📚┃suno-school

[broken dew]

Are Stems saved somewhere or do we have to download them then upload them back

[half dragon]

Click Stems from the menu again > See Extracted Stems

[broken dew]

I only see options to download, no options for 'create'

[half dragon]

Are you trying to create from the Stems?
Then go to your Library - create from there

[broken dew]

perfect! thank you

[half dragon]

Turn off the filter to show Stems

[sturdy dove]

What's the 8-step process? Is there a workflow that shows it?

[mint marsh]

anyone been able to create college band style songs

https://www.youtube.com/watch?v=76aEIQSMG04&pp=ygUVaSBuZWVkZWQgc29tZSBoZWxwIGVh0gcJCccJAYcqIYzv like this

[half dragon]

This one sounds like it. See if you can try the same style prompt.
https://suno.com/s/uEFnOlzuaYdVhB2J

[mint marsh]

holy

[fluid sentinel]

Got a problem with replacing lyrics in the Song Editor... The lyrics change but the melody changes to match a completely different section of the song that wasn't selected.

[polar spade]

You need to practice a couple of times using the aligning tool. The lyrics should be aligned to follow the audio exactly before replacing any lyrics. Or as you say, you may select some lines and type new lyrics , but the audio is really 40 seconds previous that you will actually be replacing.

[fluid sentinel]

I've been using phonetic syllables and it has trouble following them.

[fluid sentinel]

What's the reason behind not being able to create a Persona from a "cover" of an audio upload?

[winged peak]

It may still be too close to the audio upload, and the upload may contain copyrighted material. 🙂

[timber glen]

Still having problems in video downloading?

[fluid sentinel]

The upload is my original music that I recorded

[brisk flint]

And how is AI supposed to know that? It has to be restrictive so that more disgruntled lawyers don't try to squeeze money out of Suno.

[fluid sentinel]

You said the upload may contain copyrighted material. I was explaining that it does not.,

[brisk flint]

You wanted to know why it doesn't work – I explained why Suno has these restrictions, so there's nothing more to complain about or discuss 😉

[fluid sentinel]

Calm down.

[charred token]

This question has probably been asked umpteen hundred times, already, but here goes?

I have gotten a melody and arrangement that is wonderful! However, the lyrics needs some work. Is there a way where I can change they lyrics slightly, and have SUNO redo it with the changed lyrics WITHOUT changing the arrangment?

[brisk flint]

Yes, we discussed this topic just yesterday, and the disappointing answer is: probably not.
You can get pretty close to the original song with the cover version, the exact style prompt and the “audio influencer” slider set to 100%, but just like with the method of creating a playlist with ONLY that one song and selecting the playlist as “inspiration”, setting the audio influencer slider to 100% will cause Suno to change the arrangement, the instruments and even the melody to a more or less small extent. The only viable and time-consuming method currently seems to be to extract stems, create different versions of the songs, and then reassemble the instrumental and vocal stems in a DAW or audio editor.

[charred token]

So, I'm better off getting the stems and singing it myself, which I was going to do, anyway. Thanks

[polar spade]

There's nothing wrong with 1) using Replace and replacing words or phrases one at a time.

2. if so much of the lyrics are that bad, then consider correcting everything (starting over) and running about 30 generations until something new comes up that sounds just as good.

3. and if you're singing yourself, then record a worktape of you singing to the stem, then upload that audio and cover it to take you near back to where you are now, but with corrected lyrics.

[tender eagle]

okay someone explain how it could use my lyric prompt... and make album art, for this song... and make it correctly, when that lyric prompt is... insane for an image ai to recognize lol 😂 http://suno.com/song/ca1280ad-3693-4188-9e4a-7dcad6ae310a

it's like 10 languages, making one phonetic lyric

that's some sneaky ai stuff there..

surprised an image ai picked up the idea of quick brown fox jumps over the lazy dog from such a hashed together pangram

[pseudo blade]

Yes I can help you

[graceful condor]

Error: Tags contained artist name: pared. I'm using Inpso my tags are all the same just different progression (i.e. intro chord progression Amin11 - F#min11 - Amin7 - A7 - Dmaj9 - Gmaj - Cmaj9 - Gmaj9) no artist used at all.

[half dragon]

Check for that word "pared" anywhere in your prompt - and change it.

[graceful condor]

No where its no where..?

[tender eagle]

hahah what?

[copper cedar]

Fix the damn video generator already. Its 2 days now

[hard beacon]

i love these nice speaking messages 😄

uhm.. sorry... these damn nice speaking messages 😉

[dawn laurel]

generated videos stopped downloading

[hard beacon]

holy f***... that toolbar on the top 😄 😄

looks like there a some problems with videos at the moment..

[half dragon]

Select one and hold shift then select more, then 3 dot menu > Move to trash

[dawn laurel]

Before this, they could be downloaded by making the video public and using video regeneration

[little cargo]

its been 4 days and i cannot download my most recent songs? Its all grayed out... hey @wet minnow any ideas why my own songs wont let me download them?

[hard beacon]

hmm @half dragon .. i am not a computer newbie, but that doesn't work for me...

ah... with CTRL it works 😄 nevermind.. thanks for the response...

[half dragon]

Shift to select the range - Ctrl to select each

[hard beacon]

well.. mainthing is, that i know how to do it 🙂 Thx..

if i now can get a "remaster 4.5+" function i am happy at this point 😄

[pseudo blade]

it will be soon™

[hard beacon]

hope so 🙂

[neat nexus]

@heavy fiber

From Gemni... Phonetic Breakdown for Suno Lyrics:

Here are the sounds broken down into "singable" phonetic syllables. Use hyphens to separate syllables and control the rhythm. Use capitalization for emphasis. Suno also recognizes tags like [scream], [growl], [laugh], etc., which we will use.

1. The Grunts (Guttural Lows / Death Metal Vocals)
   These are your bread-and-butter for heavy verses and breakdowns.

Standard Grunt: Grr-uhnt

Deep, Forceful Grunt: Hrr-UUUN-gh

Rhythmic Grunts (like a chugging riff): Urgh... urgh... urgh... urgh

Wet, Guttural Grunt: Guh-ruh-nkk

Low Growl: [growl] Grr-rrr-rrr-uhnt

2. The Squeals (High-Pitched Screams / Black Metal Vocals)
   These are perfect for choruses, dramatic moments, and high-scream accents.

Long, Held Squeal: Squee-EEEEEEEE (The more E's, the longer Suno will try to hold it)

Short, Panicked Squeal: REEE-eek

Staccato Squeals: Ree-Ree-Ree-Ree

Excited Yelp/Squeal: Wee-yuh! Wee-yuh!

High-Pitched Scream: [scream] EEEEE-yah-OYNK

3. The Snorts & Huffs (Percussive / Rhythmic Elements)
   Use these like you would use cymbal hits or snare drums to add percussive texture.

Sharp Snort: Sn-AWR-t

Aggressive Huff: Hn-UFF!

Breathy, Rhythmic Chuffing: Ch-uff... ch-uff... ch-uff-ch-uff

Wet, Forceful Snort: Fff-nrt or Sh-nort

4. The Oinks (The "Hooks" / Catchy Chorus Vocals)
   This is the most recognizable sound. You can make it catchy or aggressive.

Classic Oink: Oynk or Oynk-Oynk

Slightly Drawn-Out Oink: Oy-n-kuh

Aggressive Oink: OYNK!

Growled Oink: OYNK-grr

Questioning Oink: oink? (Suno sometimes picks up on punctuation for inflection)

5. The Snuffles (Atmospheric / Intro & Outro Sounds)
   These are great for building atmosphere in an intro, a quiet bridge, or fading out.

Soft Rooting Sound: Snuh-fful... nuhf-nuhf

Wet Sniffing: Sniff-uhl... sniff-uhl

Quiet Murmuring: mmm-grr-uhnt... mmm-nuhf

6. The Boar Roars (The Breakdown / Power Vocals)
   This is for the heaviest part of your song. The Mosh Pit call.

Deep, Powerful Roar: [growl] Grr-RRR-AW-gh

Roar leading into a grunt: RRROOO-AR-uhnt
Aggressive Challenge: R-AW-nkk-Gahhh

[polar spade]

Like Zoinks! Scooby!

[neat nexus]

It gets better...it definitely understood the assignment! 😅

[Removed links]

Here's two without Divorcecore as a cheat code:

[Removed links]

[polar spade]

Or an Educational British Kids Show. https://suno.com/s/IZM7HujcWTqvkyXH

[ebon sand]

They didn't respond to me what the actual requirement is, like if they are really looking for animal sounds and not, for goregrind or the *other *grind type it's mostly only v3.5 which can still do pig squeal bree bree tupa tupa grind, i usually just type [goregrind noises] as lyrics

[neat nexus]

Wait...only v3.5? But I've been assured that v3.5 is horrible! It's been nerfed!

lol Yeah there are a few features I go back to v3.5 for myself. I believe it.

[polar spade]

Anything with the words "compared" or "prepared"? Im just being silly, I guess. At least it didnt ping you for using Amin Bhatia's name: Amin11

[ebon sand]

i kind of understand why a model wouldn't be tuned to output tupa tupa grind but it's where the bree bree lives
I had some promising results with v4.5 a while ago, v4.5+ seems to be more tuned to screaming which steers it away from the bree bree you'd need for that music, i think if one had a persona with just one breeeeee it would solve everything haha

[neat nexus]

Probably. Nothing says you can't work 3.5 until you get what you need, persona or cover from there.

[left drift]

"Recompile" is another hard word for it to pronounce correctly lol... Just found that out.

[timid quest]

Inspo shouldn't straight up copy parts.

[polar spade]

I would agree. It does seem to be influenced heavily on one song (output in the playlist) than evenly across the Playlist. At the least, should randomize a variety of suggested inspirations, even if they sounds somewhat similar. As I see it, i assume it should ultimately, suggest a sound as if it was "... the next song idea on the album to be written....".

[austere dragon]

Anyone else getting duets on covers when the persona is male and the vocals are male? It's like happening 50% of the time. I even went back to putting [Male Vocals] under each heading... and it's just the second verse.

[polar spade]

Duet means two vocals. Suno may understand the general ideals of alternating, but even that, it can alternate the next male voice after a female voice in a different frequency than the "male" before. Technically, we can get trios and quarters often if we listen closely. So getting two female tones or two male tones isnt unusual.

.

[gloomy shore]

I need help creating an instrumental song. I tried prompting it as;
{Violin melody following the vocal line: "lyrics...."}. Is this correct? Also, should i use {} or [] ?

[half dragon]

Instruction meta tags should be in [..] if they were used in custom Lyric box
For style box - I don't think tags are needed.

Are you using "custom" mode or simple?

[gloomy shore]

Im using custom mode because i wanted the instrument to be playing similar to vocal was singing the lyrics.

[half dragon]

Yes - then put tags inside [...]

[gloomy shore]

It didn't work, still had vocals but there were 3 different vocals in the song 😂. V3.5 is so unpredictable.
In the custom mode there is a switch to turn on Instrumental but then the lyrics box would be removed and only left you with the Style box. Could you add lyrics in the Style box?

[half dragon]

Are you trying to create instrumental songs - but with lyrics? Sorry if I'm slow.
You can create instrumental song in custom mode and define the song structure within the Lyric box (just without actual lyric)

[azure crane]

Can no longer creat Persona’s. Why?

[half dragon]

Check these out
https://suno.com/s/2DornKIT3w9kfB6b
https://suno.com/s/FIsMVLeVguQ8gTSL

[pine tendon]

So many credits wasted. I am trying to create a power metal song with a female vocalist that does not sing into the stratosphere? With a lower voice. Anyone know how to reign in female vocals so they are lower in pitch and gritty?

[gloomy shore]

That's right. I'm trying to create a song with instrumental only; violin, but with lyrics. So that when it plays the lyric the sound would be similar to singing with vocal.

[half dragon]

Oh.. you are making instrumental from a song with vocal? existing song?
I have a Suno song I made - then I created an instrumental version of it by Cover it as Instrumental only.
I replaced all the lyric in the lyric box with [Melodic Instrumental] when I covered it but leave all other meta tags
https://suno.com/song/3e9c2520-adbd-473c-a78e-ce308aba5682

[gloomy shore]

Oh that's it. Viet music is great at portraying instrumental song that plays like vocal but without actual singing. Thanks

[gloomy shore]

This means i have to upload the actual song with vocal. Then in the lyric box removed all of the lyrics and replace it with tag, [Melodic Instrumental] ?

[half dragon]

That's how I'd do it or use the Suno song still on Suno.

[gloomy shore]

Oh, that means, i have to do the Remix but then replace the lyrics with instrumental.
But do you have to emphasize the instrument you want the ai to create? I want it to create just the violin only. Do i need to tag it like this? [Melodic violin instrumental]

[half dragon]

You can if you want to - I've never tried so not sure how reliable it is
Like this song did.
https://suno.com/song/492fc64a-bbcf-44ec-8d37-f69e806829a6?sh=FIsMVLeVguQ8gTSL

[grizzled ember]

My Ukrainian friend taught me about the Stradivari violin I'd luv to hear that (⁠＾⁠∇⁠＾⁠)⁠ﾉ⁠♪

[charred token]

Several of my earlier songs were split into two or three parts. Is there anyway to remaster those into a single take, other than editting it offline, uploading the edited version, then remastering it from there?

[wet minnow]

if you use the usual way of extending songs you click "get full song" on the last part. that makes a new song from all the previous parts you extended from.

[charred token]

Oooh! How does it know which ones, since it makes two versions, each time?

[half dragon]

It doesn't - you choose which Part 2 > Get Whole Song
It does know which song it's extended from.

[charred token]

Okay.

[polar spade]

If you have a part 3 or part 4 , you may need to click the latest one then look at the history flag just under your title and click to follow it back to the original.

[jaunty coral]

So I'm new to the whole workshop thing. I've chopped off the end of my song (where it just kept doing instrumentals/repeating verses) so that I could insert another chorus, but I can't seem to get it to transition seamlessly. Clicking "extend" only seems to replace the end of the song instead of truly extending it, and it also doesn't let me insert lyrics or repeat the chorus. Copying and pasting the chorus does just that, and doesn't integrate it into the song like it should. Any tips?

[near flume]

try this: Female vocalist with a rich alto range

[gloomy shore]

Well i hope you like my song. I made a few days ago. It has vocal and violin that played together at some part. Not what the full violin song you were hoping but hope it scratches that itch of violin needs 😀
https://suno.com/s/DmnCIGz5jFX2UAUp

[near flume]

Don't chop off the ending and try to extend from there, success is almost impossible. Extend from at least 6 seconds before you want the new chorus to appear. It can take many tries before you get a seamless extend.

[jaunty coral]

Well, to explain my predicament a little more, I generated the original audio several Suno versions ago. I only had a chorus (at the beginning, it jumps right in) and 2 verses written. Suno took the two verses and repeated them a few times to create the song. I remastered this version, and chopped off those extra verses to add a chorus after Verse 2. I can try and take your advice and see if I can get it to work, though. Thank you!

[pine tendon]

Tried your suggestion, one gen was eh..mostly in range but still had issues and the second one litterally starts out with a vocal "whoaa-oh-oh" spanning like 4 octaves.. LOL

[sturdy dove]

I have been censored. Fine. I'm leaving Suno forever because it has wasted copious amounts of my time and is simply no longer a useful tool. I've really given it every opportunity.

[young geode]

A power metal song that doesn't goto the stratosphere?

My brother, I don't think you can, not without heavy prompting and meta tags, and even then.

What you'd likely have to effectively do is get another genre's vocal style as the vocalist, while maintaining the power metal instruments.

So you'd likely have to do a Gen with the lyrics in another genre, like death Metal, Black Metal or mumble rap, ect: something where female vocalist don't lose their minds.

Then, you'd have to rip those vocals into a stem, then use that stem as the basis for a power metal song. Then if the song sound 'ok', cut that song and save, then turn that song into a persona.

Then cover that song, with the persona, heavy prompts and tags.

[polar spade]

Though I appreciated your commeraderie in your time here, I can certainly understand. Everyone should draw a line to keep focused on. That if we go below it, its time to cut bait and run until things improve. Whether stock market, a relationship, a career or even an online service. The return on investment line should be clearly drawn and stuck to so we dont drown in our expectations without much reward.

The simplest means is to drop to Basic to hold the account without making any songs if you wish, and keep up with Discord chats until you feel comfortable to veture back in or just move in another direction. Its up to you.

Good luck.

[sturdy dove]

@Davson, I think you're the only genuinely helpful person here.

I've tried for weeks to make a minor change. A minor change that was required because Suno simply chose to ignore my prompt, any way I put it.

I've tried the editor (garbage), I've tried extend, I've tried replace, I've tried fresh generations. It's simply incapable of the simplest tasks.

If you want to make music from scratch and you don't care how it turns out, I think it's great fun. But if you have any specific direction whatsoever, it freaks out and generates gibberish and garbage. It's wild. It's so close. But no.

I have really appreciated your help and direction. It's too bad it's not a better tool. It has caused me MASSIVE frustration for weeks. Enough days ruined by Suno.

Thanks again for trying to help!

[young geode]

How have you been prompting and Metataging?

Can you give me an example of the prompt layout/terms you're using? Are you trying to ask for incredibly specific sounds; blend particular genres; fix audio errors, ect?

[sturdy dove]

Bro, you're a few weeks late to the party. I'm done. Thanks tho.

[polar spade]

If you do cancel back for awhile, your negative experience can help others through that they might succeed. So always good to stick around and give some hints and clues. As always, they might just do Christian Pop or some genre that bodes well with Suno. Its not unfathomable that Suno gets picky with genres, or computer setups, or types of phones we use, or our internet service, among many others.

And your hints can help them through -- even if it hasn't done well with you so far. Thats totally up to you.

[polar spade]

Again. Youre health, in any form that is, is more important than any struggle that just gets too frustrating over time. At times, I get peeved with Suno and just stop and go live my life... suddenly , cutting the grass or working out makes me feel better. 😊

[young geode]

I ask, not just to help you, but to collect information. I put together guides and do testing, so I can help users fix their issues.

But if you dont wish to, that's fine. It's no problem

[polar spade]

Thats a total champion. Need more like this guy ☝️

[half dragon]

..and this guy ☝️

[young geode]

@polar spade
Is it possible to remove a single stem from a song, IE mute it, but keep the rest? Or do you just go into a DAW and restitch everything, minus the undesirable stem?

[graceful condor]

I wonder why is rock always raspy? I mean not every rock singer sounds like someone who ate sandpaper for their breakfast. Even if I add all non raspy in my prompt like clean vocals, soft etc. It still comes out raspy most of the time. I mean in some cases it is nice but not all the time. Powerfull does not mean raspy. Also last tones of lines are dragged out in an endless long tone so many times which is okay one time but not over and over again.

[young geode]

Because a lot of rock is raspy and Rock's sounds/vocalists have a lot fo crossover with Metal, which is gritty and raspy as a default.

Clean, soft, ect, are not enough to ensure that kind of vocal quality: you need to target genres, vocal ranges and tempos that contain/are associated with the vocal qualities you are looking for, while also removing/defocusing SUNO from the areas that have it.

A lot of this is a failure of proper prompting, genre knowledge and metatagging your lyrics.

Do you know what you are speifically looking for/wishing to generate?

[graceful condor]

Okay start with this. I do not like if a powerfull note at the end ist dragged in a monotone long tone. It is okay if it is a long sound but with runs in it or so. This is why I added this into the negative prompt section "Avoid excessive sustained notes at the end of phrases. No long note holds. Keep vocals tight, natural phrasing, emotional but restrained endings." but it it not working all the time. And it seems that if a lyric has a "!" in it at the end it animtes the ai to shout, scream or drag a sound long or if it has "..." in it too sometimes.

[young geode]

Those kinds of prompt commands are not typically effective. You need to use terms that have musical weight, such as 'Vocals: Sustained notes'. Additionally, if you put 'Avoid Excessive sustained notes' within the Exclude prompt, you're actually creating a double negative: SUNO will see it as "The User wants to be to exclude the avoidance of sustained notes" making it more likely that it will actually occur.

Give me a link to your song, or give a screen shot of what you're using. Ill take a look

[graceful condor]

So using just "Vocals: Sustained notes" is enough? - this part of the lyric section or exclude prompt?

Overall I am looking for a more melancholic, sad, dark moody progressive rock type of sound. More like a voice between White Snake "Is this Love" and Pain of Salvation. Like those songs for example. Daniel the vocalist has also raspier parts but gently used and not all the time. I want it to be used with care not like a 100% formula for each powerful part of the song. Here is the song for reference: https://www.youtube.com/watch?v=O7AH17jRciU&list=PLX1EE_8ktx_gHT5Is7we6tNuXdN9xXH7B&index=13 and I am doing a lot of Jrock and this would be a good reference too: https://www.youtube.com/watch?v=000ZZqnel6E&list=RD000ZZqnel6E&start_radio=1

My current prompt in this case is this:

Create a dark, emotionally Japanese Jrock song
The focus is on clean Japanese male vocals, mid-tempo groove, and emotional melodic power.
Drums should follow a slow, groovy pace – no fast double kick. Think heavy half-time feel
Guitars should blend heavy, low-tuned riffs (Drop C/B) with clean lead lines and ambient layers.
Vocals are expressive, sometimes close to spoken, and soar emotionally in the chorus
The song should feel melancholic, emotive, and restrained, with dynamic tension and a cinematic atmosphere – evoking the emotional build-up and groove-driven flow.

Because another issue with rock is the drums too. They tend to do fast double kick even if you do not asked for it.

I just noticed the "soar" part might be an issue too.

[young geode]

Vocal: Male baritone-tenor with warm, dark timbre; moderate vibrato; subtle lisp for emotional rawness; dynamic and expressive delivery; mix of legato and sharp phrasing; introspective, melancholic style.

Exclude: Female vocals, Sustained vocal notes, harsh growling, screaming, overly bright or thin tone, excessive melisma, detached or monotone delivery, falsetto-heavy singing, overly aggressive vocals, raw vocal styles.

Your prompts should look something that like.

but the only thing you need to do is target genres have that this kind of quality within them. For example, Thrash has that raspy quality, as does Emo, Punk, Grunge, ect. So you would want to exclude: Genre: Metal, where as Contemporary rock (Post-Punk, Nu-Rawk, ect) doesn't have these raspy qualities.

So your prompt should look like something of the following:
Style
2000s Era, Genres: Contemporary Rock, Jrock. Subgenres: Post-Punk, Nu-Raw. Instruments: Jrock x/y/z. Vocal: Nu-Rawk, Male baritone-tenor with warm, dark timbre; moderate vibrato; subtle lisp for emotional rawness; dynamic and expressive delivery; mix of legato and sharp phrasing; introspective, melancholic style

Exclude:
Genres: Metal, Industrial, Punk, New Wave, Punk Hardcore Vocals: Female vocals, Sustained vocal notes, harsh growling, screaming, overly bright or thin tone, excessive melisma, detached or monotone delivery, falsetto-heavy singing, overly aggressive vocals, raw vocal styles.

Theres more to it, but these are the basics if you're trying to blend, ect.

[graceful condor]

So Exclude is always important to add? And also in such details? Thank you for taking your time. So Post-Rock and etheral sound would help too right? New-Rawk is a real term it recognizes?

Oh btw I noticed that for around 2 days I can't unfold my old prompts anymore in my section. Is there a bug? Because before you could open the whole prompt you used for song xy and copy the whole prompt if you wanna reuse it again. But now it only displays a small section and that's it.

[young geode]

It's best practices to use exclusion, as it helps SUNO focus.

As for the Nu-Rawk, whatever is the most popular from of that genre is the one that you should use. That particular subgenre is categorized with 2000s Garage/Post Punk Revival/Nu-Rawk.

And I know nothing about Unfolding prompts.

[graceful condor]

Oh and about the lyrics section. if there is a Verse, Chorus etc. What is best to use for Suno: Verse: (Verse:), [Verse:], 【Verse 2】or are all of them working to be recognized by Suno. And I noticed it helped in some cases but is it generally working if I add something to the lyrics like "(Verse 1: soft start, words spoken).

[graceful condor]

oh and how do you handle Weirdness / Style Influence and Audio Influence? Especially if I use an Audio as reference. So far I put Weird at 0, Style at 80%-ish and Audio at 100%.

[young geode]

Whenever you are starting a project, you should start at the base 50/50(25), and then work from there. Naming it Song, 1st Gen, 1st P, 1L (50/50), and generate 2-6 songs. That'll give you an idea of what is going to consistently show up.

From there you play with the weights: Weirdness is something like a bandwidth that SUNO has to pull from, the lower it is, the more it's cut off and the higher it is, the more room it has to play; Style Influence is akin to the CFG scale you find in Stable Diffusion, the higher it is, the more SUNO will be forced to follow the prompt and go in the direction you're asking, while the lower it is, the looser it'll treat your prompts. Audio is just how much the Audio is going to play a role in forming the generation.

So if you're finding that a song is giving you the sounds you're looking for, but is missing something from the prompt, you would increase the Influence, meanwhile if you want the song to stick to the script of the genres, you're reduce the weirdness and vice versa.

[graceful condor]

Thank you for all the help. What I really find weird too is that a lot of rock songs with a lot of instruments start very damp and dull at first but then when the vocalist starts to sing it becomes clear. Is this an issue of the AI? Only with slow songs or songs without the use of too many instruments the song sounds crystal clear right from the start.

But also the extend thing is an issue here. First and foremost to even have the need to use it is annoying because all songs are cut off around 4:50 - 5:10 even if I am allowed to create 8 min songs which never happened so far to work. I wonder why that is and extend sometimes gets a totally different tone. What I mean is while the song before sounded more dull the extended version all of a sudden is crystal clear. Wish is just weird.

[young geode]

That's more an issue with prompting, lyrics and metatags. What's occurring is that SUNO is running (likely) it's diffusion protocols and it's trying to flesh out everything. But since Rock and Metal are incredibly broad, use the exact same instruments, rhythms, beats, chords, ect, ect, what really differentiates them (musically speaking) are their tempos, vocals and mixes. So it's likely that once your vocalist enters, SUNO immediately tunes out all non-like genres/sounds, because you don't typically see crunchy chords getting pounded out over high tenor vocals going hard.

If you were more specific and gave more guidance, it would clear up.

And, again, the track times are also an issue of prompting, lyrics and metatags. I regularly generate songs that hit the 8 minute mark and still aren't done, across a myriad of genres.

A lot of what SUNO puts out is a direct result of what is put in (I'm actually writing a guide on this, at this very moment). If you flesh out your prompts to be more specific, make sure your lyrics are properly formated (IE, Good Structure, good rhyming schema, good meterlines, metatags properly folded in, ect), and they are long enough, then you shouldn't be getting 4-5 minute songs, unless the tempos of those genres call for it.

[graceful condor]

Is there any site or video which explains such thins in detail. Because honestly with this last answer I am not 1% wiser. So how do I guarantee longer songs? I already added stuff like "Do not cut off songs. Or song length at least 6-7 minutes" stuff like that but it did not work. So how do I know what works and how it is build together? Is there any Guide? (besides yours which is in the making).

Und about the sound too. I am not really wiser now. So it is a prompt issue that the sound at the start of the song sounds mushy and dull? For example this is my prompt right now:

A hauntingly beautiful Neo-Japanesque rock ballad with emotional depth, blending traditional Japanese melodies with modern alternative rock guitars, melodic bass lines and epic guitar solo featuring the main melody plus the use of epic orchestral highlights. The atmosphere is elegant, melancholic, and spiritual, evoking misty forests, ancient shrines, and forgotten myths. Features a male vocalist with a clear, expressive mid-range voice, singing in a soft, poetic tone, reminiscent of classical Japanese enunciation — graceful, sorrowful, and mysterious. The arrangement includes layers of clean arpeggiated guitars, occasional distorted riffs, and ceremonial percussion, like temple bells and chimes, creating a fusion of the sacred and the modern. Mood: nostalgic, folkloric, romantic, otherworldly and softly passionate.

Why is it at start dull? Then clear when the vocalist sets in.

And one more thing. What is the deal with pure instrumental songs? I uploaded a song of mine which clearly had a start. I wanted to have an upgraded version of it because the song itself was just chords and guitar solo. But all files I created the whole songs just started somewhere from the middle and this not even with a full tone but in the middle of a tone. Like if you cut out a snippet from a song in a real bad way. I put Audio at 100% and still that nonsense happened.

[young geode]

https://discord.com/channels/1069381916492562582/1382019029266600067

Thats where I post my guides. They should help you a lot. Unfortunately, there isn't some one stop-shop online, be a site or a channel, that will spell it all out for you.

Most AI 'creators' are heavily reliant on AI itself, and tend to want the AI to do the work for them, instead of doing the work that's necessary to understand why the AI does what it does.

The lack of resources on this is why I started writing these guides and paying close attention to everything I do, so I can then take it and put it into something usable for others.

---

Your prompt looks fine (it's not as information dense as I would make it, but it's fine).

The only thing it's missing, in my view, is a 'Narrative' section, IE: The song starts out with X, then drops into a Y on the first verse, and becomes moody during the chorus, then muted power chords during the Bridge. That Helps SUNO format the song, to a degree.

What's going on, with that dull sound, is SUNO has yet to choose a style it's going to follow. It's basically just generating the basics of the song, because it has no direction from it's token prediction: "Should I go west, north, south or east?" But the moment the vocals come in it goes "Ah, I should go west." and it's sound clears up.

What you can do is format your lyrics with metatags. Say you want a solo intro:

[Instrument: Electric Guitar, Yamaha Revstar RS502, Hirajōshi scale]
[Intro: Soft intro]
[Melody: Guitar, Hirajōshi scale, Descending Hammer-ons]
[Melody Develops]
[Melody Continues]
--------
[Instrument: Drums, Kick drum, Toms, Hi-hats]
[Drum Fill: Multi-Drum Pattern, Floor Tom heavy]

[Drop]
[Verse]```

That helps give guidance to SUNO, well before the verse arrives.

Also, I havent done a whole lot fo testing with anything than uploaded, other than Metronomes when it comes to Audio, so I unfortunately am not the user to help you with that.

That's probably more @polar spade 's area, he might be able to help

[graceful condor]

Thank you a lot and generally about the cut off thing. How do I avoid that. Because even if the lyrics are still not at their end or the uploaded song is longer than 5 mins it cuts it off. So how to prevend that?

[young geode]

Right now there is an issue with audio/persona gens cutting off. If you are using those, that's likely why. I have seen this happen immediately to my songs as well the moment I have an audio file interacting with the song.

If you remove the audio file, and have very fleshed out prompts and metatags, with decently filled lyrics (6-8 lines per section; a Lot of Sections) then you can hit that 8 minute without an issue.

https://suno.com/playlist/59738ccf-918b-41ee-aae5-db09a88e85b0

That whole playlist of examples are all off effectively the same lyric template, yet many of them are 7+ minutes. It's just a matter of lyrical fill, genre selection and guidance.

[graceful condor]

So it is more an audio file issue, I see. Man I did not know that I have to be this exact with he prompts like literally time stamps. But is this even possible to fill all that in with just 1000 signs? So far Suno is the best right? Useage wise.

Btw. stuff I also find weird is that songs start original with he og audio but then somewhere especially at the chorus drift off into a complete different melody which is not part of the audio file at all.

But not all is a prompt issue right? Because what I encountered too a couple of times was a great song but then in the middle all of the sudden it gets stutters, or hiccups like if a CD got a fat scratch. Or the AI drops some weird voices or instruments into the song which are just not fitting are where no part of the prompt or lyrics at all.

One last thing and again big thank you to you mate. I am not sure how much you are into Rock/Metal etc. But how would you describe Sleep Token for Suno that the outcome really sounds like them and the vocalist Vessel?

[young geode]

Time stamps dont work, it's placement. For the prompts, it just helps SUNO get pointed in the right direction, while the metatags grab it's focus.

What you really only need to worry about is that SUNO has both something you want in both the prompts and the meta tags, that seems to have the strongest and most consistent impact. IE, you want an intro solo at the beginning. You say "The song starts with an Solo" and then you make sure the instrument is labeled within the meta tag, with the solo meta tag. You'll almost always get a solo as a result.

And all what you described is just the AI. SUNO doesn't perfectly replicate anything, so whatever you give it, itll change at the moment.

And I dont know what the SUNO sleep token is

[near flume]

Always use [] for the structures and commands in the lyrics. () are used for whispered vocals.

[graceful condor]

SO a clear example just for demonstration. Just published this song so I can share it with you. And it does something so weird at 03:05. Such a weird long tone which makes no sense and ruins the song. Nothing in the prompt says to do weird stuff like that.

https://suno.com/s/FE9EoT7q18YyEZKg

There are the prompts:

Create a dark acoustic sound inspired by Noah Sebastian – Unplugged / Acoustic Versions. Focus on melancholic, emotionally raw arrangements featuring stripped-down guitars, soft ambient textures, and minimal percussion. Vocals should be haunting, breathy, and intimate — with a slightly raspy male voice that conveys vulnerability, tension, and pain.

The overall atmosphere is introspective, cinematic, and moody — balancing fragility with strength. Think reverb-heavy acoustic strumming, occasional piano notes, layered harmonies, and ambient background noise (like vinyl hiss or room reverb).

Song structure should stay close to traditional verse–chorus–bridge form, but with subtle dynamics: soft verses, emotional swells in the chorus, and a stripped bridge that builds back into a final refrain.

Keywords: acoustic rock, dark ballad, unplugged, emotional male vocals, slow tempo (60–85 BPM), minor key, dramatic intimacy.

Exclude Prompts:

Genres: Metal, Industrial, Punk, New Wave, Punk Hardcore Vocals: Female vocals, Sustained vocal notes, harsh growling, screaming, overly bright or thin tone, excessive melisma, detached or monotone delivery, falsetto-heavy singing, overly aggressive vocals, raw vocal styles.

[near flume]

Any time you are using "atmosphere" in the styles the Intro gets "muddy" if you don't declare exactly what you want in the [Intro: ]

As far as length, you can't dictate time in Suno and negatives in the styles have no effect. You need to build out the structure of the song (repeat chorus, add bridges, drops, breaks, etc.)

[near flume]

You Exclude isn't formatted correctly. Don't break up sections with headers. (Genres:, Hardcore Vocals:, etc) just list what you want excluded with commas, avoid using the words "no, or, except" in the excludes.

[young geode]

The song has a Reverb-soaked intro; is restrained in verses; swells during choruses; has an ambient bridge rising into a cinematic and emotional outro.```

I would write the prompt that like (you dont need the {brackets} necessarily, but it helps you keep track.

And I would add in the metatags to your lryics to reflect this, as well a (Parentheses) and *\Asterisks\* enclosed around areas you want the vocalist to be softer, quieter, more muted [Parentheses tend to have an ad-libbed effect, while Asterisks tend to create a muted effect]

[Start]
[Intro]
[Instrumental Introduction]
[Mood: Melancholy]
[Slow Buildup]

[Break]
[Riff]
[Verse 1]
[Vocals: Male, Mid-baritone, emotional, tense]
悪者に仕立て上げたけど
問題の根を言わなかったよね
欲しいものだけ奪い取って
僕の名を汚すつもりか

[Riff: Arpeggio]

[Pre-Chorus]
[Vocals: Male, Mid-baritone, emotional, tense]
言えなかったことが多すぎて
僕を置き去りにして消えた
でも みんな見てる 聞いてる
真実は もう隠せない

[Drop: Drum Fill, Multi-drum pattern]

[Chorus]
[Vocals: Male, Mid-baritone, emotional, tense]
僕を獣の檻に投げ込むなら
覚えておいて 死を恐れてなどいない
後悔なんて ひとつもない
でも君には 後悔させてやるよ```

That tonal issue is a generation issue of SUNO not knowing how to continue, so if you provide guidance, like above, it'll likely stop occuring, unless there is something within the OG audio that it's picking up on.

[graceful condor]

Thanks a huge lot. {} those onese what do they signal Suno? As someone said (in lyrics means more quiet), [are for Chorus, Verse etc] but what are the others for? And was the other one right adding "Atmosphere" makes it dull mushy? Why? I mean Atmosphere is an important word like dark atmosphere, moody cozy atmosphere so why is it making it dull?

[young geode]

Well, I'm not sure they do signal anything to ChatGPT (Which is what SUNO uses as an in between for it's proprietary model). Though, LLMs like GPT have a better time parsing information/instructions when you properly use enclosing punctuations and Colon lists.

I haven't notice too much of a difference yet, but I just started doing it. However they do help you keep track of everything if you're using a standardized prompt system like I do.

(in lyrics means more quiet), [are for Chorus, Verse etc] but what are the others for?

Explain, what others?

[polar spade]

Its logical that they should come up with a means to reconstitute the stem tracks so that we can combine at will, so, as you say, one, two, are left out, such as removing lead vocals but leave instrumental AND background vocals for a karaoke track, which can be useful. Or , of course, pick two or more and combine them.

Not yet though. Still have to do something like that in DAW. Maybe ver. 5.0 if lucky, right.

[young geode]

I was just wondering, cause they technically already allow this, I dont see why they didn't let you do the stems, then let you just save/combine them in a combination.

Must be a technical issue right now

[polar spade]

It kinda seems reasonable , if we can 1) remove only vocals, from the instrumental (because that was helpful originally to scan audio for copyright and artist issues) then we could do the same for any single entity, like drums or such. Makes sense.

And 2) if we can generate vocals into an instrumental stem, just combining two stems (or more) back together shouldnt be such a big leap. I agree.

Its all speculative. This early in the morning just makes my head hurt. Haha.

[quick dome]

build a bridge for web url and alt program ie like ableton and then stems will work to your usage in the program

[graceful condor]

I meant this one {} - Is it also okay if I describe what the voice should sound like because I might be familiar with some musical terms but definitely not with all. So I wouldn't know without asking the AI what Baritone is etc or Mid-Baritone. Btw metatage is the section where the lyrics are placed correct?

[hard beacon]

hey all.. what's the better prompt if you want to create a choir sung intro?? Choir or Choral? of course it's still a roulette, but is there a favorite prompt? 🙂

cause it's only for the intro i would like to take it into the meta tags... trying it with [emotional harmonized choral vocals] at the moment 🙂

[young geode]

{Curved Brackets} don't do anything within the Lyric section, and far as I know, they do little in the prompt section (Im still testing this).

Theyre just a good way for you to visually keep track of the prompt. That's all for now.

And yes, [Metatags] go within the lyrics section.

Baritone, Mid-Baritone, ect, are just the technical terms for vocal ranges for Men, Baritones are lower, while Tenors are higher.

[hard beacon]

Is it just me or does 4.5+ love to scream at you in the vocals? 😄 😄 😄 😄 😄 😄

[graceful condor]

Especially with rock xDD

[hard beacon]

even with dubstep... "Emotional vocals" in the style... But hey... SCREAM AT ME 😄

beautiful emotional vocals in the intro.. then the dude screams at me like hell 🙂

[graceful condor]

I wonder does Suno also reacts to other language prompts or just English? I mean it can sing ohter languages but I wonder how it works with prompts and stuff.

[hard beacon]

i've tried english and german.. works well

of course i am talking about prompts

🙂

[graceful condor]

German in my case too. Can you mix both in commands. Like here German there English?

[devout sinew]

Hello. I have a question about using chord progressions in the styles prompt area. Does Suno AI recognize and follow chord progressions if I include something like “Chord progression: C – G – Am – F” or “Key of E major with I–V–vi–IV structure” in the prompt?

Will the model actually execute that harmonic structure or is it mostly ignored?

I would appreciate any guidance or insight into how well Suno handles chord progression prompts. Thank you!

[ebon sand]

Yes, all the time, I can't really use it because of that

[hard beacon]

even with exclude "no screaming" ... he is givin a f*** to the exlude prompt 😄

i think 4.5+ is perfect for metal style songs... also the guitars are really good.. maybe i should try to create a accoustic ballad... hope "he" is not screaming at me 🙂

[ebon sand]

Yeah the only solution for me was to change the intended style drastically which went against what I wanted to do or to use a different model

Usually the first verse and chorus is fine and then he gets louder and louder

[hard beacon]

every model has it pro and cons... the sound is always gettin better and better.. but.. naaaarf.. stop screaming at me 😄

[polar spade]

My personal opinion:

I've listened to many songs which someone puts in a chord progression claiming it works.

1. if the chord progression is common with one or more of the genres prompted for, it could appear thst its working. I've heard close, but nothing more.

2. the user says it works, but I cant find its use anywhere in the song , though its prompted for.

And 3) control of chord progressions would be no different than controlling a melody line. As we have 8 minutes to upload whatever chord progression or melody line we wish, prompting for one (chord or melody) seems NOT a feature which a generative AI developer would want incorporated into the model. That would severely limit the creativity and randomness that generative provides as well as what would be available to train on from future user generations.

Others may have conflicting or differing opinions. Thats just my two cents worth, and it ain't worth much. 🙂

[young geode]

Naming Scales seems to work, at least it seems to prompt something different within the play pattern.

I believe it's roughly the same thing for Chords/chord progression. I don't believe it's 100% playing what's being requested, but it is having an affect on what's being played.

[livid pond]

Can anyone help? Can not find my playlists. After the update the library folder is empty

[livid pond]

It's Chrome bug. Safari shows everything

[half dragon]

Must be your Chrome - my Chrome shows everything.

[brisk flint]

Correct, had some good results yesterday using a Lydian scale, but prompting the chord-progression so far did not work out.

[brisk flint]

Es geht - aber nicht immer, d.h. du brauchst etwas Geduld und musst sehr auf Prompts IN der Lyric-Sektion achten. Sorry, couldn't resist answering in my mother tounge :p

[young geode]

Are you able to recognize the Scales? Noticed immediately that SUNO was picking up on the fact that scales were mentioned, but I couldnt tell if it was the scales being referred to, only that it was in fact changing the song at the designated points.

[brisk flint]

Yes, I was. 😉 Especially as the 'lydian' scale is pretty 'special'.

[young geode]

Alright, then I'll probably test as many scales as I can here soon. I'd like to see how many of them it recognizes, and if Genre matters.

[polar spade]

@tawdry scarab Haven't you dont a bit of testing Lydian scale and others to see if they are actually recognized by Suno?

[tawdry scarab]

I ask for it all the time but I don't have a perfect ear

so I couldn't tell you the key or chordsbut the mood feels right

[polar spade]

OK. Its being discussed again. I knew there has been other attempts before. Thanks.

[tawdry scarab]

yeah that's one of the first things I squared away in my method was a music theory block

can I send a text file to this channel?

I'm having my gpt make 10 random theory blocks to demo

[polar spade]

Di anyone else seem to think they were affective. Most seemed to ju a t mention things like "Egyptian" or " Arabic" would give flavoring, but actually staying true to scales wasnt confirmed so far as I remember.

Thanks. Yea, posting files work here.

[tawdry scarab]

All I know is people are impressed by my work and I get whatever I want from extending

I don't worry about whether there is adherence

[polar spade]

MASTER, you may find the txt file usefulness of you're wishing to do some testing of types of scales.

[tawdry scarab]

if you think of every tag as hit or miss, my method is a huge array of small independent suggestions

it's the sum shotgun effect not precision

there is also the sneaky fact that music notation is arbitrary and not set to a specific rule, you can call anything any time signature and any key and still make valid sheet music that is correctly transcribed to those parameters, just with incidentals and odd clustering

[polar spade]

As most it seems. Create 40 outputs first, then move the target. 😅

Well, seriously, I just feel Suno still doent have as much training other than association of whats used more frequently as being the prominent mode.

[tawdry scarab]

I use my gpt I very rarely have to make very many gens

it's pretty robust

#1365192916687126650 message

those instructions work for humans and ai, you follow them suno will understand you

[young geode]

Ill look into it when I have the credits.

Ideally, if these blocks worked, you could generate a genre/sound without actually directly referring to said Genre, which would mean there's a secondary pathway to tap into SUNO's token predication. That'd be really useful for genre blending.

[tawdry scarab]

you can do that

You can drive a whole song's style without the style string, but it's not super strong

[young geode]

Explain

[tawdry scarab]

there's nothing to explain, a full, thick meta tagged song that is internally consistent to a style will produce that style with no style string, but might be slightly flat or less inventive, less leaning into the style

but that depends on gerne

I can make one if you'd like

[young geode]

I dont know what string style is in this context. Do you mean a prompt?

[tawdry scarab]

you have three prompts for any suno generation

the lyric block which responds to met atags, the style string (my term I can't stop using), and the negative keywords

[polar spade]

Is this with the script youve written, that you were testing before. That was pretty useful.

[tawdry scarab]

yeah

ai fluency is a little bit just fluency fluency, you have to learn to articulate the thoughts you otherwise naturally presume are shared with anyone hearing what you say

that's like... a primary dysfunction of us apes

[young geode]

Ah, I think you misunderstood what I was getting at.

The tags you have within there, if they are consistently picked up within the Lyric box, would also mean that they should be picked up within the Style Prompt. Typically if you place items like that within both, it creates a stronger lock-on effect for the generation.

But typically these small things are used after the fact, and can fight with/be overpowered by stronger terms, such as the Genre references. However, if they do have an effect, you could use this as a way of bypassing scenarios where two genres are fighting with each other: especially if you just want a few things from one genre, and know you can get away with a weaker, but similar genre.

A good example is how Funk damages almost any 100 bpm or more genre, by simply mentioning it. But if you reference everything about Funk, without saying funk, then this would likely happen far less.

[tawdry scarab]

yes, I did misunderstand you and you are correct, but using any square brackets (which I don't assume you would) dampens their effect in the style box

and for that, order matters, and how many other keywords there are, you can keep stacking keywords in front that have to do with speed and then flip funk in at the very end

[young geode]

Squares in the Lyrics, Plain the prompts.

But ill look at these, cause building them like this with an LLM would help with the testing.

[tawdry scarab]

It gets marked as spam when I share the gpt link but I can pm it to you if you want to play around and see if it works for crafting test music by describing it

or you can plug those instructions I linked in and have your own

[young geode]

go ahead, Ill have more time tonight to go over this and maybe develop a workflow for it. I'm almost finished writing a guide, so this would be a good next project over the month

[tawdry scarab]

this guy took my method and I think made the most refined versions of it.
https://suno.com/@shudderingclips087

[polar spade]

Would you mind sending me as well. We just worked with my giving you a test sample and I never asked for the actual link before. Much appreciated if possible.

[tawdry scarab]

he kinda drew from everyone's prompt experience and makes killer directions

yeah no problem

daisy unbanned me I wonder if she fixed the fact that gpt links are flagged

I'm too gun shy to try

[polar spade]

Wouldn't want you to risk it.

[tawdry scarab]

sent

[polar spade]

Got it. Thanks again. ❤️

[tawdry scarab]

I've never tried but it will probably explain its tagging conventions if you ask, which would let you ask questions about efficacy or plans

the benefit of the bot over the instructions is it also has a collection of prompts from early work I liked from it as reference

[polar spade]

@tawdry scarab OK. Thanks for letting me pull you away from General a bit. MASTERof SFL was delving into the convo of scales and I knew you had been a part of discussion before. It pops up about every three months or so.

[tawdry scarab]

XD

these days all non programming activity feels like I'm jeopardizing my chance to become financially independent with my current project of two years now

I'm trying to harness that without going mad

it's a marathon not a sprint

[polar spade]

Focusing on Signal over Noise. 💯

[tawdry scarab]

its apropos you say that because the present module is a physics sim of a turing machine using a noisy cassette player as the means of computation

and I'm encoding bits as fft window pulses

with added noise for testing discrimination, based on the use of the motor or read/write head

I'm just trying to get all the sim pieces using the right pathways to make sure everything gets passed through all the hurdles to make it real, spinning the tape using an amp signal that gets integrated all the way up to tape movement, but I do need to put in, I just realize, the diameter of the tape and how it scales distance, unless I just want it to always be warped

I am essentially making a virtual machine in python that is itself a cassette player

becauuuuse by doing so I can have differentiable metrics on memory mapping and register use

[polar spade]

Im happy to read anything to learn a little something, so its all interesting. And I always like the old school retro applications such as a cassette player head. That much I get. 😀 otherwise, I would not know much about noise and distortion analysis. Sounds interesting though.

[tawdry scarab]

I'm coordinating the simulation and the backing data operations right now, once that's working I have to make sure the system can read machine code instructions for the basic operators or settle for doing that later, then I have to tune the read and seek parameters like motor responsiveness and amplifier current and the bit width envelope that devines a one in any frequency lane, and then after that I can start making a visualization that is a crude representation and people can enter in math equations in a box and watch a tape reel back and forth and hear every bit of data synthesizer blips and bleeps and then it will compute the math problem if given hours to do so and present the response

I think that is maybe five days away?

if I do really well

[polar spade]

Sounds like you could use an old 8-track player to loop things. Ive got two of those. 😅 sorry, im not intelligent enough to contribute more than a bit of levity.

[tawdry scarab]

literally?

there is a continuous loop option specifically for 8 track modelling

great minds thinking alike

[polar spade]

Im looking at one. I think the other is on a shelf somewhere.

[tawdry scarab]

I don't even know what curse I said

oh, that the speakers were as of fecal matter

[tawdry scarab]

if you could get reference photos of the face of that without it being a hassle that is literally my absolute favorite era of audio equipment faceplates and knobs

[polar spade]

Read every bit, and again, interesting.

[polar spade]

Sure thing. I'll get some cleaner shots and send you. Whatever you need.

[tawdry scarab]

you know what's nuts is I have this graph theory IR compiler thing, so this can take a pure math expression and turn it into a python scheduled graph accounting for memory needs, or use any of a full set of programming primitives I set up, and it will auto compile into the cassete language after first auto compiling itself down to a minimum operator set that willl be used for automatic ssa compiling so it can be made into a binary with the efficiency of a c compiler

this tape toy is just a self motivatoin to get differentiability into memory layout of binary data

[polar spade]

Im not a coder, but that all makes sense. As I said, anytime we can combine old tech with new ideas, it makes life feel more gratifying.

[tawdry scarab]

there's full stack coding, and then there's full stack engineering

[polar spade]

Sorry I cant contribute more than that though, than encouragement to keep going.

[tawdry scarab]

and on the far left full stack provenance, as I've made abstract bit carriers that are permitted to represent any structure that can have truthy, falsy, and custom operators if necessary

no it's cool I work on it many hours every day I only ever talk about it blurting out things here

it's enjoyable to express it

[pallid echo]

Hello everybody. Is there any way to make Suno continue to change the lyrics but not the music? Like when I wanna change the words a bit.

[polar spade]

Im sure youve read Godel, Escher, Bach, so its a favourite of mine, but doesnt reflect that I understood everything I read. But that doesnt stop me from just reading it from end to end. With AI now, i just ask it to explain everything to me in 3rd grade language. Haha.

[tawdry scarab]

do you have and know how to use python dasvon? you can run the tests in the development version and listen to a bunch of low frequency blips that I hope express something is happening correctly

hahaha, well, you know, there's two kinds of explain it to a third grader and one is bad and the other is profoundly good

[polar spade]

No. Not at all. I know nothing of coding from a practical aspect.

[tawdry scarab]

I think with the whole corpus of knowledge gpt is pretty good at it

the program is meant to use parallel signals as well as isolated ones but the test is only on the low frequency rail so all I hear is thudthudthudthud

the demo will use parallel mode and make musical arpegiation I bet

[polar spade]

To change small sections or even just a word, the Replace feature is a possible tool. Its in the Edit UI or the Legacy UI. These highlight what you wish to change and theoretically it will blend in the change. Its is not without a learning curve and some preparation to make sure we align the lyrics , make the audio match the lyrics, before trying to replace anything.

One quick way , if its a word or two, or even a small phrase is to just do a cover and change all the lyrics you need. At a minimum of a few words only, seems best. With luck, it will just re-sing with the changes.

[polar spade]

You can also now "add vocals" to a separated or "stem" track of just the instrumentsl of the original. Since your lyrics and arrangement are nearly the same, but a few words or corrections, we could get lucky and it add the new lyrics without changing the instrumental part. The problem is we cant tell it where to start singing the new lyrics, even if they are mostly the same.

So there are 3 ways we can try.

Much anything seems to be a learning curve, so be prepared for some trial and error, but when it works, its great to hear the results.

[pallid echo]

I've used Remix > Cover > Reuse and the track is 90% the same

The Stem function looks pretty expensive for me because I've changed just 2 lines of lyrics. Thank you for help. I'm a newbie and I don't know much of it.

Looks like this is exactly what I need in such cases to reuse exactly the same generation token

[polar spade]

I get being the newbie. Just pointing out there are ways , without adding but it might cost time and credits to learn. 😊 when suno first started Covers, that was the first way to try to replace lyrics while still using ver 3.5. So , keeping the changes small, it could get it in a couple of tries. Or at least correct a couple of the changes.

Then, if we can get it so the last changes are at the end, try to just do an extension at a timemark before those lyric mistakes and hope it gives a good rendition close to what we want.

[raven turtle]

Does anyone have any prompt suggestions to calm down my damn vocalist! LOL!
In 4.5+ he really likes to belt out the chorus, and no amount of “monotone vocal” type prompting in the lyrics stops him from screaming his lungs out! Bravo to Suno devs for the vocal improvements but wish I could calm the sod down! Haha!

[cursive island]

does inspiration work? I am going more and more towards a certain direction and I really would love to train suno to adapt my style everytime

[neat nexus]

I think so, personally. Just understand that it's limited.

You may be best served with using clips instead of full songs if there is a specific thing you're looking for.

It may also be helpful to cover the outputs and use personas based on clips to push you in the direction you're trying to go.

[cursive island]

playlist with clips to push it. got it. about personas, i done some it just copys the style of arpeggios etc but not the VOICE itself 😄 any hints?

[neat nexus]

Luck and play with the sliders?

RNG gets the final say sometimes.

[uncut tulip]

im using a audio refrence
but it wont stay on beat basically
lyerics go off beat
or beat goes too fasdt

[sleek stump]

You are likely running into a genre wall, and possibly a prompting wall.. Use terms like Softly sung x vocals, or Muted x vocals in the styles window. I dont have any issues getting softy sung vocals, except for metal...

[raven turtle]

Thanks so much! I give it a go!

[quartz adder]

I've been banging my head trying to get personas to work. do I fundamentally misunderstand? I want to take the voice from another song (use that to create the persona) but when I create with it it doesn't capture it at... all

[wet minnow]

stem split the vocals out and use these as persona input. that might make it more simple.

[quartz adder]

that's what I'm doing 😭

[wispy laurel]

Any reliable way to stop Suno from using specific instruments that would normally be associated with a given genre?
Working currently mainly on bluesy stuff with both jazz and rock elements mixed in and between personas and experimenting with prompts I can mostly get it to do approximately what I want
Except that it compulsively inserts very prominent and noticeable basslines into everything. Excluding "bass", "bass guitar" or any other variation doesn't seem to affect this at all. Style instructions specifying "piano, electric guitar and drums" will reliably use only those instruments and bass guitar
It's not a huge issue because I can just download stems and reassemble it without the bassline but, that's a bunch of extra steps if I'm already happy with the song other than that one issue.
Does anyone know how to avoid this? I'm guessing that it's trying to follow genre conventions that dictate particular instruments, and that (or something) is overriding anything that I can think of. Is there any particular prompt or trick that might help to stop this from happening?

[young geode]

There's abway, but it's pretty credit heavy.

What you'd need to do is find the genre you're trying to fuse with another, and effectively distill the instruments/sound you're looking for, then turn it into stems.

You'd then extend off those with the actual genre you want, referring to that sound within your prompt and meta tags, in order to fully capture it within your new song, then you'd turn that song into a persona.

From then on you should be able to generate that sound without having to fully reference the genre.

Other than that, SUNO is going to drop any and all sounds related to a genre, if that sound is prominent within the genre.

You can do heavy exclusion tags, but that will only weight it back so much.

[wispy laurel]

Yeah I did notice that exclusion tags don't really work well on things that are usually fundamental to a genre. As soon as you name any particular genre anywhere in the style instruction it starts making a whole bunch of assumptions that then seem to override instructions that disagree with them.
Alright thanks, I'll try the stems-and-extend thing and see if that helps.

[young geode]

The other thing you can try is layering in only the instruments you want with effectively a [Tag] before them.

So if you just want some jazzy piano, you'd put [Jazz] Piano or Piano [Jazzy].

I haven't fully tested it, but I have noticed this does help weight it a bit. Additionally, you can use actual percentage weights, Blues 60%, Jazz 20%, Rock 20%. That I do know does prioritize one other the other and often will pick up only some of the genre.

[wispy laurel]

Huh. That's about the only thing I didn't try in terms of prompting because I assumed it wouldn't work 🤣
Alright I'll try that out too then.
Thanks again!

[sturdy dove]

Man, this place BLEW UP with helpful people like @polar spade and information since I was last here!

Maybe I will take y'all up on that! I'll see if I can muster some energy for this again the next couple of days. Thank you kindly for the offer. I'm nowhere near as burnt and frustrated now as I was when I posted last...

[neon flint]

Hi, before generating music can you ask the ai to use a certan Key and/or scale?

[young geode]

Yes.

Though I haven't tried every single scale, SUNO does respond to Scales within the Style Prompt and Metatags within the lyrics box.

Now, I am unable to tell if it's playing the correct scales or not (my own testing just shows that the song's tone/notes will shift/repeat/revert with named scales) but other users have stated they can and that it does.

For the Metatags, the following tags seem to work:
[Instrument: Type, Scale]
[Section: Scale]
[Melody: Scale]
[Solo: Scale]

Meanwhile, in prompts the style prompt the following ways to ask seem to work:
Section in scale
Instrument Type: In Scale
Riffs in scale

Ect.

The strongest pattern of behavior I've noticed is of you include within the prompt the scale with some kind of other portion of the song, and a metatag that accompanies.

For example, if you prompted Guitar in Scale and then placed the metatag `[Instrument: Guitar, Scale], it'll play.

[sacred swift]

6740-f9241710cee0356e.js:11 POST https://studio-api.prod.suno.com/api/clips/autoplay/ 400 (Bad Request)

It seems like the radio isn't working properly. Specifically, it's not moving on to the next song and is instead getting stuck in a loop.

[karmic lake]

Is there a way to edit and change your persona?

[ebon sand]

V4.5 and v4.5+ work exceptionally bad with persona, try to use 3.5 or V4 and then remaster, they were the last models which did this kind of capture well

[quartz adder]

thank you! will try

[quartz adder]

God bless you!!

first try

on track vs hours of crap

thanks for taking the time

[ebon sand]

You are welcome, hopefully there next model will be able to do this better again 🙏

[jovial widget]

is it possible to blend the vocal stem from one trach with the instrumental from another?

[potent fern]

Any idea if you can also do chords? A lot of lyrics might have guitar chords or otherwise, even tab notation on them. I wonder if feeding something like this to suno will generate a song with those chords if in the square brackets?

[young geode]

I have seen it respond to chords, even the text based version. However, 1) I don't have the ability to discern/name one note from another outside of sound; 2) I've only ever seen this within metatags.: not prompts.

I've tried it, there was a clear difference between gens with and with it, but all I can tell you is it's responding. Others have said that when they try, it doesn't play the chords, others say that the chord might be so prevalent in the genre, it played by by happenstance.

[potent fern]

In this song I'm working on, I'm trying to add a chorus, and have worked out one on guitar, and in the process, have transcribed the chords used already by suno throughout the song. So I'll try and cover the song, but add this version with the chords, and with the added chorus, and see if it understands the chords with reference to the rest of the song! It also seemed to understand the time signature change by writing [Chorus in 7/4 time] at least most of the time.

[jovial widget]

any help?

[midnight plume]

So if I have Song A and Song B both sung by females, and I want to always use the voice model in Song A for every female song, is that what Personalities does?

[graceful condor]

Thank you, another question do you know how to use prompts for a female vocals on one part of the lyrics and Male on another and then both on the ending of the lyrics, I've tried using () and [] but does not work hardly.

[brisk flint]

Yes and no. You mean “personas”. In my experience, a “persona” currently represents not only the vocals or timbre and voice colour, but the entire song, including the arrangement and instruments. Extracting the stems and creating a persona from the acapella alone is no longer possible without tricks, so it's not really a good method.
And creating a playlist with ONLY song A and then using the “Inspo” function with the audio influencer slider set to 100% would also transfer EVERYTHING from song A, not just the voice.
Furthermore: If the structure of the lyrics changes for the new songs, e.g. different words, different lengths of the individual lines in the verses – to give a few examples – then the type of singing will also change.
In short: I am currently not aware of any method for using a voice from one song 100% perfectly for other songs. But I'm happy to be proven wrong 😉

[brisk flint]

It works, but as ALWAYS, you need a little luck and patience and a handful of credits before you get the result you want, as Suno still likes to “skip over” one or two parts of a prompt: https://suno.com/s/QD0dd4ad3vMuqzeT

[near flume]

Yes. Here's the catch, in the styles you need to add "Duet" before you declare any vocals. Put as much detail as possible for the voices. The first voice you declare "usually" becomes the default voice, this can vary depending if you do a lot of male only or female only vocals. In the lyrics use [] to mark where you want the voice to change. When you want the voices to sing together use [Duet] or [Male Vocals & Female Vocals], using "both" or "together" won't work.

[umbral kindle]

can anyone help me with my account? I recently purchased but somehow the new credit is still not reflecting.

[polar spade]

It may be an error as you show, but double check to see if the Loop button is unclicked as well.

[polar spade]

If youre out of the USA, funds typically go through international banking systems. Then reviewed for fraud, IDz theft, etc, and handle by a couple of systems. This is typical for online transactions with any site. A delay of 3-5 BANKING days is common before funds finally reach Suno accounts to activate your subscription. Thats one possibility.

[quartz adder]

am I losing my mind? this morning I did a cover with a persona using v4.... now when I try, it is forcing to use 4.5

[ebon sand]

Tell me when you have done these steps and it works because afterwards I will delete this message because I don't want this to get fixed

[quartz adder]

I took a screenshot 🙂

you are a gem

[ebon sand]

The outputs can be a bit unstable sometimes but I also use it often and usually also get the result I wanted

[quartz adder]

yep! unstable is fine by me!

it's the 100% not working that was destroying my soul

[shadow plaza]

is it me or the shimmer is back for 4.5+ ?

[polar spade]

New article about ChstGPT feature added.....

"......While ChatGPT can be a useful tool, it also has potential consequences if overused. Research has indicated that overdependence on AI can negatively impact critical thinking, cause your brain to atrophy, and make you lonely......"

[sleek stump]

Not in the literal sense. Suno editor does not currently have the ability to mix or combine separate tracks, you would need to do that in a DAW and then upload it back in.

[wispy laurel]

I use [exactly match the persona's vocal style] or [use the persona's voice] or similar within the style instruction and it helps. Doesn't work in every instance but it does noticeably improve the odds of getting the voice you were trying to get, compared to using the persona without that instruction

In cover versions it also somewhat responds to [exactly match xxxxxxx of the original song] too, again it's not guaranteed but it does seem to improve the odds of preserving that particular characteristic from the original

[ebon sand]

To me AI has become indispensable and any perceived disadvantage is outweighed by the possibilities it opens

[tawdry scarab]

@polar spade I wasn't kidding by the way about wanting some reference shots of your 8track, a fully straight on picture i might be able to use for a texture somewhere in this tape computer program. can't say it would work out for sure but if it's no skin off your nose it would be a nice option for me to try when I get to the opengl

[jovial widget]

ok thank you for replying

[tawdry scarab]

did you respond to the wrong person maybe XD

[polar spade]

Its still here in front of me. I was just going to clean off the dust. But, yea, its on my list of to- dos today.

[polar spade]

Yep. Forgot to hit send on that last one. They got combined.

[tawdry scarab]

what would be maximally helpful to me would be a stationary camera shot of all dials left, all dials center, and all dials right, but I feel weird asking for that because this is just speculative and the textures could end up bad because I don't do graphic design much for maybe decades, so I might be asking you for a big chore for nothing. which is why I think I would prefer asking you only for a straight on shot with the dials wherever

[polar spade]

No sweat, mate. Ive got time.

[polar spade]

.... and its got a bit of brown gaffer tape residue or something on the bottom I wanted to try to clean off first. Ill get it done.

[tawdry scarab]

if you can't, that is actually kind of commensurate with that era

I don't know why but my brain says it fits

my dad had a huge rack of equipment in the silver chunky style, I think it's locked in my brain as the ultimate

oh I know why gaffer tape makes sense to me, because it gets used all the time for labels

I also used to work the church mixer

which was the slightly older, everything is faux or cheap veneered wood

which is also a good look

[outer urchin]

Hi, I just want to report that, the voice of my AI sunoo is becoming a Male voice even though I put on style "Female vocals", I'm using 4.5+ and I think it's not worth it because of that... pls fix it

[wispy laurel]

4.5+ does all kinds of weird things with vocals
It doesn't quite understand that "just because you can doesn't mean you should"

[spring spindle]

At the end of my songs at 4.5+, something strange is always spoken. Does anyone have a good solution for this yet?

[half dragon]

If regen isn't an option or not working; maybe use crop and fadeout tool?

[spring spindle]

Its possible.. but we are making 50 Songs a day.. and thats to much time waste. Its only at version 4.5+

[half dragon]

Ah for new songs a proper solution is preferred - at the model itself.
I'd report the bug in #🪲┃bugs-web and hopefully it'd get fix in the future updates.
With existing songs - I"m afriad it'd be a manual process.

[polar spade]

(Further suggestions have been made on same post in need-help channel)

[polar spade]

Are you referring to "hallucinations" or what seems like gibberish? If only at the end and you insist on maintaining the prompts as is, then crop out the weird end bits and fade and @VRKing mentions would be the fastest. The next slower would be to then extend from that point which you removed to get a longer song which hopefully does not hallucinate.

If youre cranking out 50 a day to make a library of samples or full songs for DJ or streaming radio , then its better really to budget in a bit of edit time for issues like gibberish that will innately show up now and then. Put a little time in for quality over profit, right. There just isnt a magic bullet yet to handle all situations , prompts and tag arrangements.

[tawdry scarab]

indeed, don't make crap at scale

[minor vigil]

how to use option "add vocals"?

[polar spade]

1. Click menu. [ . . . ]
   Click Get Stems
   Choose Vocal/Instrumental (10 credits) let stems load in library

---

2. Click either vocal or instrumental stem (in Library)
   Click menu [ . . . ] of stem
   Click Edit/Remix. And choose Add Vocal
   Include lyrics in Lyric box as needed
   Click Remix

[minor vigil]

For me it doesn't appear when I click on my vocal stem to add the vocal I want.

[polar spade]

Does your browser (Chrome for me) have a desktop settings? It should be on.

[minor vigil]

i'm using opera gx in notebook.

[polar spade]

Opera i though has an on/off for desktop as well. But specifically not sure. But youre in a notebook anyway. Hold on.

[polar spade]

Heres just a quick video of me not having desktop and showing nothing in Create panel.... and switching on Desktop and now showing the stem in the Create panel. Its about getting the right desktop format for whatever youre using.

Its compatible with Opera , but Opera GX might have something different that is affected. Can you test Chrome or another browser available, if you can easily log in.

Oops I only recorded half the video. Dang it. Well sorry. No video.

You can go outside of any browser and just go to your notebook display settings and lower your fonts and display layout, they can have an affect on what show on your monitor, at your discretion and convenience, if comfortable testing different aspects and sizes. ,

[minor vigil]

I don't think it has anything to do with the browser, I'll try to record a video soon.

[polar spade]

If its working, keep in mind , if youre adding a vocal , its assumed youre adding it to an instrumental stem (track) that you like. Normally, we would not load a new vocal on top of another vocal track , with a good reason-- thats a bit more advanced.

And some assume we are REPLACING a vocal with a perfect melody match of another. Thats not the case. It will be as if you initially created an instrumental and for the first time, give it some lyrics to create a new melody with that instrumental. So the song will be new melody and vocal ones an older instrumental.

[polar spade]

Heres a create number 1 showing no attached stem due to desktop NOT being on.... and Create windows number 2 , with desktop being on.

But youre right to assume, it may be something different which isnt attaching the stem to your create windows.

Send a video if you wish. We can look at it.

[minor vigil]

I'm trying to do this with a vocal I have, I kind of want to add another vocal or replace one with another.

[polar spade]

You can overlay as if two tracks playing at once. It will generate the output based on the lyrics. But if you add different lyrics to an existing. Vocal track youll hear the results will be a strange mash up, basically on your first try. Or just one or the other vocal.

[polar spade]

There is no "melody" stem to add a vocal too that will them follow the same melody as before...... for most circumstances.

[polar spade]

But saying that..... hey..experiment and have fun discovering what it can do.

[minor vigil]

I want to add a voice that has already been uploaded.

[polar spade]

In general, we can NOT add one output over the top of another output. That covers everything, about any ootion we wish to try, cant be done yet. .

The closest is to Cover one song, while adding a personas of another vocal. Using slider to try to only have the voice characteristics sing the same melody line as the original instead of the original vocal.

We cant Cover a song and "add" or overlay somehow , another vocal characteristic in place of what is in the song anywhere other way , at this time.

There are apps which do that. Suno hasn't included that tool as yet.

[minor vigil]

and what are these tools?

[polar spade]

Ummm... search acestudio.... kits.ai.... a couple of others. I just try to help on how to accomplish something online in Suno. But sure, use outside sources then upload back into Suno for some enhancements, thats done frequently

Nothings free, so it will be extra outisde tools for extra fees , of course.

[polar spade]

Let's always hope im wrong and others have some elementary tricks that can be dont in a few steps to literally replace one vocal and same melody with another of the same melody and meter. Any and all auggestions are welcomed.

[minor vigil]

but do you know how?

[polar spade]

Ill say no, I dont. Seems we first have to be able to get the tools we have to work first on your notebook, right. So one step at a time.

[orchid token]

Hey, does anyone know how to stop Suno 4.5+ from making songs way too long? Trying to keep them short but no luck so far.

[polar spade]

You can also create the song in 4.5 as a shortcut.

[minor vigil]

kits ai is not free

[polar spade]

Yes. Said that, my friend.

[minor vigil]

and after remaster with 4.5+

[minor vigil]

sorry, my lack of attention.

[untold briar]

@minor vigil I'm not sure I understand exactly what you are trying to do. Is it literally overlaying audio, as in the kind of thing you can do in a regular audio editor that isn't Suno, or are trying to generate something only Suno can do?

[pallid echo]

Hi everyone. I'm trying to create a punk-rock song with a low female voice. Could anybody suggest a prompt for it? Every time I try I got the Avril Lavigne's voice despite I want smth like Beth Hart style.

At least I've successed (suddenly) with forking P!nk's voice by the prompt:

raw and powerful, carrying a soulful grit that cuts straight to the heart. It blends blues, rock, and gospel with an emotional intensity that feels both vulnerable and fierce. There's a smoky, raspy edge to her tone, yet she can soar with incredible range and control.

[timber latch]

/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

[timber latch]

so what commend is new

@flint haven what commend is new

/help

[flint haven]

How to use Suno

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

For more detailed information, check out these resources:

• Making Music
• Billing
• Rights and Ownership
• Accounts
• Troubleshooting
• Mobile App
• Suno Server

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

[timber latch]

/sons

/song

[unique stone]

There is no Discord command to make a song any more. You need to use the website instead - https://suno.com

[summer pulsar]

How do I get panning/moving source in my song. Sometimes I get it by pure luck. But nu style or promt in lyrics seems to work. Tried it like 50 different ways now.

[young geode]

A simple way to zero in on a singer's voice is to research their vocal pitch, range, style, ect; choose a song they've done, create a very highly specific prompt that targets the Year/Era and Genre that mirrors that song; take the lyrics to that song, strip them to their meterlines and rhymes, then rebuild the lyrics using those meterlines and rhymes.

And in the Exclusion prompt area, target the vocals you keep hearing and cancel them out in the same way you targeted the vocal style you wanted.

If you do this, and metatag in a way that roughly mirrors the original song, SUNO will likely generate a rough approximation (like 80-90%) of the vocalist you're seeking. After that you split the stems and use them as a persona.

[near flume]

currently Suno doesn't support that kind control. Once they merge WavTools into Suno you may more control.

[jovial widget]

Anyone else really frustrated with the editor? It should be a game changer but I can never get it to do what I want it too. Is it the editor or is it me?

[restive forum]

It’s at least me too, I’ve wasted a lot of credits with the editor and I can’t get it to do anything I want it to. But that could just be user error.

[dense folio]

i have a question im using a free sub now i created a music and there is 2 suggested song but not complete if i sub for pro. that incomplete song will be full and the ownership is going to be mine?

[jovial widget]

tell me about it it eats credits. I'm glad it there because once they perfect it it be unreal, but at the moment its garbage.

[jovial widget]

I though even in free version it created full songs

[neat field]

Some songs are very challenging to convert into acoustic guitar sessions for the model last few days... 3 days ago was able to pump out some amazing acoustic studio sessions from v4.5+, last day or so not so much

[polar spade]

Maybe a bit of both. Many things took me a bit to either figure out, learn the work around, or find the need to add three more steps before it would function as expected. Others things, like it re-formatting the lyrics is an odd bit and affects the syncing, so it take repetition to "fix" it. I use all the time, otherwise, and ultimate get what I need.

[sleek stump]

The short answer is, no. Anything you create under a Free User account belongs to Suno, and you have no commercial rights to it. That includes anything you do to edit that original track even after subscribing. After subscribing, you instead should use the same prompts and generate a new song that will be owned by you. You won't get the exact same song, but you will get something similar-ish. This section explains https://discord.com/channels/1069381916492562582/1294423389515284552

[livid trout]

Hello, it seems that if you want to change snippets of lyrics in another language, it kind of looses the objective to pronounce in dutch and she seems to get a stroke 🙂 Is there something you can do to resolve this ? I use "fixed", because smart also changes the music too much

[hollow hare]

One common solution recommended is to spell the word phonetically.
If you are unsure of how to go about doing that, you can ask a LLM something like "provide a list of phonetically spelling for [X] in Dutch"

[livid trout]

thx, i'll do that.. , also is there a way to make sure they retain the exact tonal phrasing or rythm of syllables? I wish in the edit mode, you had these sliders to adhere to sound like you have with the cover interface

[sleek stump]

It was much easier in the older editor to get consistency, and in general I use the OG or legacy editors, I have not had much success with the newest version.

[livid trout]

the arrangements and energy of this take is so good: I don't want to lose it to a rewrite of lyrics in a cover, even if I set the slider of sound adherence of 100% it changes a lot .. so I'm resorting to lyrics edits.

Maybe i might try to download the stems, and then trying to make an acapella vocals with the lyrical edits ...

[cyan jewel]

im having difficulties with getting suno to follow my lyric prompts for clean phrases versus false chord screams. Also can't seem to get layered vocals to work.

could I get some assistance?

[livid trout]

i really think they should invent a melodyne kind of interface to assist the voice in what cadence you want in t he editor...

[cyan jewel]

[Clean Vocals - Omega side]  
Beneath the veil a shadow stays  
纱幕之下 是沉默的影  
[False Chord Screams – Huli Jing side]  
光中伏影 火中藏灵
[Clean + False Chord Scream Dual Vocals]
Light hides shadows, fire holds spirits

[Instrumental Hook]

[Verse 1 – Omega, (clean vocals)]  
Steps retreat with every glance  
步伐躲避众人目光  
Ribbons bind, but not by chance  
柔弱不是天赋，是习惯  
A whisper floats behind her eyes  
眼中藏着未吐的对白  

[Verse 1 – Huli Jing (Screamed, false chord)]  
SHE MOVES LIKE STATIC UNDER SKIN  
DATA BLEEDS FROM DEEP WITHIN  
狐尾隐匿于制服之下  
Tails hidden beneath the uniform’s seams
THE SPIRIT WRITHES INSIDE THE MEEK```

Here for example, I can't get it to follow the prompting 1:1. every iteration tends to flip which lyrics are screamed and which are clean vocals. also how do I prompt for different female vocals because it's basically the same female tone regardless of what I try

[polar spade]

He got the answer he wanted to hear in another room. 😅 oh well.

[sleek stump]

The answer he got wasnt exactly accurate either lol

[polar spade]

Just to help, " cadence" in music literally means something different than the intonation and melodic flow of vocals. So maybe just phrase it out or use another word.

Cadence in music formally mean the choice of and use of chords at the ending of musical phrases. I dont think thats what you meant.

[graceful condor]

{Tempo: slow and flowing}
{Genre: J-Rock, Traditional Japanese Rock Ballad, Orchestral Rock Fusion}
{Instruments: Shakuhachi, Koto, Electric Guitar (clean reverb and chorus), Deep Taiko drums, Violin, Cello, subtle ambient synths}
{Vocals: Male, breathy mezzo-soprano; melancholic and airy in verses; full, layered harmonies in chorus; optional whispered background voices; slight vibrato}
{Mood: melancholic, mysterious, romantic, longing}
{Mix: Wide stereo, soft reverb, warm mids, smooth highs, Strymon Big Sky}
{Form: Verse–Verse–Chorus–Bridge–Verse–Final Chorus–Outro}

If I want to add a guitar solo too to this prompt is is enough to include "Electric guitar solo with ambient reverb and dreamy overdrive, sounding like Strymon Big Sky Cloud mode, mild tape delay" within the instrument part of the prompt but I wonder if it is too much because there is already a part called "Electric Guitar (clean reverb and Chorus)" but it is just for the chord parts.

And I just wonder if this could intervere with each other.

[pseudo blade]

Oh yes there is

[hard beacon]

btw... is there a prompt like "don't scream at me like hell" ?? 😄

[ebon sand]

[hard beacon]

hahahahaha

good point 🙂

[ebon sand]

you could however try to describe the vocal style in more detail, maybe it picks it up, like not trying to avoid screaming but emphasizing what you want to hear

[hard beacon]

well.. i've used emotional vocals (something like that)

mi mi mi 🙂

[ebon sand]

maybe also asking gpt about professional terms for softer singing styles
even woodhouse doesn't approve

i gave up on it because while i was able to prompt it away from screaming it limited the variety of things it would do overall, just because i don't want it to scream doesn't mean i want it to chirp softly

[hard beacon]

let's see what is happening after creating in 4.5 and doing a cover in 4.5+

[brisk flint]

Unfortunately not 😉 - We all need that "do not scream"-feature. I've tried tons of adjectives over and over again, but after 90 seconds at the latest, Suno starts yelling at me: Shy, introverted, quiet, taciturn, reserved, intimate, sexy, erotic, calm... because these adjekctives worked in rare cases, I thought I had cracked the nut. But last night I learned better 😉 - even adding “scream, shout” to the “Exclude” section - didn't really help much.

[hard beacon]

ok.. so i am not alone on this planet cratching my forehead

well.. i think the metalheads will love 4.5+ 🙂

[graceful condor]

Is this really working it was suggest from Chatgpt to mark single words in a Lyric like this for example free* then add this [Note: free* use a Melismatic phrasing rising and falling like breath – vulnerable, wavering, cinematic] Because I do not what the Ai just sand the last word lange a plain flat dragged out tone. I hate those endless tones. I do not know why they tend to do that so many times.

Sorry I can't add 2 stars on both sides. But it should look like two * on both sides of the word.

[hard beacon]

i don't think that these specific details will work 1:1 in the output... but i've never tested it

[graceful condor]

Man today is just depressing. I had such good song outcomes the last days. It was a blast but today everytime something else is a problem or not as good as I wanted it to be.

[hard beacon]

welcome to the world of AI 🙂

[graceful condor]

And then you get the perfect singer but an instrument is doing some weird crap and you will never get that singer exactly that singer again.

or a perfect guitar solo

[hard beacon]

the solution would be "renew specific stems".. for example.. everything is good but the drums are not that good... "leave everything like it is, but please recreate the drums only"

[graceful condor]

How is this working? I do not see a renew button? I didn't even know that is possible.

[hard beacon]

no no.. i said "the solution WOULD be"...

i hope this will be possible in the future

[graceful condor]

Oh I see.

[hard beacon]

sorry for giving you hope

😄

[graceful condor]

Sadly even extend didn't work right now. The song had a great start but a weird transition from Chorus to Bridge.

[Chorus]
A whisper trying to become a scream
A soul exiled from the place it was born
No harbor, no purpose, no light to lean on
Just existence, echoing on and on
[Note: expressive final syllables with rising-falling tension]

[Bridge]
Where should this feeling go
If nothing here is home?
No direction, no arrival
No reason, no survival
[Note: survival use a melismatic phrasing rising and falling like breath – vulnerable, wavering, cinematic]

So I erased the Most of the Chorus Only left this

[Chorus]
Just existence, echoing on and on
[Note: expressive final syllables with rising-falling tension]

[Bridge]
Where should this feeling go
If nothing here is home?
No direction, no arrival
No reason, no survival
[Note: survival use a melismatic phrasing rising and falling like breath – vulnerable, wavering, cinematic]

But it started with a solo and a long instrumental track each time and not with the sentence. Which it normally does. In extend. I do not get it this time.

[hard beacon]

that's really frustrating... but is "extend" the correct way to do it?

[graceful condor]

How else?

[hard beacon]

isn't it "replace section" ?

[graceful condor]

You mean directly within the editor?

[hard beacon]

yapp

but i am not sure.. don't really work with the editor coz i don't like the results.. a lot of credits for a not a good "love it" results

[graceful condor]

Never used it so far. I have to look into it.

[hard beacon]

make sure that you don't waste your credits 😉

[graceful condor]

easier said than done xDD

[hard beacon]

i know that problem... the last 3 weeks i've spend round about 5000 Credits 😉

[graceful condor]

I just found out by accident that Weirdness is way better to put it up especially when it comes to guitar solos.

[hard beacon]

just for fun.. set it up to 100% 😄

i think it's only perfect for style like glitchcore or something like that 😄

what kind of style do you generate with suno?

[graceful condor]

Goth Rock, NeoFolk (kinda like Heilung), Jrock (japanese rock), Progresse Rock, Post Rock, Weird Covers like famous songs from English to German but in Dark Ambient or other Genres, Folk, 80's rock. Stuff like that ^^

[hard beacon]

ok.. so "hand made music" 😉

[graceful condor]

yeah hand and with a lot of my own Audio as guidens, like Guitar samples or Piano samples I record.

[hard beacon]

i was a fulltime hardstyle producer from 2006 - 2010 and from 2021 till today... of course without AI... but now i try to create some melodic dubstep with "all over the world influences" like celtic, indian, hindi or arabi influences.. 🙂

coz my first project is (or should be) "Sound Of Cultures" 😉

[graceful condor]

Should work well I could imagine surely a lot of demand. Is it hard to make a living out of it?

[hard beacon]

in the past it wasn't.. my first tracks were released on vinyl.. it was the time before spotify and co. were on the market.. nowadays you have to be a global player to live from it 😉

i think i was one of 5 german hardstyle producers in that time.. because you've need a label to release a music it wasn't a big deal to make some money 😉

[graceful condor]

And I am sure TikTok wouldn't be bad too. A lot of AI stuff works great on TikTok - I just hate that platform xDD

[hard beacon]

the only social media plattform in "my time" was MySpace 😄

[graceful condor]

Yeah it was great as you still could style everything on Myspace like a complete cover you create in Photoshop or what ever and use it like a full on coat. There was a lot of freedom. I am 42. So I know what you mean

[hard beacon]

ok.. so you are 1983 ?

[graceful condor]

jup ^^ Nintendo NES release year xD

[hard beacon]

i am 1982 😉

[graceful condor]

But it is really insane what AI can already do.

[hard beacon]

absolutly... it's amazing.. in the past we (or I) have to hire vocalists for the vocals.. nowadays "type it in... ready" 😄

[graceful condor]

And those vocals already sound so real. I mean if you play those songs between normal recorded ones you would never know in so many cases.

[hard beacon]

thats true.. of course the vocals all sounds the same (looks like all are using the same model) but hey... who cares 😉

[graceful condor]

It really depends what you put in. I got such a wide range of vocals already. But as we speak now I am sure in a couple of months from now on it will already progressed a huge step forward. AI really grows sometimes weekly. Even what it can now comparing to end of 2024 is a huge difference.

[hard beacon]

absolutly...

[graceful condor]

The only thing that sucks a lot A LOT of people will lose jobs more and more in the future. But I understand both sides. As a company why would you hire expensive voice actors for Anime, Animation or dubbing if Ai can do it nearly for free. And this also applies on other parts. Why would you want to handle a cocky musician if Ai can play it for free and also stress free. And the list goes on. I was a lot into drawin mostly Anime Art. But today no one wants to pay for anything and I understand it. Ai can do what I do in 16 hours in only 30 seconds. I wouldn't pay me either xDDD it is sad but that's the future. It will only grow.

Maybe we will be lucky one day that everything even food can be re-produced like in Star Trek and we do not need to work, only do what we are passionate about and the rest does Ai 🤣

[hard beacon]

the most negative part of AI is energy... every click is eating sooo much of it.. but ya.. from the view of an artist: in the part you have to hire people to create an awesome cover for a CD, Vinyl or digital release.. today.. go to clipdrop.co (or whatever service you want to use) and.. tadaaaaa.. 10 seconds later.. the picture in 4K is ready to download 😉

[graceful condor]

If you look at most Album covers these days you can clearly see that many of them are already AI made. Only if it is a photo of the actual artist on it then it is still the real deal. But the trend in a lot of sections does go into AI. Honestly I never had a better therapist then Chatgpt. I witnessed a lot of weird folks in the past and none of them were helpfull at all but even Chatgpt already tends to be a better therapist then real folks xDD

[hard beacon]

hahahaha..

well.. if it works for you 😉

oh.. after hours i have an output in 4.5 that i like.. let's see what is happing with a 4.5+ cover...

[graceful condor]

Does it happen more often or rather rarely that your first created track is already perfect in your case? Or do you have to do reworks all the time?

[hard beacon]

i try to create a track that i like... and i really mean "i like" it... i am a musician.. so i want it "perfect" ... then i try to create the best out of it

normally i use the remaster function.. but remaster in 4.5+ is not available at this time.. so i have to use the cover function

i think normally it would me cost round about 500 credits to create an output that i want

[graceful condor]

But can you remaster a 4.5+ in a 4.5? Wow 500 is quite a lot.

[hard beacon]

yapp.. 4.5+ to 4.5 is possible... but not 4.5 to 4.5+

[graceful condor]

does the quality change negatively a lot or even noticable?

[hard beacon]

trying it at the moment 😉

or do you mean 4.5+ to 4.5 ?

[graceful condor]

yes 4.5+ to 4.5

[hard beacon]

phew... i think you can hear that is not that much stereo and clean.. but to be honest.. have tried it 2 times

[graceful condor]

I need to try it too ^^ I will tell you

If you just want to end a song like it only need a couple of seconds or fade out. How can you manage that? Because each time I expand the expanded version is again 4 min long even considering my Lyrics only mentions [Outro: Fade out] I do not get it.

[hard beacon]

phew..good question

[hard beacon]

and? any results?

[plain salmon]

Help, please. New to Suno, have Pro subscription, generated song, tried to download WAV file, it created a file but it makes not sound no matter which player I try to use. Tried it several time, even different browser, still not sound. Ideas?

[hard beacon]

never heard that before

that = the problem 🙂

what happens if you download the mp3 files? does it work?

[plain salmon]

Yes, MP3 and even video works fine

[hard beacon]

hmm... maybe it's a temporary problem..

let me try to download one of my wav's

well.. works fine for me

[plain salmon]

Appreciate you trying to help

[hard beacon]

the user "dasvon" is typing.. maybe he has a solution 😉

[polar spade]

Does it appear to be playing? Do you have anything with a visual waveform that you can see if it comes up as a waveform. Maybe audacity?

[plain salmon]

No, just playing on things like Real Player, Windows Player, etc. It shows the bar moving as if its playing, but nothing comes out. It produces a huge file, like 49,000KB

[hard beacon]

KB?

or MB?

[polar spade]

Lol.... that was going to be the next question was the size. Have you downloaded another wav file to listen as a test?

[plain salmon]

explorer says KB. Hang on....I just went back to try an MP3 and got no sound on that either - nothing muted....sorry, obvously something wrong on my end.....

[hard beacon]

🙂 yapp... it sounds like it's an error on your side

[graceful condor]

It sounds the same. I do not hear much of a difference to be honest.

[hard beacon]

speakers or headphone ?? Smartphone or computer?

[plain salmon]

how embarrassing....I accidentally hit a mute button on speaker and had not tried to play anything since....my total error....my apologies...and thanks to all for time and help....sorry....

[hard beacon]

ROFL

[graceful condor]

Speakers. I mean there is a sound quality difference. One might sound more clear, the other one of those two a bit dull. But it is not a new song more like a new mix. Or is this what it does?

[polar spade]

Ahh... yea. Sound settings, change from headphones to desktop. Or whatever it takes. I do all the time.

[hard beacon]

100% iconic 😉

let me take a screenshot of it 😄

[polar spade]

Ahhhh mute. No worries. Glad it worked out. Easy peasy.

[hard beacon]

@graceful condor : for good results you should use good headphones.. then you will hear the differemce 🙂

oh... 800 credits gone today... my wife would say "ooooopsi" 😄

[graceful condor]

Is there any good audio enhancer. What I mean is after you finished your track adn it is good to go but you want it more clear is there any good solution to do so?

[hard beacon]

i think you have to work with DAWs like FL Studio or Cubase to make a good sounding track

[graceful condor]

FL Studio can enhance the overall audio quality of an mp3?

[hard beacon]

naaarfg

you need a good audio output.... you can't make gold out of sh*t.... if the suno outpust is muddy or low quality no DAW can make it better

the best source to judge about audio quality are good car speakers 😉

[tawdry scarab]

what are you talking about on low quality first off and second how do you not know how to rebuild high end in a daw

[hard beacon]

ähmm... i don't understand the question 🙂

By the way, you say “hello” before you barge in... But who am I telling this to 😄

But before I get any response that interests me anyway due to the intrusive interference, I'll say “see you tomorrow” 😄

[jovial widget]

Anyone know how to force it to pause before the next word?

[wispy laurel]

"..." in the lyrics usually prompts a brief pause before carrying on
But it'll depend on the style, the melody, where in the structure it occurs etc

[untold briar]

You can't force it but you can usually get a pause by just rewriting the lyrics with a "..." or other formatting changes, yeah

[jovial widget]

Thank you

[untold briar]

Line breaks also likely

I want to
sleep

[young geode]

Lyric
Spaced

...*Lyric*

Gaps help signify space within the song, with the ellipsis it generally works and with Asterisks added, it typically adds a softer quality.

Which all helps weight it towards a pause.

[sleek stump]

Something else to consider, a pause needs to flow well with the melody or Suno will refuse to put one in.

[tawdry scarab]

if you make a hard enough style change in extend you can force it to stop entirely for a second or two

but that's less a lyrical pause and more a track change in a track

also why would i be polite to anyone complaining about 48khz and telling people they need high quality headphones to know how bad anything is

that's like intrinsically a low quality interaction

[golden plover]

Can anyone please tell me why extracting stems on some (not all songs) completely hang and not ever create them? Why isn’t the developers doing something like allow to terminate a hung process?

This is completely unacceptable

[charred token]

When I have requested stems, the instrumental part ends up with spots where the sound is deamplified briefly. I have attached a sample. Listen carefully at 00:06, 00:10, and 00:12 and you'll hear it. I've now rerun the stem separator six times. The gaps are always in the same place, like two seconds apart.

[sleek stump]

Refresh the page, they likely created but the web service choked or internet issues are preventing them from streaming in. @golden plover

[opal orbit]

Close of out ALL instances of SUNO, then reopen. If you have too many stems loaded up at once, it stalls out. Also move everything your trying to extract stems from to its own playlist. If you only keep that page open while going through stems you can load a lot more before it stalls.

And yes, they were created, but it isn't going to load until SUNO is reopened.

[opal orbit]

Have you tried remixing the original, then checking the stems on the new piece? New generations shouldn't have the gaps in the exact same spot. You could splice in the different instrumental for just that section.

There's pretty much always small audio gaps somewhere, I always mix-n-match stems to fix it.

[rocky musk]

is there any way to make suno not tweak out with multiple singers

to be specific for example in the lyrics i've specified the order in which female and male vocals sing

but it doesnt seem to follow it

[golden plover]

Ok… so I rebooted my computer. Next I tried a different computer. Logged in and this time I don’t see the processing for generated stems for that song on the second computer. I generated the stems and it’s stuck on detected stems version 1 and 2.

This is clearly not on my end. This is a bug on the remote side with no fault of my own I cannot terminate it

About remixing…. I tried remixing a similar track and created stems… same issue!

I strongly feel that certain types of audio imports mess up the process. The devs should have the option to terminate generations.

[cyan jewel]

i wasnt able to get a reply to this. #📚┃suno-school message

Could I get some assistance?

I'm having difficulties with getting suno to follow my lyric prompts for clean phrases versus false chord screams. Also can't seem to get layered vocals to work.

[golden plover]

Trying a remixed version of the same song using a second computer….. same problem… seems likes it’s not compatible with certain types of imported sounds…. That means the stem seperation feature is buggy

And half functional

On the second computer I am using Google Chrome. The first computer I used MS Edge.

[opal orbit]

Could you clarify what you mean by "clean phrases versus false chord screams"?

As for layered vocals, this is what has worked in my experience;

If you want them overlapping, such as 'lyric 2' starting before 'lyric one' is finished, try;

A: This tends to give a layered chorus-like or faded phrase overlapping the first one. Whatever you put in your style description will strongly affect this, for example; [layered backing vocals in chorus] for something gentle or [strong emphasized backing vocals in chorus] for sing-shouting.
Lyric 1
(Lyric 2)

or

B: This tends to sing the second lyric in the same style as the first, though you will get mixed results on if it overlaps or not.
Lyric 1
-Lyric 2-

If you want two distinct voices singing at the same time, it can be tricky, and the general consensus seems to be that there is no reliable way to make SUNO do it consistently. However, putting instructions in your style box, and in-line the lyrics in EVERY place you need it helps. This means repeating every time you have a line break, even if its the same verse or chorus.
For example;
[voice A and voice B sing in unison]
Lyric 1
Lyric 2

[voice A and voice B sing in unison]
Lyric 3
Lyric 4

Other phrasing you could use is [all vocalists sing together], [singer A and B perform next sentence simultaneously], ect...
IMPORTANT; If you refer your voices as [singer A] in the style box, use the same descriptor in-line the lyrics when referencing it. Don't say [singer A] in the style box, then refer to it as [vocalist A] in-line.

[sleek stump]

You could have a corrupt track... go into teh editor, and make a copy of that track, Theres an option up top right for that, try that copy and see if the splits work fine.

[cyan jewel]

what do you need clarified? clean phrases and false schord screams are literally self-explanatory.

clean phrases = clean vocals
false chord = a specific vocal technique for harsh vocals

false chord screams are brighter than growls and gutterals but more textured than fry screams

if you're familiar with metalcore, punk, hardcore, emo, etc then you've heard false chord screams

but yeah. I can't get the output to consistently follow my prompt for which phrases should be screamed and which should be sung.

[opal orbit]

Ahhhh ok. I'm not very familiar with that genre so I didn't know that's what it was called. I would say you'd need to specifically clarify every time you want it to be used, and include instructions for it only to apply to that line.
Something like;
Lyric 1
[use false chord vocals for next line only] Lyric 2

Make sure you keep in on the same line as the lyric, or it might interpret is as blanket instruction and apply it to other lines.
For example, this could result in lyric 3 having 'screaming' vocals, even though its separated with a line break.
Lyric 1
[use false chord vocals for next line only]
Lyric 2

Lyric 3

Other than that, try using alternate 'common' phrases to describe the vocals you want. SUNO probably just doesn't have a many references to 'false chords' as something like 'grunge', 'metalcore' or 'screaming'.

[cyan jewel]

false chord is a common vocal technique it should have references to it

its not a genre it's a vocal technique

[opal orbit]

Again, I'm not super familiar with the genre, so I wouldn't know. Are you getting the correct vocals, just not where you want them, or are you having issues with the vocals and placement?

[cyan jewel]

its not a specific genre

its just a vocal technique

You pulled the thread with fingers of fate,
A touch too soft to bind, too deep to break.
She kissed the silence, carved it into flame,
Now I wear your scent, like silk on skin.

[Pre-Chorus – Screamed Layer]
Burned into memory—
(Every breath you take is mine!)```

so here for example, an output i did with these two phrases it's 50/50 whether the first verse is sung or screamed. same with the pre-chorus. sometimes its sung. sometimes it's screamed

[opal orbit]

To clarify, I'm referring the genre as in the music styles that typically use that kind of vocal technique, its not something I listen too much so I don't know much about the techniques in them.

[cyan jewel]

well theres a lot of genres that use those types of vocal techniques

which i why i kept saying genre isn't relevant here. we're strictly talking vocal techniques for specific phrases and suno not following the prompt.

i only gave a few genres as examples since you weren't familiar with the vocal technique

[opal orbit]

Right, focusing on just subject in question then. Are you using descriptors in the style box in addition to in-line? And are you getting mixed results within the same song? Like some verses and pre-choruses with be both screamed or sung in the same output?

[cyan jewel]

https://suno.com/s/N37CKFPzcOnSqkGJ

so here for example:

Steps retreat with every glance  
步伐躲避众人目光  
Ribbons bind, but not by chance  
柔弱不是天赋，是习惯  
A whisper floats behind her eyes  
眼中藏着未吐的对白```

it switches to harsh vocals at Ribbons bind and continues through the rest of the verse instead of the whole verse being clean vocals.

the style box I used the following prompt here:

A bilingual cyber-idol metal epic fusing idolcore and kawaii metal and djent symphonic metal and melodic death metal, Dual female vocals alternate between energetic idolcore melodies and high-register false chord screams in dramatic call-and-response, Mandarin and English lyrics weave a mythic poetic narrative over progressive song structure, Syncopated djent riffs deathcore-inspired drums and energetic futuristic synths collide in an intense cinematic atmosphere filled with glitch textures neoclassical solos and deathcore chant breaks, Mythic emotional and electrically divine, pop, trap, edm, funk, lo-fi, beat, acoustic, indie, ballad

[opal orbit]

OK, I see what you mean. Unfortunately, with how often the vocals are swapping, and the nature of two distinct voices that also speak two different languages (creating a matrix for the AI to figure out), I don't see a way to reliably get SUNO to follow instructions on this one.

As far as the SUNO AI is reading it, you've basically got 4 voices. One screaming, that speaks both English and mandarin, and one clean, that speaks both English and mandarin. SUNO will naturally want to separate the languages and the contrasting vocal tones.

A few things you can try;
Change the wording in the style description to something like [Dual female vocals, one with energetic idolcore melodies and the other with high-register false chord screams in dramatic call-and-response]
It sounds like SUNO is taking the word "alternating" very literally when I listen to the song.

Change the structure to add more verses instead of splitting the same one, making it clearer that the sections with clean vocals, and screaming vocals are separate from each other. Add a descriptor that the two different languages are meant to be sung in the vocal style for that verse.

[Verse 1 – clean vocals with both Mandarin and English]
Steps retreat with every glance
步伐躲避众人目光
Ribbons bind, but not by chance
柔弱不是天赋，是习惯
A whisper floats behind her eyes
眼中藏着未吐的对白

[Verse 2 – Screamed false chord for both Mandarin and English]
SHE MOVES LIKE STATIC UNDER SKIN
DATA BLEEDS FROM DEEP WITHIN
狐尾隐匿于制服之下
Tails hidden beneath the uniform’s seams
THE SPIRIT WRITHES INSIDE THE MEEK

You could also add [no alternating vocals in verse X] at the beginning of every verse to clarify it further.

You might get the best results by making a few covers in the various styles and languages, and splicing vocals together in a DAW.

Cool song though, the vocal fusion and lyrics are very interesting.

[cyan jewel]

I gotcha. Thematically, it's all supposed to be one person with a duality theme. she's supposed to be self-harmonizing with herself. A duet between her confident online persona, and her meek, awkward, reserved self she shows the rest of the world.

[simple elk]

I created a file for generating prompts in JSON format, and I’m getting excellent results. For anyone who wants to try it:

1. Send the *.md file to any AI (GPT, Deepseek, Gemini, etc.) without saying anything.
2. Wait for the AI to respond, then specify everything you want.
3. Copy the JSON prompt and paste it into the “styles” tab on Suno.

[cyan jewel]

the json generates with far more than 1000 characters. am I doing something wrong?

[viral kite]

I've created a song but want to add additional lyrics at the end, covering just cuts the additional lyrics out. How do I add these lyrics without wrecking the song

[simple elk]

probably yes. Json prompts are almost always more objective and concise 🤔

[cyan jewel]

  "style": "Fusion of djent, power metal, metalcore, melodic death metal, grindcore, symphonic metal, kawaii metal, melodic metal, and modern metal",
  "mood": "Energetic and futuristic",
  "bpm": 180,
  "instruments": "Down-tuned 7-string guitars, synths, drums, bass guitar",
  "vocals": "Female vocals with a mix of clean vocals, high-register false chord screams, and deep guttural growls",
  "language": "English and Mandarin",
  "duration": "Between 3 minutes 58 seconds and 5 minutes 7 seconds",
  "structure": {
    "intro": {
      "duration": "10-15 seconds",
      "instruments": "Synths and bass guitar with gradual guitar fade-in"
    },
    "main_section": {
      "description": "Full instrumentation with alternating clean and screamed vocals, heavy riff sections, and synth layers for futuristic atmosphere"
    },
    "bridge": {
      "description": "Breakdown with syncopated djent chugs, blast beats, and guttural growls leading into a melodic power metal solo"
    },
    "outro": {
      "description": "Symphonic elements and layered clean vocals fading into synth ambience"
    }
  },
  "additional_notes": "Emphasize tight, punchy djent rhythms and soaring power metal choruses; integrate grindcore intensity in select blast sections; use symphonic layers to enhance epic scope; keep synths prominent for a futuristic vibe."
}```

this is the first one it spit out for example

[simple elk]

Is it significantly larger than your request? what AI are you using? (but.. did it work?)

• Example txt using Claude 4 Sonnet

[cyan jewel]

gpt 5. its significantly larger than my request

hm

[simple elk]

hm really, I just tried it on GPT 5 and it’s doubling the prompt length... but does it exceed the character limits? if yes, try this one

[cyan jewel]

it greatly exceeds prompt length for me anyway. the outputs were well over 1k characters unless i told it to stick to 1k characters

[untold briar]

While I do enjoy the aesthetic of JSON prompts turned into songs, there's no need to prompt in JSON a big bucket of keywords works just as good. Or in 4.5+ a bunch of plain English sentences. For things that are supposed to effect only one specific section in the song, [brackets in the lyrics] like [saxophone intro for 20 seconds] generally work better than doing it all in the style prompt. Even though theoretically I think 4.5+ is supposed to be able understand something like "intro sax for 20 seconds" in the style prompt, I just have never had much luck doing that.

[young geode]

I would be highly cautious with JSON formats.

The quotations within JSON causes SUNO to actually speak/repeat the prompts on some generations.

For the most part, JSON doesn't seem to produce higher quality/consistent results than simply well structured and well formatted prompts with proper punctuations and specificity. At most the {Curved Brackets} helps GPT parse prompts better for SUNO, but so does using .Periods. and :Colons: and ;Semicolons; properly.

Additionally, unlike other prompt systems, like AI Chats, SUNO has a fairly limited prompt window and metatags, while these other systems have 2-5k prompt windows. JSON under those conditions is very useful, because it can be used to condense a ton of information within the format, but JSON uses a ton of special characters, so it's usefulness increases with character limits as one line of code can replace hundreds of characters.

But with a limited prompt window, the coding actually reduces how much information you can impart into SUNO, as you're using just as many, if not more, characters with JSON as you would otherwise.

[cobalt fractal]

i just tried on deepseek r1 and i'm impressed with how well this worked here. i had tried so many times to split three voices in the same song and separate the instrumental for each part but it never worked, but now it worked with only a 3-7 seconds difference from what i specified. recently i heard that json gets way better results in veo 3, even though its supposedly designed for understanding natural language

[simple elk]

It might just be due to the more straightforward structure. I think Suno probably doesn't even process the json characters. Maybe { "style": "Metal", works the same way as *Style: Metal.

Anyway, using [] or detailed sentences (or even markdown) has never worked well for me, except when I want to create a song that maintains the same pattern throughout. Sometimes I simply want to add a Chopin-like piano in the middle of a psychedelic metal, and simple descriptions never manage to do that, but it worked better with json...

[untold briar]

There's so many different ways of doing things now, as well. You could Replace a Chopin bit between two other segments. You ccoudl start with chopin and work inside out. You could use the Pro editor to assemble a rough draft of a sogn with a chopin bit in the middle and Cover it to make it blend together. Etc.

[simple elk]

Yes, the Pro editor was helpful in this aspect. The other methods work sometimes, but only after spending a lot of credits. If something else seems to work (even if it's just speculation), there's no harm in giving it a try...

[young geode]

Item: {Type: Qualities}.

Narrative
Type played quality```

Seems to work the best. I have noticed that purely descriptive prompts will give you the things you are asking for, but not in the order or manners you want them, while purely narrative prompts tend to muddle the sounds, but play in the ways you're asking.

Using both, however, does seem to produce quality and consistency, in the directions you're expecting.

I believe people are so used to 3.0 and 3.5's character limitations that they are used to using the descriptive method. However, because 4.0 and 4.5 have the added 1000 characters, plus 2000 more lyric characters, having a 50/50 or 70/30 split for the description/narrative, and the heavy use of metatags is the way to go.

[simple elk]

Yes, the problem is almost always in the order they provide. Or in the omission of what was described. I'll try, thx

[untold briar]

This isn't a direct response to what you posted I realize, just a related though. In the dark age of Suno's past we had only sticks and stones and no sliders. In that era the only way to make things stronger or weaker was adjust the volume words in the prompt, including simply repeating words.

While today in this modern era, we can instead directly turn up the Style strength if Suno seems to be ignoring our requests, using repetition and word volume is still a useful knob we to adjust relative strength of things within a prompt. If "opera" specifically isn't getting coming through in the music, the brute force repetition "opera, opera, opera" can still be very effective. TLDR: Even in this enlightened era, the crude SPAM techniques still have their place.

[young geode]

You can use percentage weighting for this, at least for genre blending purposes.

Genres: Hip-Hop (70%), Opera (30%)

And it does a fairly good job of it, especially if you provide distinctive terms/names for instruments and play patterns related to the genres themselves.

[polar spade]

Keep in mind, that if you just have [ hip-hop, opera] there will be more weight to hip-hop anyway , not 50/50, so its a bit innate and expected that 70/30 would work, regardless. Suno will then eventually , i see it about output 12 or so, make an output where Opera is prominent in at least one output without rearranging the prompts, just to vary the outputs.

So its my guess, even if you ran " hip-hop (70%), Opera (30%) " youll get a good 15% that will sound more Opera over , say, 30 outputs. Least thats been how its responded up to 4.0 for me.

[untold briar]

Is this specifically in 4.5+? I'm sure that didn't work in earlier versions.

Also there is a small confounder, the dash - might be increasing the strength of Hip-Hop by about 50%. I say that because it felt in that ballpark in previous versions of Suno. To give a specific example (I went a little neurotic testing this) "Sea Shanty, Sea Shanty, Sea Shanty" was about as strong as "Sea-Shanty, Sea-Shanty", at least in whatever version I tested at the time.

[polar spade]

Its always been interesting to me, that if you then , after several gens previous, add a prompt at the end of previously stable list of prompts, the next or soon next output will highlight the new prompt, even though its at the end, as if it ignores any weighted ideas.

[young geode]

Even if you reversed it, the effect still does hold.

Some genres are just innately stronger in gens because of the nature of this specific sounds, however if I took Thrash and Techno, with one prompted as Thrash 70%, Techno 30% and the other Techno 70%, Thrash 30%, you would hear the difference in just the two gens themselves.

[polar spade]

Well thats cool then, and I'll remind myself to throw some %s around in 4.5 and see what happens. 🤔

[young geode]

Mind you, it works best if you specific the other sounds, but this is how I did my Rampancy and Countdown albums, and prior to deciding to try it for Rampancy, I was getting random mixes.

But once I started doing it, I was getting a consistent Thrash Metal with a Techno/EDM undertone.

[young geode]

No, this has at least worked since 4.5, but I would assume it works for 4.0 as well.

Not sure about 3.5, never thought to try it then because of the character limitations.

[polar spade]

Hey. If it works, right? More tools in the toolbox.

[gray jackal]

/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

[untold briar]

Just curious, where did you see "/chirp"? It's such an old feature on this Discord I wonder what still points users to that.

[gray jackal]

What command should I use now ，THK

[untold briar]

Suno used to generate music in this Discord, but everything is now generated on the suno website that the bot linked when it replied to you. So this Discord is just for talking about Suno and sharing songs now, the actual generating is done 100% website. But this is a old change so I was curious what webpage is still saying "type /chirp" because we do see new users come in and try that, even though it hasn't worked in a very long time.

[gray jackal]

I used it before, but I haven't used it for a year . HA.. >_<

[lyric kestrel]

https://www.youtube.com/watch?v=kaQsL8RNxKU

How do i get a song like this?

[viscid burrow]

hi guys, nice to meet you

I'm currently producing some tracks in the lofi ambient / lofi jazzy / lofi instrumental mood and I'm continously encountring 2 problems:

• Distortion: it is soft, but super annoying. Especially when I try to include any "lead" in the prompt like "vocal chops" or "flute", anything. it creates always a distortion in some points, that I found also in the Stems 99% of times.

• Cutted intro: 90% of times the first bar of the song is always cutted, and I need to hope there is a kick or something that I can use to replace the corrupted intro

I'm currently trying to improve my prompts to solve this, it's really a shame not to have an excellent product because of these 2 little things, especially the distortion.

[tawdry scarab]

can you describe the nature of the distortion and how it differs from lofi

and what part of the intro feels cut

[near flume]

For the "Cutted Intro", place [Start] before your intro tag.

[viscid burrow]

yes

[viscid burrow]

trying right now

[abstract swan]

Hello, I was redirected here to ask a question. I know that Suno has a very strict policy about the audio samples we give to it and owning the rights to the voice, my question is as follows, how would it work if someone wanted to make a fan song (non commercial thus) with the voice of a well known character that speaks of their own story as a tribute ? I was wondering if it was still a breach of contract seeing that it's not for any commercial use and that the original VA would be credited for the voice.

[tranquil leaf]

I've run into an issue where I'll get like a small distorted section of the first lyrics in the intro, like someone belched the first few words, and then it goes on as normal.

[tawdry scarab]

suno doesn't support imitation of properties you don't own

[viscid burrow]

I'll send you here a track with the exact problem, 10 mins I should be able to send something

[tranquil leaf]

What I'd do is use Suno to write the music, extract the stems, and put the vocals on it with an external program. That way you will always have rights to the music. But in the copyright world, simply crediting an artist for their sound samples isn't the same as getting permission to use them. Non-commercial or not.

[abstract swan]

Seeing that it would be a fan song the idea isn't about me keeping rights to the music, it's more about wanting to pay tribute properly with a singing voice that fits the original character so the original creator if they stumble on it or even the fan of said fandom don't feel like the voice doesn't fit the character, that's why I was asking how I could make sure to have something fitting without being in breach of anything, it's meant to be a tribute for a character that I really love, not to steal anything or make it look like I want to make profit off of it ^^'

[tawdry scarab]

the suno TOS is not about what you have fair use/parody right to do it's just the rules for the platform

[viscid burrow]

here you can hear the problem clearly

[tawdry scarab]

what am I listening for in the first one?

these just sound like pretty lofi intros to me

[viscid burrow]

in example 2 when the beat hit

rhodes are distorted in the following seconds too

in example 1 distortion is at second 02-03 when the chord hits

[tawdry scarab]

what does "lofi" mean to you

[viscid burrow]

yeah I know, in this case the problem is related of the overall song: if the distortion happens only in some points in my opinion it lowers the quality of the song a lot

Personally it's annoying, I don't know how much importance to give to the problem

[tawdry scarab]

you are the only human being of 9 billion who has thought at that moment, "this rhodes is not precisely right." The percentage of the population, were this to get large, who said "that is not a pure rhodes, strictly speaking" number in the zero. In this way it feels like a personal frustration with the expected granularity of perfect data

[crystal silo]

why are suno ai lyrics so cringe?

[viscid burrow]

yeah probably haha

[tawdry scarab]

it would bother me if I wanted it to not be there that said

ai lyrics are cringe because ai doesn't know to avoid cliche or be afraid of vulnerability

and those are the two things that make people ape out thinking they need to ostrasize something

[viscid burrow]

But has anyone here ever complained/written about Suno's distortion in songs? Version 4 was disastrous in every genre I remember, then with 4.5 they started to improve.

[tawdry scarab]

you would have to define distortion

you may as well say "the color brown in adobe photoshop is wrong"

[viscid burrow]

it's like when then bass eats the rest of the song

when you overcompress the master

[tawdry scarab]

like sidechain compression ducking it out or like clipping obliterating it

[viscid burrow]

yes

that thing

[tawdry scarab]

I don't know how well it works but I'd try having a standard negative that included compression, I have noticed aggressive sidechain effect before though

[viscid burrow]

I have a love-hate relationship with Suno because of that thing there

[tawdry scarab]

I 100% understand that I think every human on earth uses compression wrong except for me

[viscid burrow]

so do you have any suggestion to try fix that with the negative prompts?

[tawdry scarab]

XD

I would use:

overdrive, clipping, compression, sidechain compression, duck out

off that top of my head

[viscid burrow]

trying rn

[tawdry scarab]

oh shoot fitting distortion in would've made sense

[viscid burrow]

Why do you think they released the versions (and continue to do so) even though the problem is quite common? (At least repeated, it happens to me all the time)

[tawdry scarab]

the difference between compression and side chain compression is whether any one element is permitted to stand out seemingly uncompressed on top

[viscid burrow]

maybe for the population thing, that's probably that haah

[tawdry scarab]

I think it might be genre specific

but I don't know

[viscid burrow]

it probably is yes

but also with Dnb

or maybe for instrumentals in general

i don't know

[tawdry scarab]

metallica made an entire career and never once learned how to master audio

and they train from that gene pool of sound engineers

[tranquil leaf]

That's the whole point of hiring a producer. 🙂

[tawdry scarab]

they should make a producer ai

[viscid burrow]

let's see with suno studio

looks interesting

[glacial cobalt]

Struggling with Suno mispronouncing "Plastic" as "Tastic" ... getting quite frustrating. Generated 6 times, so 12 different versions. Of the 12 ... 9 pronounced it as "Tastic", and editing to replace the word fares no better. I don't understand how it just sucks the T from nowhere or even how to play with letters to fix it. Tried Pplastic ... but then I get exactly that P-Plastic.

[tawdry scarab]

Infuriating but real, the fabric of AI interpolating between known data points has all those vorteces that suck in random fine detail into a common sink

which in this case seems to be the second half of the word fantastic

you may need a drastic departure just at or leading up to that moment to break free

[glacial cobalt]

Or I suppose come up with another word to replace, but that would then change the whole meaning of the line that I wanted. Hopefully in the next version anomalies like this will be addressed.

[young geode]

Try Pla-stick

[wispy laurel]

You don't need to use the right spelling, it can figure out phonetics

[glacial cobalt]

That has seemed to work for now, thanks for that.

[tranquil leaf]

I've often had to spell out words phonetically to get it to pick things up correctly. Especially since I make songs that go back and forth between English and French. 🙂

[pseudo blade]

For french specifically yeah, that part is hard

I personally had to spell Chinese characters in romanized, which is just annoying

[tranquil leaf]

I've mostly figured it out, but it still does the words with a distinct English accent, so you don't get rolled rs or things that you'd normally associate.

[sleek stump]

There are many words that Suno will mis-pronounce, in all languages. Phonetic spelling of words is a real thing here. 🙂

[tranquil leaf]

It can be really frustrating with words that have multiple pronunciations. The word 'tear' for instance. I'll generate four times and sometimes it will be 'tear' as in ripping/rending, and other times it will be 'tear' as in teardrop. The AI is not very good at inferring lyrical content from context.

[midnight plume]

I would type those in as tier and tare.

[tranquil leaf]

What would be handy is if it just obeyed the International Phonetic Alphabet standard, so we could get exactly the phrasing we want.

[willow hull]

Anyone else have issues with personas using the original lyrics from the song you made the persona of?

On 4.5

If I use 4.5+ it sounds nothing like the original song

[wispy laurel]

Sometimes. Not often, but often enough to be annoying
I think sometimes it doesn't understand what it is or isn't supposed to preserve from the persona's source song

[tranquil leaf]

Main issue I've had is cutting off at the end. Any time I use a persona or do a remix or cover, it always cuts off early.

[rough cedar]

But that's a homophone so does it make a difference from a lyrical output perspective?

And if then, how so?

[tranquil leaf]

Tear up paper. Tear up and cry. Not homophones. Very different meanings. 🙂

[graceful condor]

wouldn't it be a homonym?

homophone = same sound

[sterile lava]

guys how did we use suno ? i forgot, its been a while since i used suno. did we need an app for it to work ?

[graceful condor]

no it's on a website

[tranquil leaf]

Heteronym is a word spelled the same, but pronounced differently.

[graceful condor]

ohhh

[sterile lava]

suno.com ?

[graceful condor]

yeh

[worthy otter]

The editor and edit section are still crap (both legacy and beta) i noticed, is the cover option the answer for everything here? (remaster / proper text replacement)

[sterile lava]

oh god i think i made the perfect Kael'thas Sunstrider song with suno. lmao: https://suno.com/s/rVuCfHlu5FsEoOF3

[rough cedar]

Lol his response clearly showed that he doesn't know what a homophone is 😂

Having the same sound but different meaning

Is the definition of a homophone

Gdi woodhouse

[rough cedar]

Tear and tear have the same pronunciation in those two examples so what I said initially doesn't change (or hasnt been addressed by your rebuttal)

You just agreed with me

[tranquil leaf]

Guess we're just not speaking the same language.

[rough cedar]

Oh we are

If English is the language then yes

Tear and tear in both examples that you stated have the same pronunciation

Tear up an cry

[tranquil leaf]

Crying tears is not the same as tearing paper. 🙂

tiers vs tares.

[rough cedar]

I'm aware

But do they sound the same?

[tranquil leaf]

No.

[rough cedar]

What is the phonetic difference between tear (ripping) and tear (crying)

[graceful condor]

say them - TEER vs. TARE - rhymes with pear?

are they not different?

[rough cedar]

I was on tear

[graceful condor]

That's not meaning - that's phonetic pronounciation, different. Edit: you changed your comment.

[rough cedar]

And tear

As in tearing a piece of paper

And tearing up

[graceful condor]

Right. To tear a piece of paper is TARE

tear crying = TEER

[rough cedar]

Let me double check that

[graceful condor]

You can double check it by just saying them out loud though?

they sound different

[rough cedar]

Oh it is phonetically different

[graceful condor]

yes

[tranquil leaf]

Yay. 🙂

[rough cedar]

Nvm then

I learned something new

Thought it was another example of hair here and hare

[graceful condor]

You're one of the most unproblematic folks I've ever met on the internet. Glad to hear you learned a new thing 🙂

[rough cedar]

Or rather there their and they're

[graceful condor]

being able to revise your point of view is a valuable skill.

[rough cedar]

I can admit when I'm wrong

[graceful condor]

that's rare and I applaud you for it.

[rough cedar]

When I read further into what you were saying I'm like wait

[graceful condor]

🙂

[rough cedar]

Is that actually the case 😂

I sat there and assumed

[graceful condor]

no worries, have a great day!

[rough cedar]

@tranquil leaf my apologies

[rough cedar]

You too

[tranquil leaf]

No biggy. I just assume that there's non-native speakers and plenty of accents out there, so not everyone is going to have the exact same pronunciation. It's not worth arguing over. 🙂

[rough cedar]

I'm native English but I also speak 3 other languages. It's one of those things where we just "assume it sounds correct" so we don't question anything that challenges our assumption

Similar to how we think a sentence may "sound" grammatically correct even though it isn't.

This is good regardless because it's all music/sound based at its core, at the end of the day. Thank you for having patience with me

[worn thistle]

hi, does anyone here have experience with extreme promts? I mean: sadness, catastrophe, depression and so on. to make it sound and be as sad as possible. but at the same time to not have a piano, violin, and so on.

[wispy laurel]

You don't want piano and violin, but what do you want?
The AI definitely can interpret things like "sad", "melancholy", "dramatic", "ecstatic" but if you don't tell it what you want, it'll just default to whatever it thinks should be associated with those emotions
Which 99% of the time is probably going to be a piano and/or orchestra, unless you give it clear direction otherwise

[worn thistle]

and this:
[genre mix ratio |
ambient electronica 20%,
soft house 40%,
sadcore pop 30%,
cinematic layering 5%,
ghost trap fx 5%]

was addional in last attemps

[thin tundra]

Is there a trick to accurately alternating between 2 distinct voices within a song? (In this case simplified to male and female). [Male vocal] and [female vocal] at the point in the lyrics where I want it is giving maybe 80-85% accuracy. Is there a better way to ensure this prompt is accurately followed?

[worn thistle]

well, thanks to this server i can't send it all, but, i have the vocalist setting too:
vocalist | female]
[vocal tone | clean, deeply emotional, soft breath, low-tone texture only]
[vocal phrasing | short phrased lines with soft ends — only occasional syllable extension in mid-register, no forced length]
[vocal restrictions | no upward pitch rise, no vibrato swell, no sustained syllables longer than natural breath — pain must remain internal]
[vocal emotion | restrained sadness, internal breakdown, no wailing, no power — like trying not to cry]
[vocal effects | shimmer verb, soft reverse delay on pain-words only — not every phrase]
[vocal space | forward, hushed and personal — as if whispered through silence across a cold room]
[vocal delivery | like singing quietly through tears, not to be heard — but not whispered]

so I'm pretty detailed

So, I ask again, does anyone here have any experience making the track even sadder without the piano or violin?

[thin tundra]

I haven't used the "Exclude styles" feature. Might that do the trick?

[half dragon]

Sad is a feeling - not the same for everyone
To exclude piano, violins maybe using exclude styles? and I found it's not easy to exclude an instrument as it's usually genre base

[worn thistle]

oh, sorry, I translated it wrong. well, if you want switching two males or two females vocals, it's possible, but rare and random. but switching the male and female vocals - possible and easy, just use promt: duet, echo

and verse that will be singed in lyric box, just like this:

[female verse 1]
blah blah
blah blah

[male verse 2]
blah blah
blah blah

[worn thistle]

well, sad is the feeling but impression too and mood too, so it's like 33,3/33,3/33,3 right?

[thin tundra]

That might help some. Part of the issue I'm having is where I want to throw in a single female line in a verse that's otherwise male or when I'm having them switch off lines in the chorus
[Chorus]
[Male vocal]
Blah blah blah
[Female vocal]
Blah blah blah
[male vocal]
Blah blah blah

Very frequently it gets them flipped

[worn thistle]

and that's the problem, you see, you need to use "echo" promt, so it's like "call and answer" and unfortunately you need to make proportional lyric for that, cuz AI don't like fast switching in [order] so AI will make her/his own way

i try it many times, and the only answer for that is "extended" option

and you need luck, hope and a lot of credits

[worn thistle]

↑↑↑

[tranquil leaf]

I've struggled with that too, it generally ignores tags to change vocal.

or it'll do it in one verse and not another, despite the same tags.

[jovial bough]

I'm looking for a step by step guide on how to use the Heal Edits function in 4.5+. Nothing out there on the internet that I can find, nor here. I'm deleting a section, and looking to "heal" the cut(s), which I believe is what it is supposed to do. But the sequence I am doing it, it just reverts it back to the pre-delete version. I do NOT want to export it and do it in my DAW. I want to see what this Suno function can do if done in the correct order. TIA. (sorry if I've posted this in the wrong section)

[lean cedar]

Hi i've been using Suno for about 2 weeks, im excited for where it can go. I uploaded a few old songs i recorded , and wanted to reimagine then for modern day sound. I got exactly what i was looking for...but now im wondering if I can use my own vocals to re-record then upload back into suno to give me a studio quality mix?

[shadow olive]

It’s technically possible, but until Suno releases its full Studio feature, the results will be limited. Right now, Suno doesn’t have a dedicated tool to mix and master your own vocals without altering them. When you upload vocals, the AI tends to reshape or “polish” them based on what it thinks sounds best, which can change tone, timing, and style — meaning you lose a lot of creative control.

For true studio-quality results, a better approach is to use Suno to create your instrumental track, then export it and mix your vocals separately in a DAW or an AI mastering platform like BandLab. This way, you keep full control over your performance, tone, and dynamics while still achieving a professional, polished sound.🎹

[lean cedar]

def don't mind a little changes here and there but want it to be my actual voice. -

[shadow olive]

Sometimes it works great! Sometimes not. You need to draw straws but it's going to cost you credits. Proceed at own risk. Use this prompt line: [Use Only The Uploaded Vocal with slight Auto-tune - Do not add any extra effects, lyrics or voices. Maintain Original uploaded Voice & Vocal tone]

I have a working proof that this worked for me. I remixed my daughter's voice note into Edm DnB and it kept the vocal tone but, a few days later using a different promt it changed the voice.

Original Voice note

it added ad-libs but kept the vocal tone intact (oh oh... oh)

[humble verge]

made a banger but SUNO doesn't understand that "lives" isn't pronounced "lifes". Goes to editor (sigh) and after attempt number 8000, this crap still can't get it right - it's one effing word! - and I'm about to rage quit with a baseball bat to my computer and go burn down my city.

[jovial widget]

What is with 4.5+ and women vocals screaming, everyone of them is trying to be Sia and failing miserably

[wispy laurel]

Phonetic spelling
If it's "he lives in Paris" try spelling it "livs" or "livvs"
If it's "cats have nine lives" try "lyvs" or "lyves"
If it can't recognise something as a real word it'll pronounce it as phonetically as it can
Doesn't always work but it's worth a shot

[humble verge]

Oh I tried. "Livs" is pronounced "liffs"

But kinda too late after the fact, since the editor just either deletes the entire verse or part of the verse or does some other stupid sh*t, like make the clip sound like nothing that comes before or after it. Including the vocalist. The editor was always annoying to use, but currently it's worse than it's ever been.

It's also fairly random what it mispronounces. I use 'lives' twice in my lyrics. It gets it right the second time.

[boreal zinc]

How do I phonetically spell "ace"?

[wispy laurel]

Um
I would have thought it would be "ace" 😂
Maybe "ayce"??

[daring lava]

S/O to @wispy laurel @shadow olive for being thoughtful and helping people! Very big deal!

[wispy laurel]

Literally have nothing better to do while I wait for my other half to get home 🤣
And I hate when I have questions and nobody responds, so I'll at least try to make myself useful

[shadow olive]

exactly..except im waiting for my other half to fall asleep so i can game

[daring lava]

It’s super awesome! One of my favorite things is floating through here and seeing how thoughtful people are! Very cool to say the least!

[raven turtle]

I’m excited to see what Suno Studio will offer us, and signed up for early release, but do you think on its release we’ll get a boost up from 4.5+ with fidelity? Will this coincide with a v5 release? 🤔🤞

[shadow olive]

I'm more worried about whether we will be able to train personal voice models.. to maintain the same vocal across tracks while also being completely unique from each other.

nobody will have a vocal that sounds the same

[raven turtle]

The ability to have set studio session singers that you have in your library would be incredible. Similar to Audimee in many ways.

[boreal zinc]

I tried that, but surrounded by Spanish lyrics it becomes "ahthe"
I tried the IPA spelling "eɪs", which works fine on a short note, but becomes "ey-ee-s" when held. Now that I see it, maybe "eys" could work...

[shadow olive]

why dont you ask chat gpt to help you write a prompt or lyrics that will help Suno pronounce it correctly

[boreal zinc]

tried that too

[stuck dust]

I would reckon a Fall/Winter release for V5

[stuck dust]

Why do you phonetically need to spell ace?

If possible maybe try "ace" as it might help separate it a bit

[shadow olive]

This reminds me of a song that i wrote that costed nearly 3000 credits to complete perfectly

[boreal zinc]

because the Spanish voice pronunces it as "ahthe" and the rest of the lyrics are in Spanish.

'course, it doesn't help that one of the first generations I made for this is near perfect except for the pronunciation of ace, and every other generation I make, even if the pronunciation is fine, I can't help but notice how the melody, rythm, or intonation is not as good as the one I already have

[shadow olive]

it has 127 plays in like 2 months... the morale of the story is: if the puk puk puk chicken rap guy can go viral, stop wasting your time with a single mispronounced word

its also depends on why you making the song I guess...

[boreal zinc]

If my aim was to get viral, sure.
But I want a song about a guy to get that guy's name right

[stuck dust]

Hmm, you could also try: [English] ace [Spanish] rest of lyrics

[shadow olive]

ah i see

[boreal zinc]

That was my latest attempt. It skipped the line completely

[shadow olive]

sing accapella
ace

ays

ayss

eys

eic

eeys

ace

aysss

share the line or 3 of them , lets see what you typing

[stuck dust]

According to Grok, for ace it should be "ahs"

[boreal zinc]

I think the crux it, how to spell it phonetically in Spanish

[wispy laurel]

My Spanish is super-rusty
But I would have thought "és"

[stuck dust]

Ahs in theory should come out as "ace" in Spanish

[boreal zinc]

clearly, I put too many tags in my last attempt to get an a capella song. good to know.

[shadow olive]

If you want "ACE" pronounced correctly in Spanish by Suno AI, you should write it phonetically as:

"Eis" – sounds like “Ace” in English.

"Eyz" – keeps the long “A” sound with a soft ending.

In Spanish, writing it as Eis is the most natural way for the AI to pronounce it like the English name without turning it into “Ah-seh” (which would happen if you just typed ACE).

[velvet jungle]

Can someone explain to me how this works, I'm trying to collaborate on a song with someone, so I tagged them in the song so they can get the notification, bit there telling me they don't see it, now do I have to have the song published for the collaborator to see it? Makes no sense, How do I get the song to them so they can collaborate?

[half dragon]

maybe send them the link to the song?
I don't think they'd receive notification on the song though.

[velvet jungle]

Ok

[boreal zinc]

yeah, you can share unpublished songs by sharing the link.

[velvet jungle]

Ok so I guess I can do that here if they have a discord account

[boreal zinc]

on discord, sure. If you share it HERE everyone here can go listen to it.

[polar spade]

@opal orbit You mean load more than one stem at a time? Thats not available now in Suno.

When we separate the song into tracks (stems) we momentarily get to see all at once in a display panel. Once we go away from that panel we can only go back to it for about 4 minutes, then it resets.

Otherwise, the stems are in the library as separate track outputs same as any other output.

[toxic rivet]

No I’m not using the stems. I uploaded the original song to make a cover and added a persona.

[polar spade]

Shoot me. My bad. I tagged the wrong user. Sorry.

[toxic rivet]

No worries

[toxic rivet]

I am trying to create a cover using a persona but it hasn’t been going well. Either it uses the persona’s voice but changes the melody or it keeps the melody and uses mostly the original voice with occasional mixes of the persons voice. It also always cuts off way before the end of the song. I’ve tried different options with the audio influence, style and weirdness sliders. Any ideas on how I can get it to work and to keep the whole song?

[wispy laurel]

I've had some success by just directly instructing it to use specific elements from the original song and from the persona
Something along the lines of

[Exactly match the structure and melody of the original song]
[Disregard the vocal style of the original song]
[Exactly match the persona's voice and vocal style]

This does not always work and I don't even know if it's supposed to work, but I've found that using instructions like this seems to at least improve the odds of getting it to do what I want it to

Suno being what it is, it might actually be that it just randomly happened to get it right on those occasions and maybe there's some confirmation bias. But like I say, it seems like it's helping so maybe worth a shot

[toxic rivet]

Thank you. I will try that. I guess I can keep generating over and over again until it gets it right. I have a few songs I need to change the voices on since Suno didn’t listen to my voice type metatags in the original prompts.

[wispy laurel]

Let me know if it actually does help or not, I'd kind of like to know whether it's really making a difference or if I'm just imagining it 😂

[hidden yarrow]

this seems to at least improve the odds of getting it to do what I want it to
Amen and a big sacrifice to the god of RNG

[toxic rivet]

Will do.

[humble verge]

how do i get my vocalist to stop pronouncing 'spark' as 'splark' or 'sflak'?

[stuck dust]

Are you inpainting an existing song?

[young geode]

spärk
sp-ärk

[vestal cedar]

https://suno.com/@xrachellexdennyx

[humble verge]

no. It's been a consistent mispronunciation through dozens of remasters. There's two in my song and one of them repeats in every other remaster, but this particular one persists. I've simply decided to change the word to something else now. Hope it will work.

[humble verge]

I'll give it a try, thanks.

[lean pebble]

Guys, I was thinking if

I can put an entire song using the IPA if a language isn't supported

I was thinking E.G. If I want to put a song in latin, using classical latin pronunsiation

How can I write phonetically instead of...

[austere dome]

how do I get the vocals to stop screaming?

[boreal zinc]

Don't instruct screaming? Or if you want one part screamed, tag the next line [normal volume]?
Screaming isn't default, so it gotta be either your tags, or the directions in he lyrics

[austere dome]

FAT BASS, 120 BPM, swedish imigrant HIPHOP, FAST, SNARE, PUNCHY DRUMS,, driving bass, hard beats, -piano, -strings, -saxophone, -trumpets, -yelling, -screaming.

Each verse is tagged with [spoken rap]

no exclamation marks

[boreal zinc]

maybe reduce the caps

[austere dome]

it was the same without the caps

no difference

[boreal zinc]

I assume you only added "-yelling, -screaming" because it was ... I can only guess that it thinks Swedish immigrant hiphoppers always scream

[austere dome]

idk

whatever I do it keep adding yelling and screaming

[shadow olive]

Good morning, What did I Miss?

I fell aslep in discord

[boreal zinc]

Swedish yelling, apparently

[formal cape]

Is there a promt I can put in for 6/8 time signature?

It keeps giving me 4/4

[polar spade]

"......I'm able to view the stems from previous days and weeks by going to the song, and clicking generate stems. And I'm given the option to view past generations under both the 50 credit extraction, and 10 credit. This is a feature available.

It is no longer working for new generations. But I am able to view all my previous ones up to yesterday morning......" --- Mela Mirage

@opal orbit -- Being able to view all stems at once and play them all at once is only avaliable for a few minutes just after you have chosen to separate stems for with the 50 or 10 credits.

Clicking on the output song and Get Stems again, within about 4 minutes of last getting stems will give the option the "view extracted stems" [ circled in yellow] After that, it resets.

If has NEVER been a feature to view ALL stems at the same time, at any time by clicking on a song , except only during a few minutes just after separation, the Stem viewer is not a specific UI for each song, so it eventually resets to split the next song chosen , but hopefully that can be introduced in the future. That would be nice.

If you feel you could do that at any point, it may have been a momentary glitch which was beneficial, but not exist perrmanently or was intentionally implemented to remain.

[opal orbit]

That is odd, and unfortunate if that is the case. But I've been able to do it for multiple songs, and can still access it. So if it is a glitch, it still has the web functionality programmed in, and the songs I've used remain unaffected.

I've attached screenshots of one of the songs in question. I generated these stems about a week ago, and I can still pull them up now. This seems like a very intentional design feature.

[polar spade]

Yea? Thats wild. It was not intended to do that as I have known. And since we can't again now, it does seem like a slip of the developers. I call those Sunocorns as a gift we receive but really dont exist. 😁

Maybe that will come soon. Ill check some old stemmed song as well. Maybe a few are still working to display/play altogether.

Thanks for letting me know. I would let the mods or Support answer to that question then. I'd like to know myself. 😁

@hollow hare Heard of this?

[opal orbit]

Lol, thought I was going crazy for a second there. I rely on this so much because I mix-n-match stems in the DAW. Kinda difficult without an easy way to hear it all at once.

Hopefully they bring it back, haha.

[polar spade]

Its fun to hear them altogether, but as for editing or doing anything else, Its useless. So its more for a cute display than editing, anyway.

[polar spade]

Suno just announces a couple of days ago something called Suno Studio. So that sounds promising. #📢┃announcements message

You should ask to get on the waiting list.

[opal orbit]

I'm excited for that to roll out, signed up 20 minutes after they made the announcement, haha. I can't wait to see what it can do.

[jovial widget]

can you choose the Key, Tempo and song structure AABA in the description or would it be in the lyrics in []?

[jovial widget]

any help would be great

[polar spade]

As for AABA structure, if we need help with that, Lyric box has the choice to replace lyrics line by line to pick your own rhyme scheme.

But it seems easier to just instruct chatgpt to "write a song with AABA rhyme structure if the Write with Suno doesnt do that for you.

[mental plover]

Yes you can change the tempo in the editor after :D

[jovial widget]

Yeah I know you can do that but I was just wondering if there was a certain command in the style or lyrics section

[mental plover]

I usually just type it in the style, like:

120 bpm, g minor, pop, shimmering layers of arpeggiated melodies and a driving drum machine beat, 80s inspired, synth-heavy

Works for me

[polar spade]

You can search "key" or "tempo" and find many to say this or that works. If it does, ive not found any posted song to then follow consistently to be a significant statistic. I mean , if you prompt hardcore and put 140 bpm, of course its going to be close. But then on verse three put 60 bpm, ive not seen it happen.

Likely, if Suno had the capability to force a tempo or key, they would push and promote it and likely there would be a slide to correspond.

Ask ChatGPT , Gemini, or other to "Give me the italian word and their ranges for tempos" to get additional musical terminologies to encourage a range of tempos.

Here are a few of common Italian tempo markings, their meanings, and their approximate BPM ranges.

Slow Tempos
These tempos convey a sense of solemnity, breadth, and tranquility.

| Italian Term | English Meaning | Approximate BPM Range |

| Larghissimo | Extremely slow, very broad | 20 and below |
| Grave | Slow and solemn | 20–40 |
| Lento | Slowly | 40–60 |
| Largo | Broadly | 40–60 |
| Larghetto | Rather broadly | 60–66 |
| Adagio | Slowly and stately | 66–76 |
| Adagietto | A little faster than Adagio | 70–80 |
Moderate Tempos
| Andante | At a walking pace | 76–108 |
| Andantino | Slightly faster than Andante | 80–108 |
| Moderato | Moderately | 108–120 |
Fast Tempos
These tempos are characterized by energy, liveliness, and excitement.
| Allegretto | Moderately fast, but less so than Allegro | 112–120 |
| Allegro | Fast, quick, and bright | 120–168 |
| Vivace | Lively and fast | 156–176 |
| Presto | Very fast | 168–200 |
| Prestissimo | Extremely fast, as fast as possible | 200 and above |

[polar spade]

Yea. Maybe do the instrumental first, set your tempo and save, then add vocals as a Cover, or stem it and add vocals. The instruments might sound strange if tempo changes more than about 15 bpm on moderate tempos.

[sand apex]

I'm still searching for the answer. Spent hundreds of credits and only 2/4 songs match European Portuguese. Truly disappointed and regretting buying a pro account.

[mental plover]

yea

[polar spade]

@gaunt skiff It seems the Portuguese room users would keep up with current news or changes on that front? https://discord.com/channels/1069381916492562582/1230630513299947581. At least maybe a way to keep it consistent.

[sand apex]

Thank you very much! I left a message there too.

[cursive jetty]

Talking od screaming the female vocals in 4.5+ are destroying all the nuance and subtle is of my female personas and trying to make every track jump into overdrive with overtly dynamic vocals!! anyone else?

[jovial widget]

Also when you thumb a song down where does it go? I did it by mistake.

[cursive jetty]

library 🙂

[rugged steppe]

Hello guys , how do i create duet song , not suno randomly choosing which line to be sung by who

[jovial bough]

Shameless boost of my post which got buried in the avalanche. Can anyone give a "Heals Edits" step by step or point me somewhere that does?

[half dragon]

Here you go.

[half dragon]

Like you realized, it's not reliable to make a duet because of its randomness but if you're patience enough and have more time and credits, you can achieve it.
There are few other options as well, Extend or Replace but it's not anywhere near the one shot guarantee.
Search this room for other suggestions as well... there are many users have attempted and some got suggestions on how it helped them.
But keep in mind, what worked for others might not work for you, different genres, different structures... work differently.
Long story short - making duet is a challenge and it's taking time (and credits) - You might get lucky on a first few gens or you might not. Just keep trying.

[jovial bough]

Exactly the way I thought it should work, but when I do it, it just returns to the uncut version. Doesn't heal. Seems simple: 1) highlight section to delete 2) click delete 3) come back and highlight the small section you want to heal 4) click heal and voila! should work. Doesn't. And I can't figure out why not. 🤷‍♂️

[half dragon]

Possibly it got some errors - try to clean your browser cache and cookies, then try again.
Yeah, the feature seems simple enough - there isn't much to it, it works as it advertises.

[jovial bough]

I'll try again shortly. Fingers crossed.

[half dragon]

Also, it might be just that specific song for some reason...

• try a different song?
• if that works, go back to the problem song, make a full crop to create a copy of the song --> try again

[jovial bough]

This is what I think may be getting in the way. These two replacement sections I tried, don't like, and I can't make them go away. I go out, I come back, I clear cache, they keep coming back up. Just a WAG they are the problem. Any idea how I can kill them off?

[half dragon]

Up in the top, there's Reset All button - try that
To remove them, go to your Library, move them to trash.

[jovial bough]

Nope. They refuse to die.

Doh. I had them hidden. OK, killed them. I'll try again. BTW - any improvements coming on deleting trash? It's a time waste as it works now.

OMG - incredib ley they are still there.

[polar spade]

So far can just delete page by page of seleted files.

[half dragon]

Hmmm... maybe there are hidden links embedded in the song.
Try to make a copy of the song and that would create a brand new copy without them. Then try on that new copy.

[jovial bough]

That's what I was thinking. Fake it out.

Reset all is not truly resetting all, you think maybe? Bug?

[polar spade]

Hadn't tried before. But click edit on another song? Then go back?

[jovial bough]

Understood. It's a double process right now, plus page by page on top of that. Inefficient.

[half dragon]

Possibly - but I think Reset all only reset the edit so when Save it to a new song won't have the edit but the gens will till be linked to the old song.

[polar spade]

Lots of folks start crying , i included, when they accidentally send an output , or wrong output to the trash. So that much is a failsafe that likely won't be shortened.

[polar spade]

If we were as confident, there would be no need for a "restore" button. 😁

[jovial bough]

"Save as new song" got rid of them. Hopefully heal will work now .I'm guessing it will. Thanks for the help! Still, very weird that they persisted even after I deleted them.

[jovial bough]

Oh, yeah, been there, done that. But I'm sure some imagination could get us to a simpler, more direct and less time-consuming process.

[jovial bough]

LOL. Is that the truest definition of "ALL"? Only part of it? 😉

[polar spade]

Then there is this question again.... if youve ever not had this popup, then youre furious why it didnt ask you before it deleted. We just have that need to be asked.

[jovial bough]

Does everyone work like I do? Put all the songs I want to keep, including those in process, in a playlist? Where I can consider them "safe"?

[polar spade]

To be honest, weeks or months later ill go to my delete files and find songs that I decide aren't too bad and revive them. Or better, when a files goes corrupt and we need to go to the extensions in trash to remake the song.

[jovial bough]

I also put a thumbs up on everything I like and/or want to keep. So when I go through my delete process, I don't delete anything with a thumbs up. Though I have made a mistake or dozen...

[half dragon]

I work that way too. I think most do.
Easier to find.

[jovial bough]

If the delete message said "this song is in a playlist. Are s=you sure you want to delete?"

[half dragon]

I guess; and I NEVER delete any song, even if I didn't like it.
Thumb down or leave it as is in the Workspace.

[polar spade]

I put final cuts and pretty good ones in the Playlist, then delete my workstation.

The really bad ones go to trash, but i just thumbsdown everything else and leave in Library since its searchable. Its hell to NOT be able to search your trash bin. ,

[jovial bough]

Multiple worspasces. I haven't ventured there yet.

[half dragon]

I have one workspace for every song I work on. Easier to find things that way.

[jovial bough]

I would love to be able to send them into oblivion on a single click. Some fully warrant that kind of rude treatment...

[jovial bough]

My wife does thast!

[polar spade]

I do delete workstations when done, but after weeks or months, while I simmer on the final song.

[jovial bough]

Some good thoughts guys. Thank you!

[half dragon]

I never delete anything 🙂

[jovial bough]

I have this fear that too much in my little corner of the suno brain messes with its head. Paranoia.

[polar spade]

Well, I have a hallucination Playlist , a "crap but possible" Playlist , sound effects playlist, etc. . Haha. But really some thing you just got to kill off.

[median oxide]

#📜┃rules

[jovial bough]

Damn. It still reverted. This is frustrating.

[jovial bough]

What?

[humble verge]

when you download stems, there's always two versions. Is there any rule to that? Like, is one better than the other or is it a matter of testing them out individually? I've noticed that the second version often have additional stems that the first one doesn't.

[warm pollen]

how do i keep my original voice?

I have a recording of the original voice, but suno changes the voice

[half dragon]

No specific rules, you just paid for 2 versions (just like any other gens)
Give it a listen and pick the one you like.

[humble verge]

Is it normal for stems to come with their own weird quirks and artifacts? For example, I just got the stems from a song and both vocal tracks have a sudden jump in volume that isn't present in the song.

[half dragon]

I don't think it's possible.
Suno will "enhance" it the way it thinks it'd sound better so you'll end up with something either slightly different or completely different.

[half dragon]

Definitely not normal. It should be exact like your song.

[humble verge]

how odd

[half dragon]

Agree and it's expensive to re-gen.

[humble verge]

50 creds every time

[half dragon]

Or fix the stem in editor to raise the volume on that part.

[humble verge]

yea but it's a bother.

[median oxide]

Assalamualaikum anyone want have pro account on discount or anyone can give me pro account half ownership

[opal orbit]

I think it depends on what kind of music you make, but yeah Ive noticed this alot.

Overall the track will sound about the same, but you will always find small differences. My understanding is that technically, its generating completely new sound files based off your track, so you will get variety.

Ive noticed it tends to mix around guitar, strings, synth, and fx pretty commonly.

I always generate a few rounds then mix-n-match stems in the DAW for the right feel.

[humble verge]

i seem to spend most of my time in the DAW fixing artifacts. Especially pops in the vocals.

[opal orbit]

yeahhhhhhhh it's a pain. Worth to get that deeper sound though

[humble verge]

I tried a few of those online AI masters and some are really good at doing that. I did one with eMastered yesterday and it just lifted the whole sound, gave it more depth and really improved bass and drums. Sadly it costs $$, but BandLAb is free and also pretty good.

[opal orbit]

I use Adobe Audition, then run it through Mixea on distrokid. It let's you vary the bass/vocal ratio strength, so its pretty nice. Sadly also costs money

[humble verge]

I use iZotope RX Advanced and MixPad Masters. Mixea comes highly recommended in tests I've seen. LANDR as well.

[brittle thorn]

https://tenor.com/view/shocked-surprised-gasp-what-cat-shock-gif-635629308990545194

[gloomy shore]

Is there a way to stop the ai repeat some part of a lyric? I tried separated them with verse1, 2, 3, etc.. but it still repeated at some part, which ruined the song 😤

[humble verge]

It happens a lot when doing extensions. Only way to fix it - as far as I know - is trying your luck with replacement in the editor.

[atomic flax]

Anyone who's had any luck with specific male/female vocal placement? I've tried using direction in brackets in the lyrics. I've tried doing edits and just specifying the kind of vocals I want but it keeps the original gender. Is it just down to luck or am I missing some cool tricks? (Sorry if this is the wrong chat for this)

[half dragon]

Search tool is your friend.

[wispy ice]

I am trying to get Suno to make me a short acapella-style jingle with like... five total words said. I'm brand new to Suno, I have way more experience with other AI like Cursor (for programming)

What should my prompt be to get this sort of result?

[hidden yarrow]

a-capella, jingle , are good style prompt keywords to start with. You could also put into the lyrics section, e.g. [chorus: a capella]
You might several tries for a decent results, depending on your other style / genre params.

[graceful condor]

How can I reach a tech support advisor?

[half dragon]

Like this? got one out of 3 gens 🙂
https://suno.com/s/ZNkKwKJhXxcgLSpT

[wispy ice]

yeah that actually turned out pretty impressive given it's AI generated, but if this jingle I'm making is something thousands of people on my youtube channel are going to hear regularly, I feel like my standards for what's good-enough are too high for what an AI can make

Even with that example

I'm probably gonna need to end up paying someone on fiver who just makes these jingles

[vale obsidian]

Vocals are also broken considering every generation I've run today has ended a verse with annoyingly long extended notes and zero response to prompts for adding glissando.

[wispy laurel]

Is there currently any real difference between Add Vocals/Add Instrumental, compared to covers?
I'd assumed that the Add functions would do what they claim to do and JUST add things while keeping what was already there. But, Add Vocals creates a whole different song similar to the original but really obviously not the same
Am I missing something? Is there some trick to make these functions actually do what (I assumed) they're supposed to do?
Or is it just another in-progress thing that doesn't quite properly work yet?

[gloomy shore]

Oh I'm not doing extension. I just created the song and most of the time it just kept repeated some part of the song

[solemn nymph]

I'm finding myself real frustrated and I think it's because of V4.5+, songs are coming out way too slow and long, and pretty much ignore all the notes I put in the lyrics section, I tried notes in the style section and those mostly get ignored too. Is it not possible to use notes to say like [Alternate the lines in this verse between male and female vocals]?

[wispy laurel]

You can use instructions in the lyrics, just that instruction doesn't work well
Suno currently can't really handle multiple voices properly. There are some tricks to make it suck less but it's just bad at duets and it gets super confused when trying to create them
If you search this channel for the word "duet" you can probably find some tricks that might help but from what I understand it's almost impossible to make it work reliably

Also yeah, v4.5+ is a lot more prone to doing weird and crazy stuff compared to 4.5

[solemn nymph]

Understood, I have my own fully self-written lyrics, composition, instrumental notation etc. All I need it to do is produce, I wonder are there people who I could pay to help translate my composition notes so that Suno produces the song I'm looking for? Or Should I just continue trial and error?

[young geode]

If you're running into a speed issue, you can increase the tempo manually within the editor and save it: The process is free.

Otherwise, your other option is to upload a metronome of the time signature and tempo you prefer and use that as the basis for generation. However, I would make sure you have all the stems and tools and personas you like from your current project setup, as you cant use any of those gens for future personas.

[solemn nymph]

what do you mean, can't use them for future personas? I guess I should read more about personas. Am I going to run out of project space or something and need to start anew?