instrumental]
[Chorus Instrumental]
[Interlude Instrumental]
[Verse 2 Instrumental]
[Chorus Instrumental]
[Interlude Instrumental]
[Verse 3 Instrumental]
[Chorus Instrumental]
[Interlude Instrumental]
[Chorus Instrumental]
[Verse 4 Instrumental]
[Interlude Instrumental]
[Chorus Instrumental]
[Outro Instrumental]

[tidal prism]

Can anyone with the suno gpt please put in “give me a prompt that best fits the heavy sound and of the metal band orbit culture”

I don’t have chat gpt pro or whatever its called to do it myself

[sullen valley]

@lone owl might try this when Suno's back up:

Instrumental pop with early-80s new wave sensibility,
mid-tempo shuffle groove with subtle swing, ~80 BPM,
rhythm-led structure with continuous forward motion,
tight pocket, consistent pulse, no dramatic rubato,
ensemble-driven phrasing, melodic ideas distributed across parts,
primary focus on rhythm, harmony, and texture working together,
analog poly-synth chords with Prophet-5 character,
Yamaha CS-80 pads reinforcing emotional weight,
melodic synth bass locked to kick and groove,
controlled dynamics, measured expression,
serious, reflective tone, emotionally direct but restrained,
clean stereo image, period-authentic studio reverb,
polished but grounded, purposeful, understated confidence

[lone owl]

But that everything on tag style?

[sullen valley]

Yeah, I'd slap that into the style

[lone owl]

Well hope sound like "Everybody wants to rule the world" 🤔🥀😩.

Not same but some simillar synth maybe sounds cool enough

[sullen valley]

Then I'd refine it after I heard what it produced

[sullen valley]

Alternative:

Mid-80s new wave pop instrumental,
laid-back mid-tempo shuffle with subtle swing, ~110 BPM feel but relaxed phrasing,
groove-led composition with steady forward motion,
calm, confident rhythmic pocket,
repetitive, hypnotic melodic figures,
ensemble-driven arrangement with no showy lead presence,
clean guitar and keyboard textures sharing melodic responsibility,
warm analog synth pads supporting harmony,
soft, understated chord movement,
melodic bass line anchoring the groove,
emotion conveyed through progression and rhythm rather than ornamentation,
cool, observational tone, quietly reflective,
wide stereo image with clear separation,
polished 80s studio production,
natural reverb, smooth transitions,
accessible, thoughtful, restrained, timeless

[lone owl]

That's should be synth brass or synth bass? Because i hearing tears for fears is brass synth anyway(maybe)🤔

[sullen valley]

you could always try adding in

bright brassy synth motif in the upper register,
short, punchy envelope, rhythmic and repetitive,
separate rounded synth bass anchoring the groove

[lone owl]

Alright thank you, gotta try later after these technical issue fixed

[candid marsh]

As a better idea if you want to really dive into learning about how Suno works so that you can figure things out better to your taste try this

Open Chat GPT and type in this - Research everything you can on building a song and style prompt with suno v5. then compile that information into a document for making a chatgpt project with Suno Songwriting.

it will give you a lot to read. If you then want to use ChatGPT to help write music for Suno the type in "from here forward anytime I ask for something for Suno reference this" and it will lock in that chat sequence with the Suno guide.

Then type in "I need you to imagine you are a music guru that knows everything about music theory, production, mastering and dubbing". Anytime you want with lyrics or styles type it in that chat conversation. I dont recommend using it to copy/paste into Suno but it's a stron starting point

[flint elm]

Thank you very much for giving me such a detailed answer! Much appreciated!

[raven turtle]

Dave Bascombe produced TFF “Songs from the Big Chair”. A lot of plate reverbs and big drum samples. Have you tried prompting for production style of Bascombe? I wonder if Suno understands producers names?

[quiet geyser]

Hello,
when I try to create a hook in Suno, I get the error:
“Exception during hook creation. Your account requires support assistance.”
This happens consistently and prevents me from using the hook feature.
Could you please check my account and enable hook generation?
Thank you.

[frigid orbit]

lol no

The moment you said GPT, you lost

GPT knows as much as we do regarding Suno, which is 0

[elfin lodge]

@frigid orbit @small violet

How would you put this into a style prompt

[frigid orbit]

Industrial Grinding, Mechanical Pressure are some terms I’d try

It’s a fair bit metal-esque. But I don’t need to tell you that

[elfin lodge]

industrial metalcore cyber aggression glitch textures manic energy distorted vocals chaotic structure
modern metalcore rage industrial synths breakneck tempo abrasive tone digital violence
chaotic heaviness electronic distortion panic driven riffs hyper aggressive delivery
cyberpunk metalcore machine noise erratic rhythms visceral hostility modern brutality
industrial breakdowns glitchcore influence unhinged energy distorted atmosphere
frenetic metalcore digital decay mechanical rage relentless pacing
modern extreme metal synthetic chaos aggressive catharsis technological dread

This was the prompt I was using

[frigid orbit]

I don’t like suggesting GPT. But what you could do is ask it to remove redundant terms.

It seems to repeat quite a bit of vibes

[elfin lodge]

That’s probably why it glitches alot

Or it’s cause the audio sucks lol

[frigid orbit]

Let me go through your prompt and point out the terms I wouldn’t use personally.

Cyber, (anything Metal, I just don’t like AI metal but that’s beside the point), modern, visceral, hostility, decay,, extreme, synthetic, any mention regarding tempo. My preference goes to fast or slow pace.

[elfin lodge]

I have a new prompt imma try

Industrial metalcore with cyberpunk influence. Aggressive, glitch-driven guitars and industrial synths, distorted vocals, and mechanical rhythms. Chaotic structure with sudden drops and breakdowns, driven by panic and manic energy. Abrasive tone, digital decay textures, and synthetic noise layered into modern metalcore brutality. Fast tempo, relentless pacing, visceral hostility, and technological dread.

This was tuned for Suno I guess

[frigid orbit]

Cyberpunk

[elfin lodge]

I just caught that lol

[frigid orbit]

Suno’s cyberpunk, cyberpunk the setting and cyberpunk the genre are three entirely different things.

[elfin lodge]

Oof

[frigid orbit]

Well the setting is just 70s-80s-90s media with a modern sheen dialed to 11

The genre is far more electronic iirc

Suno’s is the genre+punk+artifacting+glitch

[elfin lodge]

Is there something in that prompt that’s making the vocals glitchy?

[frigid orbit]

My whim says Digital Decay, and maybe Synthetic

[elfin lodge]

I’m starting to think it’s the audio

Cuz it’ll randomly glitch and cut randomly

The crazy part is I already have a version of this song in the album I’m just wanting it to be a tad different lol

[frigid orbit]

Oh are you trying to cover/extend from something with Audio?

I always work on low weirdness and 25-60 audio influence. If I even touch the sliders

[elfin lodge]

I have the audio at 98 💀

[frigid orbit]

Bro ofcourse it glitches

[elfin lodge]

🤣🤣

[urban zephyr]

Is there a way to add personas in the lyrics prompt so you can have multiple persona voices singing different parts of one song? E. G. Duets etc.

[earnest burrow]

Can I determine which words are sung by a man and which by a woman without Suno mixing up the roles and how?

@quick wedge ?

[quick wedge]

I’m not sure I guess something like this

[male vocals]
Lyric here
[female vocals]
Lyric words here

Just use meta tags

[earnest burrow]

so he confuses the roles

Is there a sample song or instructions somewhere on how to do it?

[quick wedge]

Here’s a sample that someone else made

https://suno.com/playlist/3199d794-0624-46a7-849f-360e9e76df76

[earnest burrow]

Please look at this song and tell me why Suno is wrong? https://suno.com/s/mhcbVEnd4EZbWOlr

[urban zephyr]

Suno always screws up who sings what part no matter how you prompt. Hoping new persona tags can fix that

[quick wedge]

Try adding layered harmonies in your style box I don’t understand your song
Or try this [verse - female vocals] or something like that
Duets are trial and error it doesn’t always work well the first time

[earnest burrow]

Is there a way to tell Sano, through the Persona's name, that the persona IVA is singing here, for example, or the persona ALI?

So the problem isn't me? It's just that the duets haven't been fixed yet?

[quick wedge]

I don’t use personas or you can add a person’s name in the meta tag

[earnest burrow]

And if I use a person's name, do I have to describe them in style?

if I have 5 personas with names 1 2 3 4 5 can I describe in the text for example this text sings persona 2, this persona 4 .... ???

[quick wedge]

I dunno anything else than that I’m sure there’s tutorials on YouTube too

[frozen dagger]

A modern metal-driven track blending tight, downtuned guitars, aggressive double-kick drumming, and atmospheric synth layers. Verses lock into groove-heavy riffs with punchy bass, while choruses explode into melodic hooks with soaring lead lines. Ambient textures add depth between breakdowns.

This is what Suno interpreted the band's name to. If this is what you were asking for.

[earnest burrow]

Is there anyone from support I can ask?

[half dragon]

I don't think you can reference a Persona in metatag at all.
You can attach ONE Persona for reference when creating the song but you can't reference it ...or used in a duet use via metatags
It doesn't work that way.

[frozen dagger]

u can not currently just use text to tell suno to use a specific persona. the most reliable way for multiple voices placed exactly where you want them is creating multiple copies with different voices and seperating the vocal stems, and then recombine and edit within a DAW.
Can search this thread for things like "duet" "multiple singers" and stuff like that to see a lot suggestionss and stuff people have already tried, but we do not have a streamlined A B C method for this right now. It requires workaround and effort.

[frozen dagger]

this make me think another option to try would be to create Covers using the different Personas and then throw a few of those together in the +Inspo. Since the music would be mostly the same as they are covers, but the voices would vary. It would stitch these together in a way that could be pleasing, but this is not a method with any control. Just an afterthought of another way to blend the voices. When not so concerned about placement. Suno, take the wheel.

[half dragon]

Yeah, it'd be an inspired creation but there's zero control only hopes 🙂

[strong marten]

How to make something similar to
yung lean ginseng strip 2002

[urban zephyr]

This is how I did duets is I just made multiple versions with different singers then cut it all together in a daw

[frozen dagger]

Suno translated "yung lean ginseng strip 2002" to:

A hazy cloud-rap production opens with lo-fi melodic synths, delicate bell textures, and distant atmospheric FX. Sparse trap beats punctuate the space, boosted by deep 808s and a swung hi-hat groove. Ethereal pads swell in the chorus, creating a hypnotic, melancholic vibe throughout.

Hope that helps.

[gusty spear]

does anyone have problem with male rock vocals being too machinery? is it just me doodooing the prompt or the heavier rocky styles have this issue generaly for male vocals?

[mental pendant]

Any1 knows how to prompt so an instrument only plays during a chorus? lets say i want piano to only play sometimes during a song, like a chorus. how do we prompt this? every time i mention two instruments they both overlap the entire song

[frozen dagger]

tried something like [Chorus- Piano Solo]¿

[mental pendant]

How would you incorporate that into a prompt? Let's say I want a guitar playing, but during chorus the piano comes in. Im rather new to suno so i dont fully understand the prompt system

[frozen dagger]

you want to just place it inside brackets within the lyrics prompt where it appliess

[frozen dagger]

[Verse- Guitar leads the section]
[Chorus- Guitar with Piano accompaniment]

[mental pendant]

no luck sadly, also im only doing instrumentals. maybe that makes it harder

[urban zephyr]

How is the midi output? Has anyone tried it? Is it accurate? For stems.

[frozen dagger]

if its all instrumental i wouldn't use [chorus] as a tag to avoid the
just
[Guitar instrumental]
[Guitar with piano accompaniment]

[remote ivy]

if your using an uploaded instrumental track, using the add vocals option to add lyrics for full song...is it possible for it to render a longer length the underlying instrumental if the lyrics permit it....i have tried a few renders...and it just cuts off at the instrumental track length...so I am guessing no?...I believe I had luck w/ tracks that weren't downloaded then reuploaded

[hardy cosmos]

Does anyone know any good Prompts for SoundCloud Rap in the style of?

[frozen dagger]

sounds like you would be wanting to do extend, and adding sections in editor\studio, or if try cover with sliders adjusted to be flexible

[frozen dagger]

try putting an artist that represents that style of music you like in the style prompt box and lcick teh magic wand button. it can give a you a prompt that gives you an idea of where to start at least with what Suno thinks. Kinda like reverse engineering, but not really. can repeat for different phrasings or interpolate your own instruments and personal flavors to ti and hit the button as well ofc

[frozen dagger]

doesn't work for all artists obviously

[remote ivy]

i don't really want to do an extend, i think i have had luck w/ non uploaded instrumental...i just didn't want to waste the credits on 10 renders to find out its impossible....i have 0 faith in the extend and replace suno from prior experience.

[frozen dagger]

PSA: dont close your Suno tabs or could get locked out like others.

[hardy cosmos]

Thx bro 👊

[raw wraith]

Hi I'm in Brazil, for some reason, suno won't open in the app, neither on website, can someone please help?

[proven forum]

it is down... again...

[raw wraith]

Oh, hope it comes back up soon, this platform helped me make my imagination come to life

Im so grateful for suno

[safe maple]

Is there a more official call for reversing tine, I happened across,"
(Reversed particle streams: .enoz emityna otni sgnidneB)
Which invoked some backward masking sounds. Has anybody a library of tangential, tertiary [commands]?

[safe maple]

Megaditto and megawatts! Lol! Whole new vistas are opening up

[half dragon]

You might want to post in #1252267444773326848 so this won't get lost in this channel

[frozen dagger]

I always appreciate when others share their findings. I will definitely look at it later and see what you came up with. JSYK it not falling on def ears...(eyes¿) Thanks for sharing.

[rocky haven]

-Sorry I uploaded the wrong version of the document before.-

Sharing a framework for understanding why Suno outputs behave the way they do

I’ve been experimenting for a while with a coordinate-based way of thinking about Suno prompts — basically trying to explain why certain prompts stay stable and others suddenly flip style or vocal posture.

I wrote this up as:

a short spec (axes + rules + examples)

and a longer paper explaining what I observed

This isn’t a claim about “how Suno works internally” — it’s a user-side framework that helped me predict failures and control outputs more reliably.

Uploading both here in case it’s useful to anyone else. Happy to answer questions or clarify parts.

[frozen dagger]

People be cutting in line. My Suno que timer for place in line has increased from 43 minute wait to a 1hour 22min wait over the last 20 mins. 🧐 🍻

[mellow moat]

Is there anyway to upload an instrumental and keep it the exact same? I made one from scratch

[frozen dagger]

not sure i follow. if you upload a file and want one exactly the same, you want to create a copy¿

[mellow moat]

Nah i want to add vocals to it but keep the instrumental the same

[frozen dagger]

I think you are looking for the Add Vocals.

Ac tlike you are going to do a Cover of the song then change Cover in the drop down menu

[ionic pendant]

I'd like to actually add my vocals, you know like create my twin voice since I load up god knows how much of it already

[frozen dagger]

I don't think we can right now. Turn an audio upload of ourself singing into a Persona would be the way to do this, but we currently can't use uploaded audio to create a Persona. So would need to use other tools.

[frozen dagger]

which is totally doable. but also can't publish on Suno, I think, if that was something you wanted to do too.

[feral mason]

Adding your own vocals is pretty far out from the picture right now. This is more a legal issue then anything. Realistically, theres no way for Suno to be able to verify if its you your uploading, or 50 cent. Plus it takes a lot of training to allow that. Which means, they would have to literally go out of there way to open themselves up to legal reprucussions

[quiet solar]

I wouldn't say it's that duets "haven't been fixed", it's more that Suno as it currently stands isn't designed in a way that facilitates individual vocalists or instruments being tracked independently and generating their own parts independently throughout the song as a whole.

For example, in a 'normal' music track if you got hold of the audio stems they would be completely separate recordings by each individual musician playing or singing their individual part. If you do the same in Suno, you'll see the stems are artificially split out from the final version. Suno doesn't typically create things at 'stem' level and then build up an overall song, it just creates everything all at once as it goes along - until we get to separate stem creations, creating duets in Suno will be incredibly challenging. Not least because prompts like "[Singer A] Sings this bit" in the lyrics are effectively suggestions rather than commands.

Duets can be done in Suno despite that limitation, and good prompting can help, but unless you're very lucky or don't care who sings what, it isn't a case of putting in your prompts and hitting 'Create'. It requires using Suno as a tool to generate pieces you can stitch together afterwards in a mixed exercise of prompting, creating and audio editing to get to a final version.

As others have said, there's lots of advice in here if you search for "duet" or "male and female" as to various approaches and workarounds you can use to create something like that using Suno Studio or an external DAW/audio editor if you're interested in taking the plunge and using Suno a bit differently. But be aware it's not a simple or quick exercise!

[quiet solar]

The easiest approach there is to download the instrumental stems, sing over the top of them and then mix/master as required.

If you want an AI 'Twin' of your voice, there are also alternative AI programs that can take a crack at that.

The only downside of either approach is that you won't be able to share the final version via Suno.

[lone owl]

Never tried that before

[molten jay]

oh no, right now i wanted to ask about that, i have 2 different voices, i need to use them at single music as Duet

[half dragon]

You can try Extend.
Generate the first section with one Persona , then extend the next section with a different Persona.

[brazen ivy]

Good morning everybody, I am kind of new on working with Suno but still have a problem that I cannot get fixed.
In nearly every song that is created, the song starts with a kind of intro and the voice is humming, singing or chanting: Oh Oh Oh. Or something similar.
I have tried all kind of prompts but cant get rid of that.
Does anybody have an idea what kind of prompt can make this disappear?
Thank you!

[hollow hare]

Unfortunately I don't think there is much that can be done since v5 training appears to have biased itself to have that at the start.

The easiest way to get rid of it is to vocal-split the track at the end (e.g. through Suno or several programs like UVR5) and replace the initial portion with the vocal-less part using an audio editor like Audacity

Another is to generate an instrumental intro/piece or start off with an earlier model and extend into v5

[brazen ivy]

Thank you!

[raven turtle]

It’s frustrating that I can’t bake my own vocals into a persona but the workaround I use is splitting vocal stem then remaster it for clarity. Then download instrumental and remastered vox stems. I then use kits or Audimee to replace the suno vox with a trained vocal of me and then pop those over the instrumental stem in Logic Pro.

[south yoke]

I use custom mode with my own lyrics. When I type to song description: "Vocals must be female. Vocals must be solo.", or "Female solo vocals", or "Female vocals. Solo vocals", or "Vocals must be only female. There must not be male vocals", or "Female vocals, not male", or "Prompt: Female solo vocals... bla bla bla. Negative prompt: male vocals." , or "Solo female vocals. No male vocals."
I usually get a duet of a female and a male. How to fix it?

[hollow hare]

Usually the odds are genre dependant, which one are you looking at?

Ideally I would avoid using 'solo' since it may simply solo verses during a duet and the pattern biases towards duets.

Fo not use any negative terms in the style prompt as the negation may be ignored (so avoid mention of male vocals). Try adding more feminine adjectives, mood, emotion, etc.

If it is still giving you multiple vocals, try labeling in the prompt with a name so that the AI tries to stay consistent (not sure how well this works in v5), like:

[Verse 1][Amy: cute peppy female]
Heh hehe hehehehehehe

[south yoke]

Mixture of multiple genres.

[hollow hare]

Hmm try doing what I mentioned. If it fails, provide me an example of your prompt so I can try on my side
This was my attempt at an example, I should have chosen better https://suno.com/s/CLCLU3QkK6bRTBik

[south yoke]

Does name play a role? Kinda what is better: Patricia, Anna, Hope, Jane, or something else?

[hollow hare]

Hm it kinda plays a role with referencing, helps gets different voices when doing a girl group, boyband, or idol genre for example. Could be placebo but helped a lot with duets in past models.

As for the name selected, I do not think so but there is a chance that ethnic and gendered names may help you reach that particular voice if the name is common/unique enough to have been transcribed in whatever source Suno used (so more likely if the name was in a play of opera is my assumption)

[south yoke]

It hard in my case, because usually I generate multiple genres in one song and need solo vocals at the same time.

[viral pond]

anyone here got a prompt that introduces unstable rhythms, uneven beats etc ?

[south yoke]

Advice with names helped to solve gender problem. Thanks. But now it started to focus on those names, and less care about other instructions.

[hollow hare]

Try redefining when needed, like
[Verse 3][Amy: melancholic asmr][upbeat bubblegum pop]

There were terms to help force genre change without extending and avoiding homogeneous mixture but I can't remember what they are. More like it was just placebo

[odd coral]

Does anyone have solution for payment bug
" We were unable to process your recent subscription renewal.
In order to keep creating, please update your billing information and retry payment."

Even though amount has been charged?

[brazen ivy]

It worked with v4.5 and uploading the song. Then same input and the intro stayed the same in v5. Thanks a lot for your help!

[south yoke]

I have did it always until today without names. Now names make them less strong.

[hollow hare]

Hm exclude the name then and just redefine the vocals I guess. The name stuff I used were for duets and idol groups

[quiet solar]

You shouldn't have to use names to prompt for a female singer. I'm assuming you're on Basic and don't have access to the gender selection in the advanced settings under Create, as that's the simplest route.

Alternatively, a simple prompt only referencing a female singer should do it 90%+ of the time - e.g. adding "Vocalist is a female alto with a breathy, haunting tone" to the style prompt. As previously stated, don't mention a male singer at all if you don't want one. Suno is much better at responding to you telling it what you do want rather than what you don't.

It will also be influenced by genre, so if you're asking for a female singer in a genre overwhelmingly dominated by male singers, you may need to take a few extra attempts.

[south yoke]

Without names I get female+male duets.

[hollow hare]

Hm sorry, can't help you with this issue without knowing what prompts you are using. Perhaps someone else will recommend a method when they see your issue.

Chances are someone may have already addressed it here or at #1252267444773326848 but the odds of finding it from searching...

[south yoke]

Do you need my prompt?

[hollow hare]

Unfortunately I don't have time to test around right now, sorry

[frigid orbit]

Just adjust your tags til it's right.

[south yoke]

Look at my first post in this channel.

[frigid orbit]

I've read most of it. Lumine gave decent pointers that should put you closer, finetuning should be you thing.

It's along the lines of most answers here regarding gendered vocals.

[south yoke]

In this case it focuses on names too much.

[frigid orbit]

Well adjust? I don't use names at all, nor structure tags, nor natural language in style prompt and I get little issue using genders. How I do it is that I focus on mainly one voice in my style prompt. Then when I need a different voice, I'll just use a simple [tag] like [Female] or [Sampled Vocals//Female Whisper] and keep it like that.

I get the logic behind it, and I give that same tip.

[south yoke]

Bro, it provides female+male duets to me.

[frigid orbit]

Goodluck.

[south yoke]

Bro, you're very polite 👍

[frigid orbit]

Well most of the tips required to avoid such issues have been given.

[lone owl]

nope it doesn't work at all

it goes overused saxophone appear again.

[neat lake]

Hi, I asked on general, but of course nobody could answer, so I'll try here. Have any of you guys noticed metallic, ultra harsh top end frequencies on your generated outputs on V5? I experienced this with previous models, but not V5 and am a bit surprised, wondering why this happens if the models cannot be altered after their release and previous results from let's say, 2, 3 weeks ago were fine? There is also an issue with unnaturally raspy vocals.

[hard beacon]

In fact, no one will be able to give you a reliable answer to that, but I've also noticed it sometimes in some outputs over the past few days.

However, this “mistake” usually occurred when I resorted to inspiration or personas.

[frigid orbit]

Could be that you just have a lot of words, that in output generates that. Unless we get a link or an example prompt, we can’t really tell you

[surreal lava]

Is there a place that shows how to create the prompts and lyric breakdowns for programming in SUNO? Been using this program for 6 months wanting to improve

[dapper blaze]

I added

[Instrumental section 1]

[Instrumental section 2]

[Instrumental section 3]

[Instrumental section 4]

[Instrumental section 5]

[Outro: Fading echoing laughter in a concrete room]
haa haa haa haa haaa

to a prompt that was working fine and still does without it and I'm getting copyright warnings, why? no way that's copyrighted.

[hard beacon]

hmm..

that's strange

[dapper blaze]

adding hah allowed it to work.. the hell laughter is copyrighted?? no way!

also tried

ha ha ha ha haa

and remoning different sections

[hard beacon]

sometimes AI is a big secret 😉

[surreal lava]

Hmm?

[hard beacon]

@surreal lava : can you be more specific with your question?

[surreal lava]

High-energy, cinematic hard-rock anthem with gritty, slightly rough lead vocals, Driving electric guitars with fast riffs, punchy bass, energetic, punch-driven drums, and rhythmic acoustic accents for grit, Verses are short, punchy, story-driven, moving fast with intensity, Pre-chorus builds tension; choruses are explosive, chantable, and arena-ready with layered backing vocals or military quartet, Spoken-word intro and mid-song narration like a commanding storyteller, Theme focuses on G, I, Joe Trident rapid-response unit, deployed globally from USS FLAGG, emphasizing speed, precision, first-to-the-call response, and defending freedom against Cobra, Mood is urgent, heroic, cinematic, and triumphant, modern hard-rock energy, faster-paced than traditional military country rock

Okay my post disappeared

Not very helpful it keeps getting deleted when I am asking for programming help

[hard beacon]

copy and paste it without empty lines

[surreal lava]

Intro Spoken – Announcer Tone, No Music]

“They are G.I. Joe’s elite rapid-response unit — Trident.
Deployed wherever trouble strikes, from the mighty USS FLAGG.
Fighting for freedom — over sea, land, and air!”

[Soldier Battlecry – immediate, unified]

Yo Joe!

[Verse 1 – Gritty, Driving]

Steel decks hum beneath our boots
Engines roar, we follow the route
When the call comes in, we don’t delay
We hit the ground, we clear the way

[Chorus – Anthemic, Chantable]

G.I. Joe! (First to the Fire!)
Riding fast, we answer the call now
G.I. Joe! (Global Ready!)
From the FLAGG, we strike them all
G.I. Joe! (No One Stands Alone!)
Speed and courage, standing tall
G.I. Joe is there!

[Verse 2 – Story Driven]

Cobra’s plans won’t see the light
We break their schemes before the fight
Jets and helis cut the sky
We bring the force before they try

[Pre-Chorus – Rising Tension]

No retreat, we never fall
Every mission, we give our all
From sea to sky, across the land
Trident strikes with steady hand

[Chorus – Full Power, Gritty Vocals]

G.I. Joe! (First to the Fire!)
Riding fast, we answer the call now
G.I. Joe! (Global Ready!)
From the FLAGG, we strike them all
G.I. Joe! (No One Stands Alone!)
Speed and courage, standing tall
G.I. Joe is there!

[Bridge – Spoken-Word / Half-Sung]

“G.I. Joe — the world’s daring, highly trained mission force.
Defending freedom against Cobra and their allies.
First to the call. Always ready. Always there.”

[Final Chorus – Anthemic, Rugged, Climactic]

G.I. Joe! (First to the Fire!)
Riding fast, we answer the call now
G.I. Joe! (Global Ready!)
From the FLAGG, we strike them all
G.I. Joe! (No One Stands Alone!)
Speed and courage, standing tall
G.I. Joe is there!

[Final Shout – All Voices]

GI! JOE!

[hard beacon]

ok.. and what is your specific question now?

overall the prompting looks good

[surreal lava]

Looking for ways to improve my work.

[hard beacon]

but what do you wanna improve??

[surreal lava]

Anything I can. I think the break down I use for the lyrics and my SUNO Style prompting need work.

I was hoping to find something to read on programming better

[hard beacon]

well.. ever tried [breakdown] in the lyric section? 🙂 in the end of the day SUNO is AI... i can just repeat myself.. prompting looks good...

[surreal lava]

thanks

So no place to look or study then Hmmm

[hard beacon]

not really.. of course you can ask services like chatgpt oder gemini.. but.. well.. don't know if they can you give your more infos

[lament holly]

Hi all — sharing a community resource that may be useful here.

SUNO Driver’s Guide: What we wish we’d had when we started

It’s a practical orientation guide covering prompting posture, exclusions, workflow patterns, and common pitfalls, written from accumulated community experience rather than official docs.

The guide is openly licensed (CC BY-SA 4.0), archived on Zenodo with a DOI, and mirrored on GitHub for ease of access.

Mods — feel free to place, link, pin, or ignore as you see fit. Just making it visible.

Zenodo (canonical): https://doi.org/10.5281/zenodo.18065688

GitHub: https://github.com/RosCollective/suno-drivers-guide

[quiet solar]

@surreal lava ChatGPT and other GenAIs can sweep the collected information of the internet, including various half-truths, nonsense and "it just so happened to work when I did it so I reckon it must be coded in" findings that are probably as likely to mislead as well as help.

There is no official documentation behind Suno accessible to us mere mortals outside of the coding and dev team. But the good news is that you don't need it. Prompting isn't a code with a specific form of instructions, it's a method of passing suggestions to Suno which it may (or may not) take into account by breaking it down into keywords and trying to look them up against its training library of music to understand what you're asking for.

So on the plus side, that means you can express the same thing in multiple ways and be effective. On the down side, it means there's no "right" or "wrong" way to do something and so it's trickier to learn from others' experience, especially since it can vary in the context of different genres and underlying randomness.

I've very quickly eyeballed IndigoWalker's GitHub above and would say you might find it worth a look. Just remember to consider it all as a source of inspiration to experiment with rather than a template for "if I put in [x] then I will always get [y]", as that's not how Suno works.

[neat nexus]

Suno is more akin to alchemy than it is to a wordprocessor for sound. And that's a good thing.

[sharp grotto]

You really have to restrict its behaviour, in case running wild

[quiet solar]

When you say it "focuses on names too much", what do you actually mean?

Speaking more generally, I and others have tried to give pretty comprehensive suggestions since you came in for how you can prompt for a female singer and avoid accidentally prompting for a male singer at the same time - this isn't something many other users seem to have run into trouble with. I'm happy to help, but want to understand which of the suggestions already made that you have or haven't tried.

[neat nexus]

One good way to work that is to specifically call the voice types in the style prompt or lyric prompt. Soprano, Mezzo-soprano, Alto (Contralto) ...

[crimson pine]

When I see (no styles) song on the front page, does that mean it's wiped from public eyes?
How do they do it and why?

[neat nexus]

IIRC that's something that can happen with Studio. Or a stem.

[crimson pine]

Oh I thought It's intentional

[neat nexus]

For some people it might be.

[surreal lava]

AI can only do so much and even creating a prompt for AI you need a better understanding. I like to know how to works to make the most of my work

[quiet solar]

Honestly, the best answer I can give is to say that all of us here have got there through experimenting ourselves to understand what is and isn't possible. And that can change from one model to another, and each new model tends to come with its own unique quirks to factor in. And it can vary by genre as well, just to keep you on your toes.

If you want some really general advice on prompting, I'd give you the following:

• Consider all prompts in the style and lyrics box as suggestions, not instructions. The more you include, the more likely Suno will disregard some - especially those nearer the end of your style prompt
• The more your prompts relate to musical terminology, the more likely Suno is going to be able to successfully look them up and interpret what you want
• If there's anything you want to call for specifically in lyrics prompts at a certain time, it helps if you've previously included it in the style prompt (e.g. multiple singers, specific instruments, etc.)
• There are no hard and fast rules. Experimentation is your friend.

It's also not just about prompting. For example, creating duets in Suno feels like something that can be done reliably through prompting, but people get frustrated if they try to specify exact prompts that Suno must do at specific times, which doesn't play nicely with how the prompting system works in practice (there's a chance any prompt will be taken into account, but it's not guaranteed).

However, if you take a step back and consider Suno as a toolbox in which prompts are only one tool, there are a ton of potential workarounds you can do to create an effective duet using editing (whether Suno Studio, or an external DAW/Editor) despite limitations in prompting.

If there's anything specific you'd like to know how to do, I'm happy to try to help. But trying to explain everything I've picked up about prompting in a year and a half of using Suno isn't going to fit easily into a single Discord post. 😉

[quiet solar]

You can do it by, for example, doing a Cover and wiping the style prompt. If you set audio influence high and everything else low, it works as a sort-of alternative method of Remastering but with more underlying creativity.

There might be a way to do it via Studio as well. As to why people do it, there could be a variety of reasons...

[south yoke]

I tried all of them, 0️⃣ benefits. If I add names it stops to make duets with one female and one male but it starts to focus on these names too much, without names I got female + male duets again.

[quiet solar]

Could you share your updated style prompt, because this is a really odd situation you're running into.

[south yoke]

Lyrics or description?

[surreal lava]

You cant name most artists and bands even when you are mimicking a style. You have to work around that.

[quiet solar]

And to confirm, you're not on a plan that lets you access the Vocal Gender setting?

[quiet solar]

Overall style prompt, not lyrics. I'm guessing your lyrics prompting only references the gender of the singer if you're calling them by name. If not, include that too.

[south yoke]

I didn't pay any money.

[quiet solar]

Yeah, that's fine. It would be an easy fix if you had it, but it's usually pretty easy to get Suno to get the right gender of singer without it. Which is why I'm surprised you're running into such difficulty.

@south yoke And also, just to confirm, you're doing a straight creation - not a cover, or using personas, or anything else in the + boxes:

[carmine gorge]

I had issues getting Duets also, I tried styles like Duet, Trio, Quartet, Band, Group, and some others. also tried things like in the lyrics putting [Verse 1 Jim] [Chorus Laura]

[quiet solar]

AFAIK, Sandy has the reverse problem - he's getting duets without trying to.

If you're looking for advice on how to prompt for duets, there's a LOT of previous posts in this chat covering the best practice - search for "duets" or "male and female". Just know it's not something you can reliably do with prompting alone, and most methods involve workarounds like multiple extensions, or editing in Studio or a DAW.

[south yoke]

I can't send it with one message, only several. Because lyrics are long.

[carmine gorge]

oh thats odd lol. exact opposite issue

[quiet solar]

Okay then, just the style prompt is fine.

I'm about 99% sure that's where the issue is likely to be anyway

[south yoke]

Celtic folk, medieval, hurdy-gurdy, pop, hip-hop crossover, female vocals only, powerful solo female voice, mystical, atmospheric, no male vocals.

[quiet solar]

Immediate advice: Take out "no male vocals". You can't prompt negatively in the style box.

[south yoke]

I tried that

[quiet solar]

I'd also suggest removing anything you have in Exclude Styles unless you're 100% sure you know what you're doing.

[south yoke]

What is "Exclude Styles"?

[quiet solar]

It's under Advanced Settings. If you can't access it, don't worry about it. 🙂

@south yoke Give me one sec, I'm just testing using an edited version of your prompt to try to replicate the issue

[frozen dagger]

Want to add my own personal finding, randomly. In my experience asking ChatGpt to create Suno Style prompts, anytime i told it to keep it at 1,000 characters or under, it's always too big. Now i ask GPT to give me 700 or under and when I paste it to Suno, Suno character count is around 900. This has been a huge time saver, realizing these counts are not equal.

[quiet solar]

@south yoke I ran two tests - one on v5, one on v4.5. Everything as default, no setting in the vocal gender. Worked like a charm. For example, this is 4.5 which should be the same model you're using:

https://suno.com/s/2Qdx0OvKFoITQMTs

The style prompt I used is a modified version of your own, based on the advice given previously about removing 'solo' and 'no male vocals':

"Celtic folk, medieval, hurdy-gurdy, pop, hip-hop crossover, powerful female vocals, mystical, atmospheric"

[frozen dagger]

4.5-all is the every1 model i think

[quiet solar]

Yeah, I used 4.5-all. The v5 test was accidental, as I had it set to that before I thought to change it. But same result. All four songs had a solo female singer.

[south yoke]

That's because all they 4 have very simple easy lyrics.

[quiet solar]

Alright, then you'll need to share your lyrics and split them over a couple of posts. But lyrics don't usually impact the gender of the singer unless you've set them up in such a way that you're trying to make them do so.

If your song is ridiculously long, feel free to DM me rather than drop them all in here if you prefer

[south yoke]

https://suno.com/s/n0W5GYkVk4exPcBY

[quiet solar]

I don't see anything being duetted, just the wrong gender of singer?

[south yoke]

Yeah, this guy is not a woman.

Here's gender problem + duet problem
https://suno.com/s/TKZlawr4HbuMMS3W

Two problems together.

[quiet solar]

Okay, let's focus on one at a time. xD

It looks like the problem you're running into is that hip-hop is such a male dominated genre in terms of artists that it's influencing the decision of singer. I had one version that switched genres on the new prompt for those sections, another (linked) that didn't. This is one of those cases where prompting can have an influence, but isn't the only factor going on:

https://suno.com/s/BQC2i8CzyGZUnsnZ

You can try tilting the odds in your favour by calling for a female singer in the lyric prompts for those sections, e.g. [Verse 2 - Hip-Hop Flow, female singer]?

[south yoke]

Also duet of a male and a female
https://suno.com/s/nDKDgvg8c7zCiFj7

[south yoke]

Every genre is dominated by females or males.

[quiet solar]

Some more so than others. The point being that Suno references previous music from the genres it's learned to understand the factors that make up that genre. If you reference a genre where male singers are dominant, it's likely to weight having a male singer more highly, even if you specifically prompt for a female one.

In addition to adding reminders in the lyric prompts that you want a female singer, you could also try building the song up using multiple extensions. If you know a genre is likely to be problematic, Extend an original song you're happy with from just before the male singer kicks in. Repeat as necessary until you get to the end of the song, then combine them together using "Get Full Song". Optionally remaster the whole thing if it's a bit clunky switching between extension points.

---

The Ballad of the Benchmaxxed Machine also has a bit more going on in the lyric as " * () and ALL CAPS all influence the singer to some extent, and can suggest to Suno that a second singer should be coming in. If this is a problem you're running into a lot, try limiting your use of those except where necessary - especially () if you're trying to avoid introducing backing singers.

[shell loom]

speed mechanics hopefully.
Gallop
canter
Trot
walk
https://suno.com/s/btpT3pdAoO1kN4GF

hey we are getting horsey here:)

[south yoke]

I'm running away from a crowd of singers but they are running for me🤣

[quiet solar]

Your second issue of duets was confusing us - am I right in saying what you're trying to do is have a single singer, with no harmonies at all?

If so, that's a bit more tricky. Harmonies are common in a lot of genres and will tend to creep in. There isn't a straightforward way to prompt for them never happening under any circumstances. Again, I'd suggest the same workaround as before - expect it's a problem, so use Extend before any harmony kicks in and try to get a version where Suno doesn't harmonise. Then use that as the basis to continue the song, and Extend again as needed until you get to the end.

As a general rule, the more you want Suno to specifically do or not do a specific thing at a specific point in the song, the harder it's going to be to get a version that works as you envisaged, and the more attempts and credits you're going to have to use if you want to generate it naturally - especially if you want to generate the whole song on a single click, and not use workarounds like multiple Extensions.

And it's even harder if you're using the free version, as you don't have some of the tools that open up other options to avoid problems like this, or fix them when they occur.

[south yoke]

I don't need situations where extra singer signs big part of a song.

[quiet solar]

I appreciate that. What I'm trying to tell you is that by switching genres and, in some cases, using style marks in the lyrics you're making it more likely that it happens. And you don't have access to the Vocal Gender setting which would help avoid it, so you need to bear that in mind when creating.

What you're trying to do is a bit more complex than a typical Suno user, and now I understand that better I can see why you're hitting some headaches. Resolving them is going to take some patience and a lot of repeat attempts. It's not just a prompting issue - you're running up against issues with how Suno generates songs and handles genre switching mid-song.

---

There's a very easy answer if you're willing to take a step back and say "Okay, FINE, I guess we'll have a male singer too" and compromise with the machine.

If not, I've given you suggestions for prompts and a workflow using Extends that should help you get around it - but it is going to take time and effort to get a final version of the song in the way you want it.

Anything where you need Suno to specifically do something (or not do something) always does, because it stops being just about prompting and becomes a question of working with various tools inside (and sometimes outside) of Suno to achieve what you want.

[proud nova]

how does suno understand promts to change vocals, for instance, i want 1 line to be female, 1 line to be an announcer styled voice and then to go back to the main voice, which is neither of the first 2 voices mentioned. however, suno cant seem to differentiate brackets very well and chooses where it wants to change the vocals.

[frozen dagger]

Wait for v6 or beyond, as it stand not a reliable way other than doing multiple gens as Covers with seperate vocal personas, or same idea using extend with different personas.
Try searching this thread (CTRL +F) for key terms related to multiple voices , duets, call and respose, harmony, etc and can find lots of suggestions and things people have tried using prompting, with some success, but nothing is 100% with text alone.

[quartz tiger]

hi i am very new to Suno, i am trying to make an Avali (vr chat species) themed Irish pup song. where one verse is sung by the Patron and one is sung by the Bartender, but no matter now often i reroll i cant get it where the right singer sings the right part the best i have gotten is one messup, after about 30 rerolls, ive tried with the help of AI to edit my prompt, (because my editing did not do anything) so.. uh, im kinda lost now that its still not working.

[frozen dagger]

see msg i sent above yours. same thing

[solemn lintel]

After the new persona update, I'm still having a really hard time getting metal to sound any good. No matter what I try, it just comes out sounding manic. Really fast, random notes, just not the "melodic metal" I keep telling it to (and which always worked before). Does anyone have any suggestions on how to prompt Suno to do better?

[carmine gorge]

trash bot it took me hours to write that

sorry for the multi message spam but the bot deleted it because "2 lines of space" so

[frozen dagger]

been there

[carmine gorge]

I'm trying to get a voice that sounds like Corpse Husband. iv tried a couple prompts. gonna take me a sec to find the 2nd one. the first one is this

````Dark trap, Heavy distorted bass, Phonk, Industrial Bass, rhythmic spoken word, heavy distortion, Peter Steele influence, Type O Negative style vocals, no high frequencies, Male deep‑bass, male sub‑bass resonance, male low‑register rumble, male gravel‑grain, male harsh‑rasp, male throat‑dominant resonance, male chest‑resonant pressure, male smoky‑dark timbre, male breath‑grit undertone, male compressed delivery, male close‑mic presence, male whisper‑weight projection, male slow‑drag pacing, male controlled growl edge, male sandpaper consonants, male subharmonic rumble, male shadow‑warm coloration, male gritty‑breathy fusion, male low‑frequency saturation, male nocturnal tone, male calm‑menace aura, male intimate‑threat softness, male low‑register dominance, male textured‑air friction, male heavy‑grain articulation```

Dark trap, Heavy distorted bass, Phonk, Industrial Bass, rhythmic spoken word, heavy distortion, Peter Steele influence, Type O Negative style vocals, clear but deep, no high frequencies, His voice hits like a demon speaking through a human chest, impossibly low and vibrating with a sub‑bass rumble that feels physical, Every word carries a rough, torn grain, like his vocal cords scrape together instead of sliding clean, He speaks with a slow, heavy drag, letting syllables sink rather than snap, There’s a breathy leak under the grit, giving his voice a smoky, intimate closeness, like he’s right against the mic, His consonants have a gravel‑scrape edge, and his vowels sit in a dark, throat‑compressed resonance that never brightens, The overall effect is a controlled growl, calm but menacing, sounding half‑whispered yet weighted, like something unnatural is pulling his voice downward, Bass-baritone male vocals, Extreme vocal fry, Gravelly raspy voice, Deep resonant whispers

[carmine gorge]

its an extreamly annoying rule that makes 0 sense. theres not even a message saying we cant use 2 lines of spaces. so people could spend idk how much time typing only to have it go poof the moment they hit send.

[frozen dagger]

woudl make gens with that prompt til you get somethign with cadence and play with transpose in studio maybe to chgange pitch and speed

[carmine gorge]

iv tried but it always comes out like an old blues singer or a cowboy

[frozen dagger]

its finding a needle in a haystack of all it's training, for sure.

[quiet solar]

Long story short, the problem you have can't usually be solved by prompting along. If you've got them switching over in roughly the right places, it sounds like your prompting is fine - you need to use other approaches to get it over the line.

Go with the one that only had one mistake (or had the mistake as far into the song as possible), Extend just before the mistake to 'bank' the bit that works and see if Suno gets the rest right. If not, Extend before the next mistake in the new extension (etc) and repeat until you get to the end of the song.

Then use Get Full Song on the final extension, which will make a new final song connecting all the pieces together. Optionally, Remaster that final song to smooth over any differences in sound between the different connected extensions.

[quiet solar]

Rather than "his voice hits like a demon speaking through a human chest", which Suno is probably going to struggle to look up in a music-based library, try using something like "Deep bass-baritone male, angry, distorted"

I see you've used something similar further in the prompt, but bear in mind bits near the end of incredibly long prompts are more likely to be skipped. Personally, I'd remove almost all of the vocal descriptions and try something much simpler and aim for a shorter prompt overall. Adding "growling, unnatural, gravelly" might be enough on their own as additional descriptors.

[polar spade]

Assuming you have Pro:

If you have one song with only one error, the the Edit UI is the thing to use. Others mention issues but I edited two songs total of about 30 times with Edit and Legacy Edit the past two days.

Its a bit more intricate, intermediate level, as there needs to be some practice with aligning lyrics. But having done that, its about highlighting as precise possible the line needing one voice or the other and add the specific lyrics and tags you wish.

Further, if you got this one section right in one song and other parts right in other songs, there is patching all of the correctly made ones together into one, in Studio.

The original question may be, "whats going wrong and how to fix it", with the answer often as, nothing went wrong, so moving to the next step and editing specificity areas by replacing or regenerating sections, or just patching bits together is perfectly normal and expected.

[frozen dagger]

asked gpt to tell me what i am trying to create here, anyone have ideas¿
had some cool stuff come out. for sure. but at a point where i think i should open it up to outside my own brain and gpt back and forth stuff. internal loop. thoughts¿
You’re creating a high-control style prompt for an aleatoric, through-composed dark trap–metal track that:

evolves continuously (no repeating chorus),

undulates in energy like ocean waves,

fuses hip-hop, trap, metal, glitchcore, and Caribbean/dub influences,

emphasizes cadence shifts, sectional mutation, and rotating instrumentation,

uses aquatic, echo-driven textures and heavy breakdowns to feel fluid yet crushing.

Short version:

A long-form, through-composed dark trap metal prompt built around wave-like energy, glitch mutation, and Caribbean aquatic texture, prioritizing motion and transformation over repetition.

[frozen dagger]

funny thing. my baby momma just texted me and said she thought one of my Suno generations I was listening to in the shower was Corpse. She has no idea about this discord convo, she at work. I was unfamiliar i still doubting this coincidence is real. lmao. but yeah if i can fifure out which gen she talking about i can share.
kinda makes sense song is titled "Octo-Corpus" and full of corpse metaphors etc. similar shtuff.

[carmine gorge]

Thats actually pretty funny lol. I finally managed to get a few decent sounding ones, trying to figure the best to use for a persona. still not perfect though. if you do find it feel free to DM it to me.

[sharp kettle]

when prompting ai image generator advanced prompting works better you will use less creds then basic.
also needs lest trys as long as you don't typo.
prompting advice 101

[pure panther]

Is there a solid prompt or set of prompts that help keep the vocals above the instrumentals? (louder, not drowned out or muddied) I've tried vocal forward and (noisy, over produced, cacophony,) in the negatives with some success, but it's not great

[frozen dagger]

my suggestion would to try the remaster function

[pure panther]

I use that, but only after I remaster manually myself. I'm trying to get better initial output.

[frozen dagger]

im not savvy enough to know what to put, but i know that it would look something like naming the Hz and BPM and any other production related information that is stating how the sound is quantified and measured -db etc

naming your audacity settings etc

[lapis sable]

Hi Guys! I ran into a problem recently, I tried to make a duet, with a male and a female voice, but I think I mess up something because it doesn't keep the roles. Singing into each other's verses etc.

Can someone help me with that?

[pure panther]

I did one recently. I made an all male version, then an all female version. Then I spliced them together in Audacity. Then I reimported the audio and used that as a cover. There were a few other little fixes, but that's the gist of it.
https://suno.com/s/nycUv4wCc6PEoo5p

[lone owl]

🤔