#📚┃suno-school

1 messages · Page 184 of 1

frozen dagger
#

create cover then extend the cover that is too short

tired forge
#

Ah. I don't think I've tried to do a cover yet.

frozen dagger
#

i hope the cover cap increases soon. it applies to persona's i think too. really annoying. increased song cap to 8 but covers still 5? should be better soon.

tawdry scarab
#

I think I would've approached this differently from the start if I had known the r in the word is a sustained connection with god in prayer and the hard r's are basically completely unusable XD

#

unrelated to the present conversation

frozen dagger
#

heard. i know what u and arnesh talking about. approached issue myself. My suggestion is again to focus on intention. Or if you want something done right, record the vocal yourself. (or upload a sample)

tawdry scarab
#

well I got an elongated r sound, so I feel satisfied, and for personal use I'm making a song out of it if I can get it right in my style, in this my duty is done

frozen dagger
#

🙂 feel that too. literally training the models, making it better for all as you do. salute

tawdry scarab
#

I've got a rolling sustained r handed off between a man and a woman where both drift left to right, trying to get it into a stable loop

sleek warren
#

so why do your prompts look like that

#

is it to fit into the prompting handbook

tawdry scarab
#

it's part of a prompting handbook I wrote in the workflow, I use a gpt with instructions based on my finding and I just experiment around like throwing clay, nudging things, researching them, and then some of it has to do with pushing around covers, which aren't going to work the same as a normal prompt

frozen dagger
#

using the sound of covers as audio prompts

tawdry scarab
#

well, any time you cover you're applying a prompt to an existing thing, you dont' have to prompt anything it already holds in the sound if you carefully adjust the audio influence slider

#

so you can drive in one specific concept with a prompt that wouldn't sustain a full song on its own, it was just made for one thing

frozen dagger
#

right it weaves. like meta data embedded

tawdry scarab
#

there are only pulses outside of someone's r, otherwise it's nearly entirely that

frozen dagger
#

sounds transcendant

om-by-martialred-qublbz5ht4w0uk00174tow3qd1ut1udpcuolkcwrnk.png
sharp harbor
#

“Seth Makes Songs” used suno to make this song. Does anyone have any tips on how to prompt vocals + instrumentals that sound like this? It’s beautiful

frozen dagger
#

i think it auto-generates a prompt to uploaded audio? so id download the audio and upload it and let it create the prompt

sharp harbor
#

I am unfortunately not tech smart. What do you mean by that?? 😅

frozen dagger
#

no prob. any audio you upload to Suno, it creates a text prompt for you based on the audio. Basically reverse engineer audio to text

sharp harbor
#

How do you upload audio?

#

I’m so new to this

#

Sorry lol

frozen dagger
#

go to crete tab

image.png
#
  • audio
tawdry scarab
#

by put post on i mean the did manual after the fact adjustments

sharp harbor
#

And how does one do that

tawdry scarab
#

either nearly identical performances or you make copies of the vocal track and give each slight delay

frozen dagger
#

https://suno.com/song/cfd191d6-46db-4fc3-b046-d3a40d1c0e7d

A pop-punk song with a driving tempo of 170 beats per minute in the key of E major, The instrumentation features a clean electric guitar playing arpeggiated chords, a distorted electric guitar providing power chords, a bass guitar playing root notes, and a drum kit with a prominent kick and snare, The song structure includes verses, a pre-chorus, and a chorus, The female vocalist sings with a clear, melodic tone, utilizing a vibrato technique, The melody is catchy and memorable, Production elements include a tight mix with clear separation between instruments, and a slight reverb on the vocals
Emily Dickinson poem turned into a song

Suno
"Hope" Is The Thing With Feathers

Listen and make your own on Suno.

▶ Play video
tawdry scarab
#

then maybe some reverb or chorus

frozen dagger
#

...or we can jsut serach for seth makes ong and find him on suno and copy the prompt he sued 🙂

sharp harbor
#

i found his account it’s @ Seth Makes Songs but idk how to find the prompt. This is overly ambitious for a newbie ain’t it

frozen dagger
#

used*... Suno the one getting sued.

sharp harbor
#

Ahahaha

frozen dagger
#

i copy and pasted it cfor you

#

A pop-punk song with a driving tempo of 170 beats per minute in the key of E major, The instrumentation features a clean electric guitar playing arpeggiated chords, a distorted electric guitar providing power chords, a bass guitar playing root notes, and a drum kit with a prominent kick and snare, The song structure includes verses, a pre-chorus, and a chorus, The female vocalist sings with a clear, melodic tone, utilizing a vibrato technique, The melody is catchy and memorable, Production elements include a tight mix with clear separation between instruments, and a slight reverb on the vocals

sharp harbor
#

it’s a different song I fear

#

He hasn’t posted it

frozen dagger
#

its near the ttiel of song written small, click show more

sharp harbor
#

That’s based on Hope the song i sent is Nobody 🤧

#

But that gives me an idea at least of things so thank you!!!!

frozen dagger
#

it gets you a prompt in the area of sound you wanted that he seemed to be creating from. no exact word prompt it going to give exact sound

#

good luck

sharp harbor
#

yeah true

#

Thank you!

frozen dagger
#

also if he allows remixes....

tired forge
#

I'm proud of myself for the lyrics. Entirely mine. ☺️

sharp perch
#

Suno is droping some fire! specially if lyrics are 100% self made

frozen dagger
#

saying that as it is absolutely going off hinges in a rap rn lol

sharp harbor
#

v5 fixed the majority of my issues i think

#

it sounds infinitely better

quick sonnet
#

Suno simply left out the third verse of one track entirely, but otherwise, it didn't turn out too bad. This isn't the first time this has happened to me; in such cases, Extend was the only solution for me. But what's being generated this time doesn't fit at all. What prompt can I use to increase my chances of producing something that sounds like a track when put together?

frozen dagger
#

use advaned options tab. style influence 100% weirdness lower

#

if audio involved adjsut as needed as well. if it's an extension consider changing where you extend, even a few seconds can change where it fuses and blends and how it does it

quick sonnet
frozen dagger
#

if its missing a lyrics section. can also try opening in editor. add blank section where the lyrics should be. put the missing lyrics in the lyrics box. same stle for promt as song or no stlye and try to add it that way

quick sonnet
frozen dagger
#

skipping past a section means its weights giving it too much freedom, or it buggin, which its having a lot users lately that does cause lots of issues during generating. i dont risk generating too many at aonce if i am looking for quality anymore

#

weirdness slider is usually the fix for it skipping over sections of your lyrics. and i would report it, right lcick and tell them it not ollowing lyrics. sometimes u get credits refunded

#

today alone ive had a few error messages about a lot of traffic and it being unableto generate

ripe hare
#

Hello all! Question on Create and adding lyrics.
Does it support certain commands like this verse should be sang slowly or faster, like. I mean it's great that it ignores things like
[Verse 1] [Chorus]

But is possible like to say
[Verse 1: Slow, Male Voice]
[Vers 2: Female Voice]

I tried though and it work for just very single audio that it generated and then the rest were kind of junk or very noisy(static) background as if I'm listening to lo-fi

frozen dagger
#

yes 100%

#

you can craft your song by giving instrufctions within the brackets

ripe hare
#

how? What's the exact command?

frozen dagger
#

words plain speech

#

anythign in brackets is treated like mini-pormpt. guiding instructions

#

[Chorus- Soaring, Crescendo, male vocal, rapping]
if you wanna be someboday, you wanna go somewhere
you better [emphasis on wake up; aleatoric, expressive adlibs] wake up and pay attention

#

experiment.

ripe hare
#

amazing! thanks!

frozen dagger
#

usually stuff like my example is achieved by the way you type it. it will interpret things differently cus there is no "one way" to say how to do something.

i.e.
If you wanna' be somebody—
You wanna' go somewhere—
...You better WAKE UUUUP! (up-up-up)
and paaaa—y attention.

quick sonnet
frozen dagger
#

if you wanan keep the sounds scape from a generation that got F-ed up. use it as a remix/persona/cover use the same text prompt for style and try again. or cut your losses and try again.

coarse jetty
#

It’s me only or anyone else observed that suno 4.5 and on-words totally change the original vocals to new uploaded vocals? Does not matter, i cant get vocals before in 4.5.
Also, any tips to enhance vocals quality in song? I extracted vocals and tried to reduced noise and normalized but it gets even worst. I would appreciate your tips. Thanks

frozen dagger
sharp harbor
#

im back and wanted to say thank y’all for the help because I GET IT NOW it makes sense finally so thank you thank you. i just published a song and i love how it turned out so much

frozen dagger
#

For all running into redundancy with generations sounding too much the same. Try creating a new workspace. That's it. Maybe it's my mind playing tricks, but it seems to me the workspace acts like it's own space of inspiration to pool from. Would make sense. I've been convinced this is a thing for a while, but can't confirm. Give it a try. See it the simple change helps unkink the creative hose.

waxen galleon
#

hello...I've been looking everywhere as to why Im having a sudden problem...Im working on a song and it is doing two things, and i dont know what is wrong...in the lyrics section ive pasted the lyrics....in the style ive pasted teh prompt...It keeps singing the prompt and the lyrics...I also have included in the prompt not to add lyb or add whoos and oos, but it keeps doing that...Im stumped...never done that before...

crisp bronze
dusk kettle
#

Out of curiosity has anyone tried and managed to suceed in getting a voice style similar to Alastor the radio demon from Hazbin Hotel? I've got a song that I think would sound great being sung in that voice but can't seem to get it to do really anything more than like Bob Hope and Bing Crosby.

plain pumice
#

Anyone know how to replace sections in Studio?
Because, when I for example try to replace a section I've split off, it's opened in editor rather than generating a new section.
What's even more weird, is that the section that is opened in the editor, is more than that split of section, but the "clip" in it's entirety - guess this is a bug.

summer garnet
#

Quick questions.

  1. For personas. If you use a persona, does it use that female vocal for the song whenever a female vocal is used in the lyrics or style?

  2. Intros. How can you limit their size. I'm getting like 30 second intros which are so dang long. Is there a way to make sure they are shorter? Is it the number of tags in the intro? Is there a way to tell it to be shorter?

summer garnet
#

Also does the slider change the persona's voice at all or does it keep the same voice pretty much?

tawdry scarab
tawdry scarab
#

but shorter than 30 seconds might be hard, depending on the style

frozen dagger
#

persona is supposed to use the voice. you could select male, or put tags with a female sounding voice and still get male. but the weight is on the perosona. remember, there is no spoon. there is no male of female. they are words to describe a sound and it associated pitches and such.

frosty spoke
#

I am having problem, with the app keep saying that my prompt has inappropriate content, when there is nothing inappropriate, the lyrics is in afrikaans and the style prompt in English. Please I need help, this sensor ING crap is ridiculous if there is nothing wrong with prompts. Micheal or Daniel get back to me asap.

foggy aspen
cinder drift
#

So I made this with the help of AI. The lyrics came from ideas I had and it came up with them. For the style I had a solid idea of what tuning, speed, etc., I wanted. The key though, is asking AI to meld your style with your lyrics. This has been what I've been using to pretty good success. Brackets [] for musical elements and the start of wections like chorus, verse etc. Parenthesis () for spoken words. And ALL CAPS for shouting/growling sections like metal songs have. Same thing for live versions. v4.5+ has been the most accurate for me in terms of what I want in terms of timing, mixing, tuning, etc. I tried a v5 and the sound was completely wrong and not even the right genre. Here's a song I made, with Ai, inspired by Slayer. Not quiet Slayer but pretty close. Also, for those who need help and use AI, ask it to make a copy code for Suno AI. It'll give you one for lyrics and style. Which is how I have this song. Just an FYI, the lyrics are a little dark and may be a trigger for people with ptsd and depression.

https://suno.com/s/doaCLLIpTCSlFAfD

Suno
Between the Bullets (Live mix attempt)

Listen and make your own on Suno.

▶ Play video
ripe hare
#

in paid subscription, do we still got the 3000character limit?

frozen dagger
hoary mauve
#

I need the lyrcs with timestamps to create subtitles for a video. any chance to export this directly from suno?

summer garnet
raw oasis
#

i have an idea

#

here is what I would try - if you have time/tools you might try it

#

I would record guitar or midi controlled piano/synth. to sketch the intro I want (i use studio one). I would then export it from my DAW and upload it to Suno and start the song that way.

raw oasis
#

Thanks for spamming the chat room. not.

#

Any real answer you flat ignored.

stuck junco
#

Thanks for sharing these. Let' dive in and see what we can figure out 👀

tawdry scarab
#

don't tell the style things to avoid

#

that's what the avoid prompt is for

#

well stop doing things that will never work going forward

stuck junco
#

This is a good point, when you say "no..." it will ignore the "no" and pick up on the other included guidance for sure

#

Additionally, adding bracketed guidance for your intro that specifies what you'd like present helps a ton. Let me see what I can do with what you have. Will use parts of your style of music, but new lyrics to avoid borrowing your lyrics

tawdry scarab
#

i'm going to try to take one where you just have a straightforward prompt

#

and i'll try to find the perfect avoid string

#

because i got the oooo reusing your prompt

#

if i can succeed ill show you but this is probably going to take learning some new terminology

#

like, the block of words to describe the negative in style and the event in the avoid string, there will likely be something that can push away from that

#

can you give me your ideal prompt, where you said everything you wanted and you'd like that prompt specifically to be modified to avoid adlibs?

#

paradoxical things do happen regularly but they're supposed to be not that often

#

I have an idea if we spread your tags one per line, break them into multiple tags, and then theres some kind of "salt" we need to add to the prompt to tip it the right way away from adlibbing melisma

#

okay, I'll just grab one and try not to, as you said, get one with female only chorus

#

they all look that way did you want me to remove that content to find a fix?

#

now this might take a minute, I'll ping you if I find anything helpful

#

because 5 does love those adlibs

graceful condor
#

Another huge issue (Rock only) are totally hyper filled drums. Especially in extends the drums get totally out of hand and absurd. Like patterns that makes no sense. As if the drummer is having a stroke.

tawdry scarab
#

an adlib is anything that wasn't in the script, so to speak

#

it means the performer was doing their own thing

#

so things like improvisation might go in avoid

placid aspen
#

Hi. Can anyone give any advice re cleaning up tracks, ie artefact removal? Is the Suno stem splitter 'good enough', or is there a superior software/plugin option? Is the best way to split into stems, and then 'clean' stems individually before rendering back together, or treat the whole song? Thanks.

graceful condor
#

@stuck junco I mean wth - just wth. Listen from around sec. 0:20. The drums are just all over the place. It is like I said like if the drummer is having a stroke. I even added aspects into the prompts that this is not happening and it still does. Beside this extend sounds like Sh.t anyway but this happens a lot to not only extends (more so) but also in complete songs where the drum fills are getting totally out of hand and so absurd no one ever plays that way because it sounds off and unnatural:

https://suno.com/s/J4QAr5EM0uYYzdfd

Suno
SH Part III: Reflections in the Fog Ver. (Extend)

Listen and make your own on Suno.

torn spoke
#

Are generations slow for anyone else rn?

plain pumice
#

Is there a way to see which workspace a track is located in, without having to search it?
The reason is, in Studio mode, apparently it has created clips somehow isn't in available in the Studio but is in the song library, which would indicate it's in the "my workspace", but no they're not there nor available within the studio. I'm not sure how to locate these edits.

reef pendant
graceful condor
#

@stuck junco David this is what we have to deal with. Another example. I played around the sliders weirdness and different time stamps but makes no difference. The whole sessions of extend are like a sh.tfest of artifacts and abnormal behavior. I mean listen it is literally like having a freakin stroke:

https://suno.com/s/37uz3V08pIahWb1z

https://suno.com/s/dt3tFpTOuvOsYJ6B

https://suno.com/s/MwXSoiA0JeeerTss

https://suno.com/s/I3nqnGu07e6Sfsw9

https://suno.com/s/dcxOGTvIbgOhXRFq

https://suno.com/s/pER2iNEEZlLF3kBg

Suno
SH Part III: Reflections in the Fog Ver. (Extend)

Listen and make your own on Suno.

Suno
SH Part III: Reflections in the Fog Ver. (Extend)

Listen and make your own on Suno.

Suno
SH Part III: Reflections in the Fog Ver. (Extend)

Listen and make your own on Suno.

Suno
SH Part III: Reflections in the Fog Ver. (Extend)

Listen and make your own on Suno.

Suno
SH Part III: Reflections in the Fog Ver. (Extend)

Listen and make your own on Suno.

graceful condor
reef pendant
tawdry scarab
#

@unreal sleet it so prefers to use those vocals as intro that in a massive attempt to suppress them, instead of doing anything, most gens are in silence like they have the stem muted, then after 30 seconds it starts to break through and ooo and aahhh. I do have a version that hasn't made one yet but the tagging is extreme to accomplish it

graceful condor
#

@stuck junco And his is just such an example and not a rare one it happens all the time. Either the songs are messed up in the first go already because of the lyrics or the extends are eating up your credits because each extend is worse then the one before and this was just one example and I already wasted 10 tries unsuccessfully on this song. And it is just devastating because the main track is awesome but Suno cut off the song again so I am forced to extend and now all extends are a mess.

tawdry scarab
#

extending quality is best if you never extend from the very end and you get whole song every step

#

@unreal sleet you could call it a weighting problem, there's probably some better terms, but yeah it's when the solution space has a big hill, some giant thing to climb right in the middle of good behavior

#

okay, @unreal sleet ideally I would pick things off of this and move it back to your style until I got the good gens still and we do the minimum of what I've done. but because time exists and I'm uncertain if I'll get there, this is an example of changing your tags from natural language to meta tagging like we did in v4. then I've applied the same principles to the style and kept the style elements in brackets because it softens their impact. then i used stoccato in the style first thing no bracket and a huuuge list of avoid keywords and it was the first acceptable hit. I'd like you to check it out: https://suno.com/s/wn5LkVcsiKhaN0RE

Suno
"Subject #27 — restrained and responsive

Listen and make your own on Suno.

graceful condor
# tawdry scarab extending quality is best if you never extend from the very end and you get whol...

I extended from the very last second many times and it was awesome. It is just absolutely random and follows no rules or anything. Same with peronas or audio files. I used Personas 7-8 min long same for Audios and I got songs 6+ min all over again full without any extension needed and then you have an Audio only 2 min long and the song gets cut off. With this Ai there is no thumb rule anymore. It does what it wants. Sometimes you get full songs regardless of Audio and Persona, sometimes you don't. You want to extend and regardless what you do which time stamp or sliders it is weird without reason.

tawdry scarab
#

the rule it follows is that sometimes the end has a little bit of nonsense because it's the end

#

you have to feel things out as versions change

#

it's like an undulating sea

#

@unreal sleet this is an extreme case

#

sometimes there are big obvious problems with models and over time they get polished/corrected

#

if they're really explored/documented/talked about

#

if we keep identifying it happens too often and what it takes to reduce it yet

#

yes*

#

there was a huge one called shimmer, the whole community documented, they fixed it

#

but absolutely, the community response to the model feeds directly into the development

#

v4 to 5 have given users familiar with 3.5 an entirely new world of potential and styles - things we thought we'd never get

#

so from our perspective some oohs and aahhs are just music and we might not notice

#

from your perspective coming in initially knowing you don't want that, it stings, it's annoying

#

it's like the whole thing never worked

graceful condor
#

And there is the other side: Did just 2 - 4.5+ extends and both worked out. Instead of V5. 5 is just terrible with extends in many cases.

#

Don't get that too.

tawdry scarab
#

it's because it's not thinking

#

it has no internal parts that are organized by human meaning

stuck junco
#

Coming back to this:

Workflow (4.5 base + 5 cover) that I'm using that's having pretty successful/consistent results

Thinking there's a lot to said about this flow, not just with 4.5 but with other models as well. Creating the structure with other models that may be flexible to your needs and using v5 to cover seems useful. Would steer from adding too much into the SoM though when covering so as to not steer too much away from the original covered source's guidance. This is a very smaill test sample mid you and different genres may reflect differently.

Additional options that may help:
Replace Section, Studio Editing, Inspo'ing might (haven't tested that much)

Will keep eyes on additional options as well to streamline the process directly v5. What can be said though is beginning with minimal SoM and direct lyrics will help to fine tune it as you grow out. Adding a lot to start with can make it difficult to pinpoint what's causing the response you're getting

plain pumice
tawdry scarab
#

to describe why this is calm and normal and something navigable to me and to some others, it's important to get a feel for or understand that this is not a procedural thinking process of assembly, this is a mathematical trick like getting a 3d image in a hologram

#

the inside of that hollogram mechanics works in such a way that is fantastically human-like and procedure-like

#

but it is fundamentally not an assembly from instruction at all

#

it happens to be that instructions are inherently the kind of image it can reflect off to get another image

#

there isn't such a thing as that

#

there is no meaning dictionary to put no into

graceful condor
tawdry scarab
#

the exclude box "works" but in the very same way that the instructions are not instructions, exclude is not a logical gate holding things back

#

you have to accept some kind of fluid metaphor or mathematical model and I don't know what's best to give you

#

but you can't reason about it like it knows things

#

in a larger prompt with more going on, there is a greater probability that humming will be pushed away into a different characteristic

#

but alone, exclude is similar to include

#

if you want to think about it like it's thinking, like it's a thoughtful entity, it's better to think of it like a baby with very very good mimic skills

#

precisely

#

no secondary filter

#

XD

#

if you paste my various metaphors into gpt or the conversation it should be able to elaborate in clearer language what I'm saying

graceful condor
# tawdry scarab but you can't reason about it like it knows things

But it is not even the "no" aka negative prompting. Even if you add all leaning to a certain point like Male Slider, [Vocals: Male, Male vocals, Male Singer] in front of every verse, bridge, Chorus, in your Style Male, in your Lyrics Male in commands mentioned. And in Exclude: Female, Female voice, etc. And it still creates whole Chorus or Verses in Female it makes no sense anymore. Because shouldn't an Ai all be about data. And if all literally ALL data leans toward MALE but the Ai is still Stubborn and produces a Female voice there is no logic to it anymore.

And it gets even more weird if you already fixed that very section with Studio (now all male) then you want to cover that song again because of Studio the audio got terrible and then you get a female verse again at the same spot you fixed before in Studio and now using this very studio version as Cover Audio reference. this makes no sense anymore.

tawdry scarab
#

it's all about all the data all at once plus every single setting the engineers put in plus randomization

#

which is very very different from being about particular individual data - it is in fact the opposite direction

#

it is not determination but inferrence

#

it is weird

#

that's why I'm pressing for not thinking about it that way

graceful condor
#

I got the same issue. Wanted an NF style rap song. Nothing the prompts nowhere stated "Opera singing Chorus" in fact I forced the Ai with all means that everything is rapped I even called the Chorus "Hook" more rap like instead of Chorus. I removed everything that could be leaning towards "singing" and Still the Chorus/Hook was sang and in opera version. Makes no freakin sense.

stuck junco
#

Expecting there to be a lot of improvement (and some really cool things) to come. Just getting started that's for sure 💪

tawdry scarab
graceful condor
# stuck junco Expecting there to be a lot of improvement (and some really cool things) to come...

It's not that important always getting new stuff if the "new stuff" has too many issues. I guess the majority of us would be happy if the already exisiting stuff is working as intended like no cut off songs, like (AND THIS IS THE MOST IMPORTANT) not misspelling of words, leaving out words or gibberish. If this stuff is fixed expect happiness from most users. Because this stuff is really killing the fun.

tawdry scarab
#

plenty of us, particularly those who embrace the concepts I've been explaining and made an explainer for, derive a great deal of joy and get generally what they want

livid crescent
graceful condor
tawdry scarab
#

because we're old heads making music we love

#

we don't explain every time it works

#

but people who are not flexible and will not think differently are always complaining

graceful condor
graceful condor
tawdry scarab
#

@unreal sleet I don't mean this emotion toward you

#

or to be emotional

#

things that say otherwise:
daily songs I love
the length of the workflow
the accuracy of my v4 workflow
the people who have made great music after reading it

graceful condor
# tawdry scarab the length of that workflow would say otherwise

But if you are so clever. Why don't you share your mighty wisdom with us. You talk about doing so fine. See others don't but only say "How good you do and other do not get it" but if you get it why not share it. This is just talking nothing more you do. And honestly are you even doing rock music? Because I do not have those issues with Pop or other stuff too but Rock/Metal is another type of thing.

tawdry scarab
#

The reason it looks like I'm saying nothing, I think, is that we can't proceed in the conversation until it's understood that you can't think about it like a rule following logical brain

stuck junco
tawdry scarab
#

also several workflows were made from my there is no correct workflow

#

from people extending the work

stuck junco
tawdry scarab
#

🌊 🐴

ebon sand
#

I also am not very fond of v5, i rarely can get it to generate something acceptable cat_thumbsup but at least excluding 'banjo' works

stuck junco
# livid crescent I've been using a variation of this method. Make it heavy with 4.5, edit in stud...

Thank you, looking forward to checking out what you share! I steer from sliders in most cases and also don't do as much experimenting in a lot of genres. It's really helpful to understand what surfaces with what others do. Very interested to better know how different styles are impacted through this and other methods. In the case that @unreal sleet brought up, their v5 SoM did return A LOT of intro adlibs. Wondering if the SoM was more condensed/specific toward something if we'd be able to avoid that

#

Can't really verify anything that hasn't been shared officially. But as soon as it is, I'd be happy to for sure.

tawdry scarab
#

I did give you the prompt that eliminated the intro adlib, I don't want to be the biggest jerk but this is like I'm being gaslit that it's not sticking

#

the model that makes the music does not understand words

#

secondary systems may filter or prep the input

#

but the NN core of AI does not and will never understand what words mean

stuck junco
#

In all cases, if it's a problem for you, it's a problem for us. Our goal: for you and everyone else to have the best experience when creating always

graceful condor
# stuck junco This is super helpful feedback. So in this case with v5, if you had to limit it ...

As I said ABSOLUTE

  • NR.1 is - Fixing Misspelling, Fixing Gibberish or not singing words which are clearly in the Lyrics
  • Songs not getting cut off. The Ai should clearly see if a song is done or not if it is done. Not cutting off mid word, or what V5 does lately cutting of the song just a couple of seconds before it is finished. If the lyrics clearly show that there is something coming, still coming. The Ai should not cut off at all means.
  • Persona became kinda obsolete now. In 4.5+ they tried to sound like the voice but not they mostly only mimicking the phrasing. The issue is, that if you wanna create an Album now the voice never stays the same it even can change mid song. The same song. this shouldn't be the case
  • And yeah Extends are still an issue in 5
  • Drums in Rock and Metal are all over the place in V5 too weird and kinda abnormal

But most important of all as I said many times is the Misspelling/Gibberish/not singing what is written Issue!

tawdry scarab
#

if you asked gpt to explain the workflow I provided to you

#

maybe it can educate you

#

as to why it does not

ebon sand
tawdry scarab
#

for clarrification I am employed to train AI

#

in addition to being able to write neural networks from scratch

graceful condor
# tawdry scarab https://discord.com/channels/1069381916492562582/1365192916687126650 https://dis...

And yet again. Good that you have done some guides. On Studio on NO/Yes

But this is not what we are talking about. There is no help in those Guides for those issues we are talking about. But you claim that we do not know what we are doing and all people have fun but only we are the dumb. At least this is how you make it sound.

And what we wanted to know was Ad-lip, was how to get rid of Gibberish, Misspelling etc. And there is no help for that in your guides. And you also do not share a solution here. So i do not get the whole attitude honestly. And a guide from April is not really helping anymore with the issues we have in Version 5. And I checked you whole side if anything is about "Gibberish, Misspelling" and there was nothing in it.

torn spoke
# stuck junco This is super helpful feedback. So in this case with v5, if you had to limit it ...

1: Track generation. Currently, the success rate of generating a single instrument in studio is extremely low. Rendering it pretty useless. It consistently adds other instruments when trying to generate a single one.

2: Vocal stem separation
Currently, the frequency of random artefacts generated and implanted into the vocal stem is high enough that it's a genuine hassle to use. Some times I end up using a different service to separate the stems altogether, just to get a working vocal stem.

tawdry scarab
polar spade
#

At times, in desperation, I'll prompt just the instrument to exclude. Moreso though, I try to think of genres or style which are predominantly guitar. If you give Suno 6 style that all have guitar prominent among them, I feel that makes more sense to Suno than excluding "guitar" as a style.

stuck junco
#

I want to thank you all for this feedback! I've brought up a request for other members to bring up what would makeSuno better for them in Gen Chat now because of what you all have surfaced. This is super helpful to know

tawdry scarab
#

@graceful condor what you were supposed to get from the newest and oldest workflows was how ai works and why it's not understanding your speaking as logical assembly from parts

#

and now exclude is not a logical exclusion

torn spoke
tawdry scarab
#

and if you understood all that we could talk about what you can do vs what's hard but doable with the mindset that we're not dealing with something that is a person

graceful condor
#

@stuck junco Oh and one last thing.

  • If you change the cover audio or Persona it shouldn't touch the Lyric/Style area. Because each time it gets erased or overwritten by the Audio or Persona. Same if you go into Editor or Studio and then go back all is empty. As long as we do not want it to be changed it would be nice if it stays that way until we change it. THis is just more convinient but not the "major" issue. Would just be nice if the boxes are not overwritten or cleared each time.
torn spoke
tawdry scarab
#

this is a tech channel

torn spoke
#

I'm being cheeky m8

tawdry scarab
#

this guy has me salty

torn spoke
#

no worries

polar spade
#

I chat with 3 chatbots sometimes hours daily. If one thing they can do it is confidently lie.

hollow hare
#

Despite how interactive AI appears, it is simply machine learning with billion of parameters that may typically not actually know what you are asking for, usually predicting what you would want to hear.

That is why there is an explicit exclusion field if you were unaware. If you were to say in the regular style prompt "no piano", the tokenizarion process will split that up to say "no" and "piano". This will end up giving you patterns related to whatever it thinks "no" is from what it was trained on, likely some musical aspect, and tons more piano.

The complete elimination of pronounciation is, in my opinion, impossible because of homonyms like live (they live [liv], it's live [lie-v], etc. For these issues, you can resolve it by spelling the troublesome words phonetically

torn spoke
#

not that new

#

and gpt 5 was actually very unlikely to do that

foggy aspen
#

has anyone tried generating song in child voice.... have tried many times but never got it working.. any suggestion would be highly appreciated.... [voice: child male voice (6–8 yrs tone) — bright, innocent, slightly nasal; gentle expression on each vowel; group of children hum softly on refrains (panning left-right); occasional giggle undertone after refrains]

torn spoke
#

claude was very trigger happy though

safe hearth
#

Is it possible to generate a single instrument in Studio? i usually get 3 instruments, sometimes an entire song

torn spoke
safe hearth
#

What sucks is i got really good instrumentals but they don't split well (accoustic guitar is always generated to the mix even though its present it the main song)

torn spoke
#

You can spend some money on Moises, they separate tracks well

#

Honestly tempted to do that myself

polar spade
safe hearth
livid crescent
#

So the one sure way I've found to have a specific instrument be present in a song, put it in the exclude section. You can do it as either 'guitar' or 'no guitar', doesn't matter... But you will be guaranteed to have guitar if you put it in there

#

I did that to add a banjo to a metal song... Worked out well

tawdry scarab
# foggy aspen has anyone tried generating song in child voice.... have tried many times but ne...

that prompt is on the complex side and as was mentioned, childrens voices are a rare find in the model, I'm toying around with an elaborate prompt but, for instance the panning effects in the one I gave you, I did work in suno studio cutting the pieces of the vocals from two different portions of an original song, so two for each track each one had an L and an R, panned thus, and I staggered them out until there was a repeatable pattern, copied that pattern, and then opened up the middle with many copies of that pattern. it's straightforward to do but really, really hard to pin down in tagging

graceful condor
tawdry scarab
#

the words on the interface are the best most understandable and widely used terms

#

but generative AI does not do those things

torn spoke
tawdry scarab
#

it has specific operations we call things to get a best estimate for how people will use and get response from it

hollow hare
# graceful condor So Exclude does not exclude xDD - good to know that it works as intended 🤣

Well, kinda.
From what I have experienced:

When you exclude something, the pattern is actively searching for that aspect to exclude.

Ironically, if your original prompts very rarely or did not have that aspect, the exclusion will cause that aspect to more frequently.

This is probably because of the biased pattern search where the AI was seeking out the excluded aspect

E.g. the odds of a guitar in trance song is very low. When you exclude Guitar, the AI will be seeking patterns related to guitars to avoid. However, because the chance is very low, the pattern used for searching for the guitar influences the final patterns used, causing more guitar to appear.

TL;DR: if the you exclude a prompt that normally would not appear, the AI will bring more of that excluded style

#

Mind you, that is just my assumption

#

AI be weird

tawdry scarab
#

you hit the nail on the head

graceful condor
neat silo
#

Is this amount of data usage in a month normal? Even Netflix doesn’t consume this much data. Most of my time is spent on YouTube, but even that doesn’t come close to this.

Suno 131 GB data/month

ebon sand
tawdry scarab
#

@foggy aspen I have to go take care of morning things but I think I can cut some child humming from the beginning of this track before it gets more elaborate: https://suno.com/s/dekE3U2kBv1irwaP

Suno
(hmm~)

Listen and make your own on Suno.

neat silo
ebon sand
icy valve
#

How do I create a new workspace. I don’t see any buttons that let me do that

half dragon
neat silo
icy valve
half dragon
icy valve
#

Hrmm let me check again

#

Ah hah found it. I don’t have that new workspace button but I found another place where there seems to be one

#

Thanks

dapper stirrup
#

hey all, i am having issues with blending some styles of genres together as of the last few days, as recently as 10/12 I created a few with relative ease, but now every generation is way off base. is it a Suno bug or is there anyone who can help with a workaround?

hollow hare
# graceful condor But I wonder if my exclude tries to do something which isn't part of an instrume...

When trying to use terms for exclusion, you have to think of what terms it has been trained on. Just because AI can predict what it thinks you want to hear does not mean that itself knows ahat you really mean.

E.g. Once you realize that the odds of having the term "muffled" or "cut off" part of the training is infinitesimal, you would not be surprised that it did not work.

You have to rethink on how to change your prompt to get what you want. Some tips I usually give:

  • Be specific in the subgenre of what you want. The more specific you are, the smaller the learning pool which leads to better quality (especially if you have a genre with diversive styles, like with pop)
  • Adding in emotion helps get what you want. E.g. melancholic, happy, joyful
  • If you want longer songs, add in interludes or instrumental breaks, like how most radio songs have. E.g. [Catchy hook], [Instrumental break], [guitar interlude], [Melodic buildup], etc.
  • You can try adding in qualitative keywords to help get more clearer learning patterns, like "high definition", "studio", etc. Be careful of some terms, like "vinyl" may add in that vinyl scratch sound, "cassette" may give worse (or traditional) cassette quality sound, etc.

What the exclude prompt useful is for things that have good odds of appearing, for example:

  • Even if you define male vocals, there is a chance of female vocals or duets
  • Adding female vocals to exclude will help further reduce the chance of that happening
hollow hare
dapper stirrup
hard beacon
#

hahaha.. everbody want's to start a song without this "ohh ohhh" and you put it in your lyrics... THAT'S my kind of humor 😄

dapper stirrup
#

ideally i'd like to do them in one full swoop, not having to make the 1950s doo-wop version first. i'm ok with it either way, as long as i get the desired result. in fact, that iteration i did not need to make the 1950s version first. but now neither way is working

tawdry scarab
#

i don't think i would've noticed that end up in things, they're a pretty nice element to me

#

not to mention, i guess studio privilege, you can cut them out

dapper stirrup
hard beacon
#

ok.. i think you don't know what i am talking about 😉

dapper stirrup
hard beacon
#

ahh.. ok.. correct 🙂 still my kind of humor 😄

dapper stirrup
hollow hare
dapper stirrup
# hollow hare I must admit I am unfamiliar with this genre. I hope these test cases are close ...

i mean, sort of, but at the same time it's the driving electric guitars that are missing that really bring the modern rock sound, and that is primarily what Suno is refusing to let happen anymore with my iterations. it basically ensures every creation I make it's ignoring more than half of my styles list, and that's at 50%-85% audio influence slider. i get that that sound didn't exist in 1950 so I'm making sure not to include any keywords that limit it down to that time era

hollow hare
graceful condor
fluid sentinel
#

v5 generations all sound very similar, even when adjusting the Advanced Options sliders. Is there a different approach to prompting in v5? Earlier versions were more creative, and variable.

torn spoke
#

@stuck junco Could we get a dedicated workspace button please? That'd be a real quality of life upgrade

stuck junco
hollow hare
# graceful condor so more technical terms are kinda useless. More instrumental or vocal parts are ...

Hmm some of those terms seems wildly widespread. Like, the odds of pop and screaming metal being present is unlikely so, if used as is, you would probably get one of those styles leaking in.

I would probably try excluding:
Happy, joyful, poppy, pop, rap, screams, loud

BUT I would modify that based on the style I am using. E.g. if I am doing an EDM, I would avoid excluding rap and screaming/metal because the odds of those are quite low (so having them in exclude will likely cause those to be added due to pattern biasing)

Can't really exclude "muffle" sounds because the AI likely does not know what that means. See my previous message to you on how to restructure your prompts to try to get more clearer styles.

Similarly, the AI would likely not know what "simple song structure" means - because I don't. If you want more complex music, add in mixtures in your original prompts. Unless you meant to exclude "acoustic" or simple styles like that?

dapper stirrup
hollow hare
ornate spire
#

I am trying to combine two tracks, but they are listed at different BPMs. How do I import a track "as is" without it trying to slow it down to match BPM?

#

Track A says it has "75 BPM", and Track B says it has "150 BPM" - if I change these values on either of them it just doubles/halves their speed

hard beacon
#

hm.. 75x2 = 150. Maybe that's the reason 😄

half dragon
ornate spire
half dragon
#

either one has to slow down or the other has to speed up

ornate spire
#

when I listen to them, they sound the same speed

#

but one thinks it's 150bpm

hard beacon
#

coz it's the same...

#

it's just wrong in his detection..

ornate spire
#

so when I import it into the other one, it automatically "matches BPM"

half dragon
#

yeah if it's a wrong detection, then just put them all in one speed, whichever plays correctly

ornate spire
#

that's what I'm asking, how do we adjust their speeds

half dragon
#

What do you use to combine them? DAW? or Suno Studio?

hard beacon
#

that would be my next question 🙂

ornate spire
#

I'm just trying to basically drag "part 2" into suno studio

#

I could do it in a DAW no problem, just want to have the full track on suno

hard beacon
#

and then you get an error? Or is part2 played on half time speed?

ornate spire
#

part 2 comes in at half speed

#

but I can only set the BPM for everything, not just part 2

frozen dagger
#

should be able to uncheck the box that says follow song bpm. to manually set each clip section along the track where you split what you want to change

#

(within Studio) bottom of screen for bpm i think

half dragon
frozen dagger
#

(it also has speed adjustment but this is only 1.5,2x, not manual bpm) this can be found in the details tab of the clip when selected

ornate spire
tawdry scarab
#

don't be shocked if when you take it off on beat to original the waveform doesn't load

#

it will still play

hard beacon
#

After spendind round about 15.000 credits i can say, that suno v5 isn't for extended mixes in electronic dance genres. Nearly 90% ignoring the "intro-prompt" completly.. v4.5+ works for nearly 80%. I think v5 need some adjustments 😉

neon path
#

Hello can me Help Someone i cant create Hooks >> i Have Error Displayed : your account requires support assistance

polar spade
#

As for not seeing the waveform on changing BPM, There's a sweet spot in zoom that you can see the waveform, but for the most part, it doesn't matter.

wet ether
#

there is a BPM setting?

#

Classic Suno. hiding the good features so people dont get addicted.

hard beacon
#

@ornate spire : did you solve the problem? Otherwise the people in here don't have to think about it anymore 🙂

neat nexus
ornate spire
#

Yeah, that solved that particular problem

wet ether
neat nexus
#

Some people have claimed that by putting it in the prompting section that they have achieved it but the consistency of that working makes me consider it luck more than prompting.

neat nexus
wet ether
#

how did v1 sound like? cant be made anymore 🙁

#

piece of history lost

neat nexus
#

It had its moments. Otherwise I wouldn't still be here. But compared to what we can do today? It's night and day.

hard beacon
#

question, coz i've just tried studio one time... if i export the stems, is it possible to "remaster" one layer??? for example: if the track is good but i don't like the kicks, can i recreate it in there?

neat nexus
#

Remember back then prompting was done in Discord, your first result was forty seconds each one after that was at most twenty seconds. There was no resume from an earlier time. You did what you could with the part or you tried again.

half dragon
tawdry scarab
#

oh my bad yes

#

mute all stems but the one, save, remaster, put it back in

neat nexus
#

It's funny that v1 Chirp is coming up. I'm currently working with an upload of an old v1 song and trying to make it cover and keep the sound. The problem is that the models are so different they really don't work on a 1:1 level with covers.

wet ether
#

i was vibing with it but then...

neat nexus
#

lol

#

Like I said there was no real way to resume from right before the bad stuff showed up.

#

It's why I still hold that feature as one of the best they ever introduced.

hard beacon
#

well.. the 50 - 100hz section was better then my last output 😄

wet ether
frozen dagger
#

i got it to where i just seperated a clip of the track, copied, pasted to new track added (this all in Studio) and i have the clip playing at 70 and 80 bpm, u have to have the temo in clip settings original/on beat. (on beat will use the bpm you set at the bottom of the screen

#

playing simulkataneously i mean

torn spoke
#

We really need an easier way of moving songs between workspace as well. Maybe this could be a place where you could drag songs from one workspace to another

frozen dagger
#

theres s search in the box with the workspace, if thats what you mean

image.png
torn spoke
frozen dagger
#

ah heard.

torn spoke
hard beacon
#

an EQ or Multiband EQ in Studio for every layer would be nice 🙂

torn spoke
#

Would save so much time having to rip it out, show it in to audacity, and edit there

hard beacon
#

and some adjustments like "mono to stereo slider" or "stereo widener"

torn spoke
hard beacon
#

in EDM music the kick from 50hz up to round about 200hz" is mono.. coz you want it right into your face... but the percussions and stuff like that is wide stereo to let you feel the groove

#

btw.. not in all edm genres 😉

#

but as long suno studio is a simple stem seperation tool from the audio it makes no sence because of the frequency overlays

heavy marsh
#

I used the **editor **for replacing sections, because there are multiple vocals or there was a wrong pronunciation. My issue is now that the style how the vocal sing is very different to the rest of the song AND I can hear the "transition" clearly. Is there any trick or prodcedure for fixing it?

half dragon
heavy marsh
#

Also covered it, not satisfied as it changed a lot

half dragon
heavy marsh
#

I thought Editor would work, but I'm not satisfied with that

half dragon
heavy marsh
#

But i got it what you mean. Will try it, if I got the transition issue

tawdry scarab
#

asking for the moon but multiband eq and multiband comp would go a million miles

#

especially if we could stack them on the tracks

#

throw in something for harmonic expansion and it would be a dream, people would be making fire left right and center

hard beacon
#

like i said: but as long suno studio is a simple stem seperation tool from the audio it makes no sence because of the frequency overlays

tawdry scarab
#

as long as you keep the same stems the overlays kinda smooth out

#

across each other

#

I know you won't get perfection

hard beacon
#

if you try to eq on layer ( in this case in a DAW ) the whole sound is going crappy...

tawdry scarab
#

I mean, if you have super high fidelity speakers and ears that might be true

hard beacon
#

that's what i mean with freq overlay

tawdry scarab
#

but I know I don't mix to a standard

#

so this isn't my thing

hard beacon
#

export stems in an external DAW of your suno track an remove , for example, the vocals.. it will not be an acapella song... coz in the vocal stem there are also instrument freqs

tawdry scarab
#

I always add too much bass harmonic and I like some distortion of various kinds

#

oh, okay, I thought you were trying to say something was wrong with eqing with the stems all intact

hard beacon
#

no prob mate 🙂

tawdry scarab
#

because if all remain present it's just like, 10% or less wet on any bleed

#

for whatever effects you add per track

#

you can get stems from a stem too but I don't know if that has diminishing returns

hard beacon
#

i just wanna say: EQ or other effects doesn't make sense until the suno studio stem seperation is a "normal" stem tool 🙂

tawdry scarab
#

if you can invent perfect stem separation you'll have all the money you could want to throw at credits doing it the long way

#

if the context completion gets good enough you could reliably generate one piece at a time

#

but the separation is just stem separation

#

as flawed as it is

graceful condor
#

@stuck junco David I did some other extends. Something is definitely broken today. Because in Rock/Metal all extends are all over the place. The drums are mad like crazy, stuff is repeating itself, like melodies are repeated like the guitar plays a solo and each small part is played twice. this makes no sense anymore. Extends are totally broken. I send you some links via DM.

stuck junco
frozen dagger
#

used the Suno a.i. lyric remix button thingy... then it flagged its own prompt for the word "swift" for containing artist name. -.- ::shakes fist at Taylor:: but yeah conflicting with itself. kinda funny.

hollow hare
ornate spire
#

Gotta love how covering a 7 minute track produces a 2 minute track

molten verge
#

Hi i have issues, i cannot start making music on my computer cause the page doesnt want to load... Can someone help me?
I use google BRAVE and removed all my cookies already

lunar crow
#

hi folks, could anyone give me any tips for how to prompt a very slow piece with lots of negative space? For example a piano piece more like Erik Satie, than like Mozart?

#

I have tried words like minimal, slow, slow tempo, (and referencing erik satie)

#

but it still goes full mozart mode

frozen dagger
# lunar crow hi folks, could anyone give me any tips for how to prompt a very slow piece with...

to get an idea of how/where to start i put Erik Satie in the style prompt and the used the suno ai button so it would take this artist name we cant use and turn it into a prompt based on what it recognizes from its own training: here is what it gave. clear the field add his name hit the button again for a new version of words for small variety add in your own wordings til you find a sweet spot:

A neoclassical piece led by solo piano with delicate, floating melodies. Sparse, open harmonies and gentle play with tempo evoke a serene, contemplative mood. Subtle string textures and a touch of celeste add depth, while the arrangement remains minimalist throughout.

#

try musical terminology etc. Andante – walking pace

Allegro – fast, lively
Presto – very fast
Prestissimo – extremely fast, faster than Presto

lunar crow
#

thanks for the tips, it's still doing basically the same results but I learned some stuff 🙂

#

i guess its training data is weighted heavily towards the style it is putting out

frozen dagger
#

try something that tells it to fluctuate between the bpms u loking for. 40-50 bpm Tempo rubato, for varying tempo

lunar crow
#

Larghissimo - extremely slow

#

cool, does 'cowbelling' work with suno?

frozen dagger
#

v5 feels like it is harder to get stuff like that. but slowwing down any track speed and use that as a insperattion/remix/cover to try and create a new gen that is now basically making the music but also trying to emulate the slowed down version warping the cadence

lunar crow
#

does repeating terms to give them more weight in a prompt work with suno? (cowbelling) And does the start/order of words in a prompt have more weight?

frozen dagger
lunar crow
#

cool, thanks for the info

halcyon sundial
#

how to force Suno to say "trace" as in english in my song in portuguese

#

i tried "tray-ce" but still saiying in portuguese pronunciation

frozen dagger
steady talon
#

trying to adapt my piano sample into church bells resulted in this. I'll try further with something like "deathknell"

#

which is cool but not what I was going for

lunar crow
#

yeah generative AI is terrible for specifics.

#

its a tool for broad stroke, flights of fancy

steady talon
lunar crow
#

same as in image generation, you cant give it super specific instruction yet

lunar crow
steady talon
lunar crow
#

good luck!

halcyon sundial
#

extensions are just ignoring lyrics?

#

how do i do extensions now?

#

i understand when it bugs at more than 10min

#

but 5 min?

#

3 thousand characters and 20 seconds song?

frozen dagger
# halcyon sundial extensions are just ignoring lyrics?

without knowing what you are experiencing first hand my suggestion is to change your extend point. it gets tricky trying to extend because it has this like ghost memory of the lyrics you had from the first gen with each new extension adding more "ghosts". finding a sweet spot for where to extend has been the issue in my experience when it gets weird with the lyrics.

#

especially true if you are changing lyrics when after you hit extend. it gets weird.

halcyon sundial
#

im making a special music

echo moat
#

So now that V5 is out what are y'alls go to formulas for perfecting songs once they're created. I mean inside of Sue no of course minus any external master anger or whatnot

tawdry scarab
#

doubling bass and vocals and setting their transpose to +/- 5, 7, or 12, and puting in extremely small delays between them

#

oh or putting a sample on two tracks, one panned left one panned right, splitting and fading between the two for stereo movement

fluid sentinel
#

I keep getting this whenever I want to extend/add vocals etc. to an existing gen 🫤

Screenshot_2025-10-15_115904.png
plain pumice
restive aurora
#

Hello all. I was trying to subscribe and I cannot pay. My credit card kepps getting declinded. I call the credit card company confirmed I was good to go and the bill was paid lol..... also I tried not useing link and still the same declinded. So I was not able to sign up for the year... and I cannot make anymore songs... dreadful

tired forge
#

Oh for frick sake. F*** my memory. I forgot who helped me yesterday or whenever. They fully and easily explained extending Songs to my butt

#

RN was in their name I THINK

#

I just wanted to say thanks again.

tawdry scarab
#

I can't find who I think you mean in the at list

#

:c

#

so idk either

hollow hare
# restive aurora Hello all. I was trying to subscribe and I cannot pay. My credit card kepps gett...

Unfortunately Suno does not handle payments, you will have to reach out to the payment processor Stripe.

What people in a similar situation does is download the official app (https://suno.com/Android or https://suno.com/iOS) which will use Google/Apple as the payment processor. I think their subscription price might be a bit higher due to the 30% fee Apple and Google takes.

Whichever route you decide, double check the total before making the final payment. A lot of people accidentally subscribe to the annual plan

hollow hare
halcyon sundial
#

now extension is "working"

#

but start to sing bullsheet at the exact middle

#

of the extension

#

became a total iliterate

#

its happening everytime

#

5.0v is over bugged

hollow hare
#

I haven't used extensions for a LONG time, so hopefully someone corrects me, but I think you need to align the lyrics first before extending? Typically works better if it had just the song structure and lyrics without further instructions when aligning iirc
(I will double check this but that will take time)

halcyon sundial
#

i tried to downgrade to 4.5 and he just ignore the lyrics lmao

#

i hate when extensions are doomed

hollow hare
#

It was made during the 3.x days so it is understandable that things broke as new models were made. Unacceptable IMO but understandable

hollow hare
# halcyon sundial but start to sing bullsheet at the exact middle

Mind DMing me the link to your song you are trying to extend, what time, and what lyrics you are trying to append?
I will try on my side - saves you some credits.

If you want to take up that offer, just make sure remix is allowed and ping me here that you sent the DM. (I have DM notifications disabled because I get a ton of messages...)

halcyon sundial
#

call me

#

is there a way to share privately?

hollow hare
# halcyon sundial is there a way to share privately?

Yep, Direct Messaging (DM). Click the profile picture > message.
If you get an error saying it cannot be sent, edit the server setting to allow DMs (both users must have this option enabled to DM without a friend request)

tired forge
hollow hare
#

I know that all too well. Depression sucks

tawdry scarab
#

if they had an emoji for a brain falling out I would use it

gaunt acorn
#

Anyone know why and can help me create “Hooks”? It keeps giving me an error message???

half dragon
half dragon
#

Vertical 9x16 aspect ratio
Duration: 10s - 2 min
Max size: 200Mb

gaunt acorn
#

Yes I have the Premium plan, the most expensive one. It continues to say I Error and that I need support assistance 🥲🥲🥲

#

I will try again brb

sudden spear
#

Extend - I need assistance on how to use extend in these new versions. Been with Suno since release, and trying to extend end of a song now doesn't work like it used to. It used to extend based on lyrics, etc. In the Suno Editor, I choose that -----> right arrow at end of song, it does nothing to extend it out. The old extend function doesn't work properly either. I can put a whole new verse and it just puts a different ending. Whats going on?

#

There used to be an option to keep the same duration or not. I dont want to keep same duration as Extend wont extend then.

half dragon
sudden spear
#

yeah and unless i put a lot of stuff like verse chorus there, it wont extend. Say i want to just add an eletric guitar solo or something at end. All it does is add 2-3 seconds and stops.

half dragon
tawdry scarab
#

move the start from point on the extend back and or take the short extend and extend on it, making sure to move the auto selection to contain at least .5 seconds

#

extends will start up again even if they have to put in a track change

#

if you keep taking free ones

sand basalt
#

anyone ever crash the edit mode because Suno parser trips on consonant combos and just smears them into one word?

#

it won’t split the words into separate syllables with the phrase “it’s last expression”. It makes a word “itspression” and crashes/freezes and I have to reload page

karmic lake
#

Has anyone else been having issues with Suno ending all your songs 4:57, cutting off the final part of your song on ever single version of the song you do?

karmic lake
balmy mortar
#

I have a question about allowing comments on songs. How do you get it to toggle without disabling it? It’s just not working for me. Every time I click on it it’s just reverts back. I mean, there are several songs that allow it but when I go back to another Song the next day, it’s a disables it

hexed iris
#

I’m trying to create a song with Suno AI V5, but I’m struggling compared to V4.5. I’d love some guidance on a few things:

  1. Highs, lows, and mids in melody:
    How to write verses so that one is sung in lower notes, the next in higher notes, etc. V5’s way of picking highs, lows, and mids feels a bit unpredictable.
  2. Highlighting key words in lyrics:
    How to make certain words stand out, either emotionally or vocally.
  3. Extended/stretched delivery:
    How to make V5 sing some words in a prolonged way, for example, turning “A city of dreams” into “A Ciiiittttyyyyy of dreaaams,” instead of delivering it flat. I want it to feel melancholy or emotional, with the AI understanding the lyrics...not just like reading poetry.

V4.5 can capture the emotion and flow of a whole song or lyrics with a single prompt, without needing word specific instructions...unless you want something specific like layered vocals, ad-libs, echo, or crowd chorus. But V5 seems more complicated. I’ve tried a lot but failed so far.

I have so much more to ask, but this post will become unreadable 😂. If someone can share a lyrically rich composition using almost everything in the song, it would be great! Any tips or sample prompts that help V5 understand emotion, phrasing, and expressive delivery better would be amazing!

Is @polar spade there?

graceful condor
#

@stuck junco

I send you a DM with close to 30 Tracks and this is only a tiny portion of what went wrong so far. I invested again more than an hour of my time. Instead of creating music.

Take this seriously. Because honestly yesterday I really felt like being in the wrong movie. Because we talked with you for hours. People where collecting issues and in the end you posted this and tagged me in it:

-Persona models from earlier versions not working or missing.
-Exclusion commands and filters not functioning as intended.
-Some features appear or disappear inconsistently between sessions.

And then after I read this post I told you that these are not the ones we all talked about being most urgent. Then you answered of course mine are mentioned are important too. NO David they are THE MOST IMPORTANT. Not important too. Because why are we even mentioning all this stuff and then the focus goes back to something else again. It makes no sense cleaning the house if the foundation it crumbling. In this moment I really didn't feel taken seriously and not just me the other too. Because you tagged me at this message but I never said that these 3 aspects are such important ones. I came forward like others too because of much more important ones.

And none of this stuff was on the "Urgent MATTER list" None of them! In fact all of this stuff is good if it is fixed but not urgent at all. Those are not game breaking. And we told you this a couple of times. Just yesterday alone 4 people @livid crescent , @rigid pond ,@modern vine plus including myself told you that the most GAME BREAKING and really frustrating issues of the current model and even effecting 4.5+ are:

-Misspelling Words (Lyrics), Absolute Gibberish, Not following the lyrics or skipping words and parts of the Lyrics
-Extends in V5 are a mess. Especially all Electric Guitar/Drums Genre types like Rock/Metal/Blues and their Subgenres.

#

And those were 4 people at the same time but I read this all the time from countless people here and there.

THIS IS URGENT! Not some QoL stuff which might be good to have but isn't game breaking or in this case songs ruining. And we waste Credits aka Money on this stuff and a lot of time. And I and others wasted their time reporting stuff here on discord, spending more time here over and over again instead of creating music.

And so far I might found a temporary solution (if you wanna call it like that) for the "V5 Extend" problem:

If you wanna extend especially Rock focused music. Use 4.5+ Instead and reduce the Weirdness to 15% just before it turns red. And the Others around 50-60% this is the only way to get at least something decent which is not all over the place and works. Not that I am saying this is a perfect solution because actually V5 should to do the Job but right now it can't as we see and experience at least not with Metal/Rock etc.

polar spade
#

Just some random brick-wall comments of mine to kareem off of VanFersens notes.... One quirk that's been prominent: There's the peculiarity or quirk of the "ooos and ahhhs" or "huh yea" in the beginning of things like hip hop or trance.

VanFersens hit the nail on the head though in knowing to use all resources, rather than just test out the beta models (v5. 0 now, but this has been true for two years with not only Suno but all other new beta models of generative Ai music companies).... and assume brilliant results above a certain percent of the total. To do that would be quite frustrating for sure.

A. I. and 4 year olds are the same. They have to build their own synapses before learning anything.

That percent may remain low until consistent glitches and bugs are reported and devs tweak these quirks out of existence (or hopefully up to 80% effective) . Thus the "beta" bit. Yes, we are the Borg of beta testers. That's just how things work to be first in line.

My Dad used to always bring home white packages contain food mostly. His company would give "beta" food to try out on the employees all the time. We kids were, I guess, the "credits". We were considered expendable. Hahaha.

Start in 4.5+ or less and once a base is set, then remaster or edit up to 5.0 beta if it helps. Not perfect, but often better... but Perfect advice!

The latest update notice by MikeB notes something interesting coming for access by 3.5 users as well. Always like a new tool. If I can have a tool that I may only use 2% of the times, it's still worth having in the toolbox. (so I'm sad 3.0 - - the sound effects model - - is disappearing)

polar spade
# hexed iris I’m trying to create a song with Suno AI V5, but I’m struggling compared to V4.5...

Hi mz. When you say "compared to 4.5" were these things in your list that were consistent in performing them in 4.5? In other words, you got these... and can still get these.... to work in 4.5 pretty consistently?

I'll add in full disclosure, some things you mention are not in my wheelhouse as needed for things I output, so it just ends up my testing to see if I can figure some of these out, as well, myself.

hexed iris
# polar spade Hi mz. When you say "compared to 4.5" were these things in your list that we...

Yup. I tested the free version for only a couple of songs and was blown away enough to jump into the paid version. 4.5+ didn’t disappoint from the very beginning until I started creating the next level of compositions, for which I took your tips and implemented them successfully. Version 5 beta has not produced a single composition that matches what I imagined in my head, though I have tested the same two lyrics multiple times. V5 has everything to like, but unfortunately, it is not working for me.

graceful condor
polar spade
# hexed iris Yup. I tested the free version for only a couple of songs and was blown away eno...

The melody variations were always more about prompting for me. For instance, I've always liked get the more emotional responses by throwing in "CCM" in the prompts (contemporary Christian music). An odd choice but it seem to pluck the chord progressions and melody in a different direction. Not a suggestion, but an observation.

To target getting higher or lower ranges just feels as if it's going to depend on the genre you're going for, then which prompt, such as "prog-rock" but oddly, maybe even "dreamy"
to get a higher range of vocals within that genre. Did you say your choice is more heavy metal?

And that's assuming we want to prompt through the whole song rather than use extentions and do a bit of roulette with the prompts over a set of maybe 5 generations per extension. I'll do that in the Editor and just choose the best of the melodies sampled for that particular extension.

Further, now we can just decide what melody we want, do the worse of singing in falsetto or bass while recording an audio, but good enough that Suno emulates it fairly well. We can add vocals to audio, right? I'd have to check that... If true, add vocals with lyrics matching your next verse.

.... Then splice that into the end of the original in Studio.

Just thoughts. As I said. I'd have to test and keep notes to see if these would work.

graceful condor
torn spoke
#

BRO

#

@flint haven why u gotta be like this

#

Ok, this time Woodhouse, it should be ok. Madge

I'm wanting to start a community research project with the goal of figuring out how many of these terms are actually effective in prompting suno, both in the prompt section, as metatags, and inside actual lyrics.

📎 Italian_technical_terms.txt
#

Finally, it wasn't filtered.

#

It'll be a lot of tokens for just one person to experiment and figure out all the terms, so it'd be economical for multiple people to pitch in

#

I wasn't allowed to link google docs here curtesy of @flint haven Madge
Maybe we could make a server or something

hexed iris
# polar spade The melody variations were always more about prompting for me. For instance, I'...

I’m more into Pop, Folk and Fusion. This song in particular has historical references and a more poetic feel. I want each verse to sound distinct...with variations in highs, mids and lows for every reference. Or simply say some to have soft vocals and others to have mild/bright/energetic tempo. I still have one from 4.5+ (pre-V5) that turned out quite well, even without a specific prompt. But I feel something is missing in that version, so I’m now trying it on V5, all my attempts so far have failed to capture that feel. Not that 4.5+ is giving a better version in my later trials.

torn spoke
#

@hexed iris

hexed iris
torn spoke
#

Try all the methods

#

see what works best

#

and share the findings

raw oasis
#

doing Broadway / prairie fusion and it is hell so far lol

lethal nebula
#

hello everybody

#

Does anyone here know the suno tool well, the cover option

torn spoke
#

Yeah why?

torn spoke
torn spoke
raw oasis
#

And I gotta up my DAW game.. to get what I am really after.. just being lazy.

lethal nebula
#

I have the subscription for the studio, so I have 10,000 tokens.

torn spoke
#

I've used suno for a while now, a lot of time has been spent experimenting with fine tuning the prompts and texts, manipulating covers and what not, if there's anything you're curious about, hit meh

#

I might know, I might not

raw oasis
torn spoke
lethal nebula
#

The reference song has 2 verses, 2 choruses and a bridge. When he is creating, he skips the bridge and goes straight to the final chorus.

torn spoke
#

a couple months ago, taking your song out of suno, into a DAW, to edit it wasn't really an option. Cause the upload limit was so low. Now that it's 8 minutes, cleaning up songs is actually viable again

#

well, not "again". It's finally viable.

torn spoke
#

or rather

lethal nebula
#

suno

torn spoke
#

what do you mean "creating"

#

do you mean lyrics, or audio

lethal nebula
#

intrumental

#

instrumental

torn spoke
#

Have you made a personality of the song you're using as reference?

#

When doing a cover, in general, if you want something that's actually has the same melody, pace, and structure, you're pretty much forced to make a personality of the song you're covering.

#

damn, woodhouse is lightning fast

#

And, don't settle for an "ok" melody btw. Burn some credits, get a melody you really like. Then start using that as a base.

raw oasis
#

For example here my intro is being totally ignored. not hearing Dorian mode per se. not hearing nursery ryhyme melodies (but that is somewhat expected), never hearing a cowboy yodel - it may not know what that is. I added Dobro and am hearing that now though so I like that addition - and helps me know that it sees my bracket instructions.

[Nursery rhyme melodies in the verses | Dorian mode | occasional cowboy yodel | Dobro]

[Intro: opening with wind and windchimes]

[Verse 1]
She’s canning peaches unaware.

lethal nebula
#

I play guitar and I made a song only on that instrument. So I uploaded this audio only of the guitar, so that Suno could do the cover, adding other instruments but following the reference of the reference audio, but Suno skipped the bridge part and went from the second verse straight to the final chorus.

torn spoke
#

^ only necessary when dealing with uploaded audio.

torn spoke
raw oasis
lethal nebula
#

In short, the reference audio has a time of 4.48 and Suno only created up to 3.40 of the reference audio

torn spoke
#

You can be pretty descriptive in there, as in you can be specific in your descriptions.

torn spoke
torn spoke
# torn spoke You can be pretty descriptive in there, as in you can be *specific* in your desc...

This style prompt worked really well for example;

An intense, defiant musical number centered on one determined voice. The vocal delivery is realistic. Smooth, yet raw and emotional, with audible grit and flashes of anger — more spoken than sung at times, pushing words out like stubborn defiance. Instrumentation is minimal but forceful: low piano chords or rhythmic acoustic strums, underpinned by a steady, tense pulse of muted percussion. The arrangement grows in pressure through subtle additions (dark strings, dissonant swells, or sharp accents) but avoids a triumphant release. The focus stays on the solitary, angry voice that refuses to give in, embodying relentless persistence and barely restrained fury.
lethal nebula
#

Can you do this for me if I send you the reference audio and the prompt?

torn spoke
#

I should create video tutorials on how to do stuff like this I realise, would save me a lot of time

torn spoke
#

20 gens is pretty low all things considered

raw oasis
lethal nebula
#

so I have to split the audio into 2 parts and then join the versions created by suno, that's it

torn spoke
lethal nebula
#

I would also sign up

raw oasis
lethal nebula
torn spoke
#

I've been using adobe audition for ages now, I'm stuck in the adobe hell

#

It interfaces really nicely with premiere, so I've gotten used to the workflow

lethal nebula
#

olaa

scarlet lily
#

Extend stopped working, it gives an error when generating

reef pendant
pliant charm
#

Hi everyone, I am trying to make a song in the same style as the song Goldfinger.

#

Anyone have any prompt suggestions please?

torn spoke
torn spoke
pliant charm
#

Thanks

#

Take your time

torn spoke
# pliant charm Thanks

Try something like this. It might be a bit too extensive, so remove things that aren't important to you.

Dramatic 1960s-style orchestral pop with big-band/jazz spy-thriller energy: lush strings, bold brass fanfares and stabs, vibraphone sparkle, occasional harp glissandi; timpani rolls and cymbal swells; brushed drum kit with shaker/bongos; clean tremolo electric guitar; warm, rounded bass. Vocal delivery is smooth and legato/cantabile with tasteful portamento and controlled messa di voce on long notes—intimate in the verse, opening to a confident, glamorous belt in the refrain. Production is vintage and cinematic with a slightly mono-leaning image, gentle tape saturation, and plate reverb on vocal and brass (subtle spring on guitar); tight, punchy low end. Keep it elegant, suspenseful, and luxurious;
#

@pliant charm Tell me if it works btw, I'd be interested to know if the italian terms do the trick, give it 20 gens or so

pliant charm
#

Thanks

#

Ill be sure to give it a go

tawdry scarab
# torn spoke Try something like this. It might be a bit too extensive, so remove things that ...

to mirror this in another style and show the breadth of ways you can use this, I've asked a gpt to follow rules made to form keywords into "atomic concepts" indivisible keywords instead of phrases:

dramatic, 1960s, orchestral pop, big band, jazz, spy thriller, lush strings, bold brass, fanfares, stabs, vibraphone sparkle, harp glissandi, timpani rolls, cymbal swells, brushed drum kit, shaker, bongos, clean electric guitar, tremolo, warm bass, rounded bass, smooth vocal, legato, cantabile, portamento, messa di voce, verse intimate, refrain belt, confident, glamorous, vintage, cinematic, mono-leaning, tape saturation, plate reverb vocal, plate reverb brass, spring reverb guitar, subtle, tight low end, punchy low end, elegant, suspenseful, luxurious
pliant charm
#

Needs to be under 200 characters

torn spoke
#

I don't have that limit

pliant charm
#

I'm using the free version

#

That must be it

tawdry scarab
#

are you in simple or custom mode

#

or do they just only let 3.5 use 200

pliant charm
#

Yeah thats all I can use

#

Curse my poverty lol

torn spoke
#

ah ye suno is a bit expensive

#

I choose to budget around those kinda things

#

pay for things I really like and cut out the things I could do without

tawdry scarab
#

never cancelled suno but my office chair disintegrated around me and I mvoed to a car seat from a van on the floor, my wife had to buy me a new chair for my bd

pliant charm
#

Lol

#

The lives we lead

tawdry scarab
#

ok here's a condensed list:

dramatic,60s,orchestral pop,big band,jazz,spy,strings,brass,vibes,brushed kit,tremolo guitar,bass warm,legato,portamento,belt,vintage,cinematic,mono,plate,tight low end,punchy,elegant,suspenseful
#

be sure to not use any spaces after commas

pliant charm
#

thanks

torn spoke
tawdry scarab
#

you can pack more in the style than is otherwise able

#

thats it

torn spoke
#

Ah

glacial scarab
#

There also does not seem to be much of a difference if you use them or not. In the past it mattered but seems like it no longer matters (might have been when it got increased to 200 characters not fully sure when it changed)

tawdry scarab
#

something to note though is if you cover or use style I think it crops

#

because it puts the spaces back in

vale axle
#

just wrote somthing in here and it never appeared ?

polar spade
torn spoke
#

HEH luls

gaunt acorn
#

Anyone else having issues creating “Hooks”? I keep getting an error message saying, “Exception during hook creation, your account needs support assistance”. ???

torn spoke
half dragon
polar spade
half dragon
vale axle
#

any better font?

#

my font was too big i think it may be ok now

gaunt acorn
#

@torn spoke Yes thank you!

supple moon
#

Hello.
How many genres have you guys been able to fit into one song?
If it’s been more than 3, I’d love to chat with you!
please, tag me if so, so I’ll get the notification

gaunt acorn
#

@half dragon I’ll give this a try thank you!

#

@polar spade No I haven’t tried that, I will definitely give that a try now! Thanks!!

torn spoke
supple moon
torn spoke
#

I mean I usually pick one main gengre and then specify subgengre elements. Do you mean completely different main gengres?

torn spoke
supple moon
# torn spoke Like, which genres are you trying to mix?

Well. I’ve perfected a Folk-Trap Hybrid.
Then was able to have it recognize and take on Metal Instrumentals that weren’t overwhelming.
But wanted to get a mix of 80-90’s Hip Hop as well and then go off the wall and have a section of Russian Limerick tribal techno 😂

I’ve tried to map it out to have a song that will go through each genre it in sections of the song with hints of each in the song as a whole but what usually happens is that it will pick 2 of the genres I want and then add a dash of one of them. (Like a record scratch that happens maybe once or twice.)

It’s madness, I know 😂

torn spoke
#

You'll have to make personalities of the sections, and then cover those sections with your diff genres in the style prompt, then stitch em together, make one personality of the entire thing, and then generate a complete cover to smooth it over. Is how I'd go about it.

supple moon
ancient moss
#

Hello, Since yesterday, the Suno website has not been working (neither on Chrome nor Edge, nor on the Android app). The screen is completely black, except for the icons on the left-hand menu. Each menu item (Home, Create, etc.) leads to a completely black page, except for: “Explore,” “Notifications,” and “From Suno.”
I use Suno every day :-):-) and I can't wait for it to come back !

viscid ravine
ancient moss
#

on the website, I don't have access to anything and I can not sign in nor out. For the Android app, I tried to reinstalled it , but it won't open; it stays on the home page .

viscid ravine
ancient moss
pure mica
#

Guys i need help, is weeks i cant create music, i have pro, always the same error with different devices, cookies deletes, different browsers, PLEASE wait for your other generations to finish, I have none of in waiting, the Suno support doesnt reply to me via email and they are ghosting me...

vital yoke
#

Is there an easy way to get Suno to give the singer an accent?, it seems that adding "The male singer has a raspy Scottish accent, deep, and gruff yet singing gently with emotion" is totally ignored

viscid ravine
viscid ravine
torn spoke
vale axle
#

is it possible to change the key of a song you have made in the cover section

pure mica
#

Somebody does it know how do I fix this error, is been weeks i cant create nothing...

image.png
elfin lagoon
#

Hmm. Any advice on how to get Suno to properly apply the right voice?

So for this song, it is a cover of a song i wrote years ago, and I uploaded the song (complete with both male and female voices) to Suno. In addition, I added a tag in the lyrics to specify which vocals are male and female. But after 12 generations, I still get the male voice singing "I am just a girl..." Every. Damn. Time. Any advice on how to prompt so this doesn't happen? Or should I just keep throwing credits at it and pray?

Here's the song in question:
https://suno.com/s/wGl4WKZkX5tM0iSm

Suno
Falling Star

Listen and make your own on Suno.

random stag
#

What is the scenes that is going away?

half dragon
tawdry scarab
tawdry scarab
summer garnet
#

I do notice that almost every intro has an ab-lib in it. Is it that common in music and don't even recognize it or is it just bias to ab-libs in intros?

river rampart
#

Hey! I am looking for tips about studio editting. Especially with respect to finite control, such as adding a pause or instrumental part.

I find adding guitar solos is around 80% effective, Suno seems to know what the words mean and most of the time adds a guitar solo, quality aside.

In this track I really want a little space between sections, but its tricky, first the vocals overlap between the pre-chorus and the hook, as in hook lyrics are registering as part of the pre-chorus. I tried replacing adding stylized tags for [pause - collapse] and [epic drum solo transition], but so far the closest I got was a collapse - pause followed by a piano intro with about 1beat of double kicks entering the hook!

I tried modifying the styles and lyrics with different directional tags, I tried modifying the original track into something that resemebled what I wanted and then tried chopping that into smaller parts and trying to replace them, I also tried adding a new section and found it really odd that it asks you how many beats you want, but then seems to completely ignore that input in its creation 😅

is there anything I am missing here or is it just keep donating tokens until it luckily resembles what I want?

elfin lagoon
tawdry scarab
#

it is hard to steer and kinda risky doing regenerations but if you want a targeted approach for a complex drum fill prompt i could maybe figure something out

#

sometimes unlikely generations take conspicuous tag compilations

#

by which i mean weirdly long weirdly focused and weirdly broad all at the same time

river rampart
#

Yeah the track is a symphonic metal track, so a drum fill would really fit in well.

My lyrical tags so far have been:

[PAUSE - COLLAPSE]

[Explosive Drum Solo transition]

[Hook - Fortissimo, symphonic, full accompaniment]
It's

The hook part is purely for smooth integration with the hook or chorus, I switched to hook because chorus kept throwing random unwanted choir sections in.

This was the best version from before where at least the structure was right, so I tried splitting it into smaller segments for more control, the above replaces [Big Build up]

[Guitar Solo - Dynamic shock]
FORSWORN

[Big Build up - drums]

[Hook - Fortissimo, symphonic, full accompaniment]

With styles, I have tried leaving it blank, tried entering things like collapse, drum fill, drum crescendo, etc. but not had any luck!

#

my next steps would have been - find a drum solo that sounds right. Replace guitar solo so it syncs well and removes some vocals I don't want and then probably remaster the whole thing. Though I often find remaster doesn't clean up the transitions like I would expect it to 😅

tawdry scarab
#

yeah a zero style full audio influence cover can clean up beyond them remastering and stays with the same feel often, but i can give you, for sure, a longer drum section

#

i mean by providing the prompts that would push it much harder

river rampart
#

ooh cool! like lyrical prompts?

tawdry scarab
#

yes, and i can't stress enough how much it helps to have the first style keyword be drumcore

#

i'm just turning on the computer and getting my gpt ready for a drum focused gen then i'll show you and you can take from that prompting how it might work to really, really press the issue

river rampart
#

awesome! You already gave me some ideas so I am genning a couple of extra samples too 😄

#

which gen style do you find is best for generating longer fills? I feel like smart should be the choice, classic seems to respect the tags more, but none of them seem to extend the duration (unless it randomly remasters the whole rest of the song!)

tawdry scarab
#

what do you mean smart vs classic, i haven't been deeply in studio for what you're doing i usually just add woodwinds and they're pretty behaved

#

does studio let you specify different gen modes? i need to try that out and understand it

#

the biggest thing to me that impacts length is the amount of stuff you give the interpreter to chew on

#

so if you ask and barely get it it glanced over the part of the prompt, if you build it out bigger and bigger it will allocate more of the song to it

#

that said drums i've had to do short piece at a time before

river rampart
#

This leads me to question how you are generating new fills 👀 are you using the legacy editor?

torn spoke
tawdry scarab
#

in studio it would be major surgery

#

i was going to show you how hard you have to push the prompt to get it to ignore metal and focus on drums for a second then that would translate to being able to regen the drum track across a gap you'd open up between the other stems, but potentially after generating the drum fill so the fill has a backing then you make it alone then you regen the rest

#

all the while hoping your stern separation is good

river rampart
#

another tempting option is to generate a separate drum solo build, crop it into the track by exporting the song and using 3rd party tools, re import, then use studio to smooth out the breaks...

tawdry scarab
#

absolutely

#

take the path of best flow

torn spoke
river rampart
#

Getting to work! Thanks for the tips! I had been working on this all day and nearly sank about 1k tokens into it 😅

tawdry scarab
#

yikes!

#

my method is not to worry if any specific tag will work too much as long as they're written as small individual concepts not big clause-filled sentences, and I just go harder and harder at what's missing until something starts to change. so if you look at the style here, the metal have been burried behind an almost full wall of drum keywords and encased in multiple layers of brackets, and then the style is so watered down there's a long almost empty intro, but then it finally gets into drum focused metal. this is because complex drum work is so much less likely than generic metal:

https://suno.com/s/5pJRYYl6uRkJQHYy

#

@river rampart so when you have time you can listen to that and make it through the minute intro there is drum focused metal with elaborate fills

#

but to do this better, I'd need to then dial the metal back in a tiny bit at a time

#

and make the section of the prompt that's supposed to be the drum solo occupy more and more of the total lyrical tagging bulk

#

this is an example of a broad shotgun method that almost requires a GPT because of the broad level of technical details HOWEVER the same rules apply in general if you do it or a gpt does it

#

it's about matching the proportions of apparent influence with broad focus but hyperspecific details

torn spoke
#

o sh! Just found a project I had pretty much given up on!!! LEZ go, oooo im hyped

#

Download and re upload will save this thing

tawdry scarab
#

then here I covered that one I just showed you with a nu metal persona with very low style and 100% audio influence and metal comes back in but the drum focus is retained: https://suno.com/s/cXK8JtHq7PxvDTnX

river rampart
# tawdry scarab my method is not to worry if any specific tag will work too much as long as they...

well this is a completely new style of tagging to me!

what doe s the | operator do in this method? I know it as or, does it indirectly join th etag elements or separate them into key, value pairs?

The format seems to be:

  • General block direction
  • silence or vocal directions
  • instrumental directions, (ordered by block presence?)
  • end block instructions.

The Header and editor guidance sections are surprising to me, I didn't realise you could put this much detail into lyrical tags, I would have kept all that in the styles block.

and instead the style is full of different drumming styles, some stylistic direction, I also notice a lot of repitition, is that a more powerful form of style enforcement than lower/upper case?

tawdry scarab
#

I'm gonna end up publishing this after a few more edits, put it in a persona I like the guitar work of

#

the | gpt keeps using when I tell it I prefer : but they both operate as a separator of two portions of the bracketed region, which I use to have topic, then list

#

my whole thing is all keywords inside a tag exist alone or after a separator that has a category before it, and every individual keyword must be an "atomic concept"

#

something that plainly is without any negation, relation, association, etc.

#

no clauses in sentences

#

and then these subject keyword tags don't always give you what you ask for but they fill out the music, make it longer, and can give you an enormous amount of what you asked for in 4 and up

river rampart
#

Good thoughts indeed, I am an Ai engineer, but I mostly work with production data and targetted LLMs, where the input and output is text.

For prompting these kinds of models, you can be quite specific:

  • General outline
  • specific details
  • hard rules
  • example output.

For maximum stability when integrating with other systems I also use pydantic models to force responses into specific shapes, so responses are always as expected, and I definitely suffer from the wanting too much control bug, in fact some of my best pieces are the ones that came surprisingly early where my expectation and the prompt result were a magical lottery match up.

However, using the principles I am familiar with, I am often surprised that we can only give generative prompts to the AI model.

Lyrics and style basically, excluded styles are simply tagged onto the style section with a prefacing "-" character and rarely work well.

That said, I feel like if I were say, Beethoven, I would also expect as part of the creative process to have fine tuned control over the output, as he would certainly be equipped with the musical vocabulary to always describe what he wants, without using a gpt model.

I might do some experiments in the stylistic guidance and lyrical tags to see if it is possible to replicate some of the prompting methods that work well in my day to day. I reckon output examples is almost definitely how I can consider inspirations. So maybe rather than passing whole songs into these fields, cropping them down into individual parts that are desired might be even better. Less negative encouragement of sections that are disliked but included because they are a part of the main track.

Geez, my ADHD is too string for this mstvKEK

tawdry scarab
#

you might be impressed how much power you have when you apply music theory and technique and subgenres and even mild directions or vague concepts can materialize in the finished generation

#

the prompts seem to be ingested whole, front loading importance, the exclude functions oddly like a negative prompt in image generation, it's not exactly removing anything it's a kind of avoidance

#

and basically any syntax it will give it a shot you can structure prompts how you like as long as you are familiar with the feel of the model from practice

elfin lagoon
torn spoke
#

Suno is a time sink

tawdry scarab
#

the sliders now let you ease the transitions

#

start style zero influence 100, creep up the style

#

and over successive small extends you can get a smooth transition

frozen dagger
#

Inspo, with playlists, these create some Frankensteins.

elfin lagoon
#

Yes, but the final cover might revert the vocals back, because covers often change up vocalists in duets

tawdry scarab
#

duets are madness

#

if you die on everest it is less tragic

river rampart
#

I have definitely experienced some of the things you mention, but also find I can easily overload th emodel if I put too many similar terms in, for example using symphonic orchestra styles seems so integrated with choirs that even the phrase "chorus" in tags seemed to generate choral effects 😅

This has definitely inspired me to be a bit more creative with my prompting.

tawdry scarab
#

in which case if you target the structure of the classical style you want, like if you put in barque it's not going to pull choir

#

baroque

#

however that word is spelled

river rampart
#

that smy issue, I am not at all musical, so I tend to rely on GPT models a bit to convert what I want into musical language 😅

tawdry scarab
#

I can remember some cool words and I'm good at looking them up but I'm not the most surgical or thorough tagger

#

there are people who have lists they've assembled of the strongest weighted terms

#

in their experience

#

I don't know where those lists are to show you

#

but they exist!

river rampart
#

I am looking forward to experimenting now.

tawdry scarab
#

it's much nicer once you get sea legs on the model

#

and feel a sense of control

#

which is pretty idiomatic a thing to say so if that's unclear I just mean, there's an instinctual component, you need time with a control and results over time and you automagically get traction on control

#

also cool to be an engineer. I want to get into something professional but I'm not very dedicated unless it's unhealthy hyperfixation. I am an annotator that is cleared to do programming related annotating but beside that I just have guesses and basic NN experience

frozen dagger
elfin lagoon
#

Duets are frustrating. The base uploaded song I'm covering has the right people singing the right things, I don't know why AI can't duplicate.

elfin lagoon
frozen dagger
#

stle prompt placement says "ok they want call and respons e in this somehwwere" putting it in the tags within and clarifying with more tags to inform will let it know how and wheere and when

#

if the voices are where you want them then id try minimal appraoch first with call and response, male and female duet, ... tags like that and id hope it would recognize the sound within the audio already. that is frustrating if its not doing that already.

river rampart
tawdry scarab
#

right? I even have as many credit hours as a bachelor despite only having two associates

#

just so almost respectable XD

neat nexus
plain pumice
#

I've noticed that the Studio, at least for me, tends to bog down fast, and the browser just hangs (only the individual tab).
Not exactly sure why this is, it's actually same with just being on the webpage, just happens a lot faster when in the Studio.
Any fixes?

elfin lagoon
neat nexus
elfin lagoon
#

Correct.

#

Not the output.

torpid haven
#

In studio, I use percussion, set style bongos and it starts to make melodies and whatsoever ...

frozen dagger
neat nexus
# elfin lagoon Here it is https://suno.com/s/3XQyWr0lcD7X6d72

It looks like some lyrics are missing. No matter....

What I would try doing in your covers is change your lyrics with [Metatags] to assist you in making the change.

Since you're going with the first verse as male singer, I'd try:

[Verse 1] [mid-range tenor male vocal]

Then in the second verse, change that to your female vocal:

[Verse 2] [female vocal]

[Chorus? Pre-Chorus?] [Male and female vocal harmony]

As I said, your lyrics are missing some stuff so I'm guessing if it's a prechorus or a chorus that is next here....

But the thing I'd also try to do is extend from the last good vocal of the cover you liked best. If it ended at say 1:06 I'd extend from there and try using the lyrics with [Metatags] and try to force the change. I don't know your style prompt but I'd also add duet to the style prompt.

Another trick is to give the singers a name and script them together...

elfin lagoon
#

Yes, when i uploaded it, it didn't capture the bridge lyrics,I had to manually input them.

neat nexus
#

That happens sometimes.

#

Let me comb through my stuff for an earlier version of my Lord is Good song. At that point I was trying to get a male lead vocal and a female verse with the male lead and a group vocal for the chorus....

elfin lagoon
#

Here's what my lyric prompt looks like:

[Synthesizer solo]

[Verse 1] [Male vocal]
I don't know
Who you think you are
But I will tell you
You won't get very far
With this heart of mine
It's been set in stone
So you ought to divine
You'd better leave me alone
Or you're gonna get hurt
You're gonna get burned

[Piano solo]

[Verse 2] [Female vocal]
I don't know
What has turned you cold
But I will tell you
You don't have to fold
Up inside yourself
Let go of your fear
You'll find someone else
When the pain disappears
You don't have to get hurt
You don't have to get burned

[Guitar solo]

[Bridge]
[Male vocal]
I don't want your advice, Leave me alone
[Female vocal]
Why don't you think twice before you go?
[Male vocal]
I've made up my mind
You're wasting your time
[Female vocal]
Why are you so blind
I'm trying to help you

[Piano solo]

[Verse 3]
[Male vocal]
I don't know
Who you think you are
[Female vocal]
I am just a girl
Trying to catch a falling star
[Harmony]
I don't want to get hurt
I don't want to be burned
I don't want to be burned

frozen dagger
#

anytime you are working with covers (or remixes/extends) that involve 2 (or more) tracks with the same lyrics, it can (will) cause problems. it gets confused. particulary with covers and extends; extend song that had all lyrics but got cut off, and then you are extending that missing part with the lyrics missing, it can get confused and do weird things. create a blank space after the song ends and extend from there.

neat nexus
#

Yeah, I'd definitely try extending. I'd also play with the sliders and see if that helped nudge it.

queen linden
#

So aggravated with the female vocal quality. One song out of almost 17 attempts was okay, not great not good- okay. And 10 out of the 17 attempts were terrible. Not following input. Has anyone got the magic key to getting a really good female rock voice? Ive tried most of the usual prompts low range contralto, husky, smokey, raw, even whiskey soaked.... I can do without the Humming moaning thing AI seems to add to the beginning of songs..... okay going back to my corner to sulk for the next couple of days.....

livid crescent
#

I will say, I've had some success using 4.5 to get the aggressive edge, and then covering it with v5. It's not perfect, but gets closer to the mark

frozen dagger
livid crescent
#

I tried several other services, and none I've found can do aggressive vocals even remotely close

queen linden
frozen dagger
livid crescent
frozen dagger
#

emotional, overly expressive, intense, the more synonyms the better sometimes.

livid crescent
#

[emotional] has a tendency to go sad instead of angry tho

frozen dagger
#

[Death Metal Screams- guttural, raw, sludge metal]

#

thats why you add more words that reflect the tone of emotion you are aiming for.

livid crescent
#

That's kinda the thing. it's actually pretty easy to get over-the-top death/black metal kinda stuff. It's the in-between ranges that can be difficult to get.

frozen dagger
#

its like describing the tonal difference between Doom, Black, Death, or Sludge Metal. Mostly the same.... but there are nuanced difference. Suno does well, imo, with these nuances... but adding things like [syncopated, galloping rhythm, polyrhythms, 3/4 signature, downtuned 8-string guitar, Djent-Sludge Fusion] or whatever that is trying to hone in on a sound or vibe

#

fuse the styles. [switching rapidly between clean-singing and guttural growls/screams; soulful vocals transition to restrained screams]

livid crescent
#

And heavy music usually isn't a problem. its just the vocals... and mostly the female ones, that are hard to give that edgy feel. I've told it screaming vocals every which way you can imagine, and gotten back whispery falsettos, often back to back.

frozen dagger
#

the style a.i. remix button that re-styles prompt is disabled?

queen linden
#

Its the sitting through the 15 or 16 bad versions lol... its like water torture

livid crescent
#

It really can be

frozen dagger
frozen dagger
livid crescent
queen linden
#

I'm old. I like the tone of a Tina Turner, grace slick, or Debbie Harry. . I think im going to try the 4.5 ... otherwise its like beating my head against a wall... a wall of Disney Princesses singing their little pink hearts out ... lol

frozen dagger
#

qustion: do u have a sound of the femal vocal you want? is ther an artist that is close or band or gorup or style specific to the sound u want?

livid crescent
frozen dagger
#

u WANT disney princess stlye? i thought u were saying it was generating that and u didnt wanttt that

#

wwait diff person now lol

livid crescent
#

lol I never want that. 1000s of tracks of mine, asking for heavy vocals, would meet that description

frozen dagger
#

youre old-fashioned as you said if you want something to sound like tina turner. dissect her style. what parts about it you like. use gpt to "break Tina Turner's vocal style and technique into musical terminology. pool from different songs cus she's dynamic. and pick the terms and things that articulate what you are looking for in a voice. describing how they [falsetto to baritone; dynamic vocal range, arias, gospel texture. soaring high notes..etc]

livid crescent
#

a few of us have been talking with David about this, it's a known issue he's trying to help us with. Currently he's recommending using 4.5 to get heavier vocals, then remix into v5

frozen dagger
torn peak
#

Is there anyway to command SUNO to make a track almost identical to reference?

river rampart
# tawdry scarab which is pretty idiomatic a thing to say so if that's unclear I just mean, there...

So I might not be great at music, but I am very good at fixing things. I used to repair audio from old cassettes by digitising them to WAV and repairing the waveforms from other similar sounds in songs...

Thats kind of what I did here, exported the files and some similar drum heavy tracks I made and then sewed the bits together.

Its not perfect, but I hope with Suno I don't need to be as painstaking as I was 20 years ago!

I am just generating the first 0 style, full audio influence renders now. Looking forward to seeing how they turn out!

frozen dagger
#

how almost? do u want to clone it?

livid crescent
torn peak
#

I want to use it to make a reference of a clavin harris song but its just way off

#

like completely off

livid crescent
torn peak
#

Exactly it glitches and goes wierd

#

When I describe the track in detail its still way off

#

Whats the solution?

frozen dagger
#

u want your song to sound just like another artist? no, short answer. but ask chatgpt to help you make a prompt that emulates Calvin (Cavlin) Harris. and you can make something similar with luck

livid crescent
#

so careful with that

torn peak
#

Yea Im not trying to copy just want to match a style that i can't seem to describe in prompts

#

NO matter how many different prompts i cget chat gpt to write

river rampart
#

eh I think it got hooked up, 4 renders all completely ignored the Intro lyrics. hard refresh go agane.

livid crescent
#

I've had some luck feeding audio to gemini and telling it to give a prompt.... but still a roll of the dice

frozen dagger
#

if i wanted to create a sound that was similar to Sleep Token this is something I would use:
Cinematic dark alt-metal fused with ambient electronica, trap, and gospel-soul undertones, The arrangement flows like a ritual: atmospheric verses erupt into colossal, rhythm-driven choruses, Clean baritone and falsetto vocals glide between reverence and fury, 8-string djent guitars tuned low (EAEADGBE) alternate between glassy cleans and percussive chugs over thunderous 808 sub-bass, Drums fuse live performance and digital precision — DW Collector kit blended with trap hi-hat rolls, stuttering snares, sub-decay kicks, and double-kick bursts, Glitch percussion, reversed crashes, and granular noise textures fill the negative space, Cinematic strings, choir pads, and distorted brass swells rise beneath, Vocals range from whispered confession to layered harmonies and distorted screams, drenched in reverb and delay, The sound is massive, devotional, and violent — sacred soul colliding with trap and modern metal, worship through impact

torn peak
#

So you have to basically code the song?

frozen dagger
#

but its not going to create the newest Sleep Token single or anything, but i can reference style, equipmetn to get in that ballpark

livid crescent
torn peak
#

But its a language model thats always changing its like learning to code when the rules change

#

and sometimes the code works and then most times just doesnt?

livid crescent
#

Yes, yes it is

#

Correct

river rampart
#

isn't that just the neat part? Its like a video game whos parameters change every time you press start!

livid crescent
#

If you have studio you can mix and match parts you like and use that as a reference. if not, can download them and do the same thing in a DAW. I usually do this for most of my songs

frozen dagger
torn peak
#

is there a prompt for reverse engineering?

river rampart
#

I see a lot of people talking about studio, just to ensure I am not missing a trick:
Studio is a sub setting in the Browserware, not a standalone application for windows?

livid crescent
livid crescent
livid crescent
frozen dagger
# torn peak is there a prompt for reverse engineering?

ya... my Suno prompt button is disbaled. i presume not jsut me. i wonder if it has to do with my tip other day >.> about converting artist names to prompts. which might be a grey area for them, but honestly what's the issue. not allowed to reference now either?

river rampart
#

This is too repeatable to be a coincidence.

Persona: 7:59
covered audio 8:00

weirdness: 0%
style: 0%
audio: 100%

4 results, all 6:43 in length 🧐

style: 5%
audio: 98%

2 results, both 6:43 in length 🧐

Some cut the intro, some cut the ending. Anyone know why this happens? bug to report maybe?

But gutting to lose over a minute of editted content there. If this persists I guess best option is to take the best one and extend it?

livid crescent
frozen dagger
#

cover? they capped around 5 mins still, right?

river rampart
#

oh man, that would be really nice to have a tooltip for, I had no idea... maybe I can take two of these and sandwich them together.

livid crescent
river rampart
#

( I am colourblind though please tell me if there is in fact red

frozen dagger
#

if you dragged the songs over from the workspace to the remix/inspo boxes double chekc your lyircs/style box wasnt replaced

livid crescent
# river rampart no red:

I mean, the red for the sliders, where you set weirdness, style, and audio. keep sliders between 15% and 85%

river rampart
lyric kestrel
frozen dagger
river rampart
#

Moving the sliders to reasonable amounts stopped the songs from cutting off abruptly, but did nothing to increase the length, so unfortunately the renders have lost all their character and pacing, just sounds like generic mush. Very unfortunate.

This has definitely beaten me for today.

I think tomorrow I will go back to extreme sliders and render the parts in halves with a large overlap.

The first half I will set as the persona for the second half, and hopefully I should get a god start and a matching end after not too many tries, I'll stitch them together out of product and import them.

Is it possible to pull two tracks into studio at one time? It always seem silly looking at a single audio track in that style of editor 😅

nn, gl

plucky leaf
#

Anyone got advice on how to get the AI to recognize parts properly? I have a man and woman character in a song and they keep swapping roles.

floral seal
# lyric kestrel I cannot find any way to make a beat like this in Suno… https://youtu.be/51qvd...

Try this prompt :

Create a smooth, melodic trap-R&B track with a confident, late-night vibe.
Tempo around 125 BPM with deep 808 bass, crisp hi-hats, and ambient synth layers.
The vocals should feel cool, calm, and slightly mysterious, with subtle reverb and emotional undertones.
Theme: loyalty, grind, focus, and always coming through when it matters — understated confidence and realness.
Keep the hook short and catchy, and let the instrumental breathe between lines.
Style: chill night-drive energy with a mix of trap rhythm and smooth R&B flow.
Structure: intro → verse → pre-chorus → hook → bridge → outro.
Mood: sleek, focused, confident, and cinematic.

haughty lynx
#

Howdy all, quick question... can anybody let me know why the Button that used to refresh the prompts in the style section is showing up like a circle/line thru it when i hover over it?

frozen dagger
# frozen dagger when wanting a duet, "call and response", works. As been said, it's very difficu...

i been playing around with some stuff and it got me thinking about the male and female vocals thingy. everyone interested in duets etc.
try using Gospel as tag as well. I know... that may not be the style you want, but it seems that Gospel is not only associated heavy with male and female duet/call and response can help when you just want male and female vocals. yes it could add other elements like chorus textures, or layered harmonies (that you may or may not want), but just an idea for anyone wanting male and female vocals on a track. can separate the vocals and add a different instrumental maybe?
Gospel isn' t a lot of people's thing, but I reccomend people experiment with it. The model has a good handle on it and creates some phenomenal sounding vocals.

frozen dagger
haughty lynx
#

?

frozen dagger
#

i think they disabled it temporarily. likely be back up. sorry the "wasn't me" was paranoia related to chat

viral pond
#

does my old v4 prompts still work if i choose v4 ? it looks like it does not

frozen dagger
frozen dagger
haughty lynx
frozen dagger
#

it was working yesterday lol

#

i think they needed to make some more changes rq is all

haughty lynx
#

hmm. i wasnt on yesterday but this last week it was disabled for me. maybe a browser issue?

frozen dagger
#

if it is then we share it. i was asking chat earlier but no one replied. you saying it wasnt working was first confirm it was not just me today

viral pond
#

i think you still can genre lock the same way in v5

tawdry scarab
#

Genre lock?

frozen dagger
# frozen dagger i been playing around with some stuff and it got me thinking about the male and ...

Dark trap base with gospel fusions: haunting choirs, preachy call-response, Blend opera swells, jazz swing/funk grooves, rock/metal riffs, rap flows, acapella/scat/beatbox, Florida drill in breakdowns, Spectrum: groovy/danceable verses, heavy/sad drops, uplifting exhale, Structure:

Intro: Gospel acapella choir, soulful harmonies, piano swells on "Atlas — breaker of worlds", Verse 1: Dark trap 808s, hi-hats, synths; whispered adlibs, drone SFX, Chorus 1: Trap + gospel choir, organ pads, sidechain punch, Bridge-Trap: Jazz swing bass, scat rap, funky wah-guitar, Breakdown: Florida drill triplets + metal guitars, explosion SFX, screamed lines, Final Chorus: Trap + operatic soprano, orchestral stabs, Outro: Soft gospel harmonies, jazz piano fade, breath SFX, Male/female leads, cinematic reverb, -8 LUFS master

telling ya. if ya'll wanting tracks with dynamic male and female vocals... play around with gospel...
creating some incredible vocals

polar spade
cedar cloak
#

So i have 224 songs left but it wont create songs any one have ideas as to why

lyric kestrel
lyric kestrel
frozen dagger
frozen dagger
#

its geryed out, and when we hover over it with the mouse it show sa circle with a line. 🚫

polar spade
frozen dagger
#

he mentioned seeing others online having same issue months ago. some kind of intermittent thing

frozen dagger
haughty lynx
polar spade
#

@haughty lynx It's hard to tell. If a paid account 4.5 and above should have the wand active to give a phrasal interpretation of the already existing prompts.

As for literally suggesting new prompts, there was a button on bottom left where now is the "bookmarked prompts library" icon/button. That button would show new randomized prompt suggestions. It's been weeks since that was there.

haughty lynx
#

im on the 2500 credits a month plan. i used to have access to it but its been a minute since i last played with suno

#

did they move that feature to a higher sub?

#

yea, the refresh prompts is what i was refering to

#

the button that was just left of the random generated prompts that usually in no way reflects what you wanna use lol

#

just not sure why they removed it/moved it. it was a stupidly useful button

polar spade
frozen dagger
haughty lynx
#

yea the wand shows up, but that just reacts to whats in there already.

polar spade
haughty lynx
#

sheeeet. kinda makes you wonder... how else are we gonna know what kinds of prompts are useful there. i mean... its not like they are listed anywhere .... like anywhere on their site that i can see. sure theres the explore page but thats just more genre type stuff. the really useful info was stuff like the vocal options, emotional stuff, all of the little details that most people dont know would be a feasible prompt.

#

maybe if they made a prompt cloud or something that would be super helpful...

#

hell, they used to have really good articles on creating but i guess the v5 update they removed a bunch of help pages? cause i cant find them anymore

lyric kestrel
frozen dagger
polar spade
haughty lynx
#

sure... but in my v5 experience... the auto generated prompts almost never give options similar to ones ive used in the past. Or even ones im remotely interested in either.

#

maybe if they let us sort of pre-feed it with styles it might be better, but man... being able to refresh/prompt shop sure was nice

polar spade
haughty lynx
#

the wand option usually generates a paragraph of stuff that makes me go.... "Well this is exactly what i didnt want" and then just go back to using single/dbl word prompts.

#

yea, i posted it. I put it in the suggestions web because this isnt a studio issue i guess. seems like they are more concerned with studio only feedback.

slate trail
plain pumice
#

If you were generating a cover, but it ended abruptly before a cleaning ending, how to extend it?
I mean, when you select extend it, it's no longer referencing the source song, so what should the sliders be set to - the same?

#

So I tried using the sliders as they were, but it garbles and makes very incoherent audio, although some of it definitely resembles the source audio.

hollow hare
rose crane
#

howdy folks. so i've tried to figure it out, but I just haven't been able to so here I am. So...I've noticed that I get way more...oompf and fullness if I COVER a song with careful tweaking, rather than REMASTERING the song. (IDK, I've remastered a few songs on v5, but they sound kinda weak compared to a cover of the same songs). However I've noticed that v5 tends to forget how the source audio sounded around half-way through, but I don't get that issue when doing Remasters.

Anyway, sorry got little wordy there. Is there a way to have more control over a remaster as compared to doing a cover, i.e, cleaning up lyrics, maybe those nifty sliders, etc. Listen I know it sounds like a dumb question; just roll with me here.

plain pumice
hollow hare
plain pumice
hollow hare
#

Unfortunately I haven't played much with the sliders, being primarily a mobile app user :/

What happens if you set everything back in the 40–60 range?

#

If you wish, you can DM a link to your song and explain to me what you are trying to do, and I will run some tests on my side (save you some credits).

Just ping me once you DM so I know (I have DM pings disabled)

plain pumice
#

Not sure, I actually tried with Weirdness at 0%, and that actually produced something coherent, so now I've brought it up to 20%, and that does produce more coherent results.

As I'm trying to extend the song, it's very useful to know, what I should do with the style prompt and advanced options.
I basically left them as it was when I was generating the cover, so not really sure.

untold briar
#

By default, stick with default slider settings. Then if you aren't getting what you want, consider sliders.

  1. If the "audio" component of your prompt (Which is the previous audio in the Extend but also things like Persona/Cover) is not having a strong enough effect, turn up Audio Influence.
  2. If the "text" part of your prompt where you describe the music isn't doing enough, turn up Style Strength. (Consider turning down style strength if your Exclude style isn't strong enough relative to your positive style)
  3. Weirdness is harder to give simple advice for. Lower weirdness can sound more weird, it depends on the context. (Consider a classic text model that just continue yoru text the most probably - with a prompt that is nothing but the letter "aaaa" then low weirdness might be more likely to continue generating "aaa" but higher weirdness might kick the model into responding more normally like "ahh are you asking a question" instead. It's weirder in that it deviates from what came before, but less weird out of context. And there are some clear things like "very high weirdness has a lot of glitchy-sounds".
torn spoke
#

AAA I'm addicted to generating covers of the same verse+chorus of the same song

#

my credits are burning

untold briar
#

High weirdness is useful in short burst, sort of like pouring drain cleaner down a clogged drain. Sometimes you just want to force a change, but you probably don't want the whole song at super weirdness unless you are doing something glitch/experimental that doesn't sound like a typical song at all.

#

For high musical weirdness cranking style and audio often sounds plenty weird, but not "chirpy glitchy" weird.

plain pumice
#

Tried with 10/85/25, and I'm getting something much closer to what I'm expecting with regards to the source audio, but though the audio it's generating is much louder and doesn't adhere to the rhythmical pacing the previous section has.

torn spoke
#

Anyone here like melodic power-rock vocals :3

#

shamelessly looking for an excuse to upload a snippet

torn spoke
torn spoke
#

That's the chorus

river rampart
#

I am desperately trying to remove a small guitar solo from a segment about 4 4/4 beats long. I think suno just can't do it. At least not in the pro editor. The file is too big to import now so exporting and removing stems individually is not an option...

torn spoke
#

If you do, why not just split the clip you're wanting to upload in two, and then stitch them together in studio?

torn spoke
#

damn

#

not good for my wallet

river rampart
#

I don't have studio, just Pro.

Also no space to update, losing my job at the end of the month, so this is emotional catharsis more than anything.

torn spoke
#

or just upload it

#

how long is it? I can upload 8 minutes I belive, otherwise I'll have to split it up

river rampart
#

I took it down from 12:16 to 8.45 so far, would be less if I could get these transitions neat 😅

I was more looking for secret passkeys that people have discovered to force Suno, or at least v5, to drop the damn guitars. Half the time it doesn't seem to know whether its a guitar or a synth anyway. So much bleed...

torn spoke
#

Nah it's really bad at guitars.

#

good at drums and vocals

#

though

river rampart
#

I think the editor is just straight bugging now anyway. I am trying to generate 8 fixed beats, it keeps adding a 9th beat.

torn spoke
#

It's generative, I've never even tried to specify bpm or beats, cause I don't wanna waste my time

#

lol

graceful condor
#

Okay I painfully found out something which might help others. Issues with V5 are a lot. And especially Extends are a mess.

But I noticed using no Excludes at all (also main track) and leaving the sliders completely alone - not touching them at all is the best you can do. I wasted easily a 100 of extends only to notice that regardless how you tweak the sliders up and down it is a mess. And the only usable version I got are leaving the sliders as they are which means 50/50/25 and never touching them at all. And sadly that's not only an issue with V5 but also 4.5+ now considering extends.

torn spoke
pearl kernel
#

hello, i am experiencing an issue with v5, and if someone knows how to fix it please let me know, i have pro version, but i created with a v5 before upgrading and it was 1min long and i upgraded thinking it would get me the full song but it didnt, instead it shows that the song is 2:46 long now, but there is no sound after 1min and extending the song isnt helping, please help me out.

hollow hare
# pearl kernel hello, i am experiencing an issue with v5, and if someone knows how to fix it pl...

Huh, strange, I would have expected max time to still be 1 min.
What you can do is extend at the 1 min mark https://help.suno.com/en/articles/2409601
Alternatively, you can Reuse the style, make any changes you want and recreate the song

How do I make my song longer?

Do you wish your songs had a different ending? You can make songs longer by using the Extend feature. Best of all, this can extend your maximum song l...

neat cairn
#

Hi Team,

Am a user trying to make a song (v5), prompting - in a local language script, however, certain words are not pronounced as they are for real, tried to explore, experimented with some changes as well. I had done the same with 3.5 as well, but 3.5 worked for those words, 5 is not. Not sure, but any clue how this can be sorted? Am not sure if there's a diff, and if there is, probably might need to find an alternate way out, from the phonetic side, since the words / alphabets are frequent in usage.

Thanks!

torn spoke
#

what language

neat cairn
torn spoke
#

there are multiple ways of getting the ai to pronounce a word.
Daddy can be written as Dæh'dee for example

#

I'm unable to get the AI to sing norwegian in my local dialect (consistently), but it is able to sing the common-dialekt really well.

neat cairn
torn spoke
#

Non-english language dialekts are still a fair distance away

torn spoke
polar spade
neat cairn
#

But the bigger question for me, as a user is - does 5 have all capabilities/same behavior as 3.5, bcz as a user, I have a benchmark set, so the next model needs to be tested, so the output is consistent!

torn spoke
polar spade
torn spoke
#

Much, much better

#

All the models are in "beta" after all

neat cairn
#

Haha, it is better surely, and aligned on that. But in the longer run, consistency is what builds trust, else will have to keep experimenting all along. 😋 so I write, and then have be unsure of how the model reacts!

torn spoke
#

You will never get the perfect song. You'll have to generate, crop, extend and so forth.

#

None of the models are void of drift

#

The best way to reduce drift is to use a personality

polar spade
#

Split (crop out) the song. Add a 3.5 extension. Splice the remaining song back onto the tail of the extension. If you have Studio. It's link icon would be in the link menu on your left - - under Create.

neat cairn
#

Cool, let me do some experimenting, will get back on the thread. From a product side, maybe we could have a repo of pronunciation (like the way wikipedia does), that could help standardise the pronunciations. Of course, styles etc would vary, and even the tone/throw of the vocals, but pronunciation can possibly be standardised.

Thanks for the help!

graceful condor
#

@stuck junco I really do not know what it is but for me over the course of the last weeks all seems so random. There are days stuff works like a miracle first try and then you spend 2 days to fix one extend of a song. Then you finally thought V5 managed to get rid of those fast sung verses, without feeling and real flow just like to get it over with - this is how it sounds. And then it is back again. Suno feels like a schizophrenic girlfriend who is A on one day but B or C or D on the other or both at the same time. There is no consistency in literally nothing. Some stuff you thought finally broke the curse and back again and you start at zero. It is like a lottery or playing gacha games. Your rates are low and try if you get lucky but most of the time you aren't I am down at 7439 credits now and I wasted 3/4 of it for wasted garbage.

So:

  1. Misspelled lyrics, left out words or Gibberish (even worse with Personas most of the time)
  2. Feeling in the voice does not replace a flat melody or just flat structure
  3. Too many songs start with "Ahhh, uhhh or ohhh's" sometimes it is ok but not always
  4. Extends are garbage most of the time (best is to leave all sliders alone)
  5. Audio quality is not stable. Sometimes songs stark louder and get muffled and dull at the end or the opposite around
  6. To many weird artifacts. Weird clicks or sound effect which have no Value and in the end only produce an unclean mix.
  7. Voices are unstable. Sometimes start male, then female without asking or mix them together
  8. Cut off songs regardless of Prompt, Persona or Audio used
  9. Rushing songs is not gone. Still present
torn spoke
#

the higher quality the personality, the higher quality the covers/extends

#

Make personalities for individual parts of the song, crop the song the instant the quality lowers

#

Leaving things up to chance is literally only something I do in the very beginning of a project, to find a melody I like

#

@graceful condor Hell, sometimes I don't get more than a couple seconds of usable content

#

before I have to crop

neat cairn
#

@torn spoke yes, that is a solution, but as we go ahead, we need to figure out ways on what are the key things that need to be consistent across models. Of course, output quality will change/improve, but maybe test the newer model over the previous one to check variance in some key output params, think that might be an important param to establish trust with the users. If this is done, this will go big bcz users will trust it with the basics.

graceful condor
torn spoke
neat cairn
torn spoke
#

Hehehe, I didn't work on suno m8

torn spoke
#

From the samples that are good

#

If I generate something, and I really really like a 4 second section of it, I'll crop it and make a personality of it, and then use that to further experiment

ebon sand
torn spoke
#

For one of my songs, I made a personality specifically for the hook. Because if I didn't it was never able to reproduce it. I always had to crop the song before the hook, switch personalities and work on getting the hook right before I could switch back

torn spoke
graceful condor
torn spoke
#

Have you not worked with personalities before?

#

oh

#

persona's

#

they're called persona

#

woops

#

that'll save me a lot of time when writing that, damn

ebon sand
#

Personas do not work on v5 for me, it never sounds anything alike, v3.5 and V4 are the only models which replicate them well

graceful condor
torn spoke
neat cairn
torn spoke
#

setting the influence slider to at least 85%

#

and lower the style slider to like 0

#

that's a good starting point anyway

torn spoke
#

Can I ask how you use them?

ebon sand
graceful condor
torn spoke
ebon sand
neat cairn
torn spoke
ebon sand
torn spoke
#

The one time I used inspo it worked for me so idk what to tell you about that lol

ebon sand
#

It fails to 100% for me, so there's that, I never got a single usable output

#

But it makes an attempt to capture the vocals, the first 10 seconds are usually alright and then it sounds like a broken record

#

Persona, don't notice if it's there or not, it never replicates it in any way, v3.5 was the best model with personas and V4 also was still decent with it