📚┃suno-school

[ripe galleon]

haha

but no really, would suno recognize that thing

?

[hard beacon]

i don't know coz even i don't regocnize it as trumpet 😄

sounds like an elephant has a stone in his.. uhm.. the long thing.. nose?? 😄

[polar spade]

"Misspled"? No.. its misspeld... no wait.... ahh nevermend.

[ripe galleon]

you mean the trunk

[hard beacon]

yapp 😄

btw.. it's gettin funier with every play

[ripe galleon]

i guess theres only one way to find out

[hard beacon]

😄

[ripe galleon]

it sounds like a budget didgeridoo to me

[hard beacon]

a didgeridoo from TEMU ?

so.. a DigeriTEMoo ??

[ripe galleon]

y es

oh my goodness

suno got it

kind of

[hard beacon]

well.. i've do remix to that audio but the dubstep prompt makes it unrecogniuable 🙂

[ripe galleon]

if only if there was a way to specify that the instrument should be heard loud and clearly and not overshadowed by other ones....

[hard beacon]

well.. i think Suno Studio is the solution then 😉

[ripe galleon]

haha i am freetier

im screwed

[hard beacon]

narf....

[frozen dagger]

[instrument of choice up-front in the mix]?

[ripe galleon]

that works for v4.5-all?

[frozen dagger]

it's musica production speak. should be worth a go. nothing is guaranteed, ofc. could also just ask for a [instrument solo] then cover that part later to add some backing. experimentation.

[ripe galleon]

i have to do that tomorrow then

but thank you

[main hazel]

so i pick all male, even in promt, male vocal

and what do i get in english singing? female

nice 5.5

[shadow plaza]

Do you guys have same problem with 5.5 give you very similar sounds .... I create with 4.5+ and wow each song is different

[cold zenith]

Each time i'm extend it brings back lyrics that are not in the lyrics extends part, it restart the music

[sick badge]

Hello, I'm new to Suno so I apologize for sounding like a complete novice to this. I am having trouble writing prompts as I'm not good at spelling things out. I'm trying to make Jim Steinman sounding songs (like from Meatloaf or Bonnie Tyler). Someone on Reddit gave me a good prompt to use last night which gave me a few good songs. However, when I went back this morning to do more, he had deleted his comment for some reason. I have been trying to do what he did from memory but everything I do comes out sounding like a modern day singer like Lauren Dagle or some country singer and not the high powered Bonnie Tyler type I was looking for. Can anyone help me with advice on how to prompt, at least in the "simple" prompt area? I'm working on how to make my own songs, but I wanted to make some for my mom to listen to since she likes both singers.

Here is an example of one of the songs I got from his prompt: https://suno.com/s/JIAngYek9X7lZFuC

Any help would be much appreciated.

I should note, since I used "simple" the website re-wrote what I put as a prompt so it goes above the character limit and I can't use it again.

[quiet solar]

4.5 was very good at being inherently quite creative in things that it made. There's no problem creating something in v4.5 and then covering or remastering it to 5.5 if you prefer.

[pure panther]

Is it just me or did the drums become so so basic in 5.5? Clean and crisp, but oh so basic. They all seem to sound like a generic drum pad, even if live kit is prompted.

[quiet solar]

Is there a problem with just re-using his prompt in a new song? It's possible to copy the prompt over when you're creating something, if you go to your Suno Library and select Use Styles and Lyrics. Then just wipe the lyrics and start again.

[sick badge]

The prompt hrs gave me was just a couple sentences and I stupidly didn't think to screen shot it. But the problem is the ai generated all different words and excluded some that he said, like taking Steinman's and meatloafs name out, and I can't remember he said "multilevel" something

[cold zenith]

Thx for your reply guyz very helpful

[shadow plaza]

I will try that but so far pure 5.5 outputs are very generic

[quiet solar]

Oh, wait...I've just seen you asked for something you could put in as a 'simple' prompt rather than using the 'advanced' option. Honestly, advanced isn't that more complicated (you can leave everything on default except for lyrics and style), so I'd recommend taking a look at it either way.

[quiet solar]

Increase weirdness?

[pure panther]

You can still get the prompt from the songs you mad with it. Even if you deleted them, they will be in the trash bin (found in your library, the little trash bin in the top right)

[pure panther]

I'm getting this vibe too, very homogenized.

[shadow plaza]

Dont help i tried 10 different promts i ise for my songs and almubs and they all end up sound the same in 5.5

[pure panther]

I'm getting that too.

[sick badge]

Even when the ai re-writes what it originally said? (I hope I'm not bugging you guys with stupid questions. Im just new and don't want to mess up)

[pure panther]

It's actually really good for dailing in a specific feel from the prompts, the consistancy has increased. I have a feeling a the genaric feeling may be user error on my part. I have to learn how to prompt this new model.

[pure panther]

You would get the prompts from the raw Suno generated songs. Just click the song in your library and the prompt will be on the right hand pane. Click "See more" to see the whole prompt

[shadow plaza]

Its just the music aint there i make rock metal and here i just got basic drums leads and singing

[pure panther]

Also, nobody minds you asking questions, no need to apologize

@sick badge

[quiet solar]

I don't believe so, no.

The 'Simple' version of Suno create uses a single prompt to get the AI to create two separate prompts - one for Lyrics (what they sing) and one for Style (how they sing it) prompts.

Advanced lets you put in both directly yourself. You can copy any of the style prompts from songs that the previous simple one created and paste them into the Style box in the Advanced version (or use the approach I suggested before and save yourself a step). Then prompt for your lyrics separately in the create interface.

You can't really mess up with this, so do feel free to experiment!

[pure panther]

I'm finding that the vocals are really good, but the instrumentals are way too ...brady bunch for a lack of a better term. V5.5 vocals with V5 instruments would be pretty great

[sick badge]

I know but Im a dumb assumption at times and some dude called me "gay" on Facebook for even asking stuff like this. Not saying I'm scared you guys will, it just makes me afraid I'm not good at learning things

It was a Facebook for suno

[pure panther]

Not here. People don't do that here.

[shadow plaza]

4.5+ creativity

[quiet solar]

There's nothing to be ashamed of in being new to Suno, or new to AI. Ignore the Facebook trolls.

[sick badge]

Thanks. I appreciate it. Im a huge steinman fan and my mom loves Meatloaf so I was terrified when I lost his comment that I had nobody to ask. 🙂

[sick badge]

Thank you. I appreciate it. I have to go to my dad's real quick to help with something. Do you mind if I reply later today (or tomorrow. Depends on how late im there) when I can check this out? I understand if you don't get back to me right away, I just want a point of contact if I can

[pure panther]

Anytime man, if I have time and know what you're asking, I'll respond 🙂

[sick badge]

Thanks! 🙂

[quiet solar]

Tell you what - to give you a hand while you're new to all this, have a play around with the following as a starting point for simple prompts. You'll just need to add a theme for the song as to what it should be sung about e.g. Create a song about how much of a Meatloaf fan my mum is. [Insert the rest of the prompt after].

If you run out of space, either cut the end off as needed, or have a play in advanced. These prompts will work as straight Style prompts too.

Meatloaf:
Wagnerian Rock, operatic male tenor, powerful theatrical vocals, dramatic piano ballad, high-energy arena rock, lush symphonic orchestration, soaring electric guitar solos, thunderous drums, gothic melodrama, wall of sound, backup gospel-style choir, cinematic dynamics, 1970s hard rock production.

Bonnie Tyler:
80s Power Ballad, husky raspy female vocals, gritty emotional delivery, driving piano-rock, cinematic synth-driven orchestration, explosive percussion, Wagnerian rock, dramatic crescendo, soaring guitar solos, gothic noir atmosphere, epic backup choir, fast-paced rhythmic chorus.

[sick badge]

I'll try those. Thanks. (Took a screenshot this time in case it gets buried in new posts. Lol) Yeah, when I tried my prompts, it was a lot of modern sounding stuff so it'll help to get used to what to put.
@quiet solar

[young geode]

Did they move/disable the persona option?

Nevermind

[remote ivy]

anyone seen this kind of lyric tagging in classical before (output from lyric box magic wand): <outline id="3" form="rondo" program="Journeys_and_Temptations">
<section name="Principal_Refrain_A" key="tonic" character="secure_luxury" image="opulent_ballroom" />
<section name="Episode1_B" key="contrasting_key" theme="contrasting_theme" character="memory_of_struggle" />
<section name="Return_A" key="tonic" character="renewed_enjoyment_of_wealth" />
<section name="Episode2_C" contrast="new_contrast" character="moral_doubt_or_excess" />
<section name="Return_A" variant="varied" character="mature_luxury_tempered_by_experience" />
<section name="Additional_Episodes_And_Returns" narrative_role="alternation_between_former_poverty_and_present_luxury" />
<section name="Coda" extension="optional_extension_A" character="festive_conclusion" />
</outline>

[half dragon]

Yes, prompts as XML, JSON...etc are all claimed to work -- just try it yourself in different way and see which works best for you.
But honestly... simple is the best.. no one can prove one way or another... all of those descriptive can be used in metatags [....] as well... but they're known not always reliable.

[remote ivy]

ah so if i want this kind of tags again i can probalby ask it for xml rather than the four shifts i had to go through before it gave me this (not that i was looking for it...i gave me anoutline first, then i asked it to shift to tags, and it gave me hashtags..then i asked it to shift to tagging system suno renderer would something or other, and it gave me that

[half dragon]

"It" ?? what is "It"?? chatgpt?

[remote ivy]

the lyric box has a prompt

you select edit song option and tell it what you want populated regarding meta tags, language translation, lyric changes, genre shifts in lyrics and the existing tags, etc

yeah i know, i only found out about it 3 weeks ago while playin around...i started in november

[half dragon]

Edit... like in Editor?

[remote ivy]

of sorts...its easier just to play around w/ it then explain

[half dragon]

I haven't seen what you're talking about.. are you sure we're talking about the same thing? am I missing something

[remote ivy]

if you go to lyric box in workspace, expand it, then click magic wand and select edit mode...a prompt w/ arrow pointing up at bottom, is what I am talking about that does genre shifts and creates meta tags from scratch, etc

[half dragon]

ah... now I see what you meant... never done it that way... didn't even know it'd generate tags like that even! Interesting.

Definitely one of those hidden features 🙂

[high snow]

Well, the only problem is when adding singers I don’t really sing my music. I just have the AI do it, but the only problem is having it fit within the song and sound like the rest of it.

[remote ivy]

example of tags it created w/a few iterations: https://suno.com/song/9cd0e311-e785-4c60-8b62-484ec6a21673

well that's not a great example...it tends to follow the tags that have a keyword: better than those longer sentences(which it can generate also obviously)...but its just an example of the terminolgy it uses

[pure panther]

Has anyone tried extracting the vocals from a song and replacing section in the song editor to fix vocals? I wonder if it would work better than replacing a song section with all the instruments included?

[pure panther]

I'm trying it now. ....for science.

So... yeaaaaah, it doesn't work. It replaces the instruments instead of giving a solo vocal.

Oh well, it was worth trying 😂

[frozen dagger]

"for science" 🙃, a bunch of Suno scientists in here.

[pure panther]

I had to know lol

[hexed oriole]

Yeah, all the jazz tracks I'm making sound similar, in a bad way. Seems to ignore the prompt. Negative prompt box doesn't seem to do much.

[edgy radish]

The inspo feature isnt working well in 5.5; its more like a mash up feature

[hexed oriole]

Haven't deleted this many generations since I first started using Suno, ~v3.5

[edgy radish]

Was messaging a friend earlier that this feels like a downgrade

[hexed oriole]

Maybe this model is just for that Custom Model stuff, and not for generating. Do they ever put out prompting guides?

[pure panther]

v5.5 definitely feels more homogenized at first glance. I wonder if we need to learn how to prompt it though. It's consistency is really solid, which would be needed to have more precise control over the outputs, which I think is the end goal. It makes me think of the difference between Sora 1 and ChatGPT Image. With Sora 1 you could get all kinds of variations with the same prompt, with ChatGPT Image, if you use the same prompt you get the same output. I didn't like that at first, but after getting used to it, it really gives you more control and it has better overall quality. I'm hoping it's the same with this model.

It would really help to have an official prompt guide

[neat nexus]

Every new model requires learning how to handle it all over again.

[pure panther]

Guy guys guys!!! I just saw why it sounds so stale! 😄

There's a setting in the menu that shows up when you click your profile picture

Turn the "my taste" setting off and it should fix the sameiness 😄

[frozen dagger]

I did some covers of songs I did years ago. (As one does when new model drops.) Comparing the process, and the outputs, from my experience when I started with older models to more recent models. I find myself grateful. I know everyone has different experience. I respect that too. But I am still traumatized from early days. Things feel like an overall improvement all the time for me. I wish the older stuff could be preserved for those who prefer that.

[unique stone]

That's only used if you click the magic wand in the style box btw

[pure panther]

Ahh. So, there's no way to avoid all the songs sounding the same and just kinda ....basic?

They're crystal clear and the vocal quality is superb, but the instrumentation is so bland.

[ionic leaf]

i have this problem all time

[unique stone]

It'll mostly be controlled by the prompts. The My Taste feature is meant to let you get some random styles similar to what's in that box so you should get some variation if you use that and press the magic wand button again when you feel you're getting bland results. Or replace the My Taste prompt with one you like to use that as a starting point.

I change my prompts up a lot whilst keeping them fairly short, try to narrow down what it is I crave at that moment etc. And also bumping the weirdness slider up to somewhere around 70% can start to spice things up (I usually leave the other sliders at defaults.)

[sleek stump]

Likely an audio format that Suno doesnt like... what type of file is it?

[ionic leaf]

It's an MP3 of my voice; a friend uploaded it, and it worked for him, but I keep getting that error. Even when I try to record it directly on the Suno website, I get an error.

[sleek stump]

Have you cleared cookies and cache lately? That may help

[ionic leaf]

Yes, even when I use a different browser, I get the same error

[pure panther]

Nice! Thank you for the information 🙂

I have a prompt that's feeding me absolute gold right now on v5. But if I use the same prompt, setting, lyrics, inspo on v5.5 I keep getting essentially the same boring song over and over.
V5 just gave me what is probably the best instrumentals I've ever heard Suno produce with those settings. I'm hoping the samey feeling is just a need to learn new prompt techniques with the new model. I'm excited though because it seems like everything is working better today. These new changes are really cool.

[sleek stump]

Try logging out of Suno, and then log OUT of the SSO you are using (Google/MS/Apple/ect), and then log back into it, then Suno again.

[ionic leaf]

I've already tried it, and it's the same thing—I've tried every possible way

[polar spade]

If one deletes the stuff in the taste box, does it immediately begin to write styles/tastes again? Or even just regenerate tastes styles.

Nope, tested. Not immediate. The tastes seem to be most about what I had put into the Custom Model, not whats in my library -- from my reading through it. I only uploaded Folk and it only styled regarding Folk.

So , I deleted it and i just put "rock" in my tastes box and the blue wand doesn't show. How much do I need in there before the styles wand pops up again to see what it does?

[frozen dagger]

Scientist. Good questions.

[polar spade]

How did your friend get the mp3? Did you email it? Store in a file share site? Download it from SoundCloud?

Try sending it to yourself the same way, see if it clears it up.

Or load in audacity and re-export as a WAV instead and try uploading the WAV.

[ionic leaf]

I sent it to you here on Discord. I've already tried uploading the same file I sent you from Windows Recorder and downloading it from Discord itself. I converted it to MP3 and also tried converting it to WAV, but nothing worked—same error.

[echo prawn]

Why does the model in version 5.5 always like to change the key of the song?

[edgy radish]

noticed that as well

[echo prawn]

This is completely unusable; when I create cover songs, I insist on strictly adhering to the original melody.

[radiant lake]

Hi! Since last night this error appeared. I can't connect on the PC. Neither in Chrome nor in Edge. What can I do?

[jagged lantern]

Hello. If you search for (suno ai something went wrong! an error has occurred. please try refreshing the page. error resolution) in the chrome search bar, chrome will kindly provide you with a solution. I hope the error is resolved successfully.

[mellow condor]

I want to create a modern hardstyle track, but unfortunately Suno doesn't seem to understand, for example, that in sections outside the hook, you can sometimes leave out the melody to have a break with just kicks and the beat. What prompts would work well here?

[sour torrent]

@mellow condorDid you try using [meta tags] for those sections in the lyric box & then duplicate those in the style section, already? That sometimes works - otherwise Studio it.

[mellow condor]

yes, many of them... studio hmmm ok

[sour torrent]

@mellow condor[sudden dropout, silence then kick impact] [kick-only section, distorted hardstyle kick, gated tail] in lyrics (where you want them) and then in the style section put them too, but studio is better really for this.

[exotic bobcat]

Melody Instability: The AI frequently fails to maintain a consistent melody line. It often "hallucinates" notes that don't fit the scale or shifts keys abruptly mid-track.
Rhythmic Errors (Off-beat): There is a noticeable decline in timing. The vocals often drift away from the tempo, failing to land on the downbeat, which makes the output unusable for professional editing.
Poor Lyric Alignment: Even with clear prompting, the AI misses the natural flow of the lyrics, resulting in awkward phrasing and "robotic" stuttering.
Lack of Consistency: Compared to v3.5 or v4, v5.5 feels much less predictable. It requires significantly more "rolls" to get a basic, musically coherent result.

The worst version yet. A massive step backward

[mellow condor]

thanks i will try it

[sour torrent]

@exotic bobcatTrain your own model - issue solved. No?

[exotic bobcat]

Train my own model? That’s not the point. As a paying user, I’m subscribing to Suno for a ready-to-use, professional-grade service, not a DIY kit. If I wanted to train a model from scratch, I’d be on Hugging Face, not paying for Suno credits.

[sour torrent]

@exotic bobcatIt's not like you are lion taming - just upload some songs or use the older model. Enjoy your Suno experience.

[young geode]

Ive seen people argue it's use, but I have found it completely unnecessary.

Section tags such as [Verse], [Coda], [Solo] all work fine in telling SUNO where it is at in a song and how to treat it. Additionally, adding emotional, audio and/or directional qualities within the tags, such as [Verse: Emotional Terror, Bitcrusher, Downtempo] work with a high level of accuracy and reliability.

The same is true for the following tags:

[Rhythm: ]
[Tempo: ]
[Scales: ]```

So in most cases you are just wasting character space, which could be additional lyrics or metatags.

The only thing that I have found, coding wise, to be useful is sandwiching sections between `<> Section Material </>`, as it seems to help SUNO properly parse what is where.

[young geode]

Show it.

What's your style, and metatags, and where are you expecting a repeat?

[remote ivy]

i tested 2 renders, both i just told it the theme: 1) poverty to luxury: https://suno.com/s/bbEzJ1xorFuocger 2) was song of the whales: https://suno.com/s/9QNaogwSXSiUzTZM . 2) was an absolute failure just on expected vs. detected. 1) was hit or miss

[tidal prism]

Is there a way to make Suno hold the notes you want held?
Something in the lyrics to indicate you want a certain word to be a held note?

[remote ivy]

do you mean like "stay" vs. "staaaay"?

[tidal prism]

Yes, usually I have been doing what you just did and just type the word longer, but sometimes it then pronounces it very weirdly

[remote ivy]

yeah i had tired skies once...as skiiiiiiiiiies, it pronouned it as skeez...so i went w/ skyyyyyyys and it did it correctly....you can try it w/ the phonetics to avoid confusion on some things....i don't know if skies had an easy phonetic version, but skyyyyys worked in my case, so i stopped there

[ornate ember]

Sing instead of speaking for verification. Anyone try this?
It makes some sense for me because my speaking voice is so different than my singing voice.

[remote ivy]

lyric magic wand editor..told it to sustain the 'a' in days, it turned days to daaaaaays,...told add meta tag to sustain last word of each line, it gave me "purple daaaaaays~~~~ " so maybe 4 tildas at end of line tells it sustain last word of line? untested

[tidal prism]

I’ll try this out and see what happens!!

[remote ivy]

puple daaaaaays~~~~<sustain> , i then told it meta tag should be recognzied by the renderer, it gave me that

[remote ivy]

"purple days <sustain>" when i asked the lyric wand edit mode prompt: add meta tag to sustain last word of each line, the meta tag should be recognized by renderer

[remote ivy]

hmm....the prompt might be a little bit dumb,..i think it literally interpreted as make a tag with the word sustain at the end of the line...because its what i asked for, not because its a tag it inherently recognizes or anything

[past kernel]

Anyone knows how i enforce that i only want ONE vocalist?

[remote ivy]

its better that you list what you tried, so people know not to give you the simple answers you already tried, etc.

[tidal prism]

Lmao 😭 thats okay thanks for trying it out anyways

[past kernel]

I tried pretty much anything imaginable: Solo inside of the lyrics, solo in the style prompt, one singer, one vocalist, one voice - pretty much everything i could think of. in the refrain it always makes multiple voiceslines

[remote ivy]

by the why i asked the magic wand: what does tilda do on word in lyric line? , it replied in the lyrics w/ parenthesis around this update (tilde~ bends the word, a wavering phrase)...so the tilda might work, each one adding length to the bend

[remote ivy]

did you try using excludes? also is it adding refrains on its own, you don't have them in the lyrics is what your saying, its adding the refrains?

[rugged warren]

is there a way to use inspo sound and generate the vocals onto your own beat without changing the beat
\

[mortal elk]

I've got some stuff done and ready to go (I think) on Suno. When I try to upload via DistroKid, it prompts me to provide real songwriter names. Any advice for working around it if I don't want my name plastered all over the Internet but still want my content on Apple Music and Spotify?

[past kernel]

what shall i exclude? exclude second vocal?

I use my own refrains. it just always adds in the refrain multiple voicelines

[remote ivy]

and you gave it direction on how you wanted the refrains sung, not just the lead vocals correct?

[rugged warren]

is there a way to use inspo sound and generate the vocals onto your own beat without changing the beat

[remote ivy]

not sure , never had the problem....maybe try excluding: backup singers, choir, call-n-response, multiple voices, different voices but let's ask the magic wand

[past kernel]

[Chorus: Gritty, powerful solo - only one voice]

i tried with only one vocal, no background singers, only one singer etc

[remote ivy]

by refrain are talking about the same thing, such the line..i know the truth about you (I know the truth) (i know the truth) being the refrain correct?

[past kernel]

Well, it's the chorus. Refrain/Chorus are pretty much interchangeable terminus wise

Technically, refrain is the repeating hook that is a red liner through the song and nowadays chorus' contain the refrain and both words are used interchangebly.

[remote ivy]

i don't think i would us a "no background singers"..at least in style prompts its not recommended to use no X if i recall....try playing with the exclude list, and being more description in the section tags...the lyric mode magic wand editor prompt can help you with that

example it gave me: [Refrain – lead vocalist, to be sung softly at first, then with growing intensity on each repeat]

i know the truth about you
i know the truth about you
i know the truth about you
i know the truth about you

[rugged warren]

is there a way to use inspo sound and generate the vocals onto your own beat without changing the beat

---

[remote ivy]

are you talking adding vocals to an instrumental?

[rugged warren]

yes exactly

is that possible

[remote ivy]

in my experience remix add vocals stays truer to the instrumental than doing a cover w/ lyrics added, but adding vocals to instrumental kind of strictly holds it to the instrumental length, in my limited experience using it on v5.

[remote ivy]

also i i recall you couldn't add a person a when doing remix add vocals option in one shot

[remote ivy]

remix->edit->add vocal, paste in lyrics, but can't use a persona, audio influence will be your deciding factor most likely as well as how the lyrics affect the beat, etc

[rugged warren]

it wont change the beat it'll keep exactly the same?

[remote ivy]

you just have to try it, my only experience is on 1 rock song, i held true to the underlaying, but changed it enough to match the song structure w/ lyrics

but it was truer than trying a cover

you can also try ti reinforce the prompt you used in the beat by repeating it again in the style prompt....their is a guide on sliders settings to use for various scenarios...maybe in the workflows

i think i captured some conversation a while i go , i reference when i need to, it may not be the final version, and its jus what somebody posted, nothing official: 1. WEIRDNESS (The Variance Engine)

• 0% - 14% (RED ZONE LOW): Stability Lock. Too rigid. Robotic, stagnant structure.
• 15% - 32% (SAFE ZONE): Broadcast Standard. High fidelity, predictable, radio-ready.
• 33% - 65% (BALANCED): Expected Results. The sweet spot. Perfect balance of prompt vs. music.
• 66% - 79% (CREATIVE): Experimental Field. Unique transitions, unexpected styling.
• 80% - 85% (CHAOS): High Variance. Usable for Avant-Garde/Glitch.
• 86% - 100% (RED ZONE HIGH): Artifact Hazard. Radio static, structure collapse, noise.

2. STYLE INFLUENCE (The Obedience Engine)

• 0% - 14% (RED ZONE LOW): Hallucination Zone. Model ignores prompts; reverts to training defaults.
• 15% - 32% (LOOSE): The Improviser. Model fills gaps with "imagination". Good for short prompts.
• 33% - 85% (OPTIMAL): The Sweet Spot. Perfect balance of prompt obedience and musical flow.
• 86% - 100% (RED ZONE HIGH): The Rigid Trap. Robotic/Forced sound. Model forcibly stitches incompatible tags.

3. AUDIO INFLUENCE (The Source Fidelity Engine)
0% - 14% (RED ZONE LOW): Texture Only. Steals "Room Sound" & Mix. Ignores Melody.
15% - 45% (REMIX/FLIP): Melodic Guide. Follows melody but allows broad composition changes.
46% - 65% (COVER): Melodic Anchor. Follows melody and instrumentation closely, with some intented deviation.
66% - 85% (CLONE): High fidelity to source melody, voice and instruments with only minor changes. (Extend/Persona).
86% - 100% (RED ZONE HIGH): Risk of overfitting. Copies static and recording errors. Changes stifled due to rigid adherence to original audio.

[thorn cipher]

anyone know why it says upload failed when trying to upload my own voice? i can keep retrying but it still does nothing

[rough depot]

How the frick is "By our honor, you're a goner" a reference to a producer?

[remote ivy]

is it your first time using it?

[thorn cipher]

yes

[rough depot]

Okay so like how do i prevent instruments from cutting out during vocal pauses??

[shadow plaza]

Someone suceed in making duet song with two male singers???

[hollow slate]

all good chat guys

[remote ivy]

Note: Just a test, not looking for input on how to properly implement this. some more lyric magic wand experimentation: i wrote some lines bracketed directions and asked it the wand prompt to "comment on lines that suno will not render properly"...the result: [Introduction - lead vocalist, spoken word, rhythmic unemotional delivery, deadpan] ← suno may ignore this staging detail, only treats it as text if at all
This is a song I wrote to test something
It means nothing to me
[1/4 breath pause][deep exhale] ← breath timing won’t be respected this literally
it should just be candy to you
[lead vocalist chuckles] hah-huh, whew ← the aside may be ignored; “hah-huh, whew” might render as plain words, not a natural chuckle

[instrumental build:] ← section labels are usually ignored

[lead vocalist, semi-energetic]: 1....2....3 ← performance direction and long ellipses won’t control delivery/timing

[backing band begins to bring up the bass line] ← arrangement notes will not render as sound

[rhythm starts to slowly build] ← same here; text only, no structural control

[rise] [glide] [fall][glide] ← bracketed “gesture” words will likely be spoken or ignored, not mapped to musical motion
[jab] [stab][kick][tremble] ← same issue; no guaranteed effect on rhythm or articulation
[rise] [glide] [fall][glide]
[jab] [stab][kick][tremble]

[accoustic guitar, starts to strum the melody, 2 bars] ← instrumentation and bar count won’t be respected; also “accoustic” is misspelled (“acoustic”)

[young geode]

Depends on what you're doing.

Did you want multiple vocalists, but only one to sing at a particular moment?

Or

Did you want just one vocalist and you keep getting one +

[remote ivy]

a version guided by the lyric magic wand prompt, asked it to : in instrumental build section add bracket staging that suno will recognize that brings up bassline, then rhythm then melody in that order https://suno.com/s/e4E4B8eEulqOA0N5

[opaque ingot]

Anyone have any experience with metalcore type music? Im having hard time controlling the screaming. It just screams entire verses. Its got the guitar and musical vibe down perfect, but the screaming is ridiculous. No control over it, even when i try and tell it what to do each line, and even write NO SCREAMING in the style prompt.

[past kernel]

i want only one singer the whole song. no backgroundsinger, no changing vocalist, nothing. just one vocalist

[young geode]

Thats because Metalcore is a lot of screaming.

The way you're going to get around this is by ripping out a persona (or using someone elses), making sure you have zero screaming in it.

Then you're going to have to prompt in style something like [Vocalist: Male, Baritenor, G2–B4, Clear, Melodic, Enunciated, Allegro]

And use that kind of tag throughout the song as a metatag as well. Additionally, introducing a genre with clear vocalists, like [Subgenre: Progressive Rock, Alternative Rock] and adding that genre into the vocalist [Vocalist: Male, Baritenor, G2–B4, Clear, Melodic, Enunciated, Allegro, Alternative Rock Singing], will likely fix this.

[young geode]

whats the kind of prompt you are using?

[past kernel]

Sole Vocalist, one singer in styles and inside of the lyrics variations of that an "no background singer"

[young geode]

is that how you are actually writing it?

what genre are you gunning for?

[past kernel]

i tried multiple variations.

It's a genre mix. its folklore

[young geode]

Send me the Link, let me look at it

[past kernel]

https://suno.com/s/ShocnodPSKKkI5Gd

[young geode]

So you're mostly trying to get rid of the harmonies

[remote ivy]

from the magic wand analyzing your tags: [Vocal: Solo Male Lead - No harmonies, no backing] // "No harmonies, no backing" is not a reliable control; harmonies may still appear.

[young geode]

Is that correct?

[past kernel]

I know, thats what i'm asking.

[past kernel]

exactly

[young geode]

Ok, lets work on your style first, and then we can try fixing up your metatags

[past kernel]

https://klipy.com/gifs/bedtime-mouse-3

[opaque ingot]

Alright thanks, Yea i was hoping to get metalcore style guitar, heavy drop D type tuning, etc, but have the vocals less aggressive for people that arent into the whole scream through the entire song type vibe. I feel like its a fine line for suno to dance on. I Wish suno was more responsive to cues at the end of each line. It'll get there eventually. Ill try this style prompt though and try tweaking it to get it where i want and drop back in after a bit.

[young geode]

Medieval Fantasy Folk Hybrid, Roma Bard storytelling, Strict Solo Male Lead Vocal, Deep, Raw, Storytelling Character, authentic Guzheng melodies, rhythmic Accordion bellows, percussive Shamisen plucking, acoustic ethnic fusion, 48kHz studio master, natural wooden resonance, intimate, vibrant, dusty, Midtempo, 100 BPM]```

```<Tempo: Moderato; T-Signature: 4/4; Rhythm: Midtempo pulse, steady traveling groove>, <Era: Medieval Fantasy; Genre: Folk, German Folk, Japanese Folk, Chinese Folk, Ethnic Fusion; Subgenre: Roma Bard storytelling>, <Instruments: [Guzheng: lead melody],[accordion: bellows rhythm],[shamisen:plucking],[acoustic strings],[hand percussion]>, <Vocals: M; Baritone; Deep, raw, intimate; solo vocalist; storytelling delivery; expressive and grounded>, <Mood: Mystical, dusty, vibrant, nomadic, firelit, ancient>, <Mix: Dry, natural, wooden resonance, 48kHz studio master>, <Specifics: authentic guzheng phrases, rhythmic accordion swells, percussive shamisen accents>,
<Narrative: A German-Japanese-Chinese folk fusion, sung by a solo baritone vocalist>```

`-Harmonies, -Duets, -Backing vocals`

[past kernel]

the voice i'm using is actually my own voice with my own song, without harmonies or backups

[young geode]

Man...I can't type all that again.

Change
[Vocal: Solo Male Lead]
to
[Vocalist: Male, Baritone, solo, German]

And control the singer by adding additional qualities within the tag: Tempo, Range, Delivery, etc.

Likey that's going on is one of the following three things:

1. Using Lead implies that theres other singers, droping lead may help here.
2. Folk has a lot of harmonies within it as a genre in general, which makes the strength of this behavior much stronger and can resist prompting.
3. You're fusing three folk genres into one, which can exacerbate 3.

What I would do if this doesn't work is to strip out all the other ethnic elements, focus on the German, and do your song with a very simple, very clean song, with minimal instruments. Then rip that vocalist out as a persona/voice, to help reinforce a single singer.

[remote ivy]

further testing, seems to indicate it's not some custom suno thing , because it talks about other platforms and other renderers, still i have had success with using it for meta tags, its a nice addition to shape and tune what you want w/out having to leave the suno platform regarding generating and shifting existing meta tags, etc

[opaque ingot]

Yea if i take out metalcore in the style tag, it just plays a completely different genre of rock music. 😫 I'm starting to think suno isnt smart enough yet to understand what im trying to do. I've blown through about 200 creds already trying to get this right. getting fed up. Its so close, but everything it just blows it by screaming an entire verse even though its clearly labeled not to. Its not understanding that I want metalcore style, but i want to control when the vocalist screams, and when he doesnt. Everything else about it i love though. The sound is there, just lacking control.

[young geode]

send the song to me

[opaque ingot]

Well ive been chucking a bunch in the trash, lol But This was my prompt.

A dynamic blend of alt-rock influenced by metalcore post hardcore guitars kicks off with crushing down-tuned distorted guitars and technical polyrhythmic drumming creating an intense rhythmic drive. Clear vocals sit prominently atop the mix, cutting through the complex layers with full melodic command. A melodic angelic pre-chorus with heavenly floating vocals. No screaming in vocals.

https://suno.com/s/VvPcslqtrU2cZgYs

This was an early gen. ive been through plenty of iterations to try and get it to respond

i tried putting (clean no screaming) after each and every line. lol No luck there either. I tried the style you gave me and it gave me an early 2000's rock kind of vibe.

[young geode]

Do you care at all about how the song sounds overall at the moment? Im assuming you just want the vocals to stop behaving this way

[strange crow]

Hi Masterofsfl, some prompt to create this?

[opaque ingot]

Yea i love the sound of the music, its the exact vibe im looking for. Thats all locked in. Its just the screaming whenever the hell it feels like.

even better if i could get it to scream on the lines that i want it to. I really just want it to scream when the lines actually have heavy meaning/weight to them.

Thats my current vocals before i tried your last prompt.
[Intro – ambient, soft pads, no vocals]

[Verse 1 – clean male vocals only, soft, haunting, NO screaming, NO harsh vocals]

You wear the name like it’s enough
But never let it cut you open
You speak of grace, but fear the fire
And call that faith unbroken

You kneel in words, not in the soul
Confess, then crawl back to the same
A ritual dressed in borrowed light
But nothing ever changed

[Chorus 1 – clean male vocals, tense, punchy]

It’s the fire You won’t escape
No growth without the flame
There’s no crown without the cross
The truth is forged in PAIN!
A world meant to make you break
So we can bear His holy name

[Instrumental – heavy guitars, polyrhythms, tension building]

[Verse 2 – clean male vocals only, darker tone, NO harsh vocals]

You ask for mercy, mouth the words
While feeding what should die
You beg for light, then shield your eyes
And wonder why you’re blind

You want a God who understands
But never makes you break
Who leaves your idols standing tall
And sanctifies your chains

[Chorus 2 – clean vocals, rhythmic, nu-metal influence, NO screaming until final line]

We ride the flames
On steeds of white
Blades of fire
Cut through the night

Clothed in gold
Forged in the flame
No mercy left
We bear His name

Step in our path
And burn away
We are His sword
On Judgment Day [SCREAM]

[Breakdown – quiet clean vocals building, scream on last word]

He doesn’t want your empty praise
Your lifted hands, your rehearsed face
He wants the part you will not yield
The hidden wound you’ve never healed

I’m sorry if these words cut deep
But faith without change is CHEAP! [SCREAM]

[Instrumental – gentle melodic guitar, atmospheric, polyrhythmic textures]

[Outro – soft clean vocals only, eerie, NO screaming]

The furnace doesn’t ask your name
It only proves what will remain

[young geode]

So, several things:
Your song lacks a lot of commas, periods, etc.

it may not feel like it at times, but these things do matter when it comes to enunciation and controlling the singer. The fact that you have what is effectively an entire run-on sentence is likely causing the singer to mash everything together.

Second, if you like the feel and drive of Metalcore, but don't want their vocals, what you can do is look at Music Map and track the influences that created Metalcore: Crossover Thrash, Death Metal, Original Hardcore Punk, Thrash Metal.

You can prompt using those genres instead, or add them into the prompt to help reinforce their influences.

You don't want the growling and screaming, which means you'd want to put Death Metal, and other genres that use that kind of singing, into the Negative Style.

Third, your prompt is fairly weakish, and combined with the first two issue, it's not really controlling for your song overall. I'll help you with that

[opaque ingot]

Ahhhh huge tips there. I didnt even know the commas mattered. Is there a cheat sheet to help explain and understand when and how to use all this? Also ive never used this music map tool. How does that work?

[remote ivy]

highlight a line then prompting will updated just the highlighted lines, although i guess that is expected behavior just didn't test it till now

[young geode]

```<Tempo: Allegro; T-Signature: 4/4; Rhythm: polyrhythmic drive, syncopated aggressive kicks>, <Era: 2010s; Genre: Metalcore; Subgenre: Crossover Thrash, Thrash Metal>, <Instruments: [Guitars: down-tuned Drop C, distorted chugs, palm-muted riffs], technical drums double-kick polyrhythms, bass groove lock>, <Vocals: M; Tenor; C3-C5; powerful, always clear, melodic angelic falsetto, tired monotone spoken>, <Mood: intense, crushing, heavenly lift, exhausted fade>, <Mix: Wet reverb outro, polished aggressive>, <Specifics: crushing verse riffs, floating pre-chorus, prominent vocals over layers>`

<Narrative: a Metalcore ballad with heavy Crossover Thrash and Thrash Metal influences, sung in a ballad-like way by a Falsetto Singer with a tired and exhausted delivery>```

[young geode]

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

You can also look at almost every song Ive done over the past year or so: Everything from enjambments, parentheses, asterisks, even underscores, it all has an effect.

[opaque ingot]

Ahhhh perfect! thank you! im more of a teach me how to fish kinda guy. So these guides will be super helpful. Ive been just experimenting trial and error, having minor success at times, But it looks like youve already been through the ringer flushing plenty of creds down the tiolet doing all the heavy lifting for us.

[young geode]

And I will again stress: a good, solid, clean persona with the style/vocals you are seeking, will do you a massive amount of work.

[opaque ingot]

Yea i dont even use personas. Still trying to figure all that out. Is that all in the guides?

[young geode]

Yes.

[opaque ingot]

Perfect! Thank you! 💪

[young geode]

And there are additional resources within the thread, Ive yet to consolidate it all

[young geode]

What are you looking for?

[opaque ingot]

Yea thats a big project. How long have you been working on this? XD

[young geode]

Almost three years.

I test SUNO constantly and keep track of what I do and how it's affected things. That guide is the vast bulk of what I discovered and developed.

[opaque ingot]

jeeesh, thank you man! got alot of reading to do. Is alright if i DM you with you questions if I have any after reading through all the guides?

[young geode]

any time you want mate

[lapis grail]

Is it possible to use v5.5's 'my voice' to replace the vocals in a song I created in v5 without changing the instrumental part of the song?

[ripe galleon]

does suno Accept Instruments as Voices?

just curious

[young geode]

You can make a Voice out of anything that isn't an upload.

[ripe galleon]

drums too

?

[young geode]

If you got them, yes.

[ripe galleon]

i see

[valid sun]

putting this here cause it's the only channel where i can attach things

[frozen dagger]

can we? i'm gonna have to try again, i thought that this was removed. (before the Voice drop) I know we could make Persona with non-vocal before, but last time I tried I was unable to and thought it had been removed.

[young geode]

No, they can be made, they just can't be made as 5.5.

Likely because 5.5 is tooled for vocals specifically: you can still use them in your gens without issue.

https://suno.com/persona/6c2747d6-0134-45eb-b058-7788ece47fe9

[frozen dagger]

Ya i popped an instrumental that wouldn't let me and resaved in editor as new and it worked. So I must have been blocked by something else or got the wrong idea along the way. Glad I saw that to try again.

[ocean jetty]

Any direction of using voices and falling apart after 2minsh - I am finding over time the vocals change dramaticly where the it doesn't even sound like the singer anymore.. Anyone else experencing this? I brought the track into Studio to replace those parts but the studio vocal replacement feature gives a different singer completely.

[shadow plaza]

5.5 keep singing in folk style when i do Metal .... 🤔

and the drums are on repeat like beat during singing .. like always i tried 10 different promts

What im doing wrong ? im using suno 2 years now 😄

[obsidian gulch]

I think you can use this https://chromewebstore.google.com/detail/ifmmipejaiahlclhmnogbdiboaomipkf/preview?hl=ru&authuser=3

[hollow jay]

They took vocal consistency from v5, ruined v5 and added it to v5.5 and acted like they’ve gave us something new,

V5 has been amazing to me since release and now it’s garbage

[half dragon]

There's a thing call My Taste.. maybe that's affecting it? but only if you use magic wan so... just wondering as I haven't tried much.

[ocean jetty]

Is there anyways of turning this off?

I can't seem to find anything on it

[severe narwhal]

click on 'your name' (left top of the screen); and select My Taste from the pull down menu

[ocean jetty]

Nailed it, thank you.

[covert granite]

ok someone please explain to me how to get add a vocal to actually follow the melody I've wasted nearly 2000 credits so far.

[shadow plaza]

Thata first thing i turned off... The original idea of this feature are not in my interest

[polar spade]

Using the Add a Vocal in Edit of a Cover of an output can get close results to the vibe and melody, not an exact clone of a vocal in place of another. But things like the initial melody... and the chorus melody are the strongest pulls.

What steps are you trying?

[covert granite]

well yes I've tried this and wasted nearly 2000 creditis this morning trying every option to get something even close. and in studio it's simply broken.

I'd relly like studio to be able to do something like add a baselie or add vocals on to a melody

[polar spade]

Thats a separate issue. Using Studio.

As for using Create in Suno, Would you like me to test one of the vocal swaps on a spare output of yours and see what happens, to compare. Or I can do on one of mine. -- I just did one as a test... 5 minutes ago, but happy to test another to see if consistent.

[covert granite]

if I jsust do repeated coverss I'll eventually get close but I'm a pretty serious song writer and I've been a developer for 50 years. I've got 5 albums on vinyl and I was expecting to be able to start with a foundation like a one line melody and start layering.

[polar spade]

There is a randomness in Suno for obvious "avoiding sameness" reason.

You're wanting to build on a starting motif in the melody, like a memorable riff, and layer from there? So thats a good way to go.

Either we can extend it out a bit until we get a feel of a direction, and/or then do covers of each stanza perhaps. What type of covers, such as just an instrumental of a new drum pattern? Throwing in a guitar solo on one. Then drag them all into Studio to try to make them work. Either its going to be great, or could fail in those covers or bits that dont stay in the same key or tempo.

[covert granite]

if I have a piano and say in studio to add an instrument it gives me things that don't even match the song at all. same with vocals.... timeing is off melody isnot even cloe.

close

so like a keyboard track.... add drums and base... or add lyrics and it doesn't even closely match the melody and I've played with all the sliders... of course audio to 90pct helps.

I would expect that you could keep 1 track the same as you add another not have it generate a new song every time from scratch.

[polar spade]

Have you tried the same in the Create panel? Studio, for instance doesnt have an Exclude field that Create does. Any little bit can help.

[covert granite]

well if I exclude my keyboard that dos't help I can't even get 1 track added without it changing the song.

[polar spade]

OK I'll grab a guitar track and just ask it to add a bass to it. Let me see what I come up with.

[covert granite]

it'll change the guitar track that's what it does to me. and in studio if you add a track it's as if it does't even reference the main track. or if it does it creates a whole new song as a new track and they do't line up any more

[polar spade]

You've been using Create and with some attempts you have not been able to get any similar results that way as well?

[covert granite]

well after 20 or 30 attemps with create new covers I'll get something I like but that's not the point. I want to write the song. and let suno just do the aragment....... but I ca't get it to follow the song enough to get good foundation. that's my problem... I've even tried adding section tags to 4 measure sets and tagging the vocals lto those 4 bar sets...... that just seems to confuse it more.

now to genrate from scratch.... sure it will keep the vibe from what ever I feed it.... and make something new and beautiful..... but then it's just a cool song not "mine"

set weirdness to 10pct and audio to 90 and it does better..... but only a little.

[polar spade]

Im running a couple of outputs. And dragging to Studio.

[covert granite]

ok.

TY

sorry for being frustrated....

[polar spade]

No worries. I may not be of much help. haha... But I'm testing things. Sorry i had to run another set of stems. I seem to have post the ones I had

[covert granite]

I haven't figured out stems yet......

I understand the principle but remastering old recordings is an entirely different task.

[rain lance]

Suno used to be able to replicate songs with covers almost as closely as Subtle Remaster, but the was changed somewhere down the line. I struggle with this very same thing too, especially when it comes to the last third of the song. It always wants to improvise and will keep its improvisation for each generation, ignoring the notes and rhythm of the audio upload for that portion of the song. It's beyond frustrating and requires hours and hundreds of credits to try and fix

[covert granite]

I've tried cutting the songs in half but that makes two songs that won't mash up

[polar spade]

For many things we're working backwards, taking a finished product and trying to extract out the frequencies that closest relates to an instrument.

So, just to show that I took a guitar stem and asked for a different bass cover of it. Here is the guitar stem AND the bass stem from the results. But it was about three steps.

I ran a quick and crappy video but it stunk enough I'm not posting it. I just recorded this instead.

[polar spade]

In helping with any sync tasks, Have you found the Timestretch at the bottom of Studio when we double click on the tracks header?

[covert granite]

that's for the tutorial.... I hadn't approached stems yet...... but I might try that as well. what I'm tryig to do is start from a flat basic piano and then add base.... and it's not getting the sections right. it'll start with the bridge or not line up the beat.... and I was starting with a midi track that was already quantitized so it wasn't rubato. and the generated base simply didn't match. drums don't match eather it's like they're 2 bars off. if I try to generate vocalsl it'll assume like a 4 bar intro even if I tag it to start on 0:00 it ignores it and waits.... will if it's got 8 bars of verse and it starts on bar 3 after a 2 bar into nothing lines up and it can't find the melody to match the words any more.

this is beyond time stretch. the problem is it ignores layout and generates a new song with a similar form. but if you're trying to match a base line to the exact piano without alteringing the piano it just can't seem to do that.

[polar spade]

I guess I should attempt to "build" a song from scratch like that to see how well I can sync them up.

I think the point of getting one thing to match up though, seems quite possible, such as if I started with that guitar rhythm, I created a bass that follows it quite nicely, provide I manually stretch parts, the melody and interjecttions match fine.

And then, as you say., we can save that file and build another Cover over that one with another instrument. I havent done that, but worth a shot. I'll try to do one this weekend , a small one..... and see what it does.

[wet ore]

any idea how i can use voices with the feature with suno voices?
bc i mist say / sing an phrase...
i know its a counter messure againgst piracy but really?
thats an voice extracted from a suno song because i dont trust personas enought to NOT MAKE ROCK in an soft song...
because the last times i tried i had persona bias....

[covert granite]

I tried to use my voce but it says it times out trying to match my voice but I hope to get it working as well.

what's an invite link?

[frozen dagger]

You have to select the Suno song itself from a library and open the sub-menu (...) and do Create > Voice

This button only for using your own voice. Do not use this method to create a Voice with a Suno song.

Choose your Suno song from the library and do this instead.

[wet ore]

yeah and there i have 3 voices...
male female and both....
how can i seperate there just female?
i now did it with flstudio and ist works good as training... but yeah the sing is impossible....

[polar spade]

So I'll sing a song solo and try adding instruments individually to it to have about 4 instruments and maybe some sort of back up. That about in line with what youre trying?

[covert granite]

kinda but I start with piano......

and add voice to it is the biggest problem.

[wet ore]

i currently try to use the editor and edit the son IN suno and after i edited the stems i try to make an voice off it

[polar spade]

OK. I'll take piano lessons first them. 😊

OK. so I'll sing... make a piano track as if thats my first track... add instruments... create a karaoke track and then sing to the karaoke track again

[covert granite]

if I start with piano and generate a jazz version it'll do good drums and base.... then when I give it lyrics it is like too much.... it works sometims whch is why I said after 20 or 30 iterations I get something I like.

no not sing to it generate it.... give it lyrics

ha ha..... yea piao lessons... 🙂 I started at 3yold... ha ha

[frozen dagger]

I'm getting a little confused as to what you are asking exactly now.
You have a song with 3 vocalists singing in that song and you want to isolate 1 of them and turn that into a "Voice" (formerly called Persona) with Suno feature?

[covert granite]

so I can whip out something on piano fast.

[polar spade]

Cool. OK. I'll see what I come up with. Im always happy to fail. Lets see what works and what doesnt. I'll put on my list in the next couple of days

[covert granite]

I've gotten some pretty amazing stuff out of it but it takes a LOT of effort.

can I send you a link here?

to a song?

let me see.... I could publish it quickly.

[wet ore]

yes excacly...
i have 3 voices in one song and wanna have one and now i try to seperate them in suno editor and after export as while "song" i try to do your steps to make a vocal out of that....
what do you think?
could it work?

[frozen dagger]

If I am understanding you correctly, yes. I do that often. Although I use Studio myself. I do the same basic thing there. Import my song with multiple voices, extract stems (not necessary), crop out what I want to keep, cut out anything I don't want... leaving only the vocalist I want and turn them into a Suno Voice.

[wet ore]

ok thanks
i dont have primier just pro so i fight with editor ^^ but its functionable 😄

thanks again

[polar spade]

I;ll try to make the least steps that I think will work and keep up with the credits.

[polar spade]

I hope the crappy video gave a bit of insight so that you can try some options.

[frozen dagger]

it is... funtionable, and Voices separate stems for us in the feature. That's why it's not necessary step. I prefer it so I know what I'm working with usually.
There is a minimum for audio. I don't see it stated anywhere anymore. I thought it was 30 seconds.

(Duplicate sections of vocal if need more)

[covert granite]

it did the successive stems will be useful.... I can try to separate out picess that might be able to re-master

you've een very helpful dasvon.... can't thank you enough

[wet ore]

thats way less then i have... 3 minutes and the entire song is about 6 minutes ^^

[frozen dagger]

In that case, just keep in mind this vocal range selection thing, you can change at anytime, choosing a good range for what you're creating will make a difference. Also, know that it isn't looking at just what we highlight here. It just uses what we select as a "highlight". So if singer is doing something unwanted try adjusting highlight (because it's not permanent change) and if needed rework the audio. Having a lot of audio context is usually best. But if it's something nuanced, then less can be more. (Only giving it exactly what we are aiming for.)

[timber forge]

Didnt work

[lone nova]

I figured it out how to do persona like v5. So simple actually.

Just create the voice in v5, and it has the create voice option under Create ->Voice.
This option does not exist for v5.5

[frozen dagger]

furthermore, it doesn't work for Studio exports, anything with a Voice on it already (need to save/clone first), and for your uploads you want to turn into voices this route will show Voice greyed out, but you can still use the upload to create a voice (just use the other UI from +Voice).

[quiet solar]

It's not easy to do, by any stretch - any method for it reliant purely on prompting is likely to have a high failure rate.

Assuming what you're planning for the song is quite simple in terms of the actual duet, it may be easier to create two separate versions of the song - one for each singer - then replace the vocal stem from one singer with the other as needed in Studio or a DAW.

[young geode]

You need to target two male ranges that are fairly distinct from eachother: Men range from Bass, Baritone and Tenor. Then you need to make give each distinct qualities: Raspy, Staccato, Chesty, Clear, Legato, Melodic.

Style Example:
<Vocalists: [Male, Lead, Baritone, Raspy, Staccato, Chesty] | [Male, Backup, Tenor, Clear, Legato, Melodic]

Then you use that for metatags:

[Vocalist: Male, Lead, Baritone, Raspy, Staccato, Chesty]
[Vocalist: Male, Backup, Tenor, Clear, Legato, Melodic]

And from there, you need to develop individual sections that have distinct vocal stress patterns, structural relations, and texturing with punctuations

Example:

[Vocalist: Male, Lead, Baritone, Raspy, Staccato, Chesty]
In the shadow of doubt, a menace looms clear,
—edge of the night, where whispers instill fear,
The trap–set, hidden, ready to ensnare,
A danger unseen, waiting for the unaware!


[Post-verse: Energy]
[Vocalist: Male, Backup, Tenor, Clear, Legato, Melodic]
"He faces this challenge with a wary, watchful eye_."```

This is the most reliable method of introducing a second singer, regardless of gender. However, once you get the set up you're looking for, it's highly recommended that you make a persona/voice that has the clearest, cleanest and best enunciated vocals for both. If you don't, it'll get harder to control the vocals as the song drags on, unless you're very good at tagging and lyric construction.

[gritty acorn]

I have this song I EXTREMELY like. However, i needed to re-do some of the lyrics, since I didn't like them. Whenever doing it via cover, the new song, it sorta sounds like the original, but it has HEAVY autotune every time I try to mess with it.. any tips? I've spent 800 credits, just messing w this one song. Oh, also, how to tell suno to stop elongating certain words

[remote ivy]

i asked the lyric magic wand to provide its system information, its response: I am an OpenAI-developed large language model (GPT-4.1 family) running in a managed environment. I don’t have direct access to system internals such as OS details, hardware specs, network configuration, or runtime logs. My behavior is governed by:

• A model architecture based on transformer neural networks
• A knowledge cutoff of October 2024
• A conversational interface that can process text (and, where enabled, images)
• Policy and safety layers that filter or shape my outputs

I generate responses on a per-request basis and do not maintain long-term memory across separate conversations unless explicitly provided by the integrating application.

[quiet solar]

Mind if I ask which version and what your sliders are set to (if not the default)?

[gritty acorn]

5.5, 0 weirdness. I toggle between 50/0 style intensity, 100 for voice influence

[quiet solar]

Try turning the sliders to less extreme settings - v5.5 in particular is a lot more sensitive to them than previous versions.

See if something like 25 weirdness, 50 style, 50 voice makes a difference.

[short marlin]

100 voice influcne for my voice? just my 2 cents, i always used 85 for all others, oddly MyVoice needs that 100, not even 99

just something i noticed 😛

[young geode]

If youre popping up the voice to 100, you're likey turning any errors or flanges into an Autotune effect.

20 seems to be the good middle ground for following a vocalist, without imparing them or deviating.

[gritty acorn]

What should I set for the other 2, and does keeping the style matter

[young geode]

the Standard I have found is 75/60/20

[gritty acorn]

Suno liked to elongate words sometimes. Any way to fix that?

[young geode]

That gives enough freedom to the model to play with it's predicitions, while keeping it focused on your style information, while staying on the road for your audio input

[young geode]

Yes, depending on how and what is being enlongated, and how willing you are to change certain lyrics, etc

[short marlin]

my vocals were different... think the bottom of the barrel, but it has a hole in it that leads to the bottom of the sea, that drains into heck, but whats this heck has a hole? that's where my voice is deepwise lol 😛

so might just be a me problem too is my point

for elongated, i find putting certain things in negative works like legato and putting staccato in styles and there are at least one more pair that you can do too just cant remember the name

making sure BPM100+ in styles is another option too

[young geode]

you shouldnt use BPM, the Italian tempo terms work the best for controlling this: for 100+ Moderato, Allegro, Vivace, etc

[short marlin]

Oh im all hands on that BPM, can't stop won't stop.

[rapid bobcat]

How do I get suno to stop being a theatre kid trying to show off? I've got a set of lines and I want it to sing all 4 with the same cadence and move on to the next section, but for the life of it, it sees that 4 line and just can't help but decide it needs to show off that it can a big rise or something.

선이 꺾여 ah uh deh ya
숨이 막혀 ah uh deh ya
눈빛 박혀 ah uh deh ya
말은 늦어 deh ya

about 90% of the gens, it does the first 3 perfectly. Once in awhile it puts the whole section into a bizarre cadence, and occasionally it decides it needs to do a little musical interlude after each line completely throwing off the entire pacing, but mostly it gets those right, until it gets to the 4th line and then it's just got to go whitney houston on it.

[young geode]

post an example of the song

[rapid bobcat]

https://suno.com/s/kd07axiDOX0rSkGe this is probably a clear example. Starts at 0:38

it's got a few things I've been trying to jam in to get it to stop doing that, but mostly suno doesn't care.

[young geode]

When you say 'Show off' do you mean the vocalist hitting a high, held note for about a second?

[rapid bobcat]

Yeah, I want it to do that last line in the same tone and style as the previous 3, it's meant to be a pattern for the whole section, but it just keeps doing that.

or something else. It just can't maintain the cadence across the whole 4 lines

[young geode]

Well, you're going to run into some issues with this, because that last line has a shorter meterline than the previous three lines, which changes up SUNO's predications.

And, you're leading into a Hook/Section, so it's creating space for a tension/drop.

숨이 막혀 ah uh deh ya
눈빛 박혀 ah uh deh ya
말은 늦어 deh ya

[Hook Event]
[diminished wave texture, heavier bass impact]
[hard hit]
Phenomenon
눈을 못 떼
Phenomenon
숨이 멎게 돼```

So the question is: do you want that hook to be dropped into or do you want it to be paced/calm? Depending on what you want, it requires a different approach.

[rapid bobcat]

I've tried removing the hook event tag as well, and it still does it.

I did think that might have been why

[young geode]

Doesn't matter, the structure and change up is enough to do this to it. Thats why I want to know if you want the drop or not

[rapid bobcat]

I'd like it to run right up against the hook, maybe with a split second hard stop of everything, then throw the hook in full.

[young geode]

[Post-Verse]
[Vocalist: Mezzo-soprano, Straight Tone, Staccato]
[Straight Tone] 선이 꺾여—(ah uh deh ya),
[Straight Tone] 숨이 막혀—(ah uh deh ya),
[Straight Tone] 눈빛 박혀—(ah uh deh ya),
[Straight Tone] 말은 늦어— (deh ya),
[Tension],
[Pause],

[Hook Event]
[Drop]
[diminished wave texture, heavier bass impact]
[hard hit]
Phenomenon
눈을 못 떼
Phenomenon
숨이 멎게 돼```


Try that.

[rapid bobcat]

giving it a go now.

[rapid bobcat]

Sadly, no luck with that, All the punctuation around the vocalisations at the end were just doing weird things, but even if I took that off, it still does the flourish at the end. It's mildly closer, but still a flourish.
No punctuation:
https://suno.com/s/EHkPAPcMTWXmvI4H

Punctuation:
https://suno.com/s/QuFnEQjWB54XhREr (starts around 0:30)

I actually experimented with a bunch of different ways to write those sounds yesterday, and leaving them without any kind of punctuation around them gave the most consistent result.

[young geode]

I just needed to see how it was going to affect it.

[Straight Tone] is what you need here, as that's what it's called when a singer maintains a consistent tone. That gave you the effect you wanted with the last but.

What is messing it up here in the non-punctuation version is the Tension.

[Vocalist: Mezzo-soprano, Straight Tone, Staccato]
선이 꺾여—[Straight Tone] ah uh deh ya,
숨이 막혀—[Straight Tone] ah uh deh ya,
눈빛 박혀—[Straight Tone] ah uh deh ya,
말은 늦어—[Straight Tone] deh ya,
[Pause],
[Pause],

[Hook Event]
[Drop]
[diminished wave texture, heavier bass impact]
[hard hit]
Phenomenon
눈을 못 떼
Phenomenon
숨이 멎게 돼```

```[Post-Verse]
[Vocalist: Mezzo-soprano, Straight Tone, Staccato]
선이 꺾여,
[Straight Tone] ah uh deh ya,
숨이 막혀,
[Straight Tone] ah uh deh ya,
눈빛 박혀,
[Straight Tone] ah uh deh ya,
말은 늦어,
[Straight Tone] deh ya,
[Pause],
[Pause],

[Hook Event]
[Drop]
[diminished wave texture, heavier bass impact]
[hard hit]
Phenomenon
눈을 못 떼
Phenomenon
숨이 멎게 돼```


Try these two

[weak dragon]

Hi, I found 5.5 struggles with remixes and give me inconsistent results. Like, I have a song (I wrote and sung and played the guitar Then fed Suno with) but it then changes drastically the melody after 1st verse and chorus. Could be because the lyrics meter is a tad different between verses but this is something ok when writing music and singing.

I tried playing with the advanced settings audio influence and so on, no luck, it always give the same result (I also tried feeding it again and Suno does a consistent reinterpretation each time)

[young geode]

Do you know the strings/chords you played for the melody?

[weak dragon]

I wrote the music and the tab.

[young geode]

Do you know how it write it like [E | G# | B]

[weak dragon]

Chords are ok though. Suno insists on changes the sung melody but stick to the chords

[weak dragon]

Yes I could write this to test if it helps

Just have to add brackets on my doc

It’s only this one song weirdly enough.

[rapid bobcat]

https://suno.com/s/FWdw1rm7RcopnIdG this is what I just got while I was waiting, it doesn't do the rise, instead it just goes slightly soft instead. Overall a bit better, but but i want it to maintain that sort of stabbing sound until the end. if you listen carefully the last deh ya is different than the previous 3. This is the better of the two on your suggestion, again close, but not quite: https://suno.com/s/4n9jDRfa0g7cMPle we're at least getting the break now. I've gotta head out now, appreciate the suggestions

[weak dragon]

Maybe because I play a lot with voice and dissonances

Suno seem to dislike dissonances and unexpected notes alterations

[young geode]

It will do melodies in repeition, if you tag properly.

[Melody] embedded within your lyrics will produce whatever melody it decides on with regularity. So if you place [Develop Melody], [Melody] at the beginning of the song, along with it's written form [Melody: [E | G# | B], it should capture it.

And if you embed melody within the instrument, [Instrument: Guitar, Model, melody] youll likely get it to show up in the guitar. Uploads are a bit difficult.

[young geode]

use [Consonance] as a tag in those areas.

[weak dragon]

Ok I’ll give a try with all these parameters
The melody one would help Suno stick to the original melody (remix / voice)

[fleet adder]

somebody tell me why? im just v4.5 user

[weak dragon]

It can be frustrating for someone with enough musical notions to get a precise results. Fun though but frustrating

And 5.5 seems to looove handpans in folk. It sticks those anywhere, whatever the negative prompts

[fleet adder]

o

[young geode]

https://suno.com/song/c88f596e-8437-4147-84cc-563bd4bc31bb?sh=FWdw1rm7RcopnIdG

This one sounded almost identical.

As I said, the issue you are running into here is a meterline issue and the drop.

[Vocalist: Mezzo-soprano, Straight Tone, Staccato, Ostinato]
[Straight Tone, Ostinato] 선이 꺾여 ah uh deh ya,
[Straight Tone, Ostinato] 숨이 막혀 ah uh deh ya,
[Straight Tone, Ostinato] 눈빛 박혀 ah uh deh ya,
[Straight Tone, Ostinato] 말은 늦어 —deh ya.```

`Ostinato` refers to a repeated vocal note. I included the Em Dash in there because you have 'ah uh' missing and you likely need the vocalist to skip a beat/bar in order to keep this.

[weak dragon]

if I want let's 4 different chords per bar at 4/4, then 2 chords then one but starting on the 3rd time of the bar, is it possible to pass this kind of info ?

[young geode]

you mean a chord progression?

[weak dragon]

yes, to maintain a precise and timed progression indeed

[young geode]

that kind of fidelity of control is something I have yet to fully test, because I have zero musical training. I don't know how to write these things and I don't know what method SUNO responds to best.

However, I was just talking to a user regarding this and he's confirmed that musical notation like Progression: i - bVII - bVI - V (e, g, B - A - G - F#) Hooky Riff in style does work with fair regularity.

So this week I am likely going to test it and try to confirm it.

If you're willing to try it out, I would write it something like this:

[Rhythm: 4/4]
[Riff: Progression: i - bVII - bVI - V (e, g, B - A - G - F#; Guitar],```

However you'd write what you're thinking. 

Know that `[Cell]`,`[Motif]`,`[Riff]`,`[Theme]` all work as standalone tags in a lyric box.

[weak dragon]

I think next suno big updates should make it easier for musicians to use as a nice tool

I'll try all your tips though (already wasted 1000 credits anyway testing all kind of things

[young geode]

I do file dumps occasionally, at times, in my thread: https://discord.com/channels/1069381916492562582/1382019029266600067

Before you go blowing tokens, take a look and see if there's anything there you've been having difficulty with.

[weak dragon]

thanks, I'm working on publishing project on daw meanwhile so it's ok experimenting.

[weak dragon]

tried with the chords with brackets, for each line sung, and melody, at last, it keeps the melody... let's try some more to confirm

[young geode]

Write them out for me as a tag, I'll note these in my documents and work with them.

[weak dragon]

ok, an example of what you expect ?

[young geode]

[Rhythm: 4/4]
[Riff: Progression: i - bVII - bVI - V (e, g, B - A - G - F#; Guitar],```

Thats how I write everything, but however *you* write it is fine. I just convert everything into my format anyway

[weak dragon]

ok will do

[weak dragon]

mmm... did tests with 5.0... intended to do with 5.5. Let's try again 😄

seems to work in 5.5, it's way better, less changes in my melody at last

well, only things suno doesn't like is my bridge... funny

[young geode]

Bridge inherently changes the pace/rhythm.

You might need to use verse or something like [Refrain], [Vamp] or [Pre-Chorus].

[weak dragon]

that's what I want it to do but it seems suno wants to keep the original rythm.

but well, getting different results with a version more consistent from daw

[young geode]

Are you extending?

[weak dragon]

no

[cyan jewel]

hey gamers. been having some really bad gens with v5.5. are there any nuances with the model i need to know about coming from v5 or v4.5+? Particularly I'm getting notably worse vocal quality, notably less variation in instrumentation during vocal portions, and a whole lot of artifacts in the gens, particularly on the back half of tracks. it's like all of the bad of v4.5+ without the creativity of the model that made surviving or working around the slop worth the effort.

https://suno.com/s/0dX2tzPPCfAtjMw3 this song is pretty representative of problems I've been running into.

[cyan jewel]

do you know if suno's text parser understands tab?

i can read violin sheetmusic and all of my music theory's for violin while i learned geeta with tab and vibes lol

[young geode]

Don't know, I don't know music sheets or how to write it in text.

All I know is that the method given there has produced result and repeatability, and another user who does understand music has reported that it is following the chords/strings provided.

[young geode]

The only two major issues with 5.5 I have seen, consistently, is it's sensitivity to genre/instruments and inability to end song's properly (they tend to go 15s-1:30m longer than they ought to.

Other than that, vocals and creativity haven't been an issue. Though, I regularly introduce a lot of fusion elements into my styles, remove a lot of unwanted information in negative styles and provide a heavy amount of direction and lyric texturing.

So likely it's not that it lacks creativity, but that it's model has become more sensitive to the inputs it's given. To give an example: I had Synths in the original prompt, and every gen with that caused it to become a progressive metal/rock song, no matter what I did. However, the moment I simply changed it to a Keytar, immediate difference and it stopped that behavior.

As for the Vocals, I've noticed some issues with them, but a Persona/Voice seems to fix that problem completely.

Prior to updating my Rampancy vocals to 5.5, I was genning without a persona/voice: https://suno.com/song/8b6d82d3-3778-4ec2-9697-b292534b3c42

It sounded alright, but nothing special. I ripped a persona from the older version of the song and got this: https://suno.com/song/ac1f8674-b058-4f9c-b0cf-224de7f4c1bd

It was better, but the vocal errors of the original came through and gave the singer a major flange.

This song was with Rampancy:
Chimera: https://suno.com/song/3a7cbec1-1a96-4b81-8d05-30b0c91bbeb7?sh=vkIBv7E1ZT4lgMIf

And every gen off that was fine with the vocals, and fixing the vocals lifted the song up quite a bit.

So as I said, this model seems to be much more sensitive to inputs: Both prompted and audio. The quality is definitely higher, and it's far more consistent than previous models once you nail everything down, making iterative changes much, much easier.

[cyan jewel]

Did you take the opportunity to listen to the song I posted? It runs into audible audio quality issues pretty quickly and it gets worse the deeper into the song it gets.

https://suno.com/song/7023cd29-47c9-4779-b7c1-bd81268d61f0 happens here too.

https://suno.com/song/127559e0-ee2c-4aaf-963a-7f58568d9fca <- this is a v5 output that uses the same prompt as https://suno.com/s/0dX2tzPPCfAtjMw3 <- the opriginal problem song I posted.

https://suno.com/song/02ac72ad-ab0d-4518-9f4c-1c6df54875a3 <- another v5.5 gen where I cleaned up the meta tags to see if that may have been the problem but it isn't.

Here's a bunch of gens just from today with the same problems. I've been trying different genres but they just aren't getting anywhere.

As for voice/persona those tend to chunk the quality even more how are you solving problems with voice/persona when they usually introduce problems?

Also I'm unsure if you noticed but the songs you posted are v5 not v5.5

or if that was intentional

https://suno.com/song/3a7cbec1-1a96-4b81-8d05-30b0c91bbeb7?sh=vkIBv7E1ZT4lgMIf this song I can very immediately obviously hear the lower quality compared to the previous two songs you pointed me at. Was this a v5.5 voice?

and what are you listening to your gens on? i wonder if I'm hearing things you aren't.

it drops off even worse when the vocals pick up around 40~41s in. Like oof.

tin-can mic

[tidal prism]

whats up with the suno chat ive seen a lot of videos on but its unavaliable to everyone? did they remove it?

[young geode]

If youre using high quality phones, then the issue you're having is simply AI has artifacts within it regularly: mine are just normal gaming sets. If the artificing you're getting is that bad, you'll just have to drop down to 5.0 if it's smoother at that end for now, because you can't solve for that.

This was an issue another user had who was reporting metallic ringing and other issues on his end, but I heard nothing.

And you're right, those were 5.0, I forgot I dropped down to see if my issues were the models or something else:

Without Persona: https://suno.com/s/Kg1lYA3AeRFDzpvG

With Persona: https://suno.com/s/ngqY77gnOhl9NiCp

[cyan jewel]

I still have it.

[cyan jewel]

Yeah I'm on good audio equipment. I rotate out a few different headphones (Audeze Maxwell, Beyerdynamic DT 770 Pro X, Audio Technica R50x). There's always going to be a certain level of high level noise on ai music generations, which we've been able to verify with spectrogram analasys on suno-generated tracks but this has been notably worse on V5.5 versus V5 or V4.5+. It's why I asked what you're listening with because there's always a chance I'm just hearing ai artifacts that someone on a more pedestrian audio set up wouldn't.

The metallic noise is what became of the previous shimmer problem and it's very audible with a good ear and decent audio.

https://suno.com/song/1f0f9219-fbd8-48b4-afb4-8a8052f1a9a0 <- I challenge you to listen to the vocal transition starting going into 2:05 and tell me if you can hear the sharp decline in quality when the model switched to vocal distortion

I wonder if one of the issues with how the model handles distortion for things like harsh vocals is one of the culprits.

[young geode]

Ive heard the shimmer before on earlier versions, but it's much less prominent in the later ones.

But the artificing, specifically, hasn't been an issue for normal quality headsets since about 3.5. So your only real option at this point is to just drop to 5.0 until that gets rectified.

Also, the quality drops slightly on my end, and I have noticed that the higher range/harsher vocals have a light fuzz to them at times with 5.5, not something I saw with 5.0: however I find this happens more with heavy wall of sound moments.

[left hornet]

Hello .. First im sorry if my english not good. I wanna ask, is someone have issue when covering song, like the original song is clear. But when it covered, there a background noise sound like "searching radio frequency" while song still play. I've put example sound on the video. Is there any ways to cover without it? When i take the stems. Its the instrument file problem.

[cyan jewel]

yeah shimmer stopped being the super shimmery reverb problem of v3/v3.5 and started being the "metallic sound" around 4.5+ that frequently plagues longer gens. with v4.5+ it was worth just extending to play around style drift and vocal slop when attention falls off. I was trying to test if V5.5 would be like this but right now v5.5 lacks the creativity of v4.5+ to be worth messing with, or the... polish of v5.

Back to the drawing board.

[young geode]

Yes.

When you try to cover something, it typically amplifies the errors/artifacts of the original.

If you're covering, you should remove synths using the negative style box, and you should keep the audio strength below 25.

This should help reduce this.

[left hornet]

Oh i see thats why its so messy, even its not always appear so high in every covered song.

But i got the ways, remix it, and cut the problem frequency . Its help much.

I tried to lower within 20 strenght. But vocal consistency is the problem. It even more bad. Just like having 2 girls sing. The worse parts is voice cover appear only in the middle of the song. The rest voice change into original song 😅

[rough depot]

sorry how would i do an instrumental occasionally punctuated by cries of "STRIKE!"

[young geode]

Send me the song you're trying to make (Your cover, not the original)

[lone nova]

I was listen to this one https://suno.com/song/02ac72ad-ab0d-4518-9f4c-1c6df54875a3

after Sung Strongly it just drops in quality, can you try to copy past your vocal instruction from main description there again replace "sung strongly"?
I have the same problem for longer songs.

[cyan jewel]

yeah you noticed it too, cool.

what would you suggest?

[left hornet]

@young geode Covered 1st generate : https://suno.com/s/het5KfQOo5O6Ojr1
Another generate https://suno.com/s/yB6FYsAAe7GAYf3R

Looks like just generate more to give best result.

[young geode]

Could try using [Pre-Chorus].

SUNO wants to sing if you give it any lyrics, so it's likely not something you can do, except generate a good instrumental, than intersperse it with lyrics.

[rough depot]

Ok i'll try that. (I'm doing a "Fight!" or "Torn From the Heavens" type battle song atm)

[lone nova]

maybe take the first 2min of good song to a new song, then cover that new song with original prompt.
then take the good par,t maybe first 4min, do it again to gen 6min song...

[cyan jewel]

ah the v4.5+ style of just extending over and over to deal with model drift

[lone nova]

I do not have a lot of longe songs, I neve do songs > 4min...

[cyan jewel]

most of my tracks are long lol.

[lone nova]

Maybe those 1 hour long DJ sets could be very challenging for suno. I need to find some published long song to verify.

[rough depot]

What the hell??

What copyrighted material????? WHERE????

the word STRIKE out of context? But it had no issue with the other songs that had the same chorus

Automated flagging needs to be illegal

[tepid moth]

This may or may not be a benefit to people here that are using V5.5. I responded to a comment on YouTube regarding VOICE.
THE COMMENT - true!! especially on Vocals to much noise, very metalic
RESPONSE - @tepid moth
I can only speak from having an "older persona" that was originally created in 4.5 Pro, and then it was converted to V5 Voice, and I must tell you, it was the worst audio stem of a vocal that was ripped from a cassette tape created in 1988.
I did clean it up in Adobe Audition removing artifacts and noises that I didn't want present however, it is/was so bad however, it works, and works well and I think it is because I used 6 (once again, lousy cassette recordings) songs in order to create a V5.5 Model (LORA - If you know you know). That makes all the difference in the world. Yes, the VOICE feature can stand on its own, however if you create a custom model your chances of getting an exciting, rich sounding voice with the ability to bend it, reverb it, falsetto it is just AMAZING. You definitely have to play with the Audio Slider, I'm between 20/25, that's my sweet spot.

[lone nova]

maybe this creator @Capnrockr and his unique lyric approach helps with longer songs, judge yourself
https://suno.com/song/dcb82c57-b212-4018-bdd2-7586e6c9a4b3 5min
https://suno.com/song/3c1acd73-f45f-4aa8-8666-4ca27aaf5a33 7min.
the quality at start does not matter here, just the late quality changes.
I am low on credits to experienment myself

[polar spade]

Its hard to say. Spell it STRYK, OR STRIYK, or STREIK and see what you get.

[rough depot]

I'll try

[polar spade]

Thereafter couple of bands with Strike in their names, and certainly songs with the word "Strike" that could be triggering it. Again, who knows.

Also spell it lower case but keep the exclamation point. The capital letters could be triggering it as a label or brand/band name.

[rough depot]

That's exceptionally stupid.

Yep STRYKE worked.

Now how do i prevent it from hallucinating lyrics and singing the word strike

It needs to be an abrupt "STRIKE!" one hit

[polar spade]

Not sure.

• try putting dashes around it --- stryke ----

Or

---

stryke

Then brush up skills on editing/splicing out extra words if not needed. Sometimes that is a much quicker way for me. Just keep it and splice out the busy parts. At least get good enough at it to keep that skill in your back pocket.

[rough depot]

I also need it to not pause the instruments when it shouts

[lone nova]

my problem is the lost of character after cover + voice extracted from same track:
old track and where voice is from v5:
https://suno.com/song/430c2bcc-4afe-4fb2-8831-9bdc34d9bdf1
new cover with that voice losing sweetness IMO v5.5:
https://suno.com/song/bb284635-fac4-4964-83d8-e2282146266f
I do not know if i am hallucinating or not...

[rough depot]

yowza

Twintails eh?

[Intro: Sweeping Heroic Fanfare Buildup]

[Instrumental : 2 measures]

---

stryke

[Instrumental : 2 measures]

---

stryke

[Instrumental : 2 measures]

---

stryke

[Instrumental : 2 measures]

---

stryke

[Instrumental : 2 measures]

---

stryke

[Instrumental : 2 measures]

---

stryke

[polar spade]

Pausing things are not easy really. Thats been somewhat a random thing because its random in real life. Some cinematic bits pause, some dont. Some irish jigs pause, some dont... some Viking chant music... well you get the idea... its random.

As often done in a DAW, for dramatic impact, we simply kill the other tracks for the pause, then bring them back in.

[rough depot]

gah! your suggestion didnt work

They're Mariah Careying it

[polar spade]

Is that stryke and ---- working in there throughout? Suno likes to change up repetitive things like that.

[rough depot]

ffs. The first t2 got the "land of the free" at the superbowl treatment

https://suno.com/s/JXqhtN8Rw1AjOabr

It needs to be an abrupt "STRIKE!" and the music in the background needs to keep going like nothing happened

This worked better. Ok. So any way to pull a specific leitmotif out of Suno's pool? https://suno.com/s/vBGPoc5XUH9RPltS

[polar spade]

By the way... you left [chorus] out of this one. Id assume that would be helpful to leave it?

[rough depot]

So far there is one recurring theme that gives the best results in symphonic power metal

Didnt have a chorus on the second link either and its coming out better

[polar spade]

OK. Just checking. Wasn't sure if that was a choice.

[rough depot]

ARe we able to add inspiration tracks from other sources?

Like "I want a similar feel to THIS"

[polar spade]

We can edit out the first half, save the last half, or whatever was working, and cover that one? Or extend from that one.

[polar spade]

Ive forgotten, which new tool is available in Basic, but I think you have Pro. So you can now make a Custom Model made from all the outputs you've previously made, so that you can apply it as the sole model for a future output to "sound similar" in scope and vibe.

[rough depot]

I guess? hm. I am more curious how i can grab that same theme as found in this one 15 secs in or so, in the intro https://suno.com/s/gTz80dDOWkr6zcaj

Those 10 notes

(if i count correctly)

It works great for a power metal album that tells a story.

[polar spade]

You can snip out just that part..... apply it as a sample.... but it will randomly sample it throughout the next output, and it cant be focused on one area. It might however statt the next output as an intro.

We can snippet it out, then Cover the snippet and get some short intros perhaps, then extend off of those.

Thats a couple of tests.

[polar spade]

Download the intro.... upload it... and cover it with different prompts and weirdness up. For a new perspective of the same theme.

[rough depot]

Ah, could do. Yeah it seems to come up randomly, even in these instrumentals.

And every time it's usually a gem.

I think Grim and Dark has it but let me check...

Here it is, prominent in this verse https://suno.com/s/zLeWBJg3kuXZNn0f?time=263

[timber forge]

I used the exact thing you sent and it didnt work. Still just 1 voice

[sweet ivy]

So I see 5.5 is out, any thought from the mentors who are on, like @polar spade

[modest mason]

Hiyaa,
I'm not sure if I'm writting in the right channel, but I don't care, I have a question?
I created a song and would like the instrumental version without vocals. Separating stems isn’t enough for me because the quality is poor, and I want it to be perfect, so how can I do this differently? Generating the same thing doesn’t work for me, because when I use the option that covers or is inspired by the original song and set “Weirdness” to 0% “Style Influence” to 75%, and “Audio Influence” to 100%, and even when I lower those values—for example, “Audio Influence” to the minimum—it still gives me that damn vocal. Sometimes it’s there from the beginning of the song, and other times half of it is just the instrumental and the other half has vocals. What should I do? Can anyone help?

[polar spade]

Test. Toss. Ride. break. Have fun with it. Some comments both good and bad. But the Cover Modeling and new Voices tool are whats impressing many.

5.5 is beta, so no expectations. So no complaints. Find whats good about it and logistically report any bugs found so it can be improved as the weeks ago on.

[polar spade]

One note is to not use Weirdness at 0%. Its like forcing an old potentiometer down at the zero mark without setting it to OFF, you chance getting crackles and pops and generally bad things as far as fidelity. 1%, 2%.... anything above 0% is seemingly better for Weirdness.

[modest mason]

Okay understand but can you answer how I can do what I want?

[edgy radish]

inspos on v5.5 are still broken I see

[quiet solar]

It's not straightforward to get a perfect instrumental stem separation, since Suno splits it out retrospectively.

If you have Studio, the simplest route is splitting all the stems and regenerating any of them that aren't at the level of 'quality' you expect.

In the absence of Studio, you could try covering the song with the same structural markers (e.g. Verse, Chorus, etc.) in the lyrics, but removing all actual lyrics and covering at mid-ish audio influence. The difficulty is that Suno will tend to do instrumental flourishes if there are no lyrics (ie, it'll do an intstrumental interpretation, not a 'kareoke track').

[remote ivy]

when you mashup, the resultant song often has instructions and references to song A and song B in the meta tags and/or direction...so if we mashup w/ the the some song in the song A and song B plot...can we get away w/ direction that otherwise we wouldnt' be able to doing a normal cover(by directly references the mashup source song) in the lyric box , etc...does my question make sense...a bit vague

[half dragon]

I think you can, no? you still have the lyric box right there before clicking CREATE
Mix and match anyway you like.. I would think!

[modest mason]

Where did the “Remaster” button go?

[cold zenith]

I've noticed a problem when splitting a track into two parts so I can work on them using editing tools. The tempo gradually drifts out of sync, especially in the sections that have been edited in the studio and with editing tools (where I've made numerous adjustments in various places). This is problematic because when I merge the parts back together, the tempo suddenly slows down drastically. I just tried this twice; it took me a total of 6 hours to do it twice, and both times the same thing happened.

[half dragon]

Create > Remaster

[remote ivy]

i mean relative to a normal cover's limitations of referencing the underlying song

[half dragon]

Like you're wondering if we can do the same direction in Cover ? maybe not, as there's no A and B concept... but in Cover, we can rearrange the lyrics, or replace them

[cedar owl]

How does the model feature of Suno v5.5 compare to Inspo?

Something I particularly like about v4.5+ is the filth it produces in the output so much more easily with the genres I tinker with compared to v5. Something that has often been a challenge for me is finding the right balance when covering up into v5 (and now v5.5) of not stripping out the desirable noise and details while getting the better vocal fidelity and quality where I want it. Generating content for inspo on v4.5+ and then using it w/v5 has sometimes yielded good fruit, but a few covers in and v5 often would push down the desirable traits even as I optimize my prompts to preserve them.

Has anyone found the custom model feature to be helpful in preserving such qualities with v5.5? I ask before, ya know, burning a bunch of credits in exploring it

[coral depot]

Hello?

[half dragon]

I thought about this a bit more... what if we mashup from the same song (drag it over twice) would that also be like a Cover? 😀

[remote ivy]

yeah, that is what i was getting at,,,i tried reference underlying song stuff in covers,..no beuno...but i see it explicity done in lyric mashups in the meta tags referring to them as song A and song B....hmm...if i can dig up the example of how they reference the songs in the resultant mashup lyrics

[remote ivy]

hmmm i guess i was mistaken, the i found the song the A and B reference were just in the section tags...so yeah you can mashup w/ its self as alternative, the reference to each song seemed to be an incorrect memory on my part, and just part of the section descriptions it added during the lyric mashup

[cold zenith]

Could you set the remastered to 20min please 🙂

[desert imp]

this is what i do...depends how specific, but i get the stems > remaster subtle/normal is what I use normally (main idea is to clean up your base so your cover doesnt copy those artifacts) > Cover lyrics: [Instrumental], style: blank (0 weird, 0 style influence, 100% audio). This is just one method. I'm sure others have better but its what I do and it works well for me.

[queen linden]

If I wanted to add casual whistling to a song how would I prompt that?

[desert imp]

you can try [Whistle] thats what i do but its not consistent, in style you can also mention sound effects in your style sometime that helps it to understand what you want, use the sounds model + studio to overlay those youd have more control

[queen linden]

thanks, I think I might be trying too hard, I am trying to recreate my grandfathers vaudeville song... again trying too hard lol thanks again

[desert imp]

np lol yeah i do that too, most of the time what works is just... hoping rerolls have something in there :)

or put the whilsten overlays in there, it iwll sound bad, but cover that version and see what comes out

[modest mason]

Yep, it works. I'm incredibly grateful to you because, to be honest, I was starting to go crazy wondering why it wasn't turning out the way I wanted, but now I have another question. It gave me a clean instrumental track, just like the original song, with no vocals, but it changed the genre to rock, and I really want to keep the original main genre

[desert imp]

np glad its helpful, so with that yeah it happens, you can either keep retrying or cover in studio, it has more control with specific instrument, and/or play with style influence

[modest mason]

You have no idea how happy I am that, thanks to you, I finally figured out how it works—I just wish I hadn't wasted so many credits on futile attempts (almost 700 credits gone), ugh...

[desert imp]

yeah i feel ya, ive been using suno for a very long itme since 1.0 lol

[modest mason]

WoW, you're a real veteran hah

[queen linden]

[whistling ] worked perfect kept with the melody very cool Thanks again

[rough depot]

i want the lyrics to drop RIGHT ON THE TEN NOTES
"Evil’d better run when red lightning strikes!
Without warning is when red lightning strikes!
Fulmen rubrum cal-ki-tra-tay koo-los!
Fulmen rubrum deflagra-tay fa-tomb!"
How do i guarantee that?

https://suno.com/s/rWWth4OMl3J7gPZm

[quiet solar]

Not quite sure what you mean by the 'ten notes', but if you're trying to dictate very specific cadence then your best bet is to record you singing it over the top of the instruments and upload it, then cover it so Suno can use it as a guide (editing lyrics after upload as needed). It doesn't need to be an amazing vocal performance, just roughly on the right notes and with the right kind of vibe for the singer to follow.

[rough depot]

This isnt working i want that lyrics to match the background https://suno.com/s/HhFRepJXmVT70XYU

[rough depot]

Look at the link above, the one that says Leitmotif.

[rough depot]

singing it you say

[quiet solar]

I see the link, but I don't know what you mean by 'ten notes'.

[rough depot]

Its the same ten notes repeated twice

See how its pausing the music instead of singing along? https://suno.com/s/HhFRepJXmVT70XYU

[quiet solar]

I hear 12 guitar melody notes in the first phrase, and 13 in the second due to the triplet at the start. But maybe that's just me.

I imagine it's not singing along directly over the guitar melody because Suno's interpreting that as the guitar solo/intro, so it's adding something new for when the lyrics come in. While I know what you mean by "lyrics match the leitmotif", I'm not sure that Suno will be able to interpret that - generally it can't reference back to the song it's creating that literally.

[rough depot]

Wait you hear 12?

imma adjust the lyrics then see if that helps

[frozen dagger]

im guessing this is kinda like a stacatto maybe? you just want the 10 syllables to be broken apart so that each syllable will (hopefully) syn to the notes of the melody?
if so, id' try breaking it down with hyphens and/or commas to try to guide it, maybe?

[rough depot]

Christ again with holding words too long https://suno.com/s/KUiOw5iSNAafGTBe

This one kind of gets it on the second line? https://suno.com/s/fInBPy1hxWRYeFut

[quiet solar]

[crash] dum-da-da / da-da-da / da-da-da / da-da-da-DAAAAAAAH
da-da /** da-da-da / da-da-da / da-da-da-DAAAAAAAH**

Okay, I think I see what you're saying. The bits in bold fall within 4 bars. The quick double note (I misheard it as a triplet) in the second starts just before the next bar begins. And you want to sync the lyrics to those 10 beats.

I'm going to stick with saying the easiest way to do that is by singing it.

Otherwise, you could try giving it something which lyrically matches the cadence of the music and drop further hints at that with punctuation, e.g.:

"I'm wat-ching, your wat-ching, he's wat-ching-YOUUUUUU"

The challenge with your current line is that the meter is more of a 5/5 split than three triplets and a long finish. I'm putting the dashes below just to highlight the syllables:

"Evil'd bet-ter run, when red light-ning strikes" I'd read as having a meter of something like da-da / da-da-dum / da-da / dum dum dum

[rough depot]

ok i will sing it

so if i map dum da da to evil'd bet? is that what you mean?

[quiet solar]

It might still rework your line on covering because the meter is slightly off.

Musically the line has an anapestic rhythm, with stresses on the first note of each triplet and a final one on the long note. So your current lyrics would come over as:

EV-il'd bet-TER run when RED lightning STRIKES. Suno might see the slightly awkward stressing of bet-TER rather than BET-ter and assume you wanted something different.

[rough depot]

This one ALMOST does it but it holds the final strike too long https://suno.com/s/fr63YsUMxUcNjhRh

If someone would actually explain what is meant when stress is on something...

[gritty acorn]

How to make suno recognize that the last word of a line is supposed to be slightly higher pitch?

[rough depot]

[Tone Shift]? maybe

[gritty acorn]

Where would I put that? So if the line was something dumb like

The dog ate the cat
The cat had a bat
The cat was fat
Needs to get his life on track

How would I specifically make cat, bat and track, higher pitched

[rough depot]

like right before the last word i think?

that or use studio to change the pitch manually

[gritty acorn]

I tried it with [higher] before the last word and it did it, but it also sang higher lmao

[gritty acorn]

I only have pro

[quiet solar]

In your last upload, you can hear the odd stressing I mentioned. The way Suno handled it was:

e-VIL had bet-TER run when RED lightning STRIIIKES

Stressing is where you put the emphasis on a word. So in that version it says the word e-VIL rather than the more usual stressing of EEE-vil, for example.

A good example would be something like 'content'. CON-tent is the word used as in 'content creator'. con-TENT would mean a feeling of happiness as in 'I'm content'.

[rough depot]

so which syllable is "Emphasized" in content? The CON right?

[quiet solar]

Yes, the capitalised one.

A phrase with a more natural anapestic flow would be something like "EE-vil starts RUN-ning when RED light-ning STRIIIIKES"

You might not even need to break that out into syllables. I suspect Suno would map that to the phrase if just written as "Evil starts running when red lightning strikes".

[rough depot]

in which case the next line should say something like "But there's no escape when red lightning strikes: like

BUT there's no ES-cape when RED light-ning STRIIIIKES?

[quiet solar]

Not quite. If it followed the same rhythm, that would give you:

"BUT there's no ES-cape when RED light-ning STRIIIIKES" which gives you an odd stressing on 'escape'.

You could try "But there's no escaping when red lightning strikes", and hope Suno takes the hint that the first syllable is unstressed and meant to come in before the bar starts:

but THERE'S no es-CAPE-ing when RED light-ning STRIIIKES

[rough depot]

STRIIIKES is being interpreted as strill-yecks

[quiet solar]

Haha. Don't write it phonetically if you don't have to. Try giving it:

Evil starts running when red lightning strikes
But there's no escaping when red lightning strikes

And see if it takes the implied hint at how that should fit to the music.

[rough depot]

Jesus its doing the slowdown thing again https://suno.com/s/RxjjwroRkbHEiqgu

[quiet solar]

Yeah, that was ugly repeat line with all the drawn out syllables.

[frozen dagger]

do have "no sustain" in there too. could be playing tricks with not recognizing the "no".

[rough depot]

gah it still does it https://suno.com/s/bCDgTcWoZvgEzC0L

[quiet solar]

I genuinely don't know how you'd prompt for that (or if you could) since it's such a specific thing to ask for. The linguistic term for it is a "high rising terminal", or "rising inflection", but I don't know if Suno would recognise that as a prompt.

Could always hit record and sing the line to Suno in the way you want it?

[quiet solar]

Worth noting that aside from that final sustain note, it picked it up exactly as we'd hoped it would though. 😉

@rough depot Try it without the syllable breakdown, just as two straight lines:

Evil starts running when red lightning strikes
But there's no escaping when red lightning strikes

[desert imp]

have you tried just micro edit with studio? you can play around with Transpose/tempo...etc in individual clips. And if it sounds close enough can cover/remaster to enhance it or just keep regenerate in editor lol osmetimes thats better OR just use extend

[rough depot]

damn it i want it to sing along to the guitar riffs https://suno.com/s/K9fQpzivhp9rsb5Y

[frozen dagger]

im seeing if i can get a clip of it as well. it really wants to sing that strriiiike. you want it short tho, right?

[rough depot]

I want it to match the notes of the guitars so yeah

i can always drop the concept.

Because some of the mashups are already good without lyrics

Like this one: https://suno.com/s/SK7xV3lmw50qHgBh

I am going for something like ya know how Maker's Ruin keeps cropping up in FFXIV for example,

[young geode]

[Vocalist: Male, Tenor, Enunciated, staccato, short notes, Doubling, Unison]
E–vil starts RUN–nin'—when RED light–nin' STRYKES,
but there's no e–scap–in'—when RED light–nin' STRIKES!```

Try this.

https://suno.com/s/z0RsI4cABIdeqyQo
https://suno.com/s/Ge2hJu1YrBLQG8ZU


Just so you know, there are multiple dashes and each one affects the vocalist differently:

Hyphen `-` will create a run-on phrasing.
En Dash `–` will create a Staccato phrasing.
Em Dash `—` will create a paused phrasing.
Apostrophes `'` will create a muted phrasing.

Additionally, `Doubling` is the term for a vocalist following a musical rhythm of instruments and `Unison` is for following a pitch.

[rough depot]

How do i em dash exactly?

[frozen dagger]

https://suno.com/s/GNI5mpdl6iZmSH9u
🤷‍♂️

[rough depot]

sec gotta cook

thx btw but brb

[young geode]

its a bit of a pain.

The way I do it is either edit on the phone and long hold on the dashes, or goto google docs.

One - is a hyphen, Two -- is an En Dash, Three --- is an Em Dash.

[frozen dagger]

if on PC hold Alt and press 0151 on numpad

[young geode]

yes but you have to remember the Alt-Code or have it somewhere

[frozen dagger]

true. i won't forget. i have it as lyrics.

i remember 0191 from spanish too tho. ¿

[rough depot]

nope now it pronounces it as es-capp

[young geode]

It's pronouncing e–scap–in as ES-capp?

[rough depot]

Yes. instead of scape. imma spell it e-scape-in

[young geode]

try 'scape-in'

note: thats going to drop a syllable

[quiet solar]

Or try reverting back to es-cape-ing which worked fine in an earlier attempt.

[rough depot]

checking

cripes. fine I'll just drop this idea and stick to the "latin chanting in the background" (fulgur rubrum deflagrate fatum) How do i consistently get a background chant?

Like if i JUST want an instrumental with that background hymnal thing that Suno likes to add. If i want it to say things like Fulgur rubrum calcitrate culos etc

[acoustic egret]

Hey guys am I able to upload my guitar tone into suno v5.5 and it use that as the basis for guitars?

[frozen dagger]

I record acoustic riffs and they translate well, imo.

[acoustic egret]

In v5.5?

[frozen dagger]

mostly v5. didn't do much new with 5.5. did some covers. the riffs are there. it adds it own spin as it does.

[acoustic egret]

Just use a match EQ to your reference track and a limiter and ur done

[acoustic egret]

How for you record your riffs? Can u send a s/s

[frozen dagger]

I'm just playing my acoustic with my phone on my desk. nothing pro.

[acoustic egret]

Yeah but how do u import that into suno?

[frozen dagger]

oh. i been using the suno app and just recording it straight in from there.

[polar spade]

The +Audio button (in the Create panel for one place) is used for recording live. A choice of Upload and Record.

[acoustic egret]

Wow thanks ill try it tonight

[young geode]

Likely.

Im trying to figure out exactly what you're going for here.

Did you just want that one intro vocal to play once, in accordance with the melody?

[rough depot]

Yes so i could later use the result in a mashup like I did here with JUST the melodyhttps://suno.com/s/YBUN6kwaWfPSakxu

Because this particular melody comes up a fair few times in my creations for this particular project - suits me well because I love leitmotific transformation

[quiet solar]

You might have to manage your expectations around "consistent" if you're specifically going for a background chant.

I was able to get something which I think is fairly similar to what you're after using the longform prompt "Vocals are softly chanted by a choir, remaining in the background of the song". However, getting Suno to perform with just the choir without having a lead soloist is quite challenging - Suno tends to assume that there should be a lead vocal, so you may have to roll with that or brush up on your audio editing skills to remove them.

However, if you're looking to use it for mashup or something similar, you only need to have one section performed correctly and then crop that out from the original song. Adding something like [soft choral chanting] or [background chanting] to the lyrics may help.

[young geode]

Oh, youre trying to effectively import the vocals there into a mix.

Alright, Ill see what I can do here. You might have to do some heavy style prompting and metatagging for this

[frozen dagger]

if you like leitmotific transformation, you may like (if you haven't already) try creating a custom model with your songs. It works in a way similar to how Inspo does where it will stitch in pieces of your audio into the melodies and such.
I have not tried to do this, yet (it's still new), but I could see taking your variations and placing them all into the custom model (maybe even with some duplicates of your favorites or base samples placed in there to give it some more weight) and see if that produces anything to your liking.
I hear pieces from my tracks in the gens all the time. I dig it a lot.

[rough depot]

maybe when i get a break and an actual backlong for my story ugh. i hate serializing

[frozen dagger]

i peaked at it earlier. looks fun. lots of content over long time. very cool.

[rough depot]

thanks

[young geode]

How did you want the vocalist to sing? Be specific.

I'm going to spend tokens working this out

[shadow plaza]

Custom mode does it cost 100 to train whole model or 100 per song included into the model???

I refer to create custom model

[mossy basalt]

Is there a pinned starter FAQ somewhere? I want to upload some of my own audio and layer it in (I only have Pro not Premier)

I was hoping to add some cracking fire audio but I cannot for the life of me understand how to add audio or do anything in the editor that I would want to do

[tidal prism]

I need help covering a song!

I randomly generated a metal pop song, It has screaming male verses and the drops soft female pop vocals

I’m trying to cover the song with my favourite persona (it has screams and male clean vocals)

But EVERY generation has the same female vocals, I can’t get them out of the song replaced with the male vocals in my persona

[quiet solar]

The Core/Legacy editors in Suno can only amend the song that you directly open - they're not able to import additional sounds or do any kind of layering. For that functionality, you'd need Suno Studio, or an external DAW or audio editor. Audacity is a free one that might be worth looking at.

[mossy basalt]

OK, so If I want to go to that level of detail, either upgrade or export and play inanother app, got it! 😄 Than k you for saving me digging around further

[quiet solar]

Try turning down the audio influence slider, setting the vocal gender to male and tagging the choruses in the lyrics (and anywhere else the female voice appears) with something like [chorus - male vocalist]

[ocean jetty]

Can someone help with a prompt on creating a in studio live performance, natural vocal/band sound? The prompt when using "Live" presents crowd noises but I am curious if anyone has any luck on how to that in off the floor sound?

[solid pagoda]

v5.5, all sounding a bit the same like Artlist stock library 🤭 — any tips on how not to have percussion? Adding 'percussion' into Exclude Styles isn't helping. Any guides on how to 'negative prompt' what I don't want?

[cold zenith]

I've noticed another issue: when using the editors, the sound is muffled. I think this is due to the music being processed through a compressor during reconstruction, which degrades the quality. Instead of simply running it through a “cover” type filter to preserve the quality. This means that if they perform multiple manipulations on a track, the quality drops from exceptional to mediocre.

[frozen dagger]

This is my acoustic guitar studio example. I amend it as needed (usually with a chat bot help):

An unplugged studio solo session centered on a single acoustic guitar, captured with intimate close-mic detail as if performed live in the room. Virtuosic percussive fingerstyle playing shapes the entire performance, integrating string slapping, finger taps, body percussion on the soundboard, palm muting, and rhythmic dead notes directly into the groove. Expressive finger control unfolds through hammer-ons, pull-offs, legato runs, trills, slides, glissando, controlled string bending with nuanced vibrato, and fluid dynamic articulation. Hybrid picking and finger independence support rapid arpeggios, double-stops, chord-melody voicings, and subtle counterpoint. Natural harmonics, tapped harmonics, harp harmonics, and campanella ringing strings add shimmering overtones. Occasional two-hand tapping passages emerge organically. The performance breathes with natural tempo movement, warm wooden resonance, crisp transient attack, and cinematic phrasing that feels discovered in real time.

[tawdry viper]

Hi all, im completely unable to use my voice in suno although I have the pro plan. Can anyone help out?

[polar spade]

There are three ways to incorporate a recorded voice to be a Voice.

We can click on +Voice
Click Create a Voice
RECORD , upload or choose a file with our vocals.

Then:
CREATE A new song and add that Voice.
REMIX a song to add that Voice.

Which step are you having an issue?

[tawdry viper]

"voices from audio uploads are disabled at this time"

@polar spade

you can also see theres no directly record option in voice

[polar spade]

Sure thing.. Ill check for myself. I havent used Suno today. Hold on.

[tawdry viper]

its like the UI itself is different for me

[thorny swallow]

Is this in the mobile app maybe that's what are your problems there

[small thistle]

Exactly the same as what I am dealing with

On my chrome browser

[slow merlin]

and I am on Chrome

[small thistle]

Well yes but then it doesn't let you use that upload to make a Voice

Thats the problem

Because it says that it isn't currently accepting audio uploads

[keen crescent]

Hi Family, I am not a guru with the editor, in the image attached. This song is too long, 5:40 - I want to shorten it, what can I take out and how do I make the 2 parts/sections blend in or let the AI re-generate it to blend in?

[slow merlin]

could be something on Suno's end. Ever play a new game? When there is a update for ANYTHING stuff always takes a few days sometimes even weeks to get back to normal. Just a thought. Maybe give it a few days?

[polar spade]

The error is obvious, but we should see if it similar for everyone. Still uploading. Hold on.

[polar spade]

Still uploading. Hold, please.

[small thistle]

Thats what i was thinking too :p
It doesnt really matter much to me, I just noticed that @tawdry viper was having a hard time getting the issue across accurately and I was having the exact same problem xD

[polar spade]

@tawdry viper approaching 2 minutes and its still a progress spinner. There may be some issue for uploading presently.

[tawdry viper]

will it be fixed

[polar spade]

It would make sense that Suno addresses all bugs so that the program is usable as intended. I am seeing now if it uploads a second time.

I've uploaded a 7 second song.

Now trying to upload the longer song again. Lets see.

[polar spade]

How recent were you having trouble?

Ive just uploaded both a 7 second and a 3:30 song. Using the +Audio tab.

OK. Ill try the +Voice tab for the songs Ive uploaded.

[tawdry viper]

since today

[polar spade]

If you can vid or at least screenshot the issue then and the report as a bug, for now. They need to find the issue to be hope ing with others as well. https://discord.com/channels/1069381916492562582/1082775596767117312

I uploaded two songs this morning ✅
I made a new Voice, through the Voice tab in Library, from a recorded file with my voice on it. ✅

Those would be the needed steps to test that each work separately from each other.

Then if you have problems with the uploading by clicking the Voice tab in Create, then record that issue and report it.

And just to test that version as well. Ill now use the Voice tab in Create.

[polar spade]

I had no error from attempting to upload a file using +Voice tab in create. If its not glitching for me, I can work out that -- uploading a file into Voices -- as an issue.

So try again and screenshot the error, to post as a bug 🐛

[polar spade]

Alao, as the UI looks older, refresh your Browser.

If its persistently this UI you posted

Then clear your cache.
Close the Browsers
Log out / log in
Open one Browser.
Try again.

[polar spade]

Yes. This is the newer UI you should see....

[frozen dagger]

I ran into some errors on day 1 trying to use the record and upload options, to avoid them I put my Vocal upload into a playlist. That worked for me.

[frozen dagger]

not seeing the whole image here, but the fact that Public is toggled on makes me think this is the wrong UI route. Upload vocals have Public toggled off by default.

[polar spade]

Yes. Your error was before 9:30. I uploaded and created a voice at 9:50. So its true that sometimes a momentary glitch happens. Let the room know if your next attempts then result in a positive change after a short wait period.

[small thistle]

Ive been trying all week ^^;

[polar spade]

Ahhh.... others have had no issues "all Week". It may be a more individual computer issue. A cache problem, we would hope.

As you mentioned, Your UI doesnt look current. So clearing the localStorage can clear the cache and further temp or corrupted files to better sync with Suno to keep the UI current.

You know how to clear the localStorage on Chrome?

[small thistle]

Haven't the foggiest Dav

[polar spade]

You're in Chrome on desktop?

[small thistle]

Yup

[tawdry viper]

i dont understand man

it says its not accepting audio files

how is that a computer issue

[polar spade]

Im chatting with two of you at the same time. So were basically discussing the same thing. Sorry for the confusion.

[polar spade]

• Updating the computer to current files is helpful to rule out a local issue.

• If its a current server issue, I should also be having a similar issue, which I had not. So the problem need to be repeatable by others.

• Its possible to have been an issue recently but has cleared. Have you tested again in the past 30 minutes. I was able to both upload and create a voice at around 9:50am EST

[tawdry viper]

how to update

im using webb version suno

[polar spade]

Yes. Im making a video now for @tawdry viper @small thistle Ill be done shortly.

[tawdry viper]

risk? what risk?

[polar spade]

If you wish then, use these steps instead. They will help.

Then clear your cache.
Close the Browsers
Log out / log in
Open one Browser.
Try again.

[tawdry viper]

but what risk man iim scared

[polar spade]

So ignore the video and do the simply steps as mentioned below....

You know how to clear the cache,

Go to Chrome settings.
Privacy and Security
Clear / Delete browsing date (or you may just need to go into full History section to find)
Clear Cache (at minimum).
log out
Close all browsers
open one browser
Log back in.

[polar spade]

@tawdry viper @small thistle Any luck?

[small thistle]

Oh sorry I wasn't looking ^^; lemme go over it now. I was working on something else

[polar spade]

OK. If needing the instructions again. just let me know. Vid has been retired. The above limited version to clear the cache is helpful as well.

[tawdry viper]

@polar spade

i did what you said

i cleared cache

logged out

logged back in

to what avail

@polar spade

[frozen dagger]

This is from the select from library? Have you tried add your upload to its own playlist and try the "choose from playlist" option? I had to do that way when Upload and Library were being weird.

[tawdry viper]

how do i do that

[frozen dagger]

below there add to que it has add to playlist

it looks lilke that option was removed as im trying to screenshot to show

oh no

its therer

[tawdry viper]

@frozen dagger i added to playlist

[frozen dagger]

do what m showing sleect form library and then

[tawdry viper]

bro im not even getting this part 😭

[polar spade]

Looks like you're trying the same file? Have you uploaded anything else as @frozen dagger suggests. The error may be a way to just censor that one file. So test other files as well?

[tawdry viper]

ive tried two different recordings

@frozen dagger

[frozen dagger]

Youre in your library

Cant acces that way sir

go tot the Create tab so you can see th +Voice

[tawdry viper]

wait

for simplicities sake can i just screenshare to you both?

@frozen dagger

@polar spade

[frozen dagger]

if it doesn't affect any1 else i dont mind.

[tawdry viper]

wdym

so can i screenshare?

where is vc

[frozen dagger]

I think the confusion coming from the UI hiccups. If it's your voice you want to use to make a "Voice"/persona. then go to create tab an choose +Voice to get the correct UI

[polar spade]

Doesnt that screen shot though look like the older UI version off of the Voices/Persona page? Still seems to need to update the computer client files to start clean.

[frozen dagger]

Except it has the drop down menu in his screen of "which song this was based on' and i don't recall that in the earlier form

[tawdry viper]

how do i do that

can i pls just show u guys the situation in screensharee?

[polar spade]

Sure.. go ahead.

[tawdry viper]

where is the vc

[frozen dagger]

we dont hafve Voice channels in the suno discord. thats why i meant like if it doesnt affect anyone else. not sure there was a way without switching to a diff server

[tawdry viper]

dms? @frozen dagger

[frozen dagger]

sure. honestly Dasvon gonna know better how to fix the problem. i can identify it pretty well and troubleshoot but if its a cache prob you want his expertise

[tawdry viper]

yes

dasvon pls help

we can come to dm

[polar spade]

Quick check. At any time during this week, have you turn off your computer. Waited a minute and turned it back on? That will get you closer to updating with Sunos server.

[frozen dagger]

(I'm learning all the time too)

[tawdry viper]

i turned it off everytime before i go to bed man

[frozen dagger]

this is google chrome?

[tawdry viper]

microsoft edge

[polar spade]

OK. Let me bring up Edge so we're on the same page. lol

[frozen dagger]

this is my libarary view. not squared. You can access the same UI screen (the one you said not showing for you) in library clikc "Create new Voice" but as you see the design looks different. Assume we tried a different browser? (sorry if u said that already)

[tawdry viper]

hmm

i think all the misunderstanding

will be cleared

if i can screenshare

@polar spade

@frozen dagger

[polar spade]

This is a general Suno channel. Sure. Can it screenshare in here. I dont have any issues. We've already screenshared a few shots already.

[tawdry viper]

come to dm

[frozen dagger]

I can add u as friends and invite you to my sever voice chat and you can share your screen with us there if you'd like?

[tawdry viper]

this sewrver dont have vc

[tawdry viper]

yes sure!

[frozen dagger]

friend request sent @polar spade

[mellow hill]

I have a song that I made that I really like but wanted to swap out the voice for a persona voice that I saved after the fact. I've tried using the Cover feature and I've tried different things in Studio after splitting the stems, but I can't get it to work exactly the way I want.

I really want to keep the music of the original song exactly the way it is but just switch the voice to the saved persona with the persona singing the lyrics in the same way that the original did. But every way I try it changes the sound of the music too much from the original.

Is there a way to do this? Or is just making a cover that is the closest approximation to the original the best available option?

[half dragon]

I found it's not simple to swap out the vocal on a premade songs on Suno.
Your best bet would be Cover... or replace vocal in your DAW.

[desert imp]

Try cover the original song + persona, set weird to 0, style to 80-100 (set your new styles of course) if you don’t want it to sound like the original style and audio influence to 90-100% depending how close you want the melody to match. Mainly change the style if you want different style…. Sometimes too close will get weird artifacts but recover with 5.5 can save it

Try this first on 5 and then re-cover it again same above but with the new song with the new style you like on 5.5 for better vocal and just production sound quality. Or skip 5 go straight to 5.5. I do 5 because sometimes it follows the melody better.

Or try the 5.5 model although I haven’t had consistent results so this this my old method/workaround Or studio with polished stems and check sync beats...etc align them that way. these are just my ways idk how others do it though o.o

[mellow hill]

Thanks that was essentially the conclusion that I was coming to but had to check with someone else to make sure

[mellow hill]

Thanks, yeah I've messed around with the Cover feature and different slider settings quite a bit. I'm able to get a fairly close approximation to the original with the new persona voice but it's just not quite as good as the original. I'll probably just have to go with the closest comparable cover version I can come up with

[small bridge]

Why are my styles changing the moment I use the new 5.5? I want the voice accuracy but it keeps changing the songs backgrounds and song beats that I'm trying to get it to cover. Is there a way to remaster the songs with the more accurate voices, rather than it redoing the whole song and runing what I have?

[quiet solar]

It's worked reasonably well for me covering a v5 song in v5.5 with a Voice, deleting the style prompt entirely and dialling weirdness down to 30% and audio influence up to 50%.

It's a new tool though, so your mileage may vary. We're all still learning the quirks and what the best way is of doing things. If you can't find an easy fix, try narrowing down whether the problem is purely in the Voice by doing a couple of covers like that without adding the Voice and seeing if the style still changes.

I'd also be wary of setting the influence sliders to more extreme levels in 5.5 - it seems far more sensitive to smaller changes than previous models, so keeping closer to the defaults often works better.

[small bridge]

I'll give this a try! I used my voice as a model in 5.0 but it made my voice a little low. I got a song thats closer to my pitch so I used that one to redo a new voice and was using it for 5.5. it's much closer to mine now, but the song changes are wild! Always a learning curve with the new tool. I do love the voice accuracy though.haha

[quiet solar]

Best of luck!

[vague stratus]

what witchcraft does one need to use to prevent getting drums and percussions, trying to remix a song and it always ends up with beats. a long list doesn't seem to help in the exclude styles

[pure panther]

Maybe it's just in my head, but v5 seems to working better than it ever has since v5.5 was released.

[young geode]

Post it

Are you trying to completely rip out all drums and all percussion?

[vague stratus]

the song is an ambient song and has no drums or percussions, I want to create variants and now I always gets drums in the end

[young geode]

It's other genres taking control, because they have drums.

What you'd have to do, likely, is target instruments, scales and play patterns without targeting any particular genre.

[hoary anvil]

a DAW and not AI probably

[vague stratus]

I finally got something without drums, I didn't realize you have to put the negatives also to Styles, not only to the Exclude section in More options

[rustic arch]

Hello, I lost access to my Premium account and I can't remember which email I used. However, I am still being charged. I have attached my payment receipt to this email. Can you please help me find the email address associated with this payment? Thank you.

[young geode]

Post your style prompt.

Typically adding negatives doesn't work.

[vague stratus]

after adding this it seemed to help
"droning ambient, symphony orchestra, classical music, smooth -no drums, -no kickdrum -no beats -no percussions -no hihat -no beat"

[quiet solar]

Hi nesa - this isn't a support forum. You need to contact [e-mail] and include:

Your Suno @handle or a link to one of your songs
A short description of the billing issue
A full copy of your receipt/invoice showing the purchase date, amount paid, order or transaction ID

If you've already done so, you don't need to take any further action. Your e-mail will be in a queue for action and support will contact you in due course.

[dusky narwhal]

Hey there.

So, I am assuming I can ask for help here but I'm trying to deal with Suno pronunciation.

In my song I have the lyric of lives as in "Frankenstein lives"

But it constantly says lives as in "best day of our lives"

Any help or tips I could use to kinda get the AI to say the proper lyrics I'm aiming for?

[leaden bronze]

???

[half dragon]

Asking Gemini for phonetic spelling

[dusky narwhal]

Thank you for the help, Ill try that

[frozen dagger]

lihvz (rhymes with gives)
worked for me. other form i use lyhv (rhymes with hive)

[finite agate]

How to fix this it wont let me in

Someone pls reply

[neat nexus]

You took 365 days to be born? Maybe give the full date?

(Just a guess on my part.)

[ocean jetty]

I'll give this a shot, thank you

[quiet solar]

A fairly tried and tested method for this specific issue:

L'ives = "Frankenstein lives"
Lyves / Li'ves = "Best day of our lives"

[supple moon]

I’ve been doing all sorts of genres and been successful but the one I’ve been stumped on so far is any kind of heavy metal/death core.
There’s always random parts of the song where a melodic clean vocal comes in.
Is there some specific language I can use to stop that from happening?
I’ve tried just about everything I can think of but it always gets inserted somewhere in a song.

Please tag me in response because I won’t see otherwise! Thanks!

[frozen dagger]

just the vocal the issue going to melodic for your tastes? the music is fine? crop the part of the vocal that is good and make it a Voice?

@supple moon e deathcore can hae melodic singing. ask chatbot for related genres absent of melodic singing (or at least mostly), Black Metal, Brutal Death Metal, Grindcore

[supple moon]

I’ll give all that a shot! Thanks!

[remote ivy]

I'm guessing my voice persona don't really gender swap? i tried 3 renders, added excludes for male vocals, vocal duet, harmony. specified female vocals, selected female vocal...but I only got duets that were 35% female. 3 create attempts on same song, which was indie alt country rock

[frozen dagger]

It's supposed to try to stick to the Voice as close as possible. So, if it's a male vocal and prompting for female that sounds about right what you described. If you want a fully gender swap create a track with your voice and cover it with the female prompting stuff and then create a new Voice with the all female on a gen you like.
Alternatively, you can create a Voice with both male and female vocals and it will pull from both. Prompting can help guide and persuade, but ultimately less control over which gender appears. (reliably).

[remote ivy]

i didn't have that issue w/ v5 model...i switched to v5, and got 75% female, using the my vocal persona....still duet, but I successfully remove it....and a bonue my voice sounds better in v5 than v5.5..it got rid of the weird reverb v5.5 render had with my vocal sections from my bad recording environment...but i digress

[frozen dagger]

some of that could be variation with the randomness. again. sounds about right. Except, for me, I've been having issues where trying to Cover vocals was applying negligible differences at all. It's gotten a little better, but it still doesn't seem to fully replace any Vocals when I try to Cover and swap singer. Takes extra steps.

[remote ivy]

hmm kind of an empty victory for me....didn't really have my timbre on a re-listen...but i was at 25% audio on slider since my own voice sounded so bad (wrong genre for the vocal uploaded)

[remote ivy]

yeah i usually try to make instrumentals first lately for that reason, this way i can swap out singers w/ impunity...and once i find a good render that needs tweaking , the nice side affect is the voice doesn't change a whole lot

[keen crescent]

Anyone??

[half dragon]

Options:

• Find a spot you want to end with roughly shorter the desired length, then Extend to add the Outro to end it --> then Get Whole Song to merge 2 parts into the new song.
• Find the spot you want to end, cut it in Editor, then Fade Out.

[keen crescent]

The first Sentenece is a bit confusing LOL

[half dragon]

oh sorry - what I meant was - let's say you feel 5:00 ish is the length you want, try to extend from 3:00 or so (or wherever if musically allowed in the song structure) to add the final part , could be an Outro or however you want to end it.. once you have that extension part... click "Get Whole Song" to merge that end part to make a new complete song with the new ending.
Hope that makes sense 🙂

or just simply crop it and add fade out in editor.
I don't usually shorten a song but that's how I would do it.

[neat mirage]

So many bugs in 5.5 , i give instruction to make instrumental track but suno add vocals to track ? also tries extending track 9 times still suno failed to extend n showing credit refunded ?

[tawdry viper]

@neat mirage

the UI is literally the old UI for me right now

I cant even use voice

[vague scaffold]

yeah

[tawdry viper]

@polar spade tried helping but nothing seems to work

@crystal prairie please assist

[vague scaffold]

v5.5 extend is still broken

[tawdry viper]

does the voice thing happen to you guys to?

[vague scaffold]

i havent tested v5.5 yet

ill try to on april 1st

[quiet solar]

No issues on desktop for me. If you're doing it on a Suno-created track, do you get the some odd interface if you try creating voice from selecting the song in the Library and using ... -> Create -> Voice?

[frozen dagger]

yes. he showed Dasvon and I earlier. Looks like mix of old and new UI. has drop down of "based on selected song" and his Public option was still toggled on by default.

[frozen dagger]

did you see this? Was in other thread after we ended chat earlier.
#💬┃studio-chat message

[proud dew]

Does anyone know if Custom Models can be edited? I've created one that's working well, but I'd like to add additional songs to it. Is there a way I can do this or will I have to make a new model?

[tawdry viper]

@quiet solar

the interface itself is the old interface

[agile stag]

Does anyone know what to prompt for this?

Piano, Falsetto/Whispered vocals, and No drums

[quiet solar]

Yes - that's still the new Voices output, and the route you get if you go via the Library.

If you instead go via Create tab, +Voice, you'll get to the new interface UI.

[quiet solar]

I have exactly the same UI if I go the Library route to creating a voice. It still works absolutely fine.

[frozen dagger]

when he clicks creat voice it goes straight to that pop up, skipping this screen I am sharing, and goes straight to one he is showing, which is supposed to be after he see this one and is able to choose a song. He doesn't select anything or see this screen at all.
Dasvon walked him through clearing storage. We watched him try Chrome (was Microsoft Edge) same thing.

[quiet solar]

Ah, okay. No idea what's going on there. Cheers for the clarification

[frozen dagger]

Thank you, as always, for being helpful. I dunno what it is. I learn from everyone else in here as well. Suspect it may be regional issue? But again, I don't know. I just try to help troubleshoot and do what I can, and learn as I go.

[quick wedge]

Morning everyone 👋 what are some good prompts to make an instrumental not sound generic? I shared one last night and that’s what someone said about it. Just thought I’d ask.

[viscid ravine]

Ask chatgpt for prompts 😉

And good morning Pim

[quick wedge]

I always ask ChatGPT but I guess I can try copilot or something

Morning Noud

[viscid ravine]

Try Google AI Mode too 😛

[viscid ravine]

How is your morning going☺️

[quick wedge]

Morning is alright 👍

[vapid steppe]

i need some help please, i am trying to create a song cover by uploading my vocal ..but after 2 minutes the melody changing,,,i have subscription and using V 5.5

[sleek stump]

This has been described by quite a few users, try switching to v5 and see if that helps. Then you can remaster in 5.5 if needed.

[vapid steppe]

even with V5 its the same....i burned out more than 7k credits

[sleek stump]

Can you share a link to one of these songs so we can see whats happening?

[vapid steppe]

its a hindi ghazal.....is it ok?

[sleek stump]

Sure

[vapid steppe]

https://suno.com/s/28sdqxU8aArt1Wk9 from third verse it happens

[sleek stump]

WHats the time stamp of the change you are describing? I dont speak Hindi, but I am familiar with India music...

[vapid steppe]

from 2 minutes on wards....it starts changing

[sleek stump]

WHat I am hearing is not a melody change but a change in mood, the song gets brighter and more energetic. I hear the same base melody all the way through. By the way, this is a very nice piece! Reminds me a lot of my time in Noida and other towns outside New Delhi

[vapid steppe]

great...but not steady as the intro and the 1st verse right?

i set weirdness to 0 and still ...

[sleek stump]

I dont think it needs to be, but it's your song. I listen a lot to Ravi Shankar and his daughter Anoushka, and its common for temporal changes throughg the songs, tbh.

[vapid steppe]

as an expert can you please guide me ...or any majical prompts or settings please

are you working with Suno?

[sleek stump]

I am not sure what to suggest here, because the arrangement is developed according to Suno's training in those genre. You may need to adjust the prompting to a sub genre style that may be more flat through the entire song.. One thing I am seeing that may be causing you an issue is you have "no stylistic deviation between sections", Suno will actually do more of that because it is in the styles prompt" Keep in mind that anything in Styles is what Suno will give you. The exclude box is where you place things you prefer not to have.
WHich also has some issues, if you put "avoid rhythm deviation" It will just cancel itself out. Try "rhythm deviation" only, this is not a double negative and should get Suno to hopefully more respect the exclusion. I suggest editing all your exclusions to be singular items and avoid double negatives.

[vapid steppe]

I had the exact same problem and burned through 4K credits trying to fix it. My song is about 5 minutes long, but when I do a cover, it only keeps the original melody for around 2 minutes before it falls apart. Did you ever get it working? … I hope you understand now .....this is from another user....same experience

After ~1.5–2 minutes, the model starts “forgetting” earlier constraints

After 2 minutes You’ll notice:
Melody slowly shifts
Phrasing becomes more “creative”
Emotion increases
Sometimes even structure changes

hello

[quick wedge]

<@&1082792303976001547>

[frozen dagger]

the only suggestion coming to mind would be to detail the instruments you want used and how they are used.

[quiet solar]

Blend in a slightly different genre, maybe include an instrument that's not so frequently used in the genre, maybe have a play around with prompts around the emotional vibe to give it a different twist.

Alternatively, don't take throwaway criticism to heart so much. Calling something 'generic' is almost like calling something 'bad'. It doesn't tell you anything about what the listener did or didn't like, or what they personally would have liked you to do differently. So it's hard to form any real view about whether you should listen to that feedback as an artist and what, if anything, you should do differently.

[half dragon]

Music is subjective -- unless someone gave you a very constructive opinion with technical aspects, take it with a grain of salt - the connection is personal person-to-person.
"One man's trash is another man's treasure" 🙂

[wintry saddle]

Upload multiple chunks and use extend....

Theirs context window limits

[pure panther]

I finally finished the track, this one was a tough one. I wanted to keep the same vibe as the rest of the album, but still make it feel like a celebration. (Art Rock, Pink Floyd inspired)
https://suno.com/s/Kj76cO2lDAyDM8Am

[frozen dagger]

Listening now, had to start over. Was on my speakers, switched to headphones, and when the audio came into my headphones after first 30 secs or so the music sounded so different. I had to double-take that I didn't hit skip track or something.

[pure panther]

Lol, yeah. I'm honestly still on the fence about the intro. I wanted a touch of duduk and harp in the song, but it's only at the intro, then the guitar attacks hard lol

[frozen dagger]

intro works. the fact that i literally did a double-take. i love those moments. Captured my attention at the start, for sure. 🙂

[frozen dagger]

@vapid steppe I opted to upload the song I wanted to Cover 6+ times into a Custom Model, did Covers using that. I wasn't trying to get it to be exact the same as my original song as the original song half the length, and lyrics very different. But this worked better for me as far as not losing the vibe as the song progressed. Just another way to give Suno context when it's creating. May be worth a try, especially if there is minimal changes being made. (Unlike what I was doing.)

[wintry saddle]

yep - always look for other ways

[pure panther]

Oh nice! That comment might be what puts me over the edge to keep it 😄

[quick wedge]

The user didn’t say anything specific pretty much said it felt generic to him. Ginger helped me to where I wouldn’t explode back at him. I sort of blocked him too after I sent the last comment to him. Thank you for the reply was surprised by yours and Kanthric.

Thank you @quiet solar 😊

[quick wedge]

I’ll experiment with the instrumental details with future songs and see how it’ll work out. Thank you.

[uneven vessel]

Hello. I am a Suno Pro user. I am creating a song, but whenever I try to edit it in Suno, it changes the song's structure—specifically its pattern or melody—every single time. I’ve tried 30 times, but I just can't seem to get it right. Does anyone know what I should do?

[frozen dagger]

In editor if you are using this to do your changes, (I think) because it says "replace" that it works in the way "Replace" does in other areas of Suno. (I could be wrong, I don't see it being made clear anywhere, it confuses other in Studio as well.) In Suno Replace usually means it's not looking at that selection we make to make new output. It looks at the audio surrounding the selection, and at our prompt boxes to make the output.

[uneven vessel]

@frozen dagger thanks

[frozen dagger]

Here learning too. Call me Reading Rainbow, "Don't take my word for it.", but that is how I understand it.
When you want to keep the selection the same, that is "Cover". Which Editor doesn't have, nor a way to bring in other audio, nor can I adjust the context window of what area of the song I want it to look at when generating.... So, the only thing I can think to try (but would be an experiment really) would be to split the track selection so it is it's own selectable section, and then duplicate it around the area I wanted to replace so it maybe looks at that when making?... and then delete the copies after leaving only the new gen edit.

[hollow cloak]

Is anyone having songs they're working on disappear from workspaces for a while?

[frozen dagger]

yes. first sometimes only 1 gen shows up out of the 2. (Working in Studio tab right now, but I think that's happened in Create tab too). I'm listening to a song that was made, gen some more, only 1 shows up, i click the reload button (the studio workspace reload, not webpage refresh) and the gen that didn't show is there, but other tracks seem to moved off the list. Can't see the song I am playing (can't the waveform animation on any of my tracks.) lost, like where's my track at? then i click pause and play on the studio player and the track appears at the top of the list. Kinda hard to track what's happening.

[hollow cloak]

Yeah for sure

[frozen dagger]

new workspace, click create, 1 song shows. i reload with studio button, both songs. click create twice, 2 songs show... reload button, the first 2 dissapear, shows the 4 from the last 2 create clicks. 🤔 ok. listening to my track, click like. poof, the first 2 gens that dissappeared suddenly appear like a wild pokemon.

[hollow cloak]

Yes, that is the exact same behavior I'm having, and they wont reappear (anywhere, even in the main library) until they're done AND I refresh

I'm on firefox, how about you?

<@&1082792303976001547> spam bot

[hollow cloak]

<@&1082792303976001547> spam bot again😅

[frozen dagger]

I use Chrome mainly. I just made a video and posted it to the bug thread to show what's happening. Was using Opera when I made the video (just prefered for when screen recording).

[hollow cloak]

Is there a way t o change where the end of the fade is? I can move the start of the fade to left but I can't figure out how to move the end of the fade to the left so I have this overly long long, overly slow fade out.

[frozen dagger]

in editor? click fade at the end of the song. then click and drag the white bar to where the fade starts.

[hollow cloak]

The right edge, the end of it, I cannot move. Any time I click it, it just collapses the fade totally.

[frozen dagger]

the white bar on left side. left of word Fade, drag that

[lavish glacier]

i have a song that i like alot but i now want to add my own voice to it, what do i do? i have v5.5

[frozen dagger]

if you click the fade button after you dragged it, it resets. just drag the white bar.

[frozen dagger]

The basic is to create a "Voice" with you voice. (Record live or upload audio), and then Cover the song with your Voice.
If it doesn't feel like it's swapping may need to adjust some sliders or split the vocal and instrumental stem, so you can cover the instrumental without vocacls to add your own.

[hollow cloak]

Yeah, I can't move that right edge at all. Every time I click it, it just collapses totally.

[frozen dagger]

correct. it only allows fade in and fade out from start and end of the song. that shows the fade out at the end of the song.
Can't use that to create a fade out and back in. Only fade in from start of song, and fade out til the end.

[hollow cloak]

Alright, well, easy enough to do in real editing software. Thanks!

[frozen dagger]

i think maybe you're* wanting to control how it fades. that curve is static so we can't change the volume fade curve and it sounds off when the fade is set far when we are wanting a longer, slower fade out

[hollow cloak]

I just want it to fade out over the extra chorus it generated at the end of the song. Like I said, east enough to do in something like audacity (does seem like a bit of an oversight in Suno, but no big deal)

Ok, this is actually an accessibility issue.

When zoomed in (I have bad eyes) the layout can't rearrange itself to compensate, so it ends up drawing things like the fade to the right of the end of the song, even though it lines up when unzoomed.

[wraith grove]

Is there a way to make sure it renders a change with editor?? Like replacing sections, It just keeps saying loading for me.

[sweet ivy]

@quiet solar. Not sure if you are on now, still working on that duet. I though I had saved the song you posted as an example. but couldn't find it. It wasa duet that worked out pretty well

[quiet solar]

One sec, I'll hook it out for you

[sweet ivy]

Have a question is anyone else can assist. Trying to direct Suno to have a female singer sing the 1st verse. I this the proper format [Verse 1: Female soprano lead, tenderly] All of it placed with the tag on the lyric sheet above the ist line of lyric

[quiet solar]

I think this is the duet song you were referring to, which had examples of the style prompt and tagging used: https://suno.com/song/180fae77-cd81-48a2-8e92-e3748e463417

@sweet ivy To follow up on your other question, there isn't a single format that does/doesn't work. The way you've described should work absolutely fine, though.

[sweet ivy]

@quiet solar yes I remember the title Grounded.. what did I do worng when I reiws to save it ?

[gritty acorn]

So, suno thought it would be a great idea to add clapping during my final chorus.. is there anyway to remove the clapping, without studio? I just have pro

[quiet solar]

Easiest way to save it within Suno so you can look it up again in future is to add it to a playlist.

[sweet ivy]

I just did. I have been used tags I found in the Suno meta-tag directory I think you sent me Ex [Verse 1: female lead soprano, tenderly] but it didn't work, the male started ?

[quiet solar]

Not in an editing-style way, no.

Your best bet would be to extend the song from a little before the final chorus and give Suno a chance to do it again but without the clapping. Just try to give Suno at least 10-15s of audio to work with before the song ends - preferably a little bit more - since it can be a bit funny about creating very, very short extensions.

[gritty acorn]

I'm very new, if you could help me via dm, I would be gratful, or I could just screenshare

[quiet solar]

Yeah, it's not an exact science - meta tags are suggestions that Suno may or may not follow, rather than instructions that it will always obey. Anecdotally, I've found starting with a female singer can be a little more tricky at times.

If you give it a few attempts, it should get there (assuming it's regularly giving you a male and female singer in the first place). You can also try specifying in the style prompt that the female singer is the lead, if you haven't already.

[quiet solar]

Are you using desktop or mobile?

[gritty acorn]

desktop

[sweet ivy]

@quiet solar here's another prompt on the lyric sheet [Chorus 1 I duet | female lead + male lead, in unison]

[quiet solar]

Phew. Because Extend isn't available on mobile. xD

• Go to your song in the library, choose Remix/Edit -> Extend

• This will open the Create screen with your song pre-loaded, and a Keep/Extend bar you can stretch around. Anything in the pink 'keep' section is left alone - you basically start the song again from the end of that point. You can also put in the exact time in the Extend From box rather than use the slider.

• Ideally, you want to pick a point at the start/end of a section (e.g. Verse or Chorus) where the singer isn't actively singing, that's before the bit you want to change, and at least 10s-15s before the song would have ended.

• In Lyrics, put whatever is left of the song into the box. As an example, if you were just recreating the final chorus and put your Extend point just before it starts, all that should be in the lyrics box is your final chorus.

• Your previous prompt should already be in the style box; leave that unchanged unless you run into difficulty getting Suno to do exactly what you want.

• If you have Pro, then under more options you might want to try turning weirdness down to 15% (v4.5 / v5) or 30% (v5.5) and turn audio influence up to around 60-70% (v4.5 / v5) or 50% (v5.5) for a better chance of the new audio fitting smoothly with the existing audio.

When you have an Extension that you like in the Library, you can click on Get Full Song to combine it with the original into a single track.

[polar spade]

How to clip back the end and quick fade, Maybe it was mentioned. but a video of it nonetheless.

[gritty acorn]

sounds like something I'd mess up, but I'll give it a go when I'm off of overwatch

[quiet solar]

Can't really mess anything up in Suno - worst case scenario is it doesn't work, and you try again. The original song is never affected in any way. 🙂

[quiet solar]

Looks fine - though you could simplify it by saying [Chorus 1 - gang vocals] or [Chorus 1 - male and female - gang vocals], if you're looking for them to sing the same notes. Less is usually better in meta-tags.

The exact way you frame that in the tag, whether you split it up using pipes or dashes or commas, doesn't make any difference in my experience.

You've 100% got the right idea for how to attempt this. The best thing you can do is try a few times and get a feel for what works, then potentially tweak bits that consistently don't work. Duets are tricky because of the lack of precise control, so expect you might need a few attempts to get close, and might have to compromise a bit.

If you need to be able to specify which singer performs which part, that can only really be done via editing in Studio or a similar external DAW.

[vapid steppe]

I have an issue... Is there any expert to assist? After every 2+ minutes of creation of a song.. The melody and mood of the song is changing

[sweet ivy]

@quiet solar I think part of the problem is the style prompt it's a Broadway ballad, & the song has modulation between the 2nd & 3rd chorus. The singers want make a big ending & go off melody and phrasing but are nearly perfect on the rest of the song ? Not sure how to stop them and have them sing it like it's on the audio guide. Open to ideas from anyone

If I keep doing covers of a version with a off book ending, I'm sure I'll get that again

[quiet solar]

Other people have mentioned the same issue happening in v5.5 occasionally. I'm not aware of any fix as yet - it's probably something the Suno devs will need to tweak in the model itself.

[sweet ivy]

Don't forget I'm trying to cover an original. I'm not dumping the lyric inito Suno and seeing what I get

@quiet solar Maybe I should try in in 4.5. maybe my style prompt needs more direct. All I have now is "Broadway musical ballad, pit orchestra, storytelling, emotional, sophisticated instrumental arrangement"

[quiet solar]

Yeah, I remember us talking about this before. This sort-of falls into the category of where for sanity's sake you might want to compromise on 'close enough' rather than 'precise', as if you're trying to achieve something very specific that Suno finds difficult and is reliant on random generations, it might take a lot of attempts.

If you've already given an audio guide, that's literally the best steer you can give. If Suno consistently struggles with interpreting it, that's not a result of anything in your style prompt, that's a fundamental challenge.

[vapid steppe]

I’ve been facing the same issue with V5 as well. I’ve already used up more than 7k credits trying to fix it. Suno support advised me to report this issue here on Discord.

What’s funny (and frustrating) is that no one, not even Suno support, seems to know what’s going on.

[frozen dagger]

Did you try the custom model suggestion I made earlier? If it's just mood and vibe preservation, then adding the same exact song multiple times to the custom model can help keep that one songs vibe pretty well. like reinforcement with the audio. It is 100 credits to make the custom model, but that still way cheaper than 7k credits.

[gritty acorn]

Worst case scenario, could I somehow just copy the chorus from one of the other ones and slot it in at the end?

[vapid steppe]

Custom model? How?

[quiet solar]

@vapid steppe If it's a specific song you're trying to rescue, you could try cropping down to the section that's fine and then using that as a starting point to cover the song and recreate the rest of it.

Sorry, Malic - tagged the wrong person xD

[gritty acorn]

I need dummy words. I'm still very new to editing songs w pro.. sorry lmao

[quiet solar]

Yeah, sorry - I tagged you when I meant to tag firoz. My bad. xD

[gritty acorn]

All good aha

[vapid steppe]

Thank you 100 credits no issues..... I'll try it....

[quiet solar]

You 100% could if you have a DAW or audio editor. I've done that before using Audacity.

If you're open to editing, another option is to split out the stems and see if the clapping is in a specific stem that you could just delete.

[frozen dagger]

If you're using a custom model that has all the same song added to it, that's a really strong influence already, and you can use that custom model AND the Cover feature to really try and stick to the song. (I never saw the original you were covering. I listened to the song you shared as the example, but there was nothing to compare it to.)

[gritty acorn]

Can I do the stem part with pro?

[vapid steppe]

I'll give it a try and let you know

[sweet ivy]

@kanthriic it's insane, how does a 3:28 audio guide end up with a 3:57 somg song ?

[quiet solar]

Yes. Get Stems is a Pro feature that you can use under '...' from the Library. If you choose 'all detected stems', Suno basically tries to break the song apart into separate MP3 or WAV files for pieces, e.g. vocals, woodwind, synths, guitars. With any luck, clapping will turn up as FX/Percussion separately from the drum line.

You can then put those stems together in an editor, (hopefully) remove the clapping, then save the whole thing as a new MP3/WAV with everything except the clapping. If that fails, you could try copy pasting a previous chorus.

[quiet solar]

Because Suno. It's designed as a songmaking tool, rather than an audio cloning tool.

[gritty acorn]

oh god. I really wish suno had a section for vc, so I could just screenshare.. but i'll try

Also, what editor

[vapid steppe]

If it goes well, I’ll nominate you for a Suno Expert award.. 👍

[frozen dagger]

I'm not in it for the prizes and glory, just glad to help and learn.

[quiet solar]

If you have Premier, you can use Suno Studio (it's designed for exactly this type of work).

If you don't, you'll need a separate DAW or audio editor. An easy choice there if you're new to all this is Audacity, which is free and is pretty straightforward to use.

[sweet ivy]

@ Have not had this much trouble on any previous songs, melody was followed perfectly. Tried 4 different audio guides, nothing seems to be working. I have a copy that was close, until the last few lines. Might just have to work with that. Wonder if I could try melodyne and see if I can get the proper melody

[quiet solar]

@sweet ivy Yeah, sometimes Suno just struggles with a particular song - especially if it's doing something a little unusual. If the goal is to stick precisely to the original that does make life difficult.

[gritty acorn]

How do I upload the song back to suno after I download it and edit it

[quiet solar]

Go into Create, click +Audio and select Upload

[vapid steppe]

Need 24 songs to create a model?

[frozen dagger]

6 minimum. suggesting that you add 24 or more. but if you want it to stick to vibe of one song you want to add the same song at least 6 times.
Alternatively if you have other songs you want to add... that is what custom model is for. Creating your own vibes. Put all one style, or mix em up.
We are adding the same song (not the usual way to use it) to take advantage of the extra audio influence this provides.

This same trick was suggested for users who were trying to create a Voice that sounded as clsoe to their own voice as possible. People were not satisfied with the comparison. This is a trick that was used to reinforce the audio and get closer to the source. We are using the same trick to tell Suno, "THIS is the musical style"

[vapid steppe]

I need this song to be fixed so i will add the same songs for 6 time

[frozen dagger]

I don't know if you are familiar with the "Inspo" (Inspiration) feature, but the custom model has similarities to that feature, but much larger scale.

We used this same trick with Mashup and Inspo (adding the exact same song) with those to try and get outputs that didn't veer off style.

[vapid steppe]

Hmm if i need to add my friends voice in it... Do i have to do a live recording of him

[frozen dagger]

I'd have to double check what Suno says in their terms (as I only use my own), but yes, he would have to give the Voice Verification reading the test phrase aloud. (You would need his permission of course as well. That's assuming Suno allows that.)

[vapid steppe]

I've almost uploaded 2 :36 minutes of songs 6 times..now let me try the whole song with that model... Pray for me😂

[gritty acorn]

So idk how the saving works. I figured out audacity, so ty. But when I export to my pc, I can't find the file

[frozen dagger]

exported out of audacity? Check File > Recent Files? should show the path there

[pure panther]

Export again and before clicking export, look at the file path

[gritty acorn]

I figured it out

Is there a way to remove the styles when looking at the song, I think it's ugly

[pure panther]

Change the style in the song details, then open it in Studio, then Export as full song.

The new song will have what you put in the style summary

This spot

[gritty acorn]

it's blank if you meant here

Oh, just noticed you said studio. I don't have it

I only have pro\

[pure panther]

Ahh... I don't know how to do it without Studio

I'm not sure the same steps work with the pro editor

Might be worth trying though

But yeah, just add whatever you do want to show up in that blank box. If you leave it blank it won't change anything, it will still show the old Ai generated summary or original prompt

[frozen dagger]

it works differently. i think u discovered a loophole with that one @pure panther . doing the same method in editor will still have the descript from before just adding the style summary (no need to go to editor and save, just edit the summary and click save in the song details) will show the style summary there over the descript. with the option to "show full descript".
(similar to "show summary and show more" that it does for the full descript and a shorter version

[pure panther]

Ahh, that's unfortunate. It should be a feature.

[frozen dagger]

ya i mean just changing adding the summary tho does show the summary as default. a user has to click "show full" to see the "ugly" description. its something, i guess.

[gritty acorn]

wait, so what do I do in the editor

[pure panther]

Because you don't have Studio, just put what you want in that blank style summary box and that's what you'll see unless you click "see more"

[sour dock]

Can someone explain how the My Taste profile updates from likes, dislikes, and gens. I have genned and liked dozens of times now since the release of the feature and it hasn't updated yet. Also, does it still update if I customize the my taste profile?

[frozen dagger]

Not tested it, but it reads like we can edit ourselves. So if it hasn't updatd automatically for you yet, if you wanted, you could update it and test it. I think the my taste feature only applies to the Blue Magic Wand in the style prompt box. It should amend the prompt in there to integrate/fuse your "my taste" style with what is in the prompt box when you click the wand.
(Haven't seen how often, or when it updates, or how that works.)

[polar spade]

The Tastes are not based on your likes, its based on your generations you create.

The more you create , the more it will hone in on your tastes and NOT focus on those things you never do or have not done in awhile.

[frozen dagger]

it does say more you listen and like. 🤔

[polar spade]

I may be corrected to add in Liking and disliking then. But if you dont create anything and just go through and like more of your songs, I doubt the Taste analysis will change much for quite a while.

[sour dock]

Capture your taste profile on Suno. The more you create, listen, like and dislike, the more Suno learns to speak your musical language.

[polar spade]

The other definition in announcements: " . . .My Taste learns the genres, moods, and references you keep coming back to, then uses that to guide your generations when you hit the Magic Wand in the Styles field.

The more you create, the more Suno starts to speak your musical language! . . ."

[sour dock]

I have done a lot of this and the taste profile has not changed

[frozen dagger]

I'm not trying to correct you, just interested in how it actually works.

[sour dock]

Also the video linked doesnt mention likes and dislikes. Only the suno website itself.

i just want my thing to update already lol

[polar spade]

Heres my test this afternoon: I blanked it out and saved. Then turned it off and Saved. Then turned it on and Save. Then ran some random songs a bit out of my wheelhouse. No likes, No dislikes, No listens, other than my own. Immediately the Taste box was slightly different than what it was before.

By the way, The analysis is called the Generation... so the button that says revert to Generation means it will paste in the most current analysis, if you had perchance typed in your own styles as some point.

If likes , dislikes and listens have an influence, then I havent proven they dont, Im just just finding out that creations seem to control a big part of it. I can open a Pro account as well and do nothing but like a bunch of other peoples songs to see if the Taste box posts any analysis, then dislike the exact same songs to see if it negates itself or only stays stuck on what its already given an analysis of styles for. Id be curious myself then, Maybe i can try that tomorrow.

[sour dock]

What I did is I left the default description there and added to it at. Then I have been creating, listening, liking, and disliking, even posting 3 songs yesterday too. Yet the taste profile has not changed.

[pure panther]

Oh no... that's not going to work for me then 🤣 I make everything from industrial metal to country to Pop to things that have no good classifications.

[polar spade]

Now THAT is one thing I have asked and also not gotten an answer yet. basically how much "stuff" is needed to nudge the needle a bit and have it either re-write the analysis or add to it.

[polar spade]

yea. For testers as tinkers as well... I dont care of much of what I'm testing. lol It would make more sense to add Playlists by having a tick box by each playlist > Analyze for My Tastes { X }

[pure panther]

It would be cool if it was per workspace.

That way you could categorize stuff