📚┃suno-school

[night dirge]

This ai webpage could help you. If you put a song there, analyze a lot of things, including the words that describe that song in particular

https://suno-ai.notion.site/Using-Metatags-34944efe09ec41d693e314f13af44695

Would be amazing to have something like this but updated, like the last version

[static wigeon]

hey thanks bro a guys befor you, give me this, is the perfect link for me yes 😄

my software is Ultimate vocal remover UVR i dont remeber but idk if is the best method

[hasty crescent]

is there any way to upscale the sound quality? currently the quality is too low, seems are only 192kbps/s

[final hazel]

Hey rydensun, yep, Suno's current max is 192 kbps. For now, that's the ceiling, but I get how higher quality would be awesome.

[hasty crescent]

Could you please let me know if there are any plans to increase the audio quality to around 320kbps in the upcoming v4 or v5 updates?

[final hazel]

The team is pushing the limits to improve it. It's a complex task, so getting to higher quality like 320kbps is still in the research phase.

[hasty crescent]

Got it, thanks for the update. I might need to use my laptop speakers instead of the earphones for now..... Wishing the team all the best in achieving this breakthrough!

[neat nexus]

You mean something like this:
https://www.suno.wiki/faq/metatags/

[shy cosmos]

Hi, guys! Thank you for amazing product! Just registered yesterday evening and already burned 600 tokens experimenting. Quite often though i have this problem. I generated a song using my own lyrics. And it sounds perfectly, but there is one single word which just does not sound right. And if i change this word and regenerate - the sound is completely different. And i have to regenerate over and over again(last song took about 47 regenerations) until i get result that i like.
Are you planning to add seed to generation parameters? So it would be possible to get similar results.

[fresh hemlock]

i want to sing over the voices, is it possible to generate voices at a very low volume?

[final hazel]

Currently, Suno doesn't offer a feature to set a seed for generation parameters, which would allow you to obtain similar results across different generations. Every generation in Suno is unique due to its intrinsic randomness and doesn't rely on a fixed seed. This means each regeneration, even with minor changes, can produce a completely different outcome​​.

However, if you’re facing difficulties with specific sections of a song, Suno is developing a feature, somewhat akin to a "paint in" function, that would let users regenerate specific parts of a song while keeping the rest unchanged. This could help with issues like yours, where only a small segment needs alteration. But keep in mind, this feature is still under development and not yet available​​.

[shy cosmos]

thanks, looking forward to check it. thank you againg for the amazing service. having a lot of fun with it

[final hazel]

Directly generating voices at a lower volume isn't an option in Suno. However, you can use audio editing software to reduce the volume of the vocal tracks in your generated song, allowing you to sing over them at your preferred volume.

[fresh hemlock]

but it also break the general sound of the song because of the reverb of the voice and the autotune, and also choirs is difficult to reduce without break the sound

[final hazel]

I see your point, Manu. Managing the volume of vocals without affecting the overall sound quality, especially with effects like reverb and autotune, can be tricky. Suno is not perfect yet.. but it will be in the future

[fleet zephyr]

Sometimes SUNO stresses the wrong syllable in the lyrics, especially for non-english languages.

How to put accent on a specific syllable?

[final hazel]

To direct Suno to stress the correct syllable, try using phonetic spellings or punctuation. For example, spell out the word phonetically or use capital letters to emphasize the syllable you want stressed. It requires some trial and error, but these tweaks can guide Suno’s pronunciation more effectively.

[graceful condor]

It's saying I have generations in line

Bit i checked

And there aren't any

Help

[final hazel]

Hey Outsider! It sounds like you're facing a bit of a hiccup with queued generations in Suno.
Should be working again.. they are monitoring the severs atm.. it's busy

[night dirge]

YESSSS, thanksss

[willow wren]

You can use opensource DL Models like MT5 for stem separation so you can throw away sunos Voice

[past trellis]

question: does Suno accept input of flat and sharp symbols? is it ascii or html?
i.e.:
The Unicode character '♭' (U+266D) is the flat sign. Its HTML entity is ♭

for context, any prompt with 'blues' is almost always B♭ or E♭, want to get to A♭

[final hazel]

It will accept, but if it will work is the main question. Why don't you try it and let us know?

[upbeat depot]

Is there any plans to reveal more advanced parameters for users?

CFG, seed?

[final hazel]

You are comparing it with images seeds.. Suno works different

[upbeat depot]

There would certainly still be similar parameters

[final hazel]

so Suno never set seed
that's a common misconception from users that there is some deterministic way they can do their generation
deterministic way exists but not in production ... (not yet at least)

[neat nexus]

https://tenor.com/view/judy-garland-somewhere-over-the-rainbow-singing-gif-14312924

[pliant cave]

They have mentioned from time to time that they may expose some of the control knobs at some point.

By the way, I have to say I'm finding v3 incredible compared to v3 alpha. It seems more capable of handling a much wider range of styles and genres and seems to nail them on the first try. Extremely impressive. I'm not sure how they managed to do it unless they did a full retrain or something. It's great.

[untold briar]

Yeah, absolutely. People asking for a seed feature are usually looking for a way to make their generations more consistent when they want them to be. Which just happens to be a seed in some things but could be a feature implemented many ways.

[teal raft]

georg's mention of v4 inpainting will probably end up solving a lot of people's requests as well

Since if you have a song you like, inpaint over the vocals

the chorus, whatever

No seeds needed

[lusty violet]

are images generated by the model used in the music generation process?

I can imagine each set of users lyrics could be displayed as image data for the model to read it as a particular latent space which is associated with the traiing data

or something

[untold briar]

No. But that is a super fun idea!

[lusty violet]

I was thinking about it when I saw this fork on musiclm, meow something, let me see if I can find it

I thought it was called meowv2 but now I cant find it. anyways it was a video to music model with the concept based around creating music scores for videos/movie scenes/what have you

and I was thinking what if you reverse engineered it to just work with images

@untold briar oh hey I found it. https://arxiv.org/html/2305.06594v2

[untold briar]

No code... 🫤

[lusty violet]

dang

[lusty violet]

Do you have any advice for trying to build a music gen model ?

[untold briar]

Not really. Discord mods are just Discord mods, the Suno folks are the ones who built the music model. I did make a funny poetry only no prompt TTS model recently but that is just my amateur silly experiments.

I am interested in the idea of using images as an condition for audio models or even text models.

[lusty violet]

ah i see. yeah I think it would be super cool!

I thought I was on a github page for v2meow but i cant find it now

kinda mad about it

could have just been a readme

i was on my phone I forget lol

[untold briar]

I would probably start with retraining meta's musicgen but maybe using something newer than encodec, I think

[lusty violet]

I had one other question i was wondering if u knew if suno used spotify's data at all when tagging songs?

[untold briar]

Sorry no idea how it works.

[lusty violet]

haha ur good its mb

[lusty violet]

i wouldnt even know where to start so this is cool info thx

[indigo basalt]

I have two questions. #1 is does Suno understand the genre J-pop? And, if so, can it make a J-pop song with a prompt that is entirely in English? I just love the J-pop genre and thought it would be neat to try making a song in one of my favorite genres. Any and all help/information is appreciated! 🙂

[limpid spade]

Image to music? 🤯

[gilded mist]

With the next NVidia chips being AI monsters, I wonder which additional features Suno will introduce.
I know people most likely said it a thousand times already, but I'd even love to see a local version of Suno eventually.

[shy cosmos]

Local models do not bring money to it's creators. Doubt it.

[gilded mist]

Since Suno watermarks it's song, they could easily require to buy a license before handing out local versions.
But you're most likely right.

[shy cosmos]

Didn't know about watermark. This realy makes sense. Have no idea how to track your songs in world though. Unless it was played on some popular radio station, you have no chance to say that someone was listening to something.

[indigo basalt]

Well, if you post it on something like Spotify or YouTube you could probably track viewers/listeners.

[shy cosmos]

Good luck finding it from millions of other songs

[indigo basalt]

That is very true. If you would be able to make it big, that would be like a one in a million chance, but it doesn’t hurt trying. If anyone puts in the effort it might be found and get popular.

Also is there anyone able to solve my questions from earlier? If so, that would be great. I know people can get busy and I can wait forever, but just don’t want it to get buried. Any and all help is appreciated! 🙂

[tacit chasm]

When does the daily timer reset?

Like the timezone or something

Please ping me if someone knows the answer

[gilded hound]

1. yes
2. yes

[formal shoal]

/chirp

[flint haven]

The /chirp command is not available in this channel

Go to the #🐣┃chirp-beta-1 channel to use the /chirp command, and check out the #🐣┃how-to-chirp channel for step-by-step instructions for making music using Chirp on Discord.

[gilded hound]

adding on - jpop is influenced by english pop which makes it especially easy for english lyrics to work well

[indigo basalt]

@gilded hound So, I created my song, but it sang it in English, how do I make it sing it in Japanese? Or can it only do English singing from an english prompt?

[supple rapids]

Daily at <t:10800:t> for your time zone.

That's a tag that shows the time adjusted for your location.

It's 3:00 UTC

[gilded hound]

oh i thought you wanted to make it sing in english! if you want it to sing in japanese, you should translate the english to japanese via google translate or a real person and post the japanese text - unfortunately you cannot make it sing in japanese via an english lyric prompt

[hasty crescent]

Is there any way that I assign the gender of the song's voice?

[rose ore]

I just discovered something super interesting. Get this

If you're planning on using mastering tools like Ozone, use the "unprocessed" tag in your style box in the custom view. It'll give you vocals that sound rougher at first, but then when you process them through your DAW, they're freaking perfect. Noticeably better than the default.

makes me happy

Maybe something like: {C,E,Bm,A, female, unprocessed} rock upbeat, 90bpm,

[exotic fiber]

any difference on using {} instead of () pr [] ?

[shy cosmos]

to specify voice, i'm using "male fronted vocal" in the beginning of style description. works every time so far

[rose ore]

you could probably just get by with male.

[shy cosmos]

haven't tried {}. but there is a significant difference between [] and ().
() is usually treated as back vocal.
[] is more general hint.
back vocal will try to sing literally what you have in ()
[] will will try to set some sort of style from what you written in there.
From one of my last experiments, if you write
(sound of starting engine)
then you will hear it literally word by word from back vocal
but if you write
[sound of starting engine]
the main singer will sing 'rawr, rawr, rawr'

[shy cosmos]

maybe. i copied it from other peoples work. what have i tested is that male in the end of the style does not work that well. if you get unlucky and chirp will select female voice - it will try to make it more male-ish in random places.

[exotic fiber]

Thank you...!

[supple rapids]

A dev told us earlier that it is a viable move to repeat style prompts, & gender can be forced by prompting something like:
"male, male, male, male, male, male, male, male, male, male, male, male, male, cute waifu"

it should still be a dude then lol

[exotic fiber]

where can we find all devs tips?

[supple rapids]

It's like underlining it on a post it & slapping it on Suno's face.

[supple rapids]

You gotta Discord with us, we will guide ya

It is not well documented right now

they are actively preparing better documents and ways to communicate

they are 12 people serving about 20m visits to Discord & web combined, daily....

[shy cosmos]

yep, but i 90% sure that you will loose all the style information with all these repetitions. you can get a country song instead of death metal with 'male, male... death metal'

[supple rapids]

They're also largely the ML Engineers, & not really the community management & user relations guys, haha, but they've all been so very nice & friendly when they have a chance to speak

[supple rapids]

So I believe you aren't using descriptive & effective styles......

Like what are some of your style prompts?

[rose ore]

makes sense, none of the bases are that good by themselves

[shy cosmos]

it was just a theory, based on LLM and StableDifusion experience. that's why i said 90%))) i understand that this experience is not always relevant to suno.
from my latest test groovy emotional rock, alternative rock, male voice gave me a combination of male and female voices. so this makes me think that the later style appears in the list, the lower weight it has

[supple rapids]

You should use emotional language & specific music terminology. T-Pain isn't "rap", T-Pain is "confident male, southern hip hop, chopped & screwed, club rap, house"

My big big big tip is ask ChatGPT or Gemini to describe an artist you like

ask for subgenres & what specific musical terms are

[shy cosmos]

maybe you could share a prompt?

[supple rapids]

ok for that prompt i would say maybe add more emotional language to the male part, and repeat it if needed

[supple rapids]

sure thing! i share tons!

wanna see brainstorming chats with gemini?

https://everynoise.com/ is also worth a look

[rapid bluff]

I think you may have somewhat solved the problem!, doing the opposite of this, littering the word processed all over the style and in the tag almost nearly removed the effect all together

[shy cosmos]

sure! love this stuff. though i more gpt and mistral kind of fan)

[shy cosmos]

wow! may be usefull. will try it right now)

[supple rapids]

ok, wanna do a deep dive on a track i did earlier?

hm, do i have permissions to make a thread anywhere????

[shy cosmos]

sure. maybe a call?

[supple rapids]

hey, @dim glacier ? i see you online.... would it be possible, if you have perms, to just make a public thread so i can have a long explination?

[supple rapids]

i want it to be public for others too! ❤️

[shy cosmos]

thear will get lost in messages the same way as any other message

[supple rapids]

you people don't have any clue how fast & powerful Discord search is, do you....

Eh, I'll just do a long bunch of shit, I guess. nbd

anyways

so.

earlier someone was asking for ideas to get a bass drop in this song:

https://app.suno.ai/song/e440d229-ff07-4589-8511-fd7722c7aa91/

so this is their prompt:

[dim glacier]

I don't think I have the permissions, sorry.
A long explanation of what though?

[supple rapids]

which continued this one:

[supple rapids]

i'm just giving him an idea of how i approach adjusting a song with style words

so you know thaty you can continue and then change the style?

right?

[dim glacier]

Right. Sometimes. It doesn't always work though

[supple rapids]

so 90's rave jungle drum n bass is being interpreted as one thing

[exotic fiber]

Something to notice, the App is detecting Latino in the style as an Artist.. So we cant use something like trap latino because it gets flagged :p

[supple rapids]

i follow with this, & tag a key change, and use an EMOTION to header a verse

https://app.suno.ai/song/3208f170-7ffd-4dbb-b871-7faf930b28d9

she's not just female, she's powerful, and singing loudly

[frozen umbra]

Sadly, it doesn't work for me. Still washed out harmonies with tons of grainy reverb.

[supple rapids]

they are aware of that; it is flagging a lot of Spanish language stuff for copyright. fixing as soon as they can. 👍

[rose ore]

no kidding?

[supple rapids]

i sometimes add "remastered"

[rose ore]

that one does not sound good at the moment for me

[rapid bluff]

yeah, its definitely improved: https://app.suno.ai/song/23a60f1e-478c-4a8d-9ceb-f63903692c12

[supple rapids]

hm, personally, i haven't noticed any quality issues....

[rapid bluff]

Ive gotten this normally: https://app.suno.ai/song/aef93a1e-e084-4373-8cc6-3c36667eeaac

[supple rapids]

i imagine it is heavily related to genres we craft?

oh shit, yeah, i hear it!!!!!!!!

[rapid bluff]

its heavily genre related

[supple rapids]

those are night & day!

That's notable. they should possibly ask the people with quality issues to share the genres they have issues with?

[rapid bluff]

Ive been at a standstill, no matter what i did I couldnt shake the grainyness

[supple rapids]

oh woooooooooooooooooow the second one degrades though the song soooooo much!

do you do rap hip hop a lott?

[rapid bluff]

its all i do

[supple rapids]

yeah. ok. let's find out if hip hop & rap got some shit quality training data.

[shy cosmos]

wow! new keywords) i'm not a musician, don't know all of those. maybe you can help me out with my linkin park style song? so male, rap metal, post-grunge, nu metal, alternative metal, rock style works perfectly. but i wang the chorus of the song to be in rap style.
addin [Chorus. Rap] before chorus does not help. any hints for that?

[rapid bluff]

my entire streaming platform is almost 60 songs of dark comedy hip hop

V3 Alpha was perfect, minus the bugs

V3 Alpha:
https://app.suno.ai/song/ee29f448-8164-4009-bf9e-b3204ebdd2af

V3:
https://app.suno.ai/song/35facc3e-91b5-434b-a3d6-89fa4a74ba31

[supple rapids]

so I'd say continue the song at that point, change the damned style to "rap, male, etc.", & head the chorus with as well with [rap chorus]

you might wanna make the style "hip hop, rap, male rapper, west coast rap" or whatever

[supple rapids]

like say rap repeatedly lol

[supple rapids]

this set is less obviously worse for me

it degrades a lot for suuuuure though

ugh, actually, i'm half way...... it isn't gonna hold up much longer......

yeah. problems.

@shy cosmos check out the power of emotional languge in this one:

https://app.suno.ai/song/cb504129-498f-440b-ba69-6b2af8372990

[rapid bluff]

yeah and im sure you know once you continue with a grainy voice, that grain only amplifies tenfold

[shy cosmos]

hear no difference except for [Yearning]

[supple rapids]

oh, for suuuure!

almost none of those worked, but it took a little of each of them

[frozen umbra]

All the tricks didn't work for me

https://app.suno.ai/song/cddf61d0-d9b8-4ef4-a313-f3b3a639f4c2/

[supple rapids]

i think i got all of em, this is all the parts after

i actually didn't even change the style on this one

from each part

final track of that:
https://app.suno.ai/song/666a13fb-745c-4c62-bc2c-d6661a0c159b

[shy cosmos]

and i though you MUST have [Verse 1], [Verse 2] and [Chorus] in your song. Songs without it sounding like some one in dying(Elegy for Jon is a proof).

[supple rapids]

yeah i totally got rid of that bullshit.

[supple rapids]

rofl i have more structured ones

he still does a chorus though!

these are also me still learning, to be clear.

this is just like a pile of raw data. not everything in all of these is gonna work. but some of it is on the right track

[frozen umbra]

I can hear it throughout all genres, it's just, in some it's more annoying than in others.

[supple rapids]

i wish we had user profiles already so i could just like click you and check your tracks haha

what genres?

[shy cosmos]

also think this would be a good idea. maybe not user profiles, but public playlists

[supple rapids]

happening. soon ™️

[shy cosmos]

@supple rapids [Rap chorus] did amazing thing. no rap did not appear in the song, but the tempo changed, making even more stronger accent on chorus

[supple rapids]

So, also, idk your process, but here is a current conversation with Gemini, writing a song:

https://g.co/gemini/share/c721d062f08e

It isn't just asking for lyrics. It's asking for adjustments & collaborating.

[shy cosmos]

hm... quite interesting) thanks)

[supple rapids]

https://g.co/gemini/share/1d2eec1e3bc5

i got some good stuff just there

i think i finished that half of it. now i move to suno

ok, i will stop after this, but i think this is really powerful to demonstrate the need for emotion in your styles:

here is Gemini's lyrics before & after this prompt:

"The words should be angrier! It isn't his fault!"

Such a simple thing. It changes a lot.

[shy cosmos]

@supple rapids thanks for everynoice! the results are SO MUCH better now!!!

[bright vector]

hi

anyone pls let me know how to create songs with lyrics of indonesian or spanish

[misty dust]

Is it just me or are the lengths of generated clips inconsistent?

[broken lotus]

use Custom mode.
Lyric prompt and Style prompt are separate.
you can add your own lyrics.

[bright vector]

can you pls type one example for it

like for a indonesian lyrics

[broken lotus]

I use chatGPT to get lyrics in other languages, then paste them into Suno Lyric Prompt.

[bright vector]

so if we generate lyrics in spanish,then we wll come here and use /chrip ,then it will generate songs in same spanish language right

am I correct

[broken lotus]

I suggest using the website...
app.suno.ai
generations are longer and higher quality.

Here I used ChatGPT to write a song in French (I do not speak French)

I copied from chatGPT, and paste into Suno Lyrics Prompt.

it generates a song, sung in French.

[bright vector]

ok thanks,so May i know what genre we can give for kids nursery to sing like kids channel

[broken lotus]

maybe...
'children's sing-along' 'toddler nursery rhyme'

or combine with pop genre for more modern sound:
'synthpop sing-along' 'children's K-pop'

something like that...

[bright vector]

ok thanks

for your help

[thin flint]

What's Everynoice?

[south palm]

有汉语吗

[thin flint]

I like everynoise, but how does that help us with style? I go there to find new musicp; I guess it helps with new genre keywords to try.

Is there anything V2 is good at over V3? Not sure why a model selection is a thing, just make V3 be the default. I mean it is, but phase out V2.

[pliant cave]

Anyone who has used GPT-type models knows there is always something the previous model can do better than the new model. Giving people more options on how they want to spend their credits rather than fewer is a good thing. If they phase it out, it will probably have a depreciation date, I would assume.

[broken lotus]

I haven't heard anyone say they were sticking with v2, tbh.
I thought I would..., but I never select it...

[thin flint]

Curious if the new V3 official sticks to style prompts better, E.G. for making style medleys that @rose ore started in the V3 alpha days. I like seeing what styles it can transition into. I made a 15 minute or so song intended for stream breaks, but the longer the song is, the more artifacty and noisey the sampling gets sadly.

[rose ore]

At the moment, compositional decay is a lot heavier than it usually is. So far on the v3 final, compositional decay starts to hit around 2ish minutes? You can ride that though. Make it interesting

Also, I have eras? lol

[thin flint]

Lol. Hard to believe that v3 alpha only really existed for a month.

[rose ore]

What's funny, is that you're right. AI time is so weird. I'm working on my fifth album using a technology that didn't even exist this time last year. This is all crazy

[thin flint]

What are you adding to your album to make it your own? Or are you just publishing the Suno tracks and calling it an album? Something that I don't think is a bad thing, I'd just like to reduce the artifacting. I wish they had something in place like Boomy where you can just publish right from that UI and they handle the logistics from there, but I only see one issue with their method: you don't own your stuff.

[rose ore]

I'm using it to stretch my song writing skills, mainly. I'm also in very tight control of my timing and chord progressions these days. And the way I work, I'm either playing to with or against the performance of the machine. Think of it like using a very very smart paino.

So pretty much everything I do is bendy in some way

Makes everything more engaging I think

Lately, I've been experimenting with jazz and country composites

[lilac harness]

Is there a list of all chirp tags? I found you can use [man] to have a male singer but I assume there are more possibilities to maybe change the instruments for the chorus or select specific voices?

[shy cosmos]

here isa example: i know that everyone saying that Linking Park is nu metal. Goggle says that it's genre is alternative rock. With both of those style prompts i could not get something similar to Linking Park. But everynoise gave me rap metal, post-grunge, nu metal, alternative metal, rock. And it sounds VERY similar(though voice is different, of cource).

[shy cosmos]

from my experience this depends on genre you are using. for example, if it's drum and bass - you can use [Beat Up] and [Drop] to speed up and slow down the rythm. They work perfectly here. But for something like rock they just create a bunch of noise or hallucinations.

[tight zephyr]

It works fine in standard mode prompting the language: "A lounge song in Spanish about the beach"

[gilded mist]

I know they said they are thinking about doing it in the future, but is there a rough ETA or any news on if and when they would expose more knobs for us to tweek? So far it feels very random. The same Lyrics and Styles Prompted 5x will yield a huge variance in results.

[fallow gull]

I would like if Suno gave some insight into the tags it used to train its model, but it may give away some if its proprietary process.

I have looked at Cyanite which is a different audio classifier and has an API. It is interesting to see its genres, instruments and mood classifiers for inspiration.
https://api-docs.cyanite.ai/docs/audio-analysis-v7-classifier/

But Suno seems to understand much more. With v3 they managed to fill huge gaps in their model so it understands more "world" and "folk" sub genres.

[kindred minnow]

i would liek to know if Suno will get better in audio quality in further version.. also maybe the possibility to get wav format...I noticed that often the part 1 with v3 sound great.. when you get to part 2 its still good but often in part 3 there is always some fuzz and noise that prevent to hear clearly the vocals,, often look like the 3rd part is more lof-fi even if i try to render 40 version all the version have this noisy low quality fells to them.. witch prevent to have a high quality song at the end.. Even the songs that are in the most popular on the website have this problem too

basically we can all tell its ai generated due to audio quality thats is not the best all the time

even using tags like studio quality or hi-fi or mastered

[gilded mist]

I also wonder if they will allow us longer Lyrical prompts.
With fast Genres like Nightcore/VOCALOID etc. I can completely fill out the prompt but the songs only generate to ~1m 30s. I could easily squeeze in another Verse in there with how fast it sings.

Or even negative prompts.

[kindred minnow]

it sometimes get to 2 minutes.. but than the subsequent continuation are on 1 minute or less.. but when you do a continue option you can add new lyrics to continue

i usually do (intro) (verse1) (pre chorus) (chorus). i choose the part i like for second part verse 2 pre chorus 2 chorus

im already ready to test V4 alpha

[gilded mist]

yeah but as you said, Part 2 already has diminished Voice Quality, so I'd love to get as much Lyrics in the Original part.

[pliant cave]

V3 genre adherence is really incredible. I'm not sure what they did compared to V3a, but there's a noticeable improvement in what it can do now.

[kindred minnow]

v3 is great overall.. anyway its the best outhere so.. even if not perfect still the best Music ai generation... there is other like soundfull, butthose you need to select a style and there is no vocals, no lyrics is just random

[pliant cave]

I believe the problem may stem from the way transformer-based models work, in that they essentially presume the input context to be accurate and correct, and build on it. So, if there's a flaw of some noise, the model will run with it as if it's meant to be there. I'm not sure how they could address this issue.

[pliant cave]

Riffusion is the closest, they do vocals as well, but they have been stagnant for a whole year, offering only 10-second clips, so it's more of a novelty. I don't consider Suno to even have any competitors at this point.

[kindred minnow]

well, the training data base should only use very high quality studio master on eveyr style

sometime sits not even noise, its like some residual reverb that overpower everything else

[pliant cave]

Yeah, that kind of thing is hard to come back from.

[quiet mountain]

How do we do a complete song on Suno with an ending? Will it always just cut off?

[kindred minnow]

i can fix some of the problem with using mastering software or use Emastered

[pliant cave]

v3 is essentially at a point where the average person wouldn't be able to distinguish it from a real production on FM radio, provided it's a good generation.

[quiet mountain]

I agree - I just want an entire song

[kindred minnow]

also

[thin flint]

Google has this Dream Track thing that the Rolling Stone article talked about, but I don't know how good it is.

[pliant cave]

Do a continuation, and then in the lyrics box, put [outro]

Try that

[pliant cave]

It could be the model they are using for youtube creators

[kindred minnow]

whats often the problem is how the lyrics flow on the song.. some words sometimes dont have the best flow... to ensure the best flow when you generate lyrics using chat gpt or gimini ask to write it in the similar stuckture of greatest hits.. and i also found another coold trick.. becaus e when you ask for a song.. i feel that lyrics are often generic even with chat gpt.. example a synthwave song will always most likely ahve the words neon lights... streets lights... or stuff generic like that.. so my trick i ask chat gpt to create asmall story about something example i asked to tell me a short story about a ai female robot that become counscious of the world... thans i ask it to take that story and convert into a song and too chose each words with the most impact because a song is shored than a story

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

still the ending sound weird becaise of noise

[pliant cave]

Ah yes, the infamous "neon lights" in every song.

[kindred minnow]

but all words are not that generic and flow well

[quiet mountain]

I send people songs and you'd think they'd be more touched -but I think they think it's easy with AI. I wrote it and picked out the genre and did tons of retries to get the right song. It's an odd response. Still trying to find best use for this. Any suggestions besides a youtube channel?

[kindred minnow]

dont forget to ask chat gpt to put back vocals... and onomatopé netween () in ket points too

[pliant cave]

One thing Suno could "open source" is their lyrics generation prompt for OpenAI, allowing the community to improve it. I have an idea of what the prompt is, but not exactly.

[pliant cave]

Anne, I've been in the AI industry long enough to know that it doesn't matter if it's art, music, or text; if you even so much as mention "AI," people immediately have a negative or dismissive bias against it. Or, like you said, they think it's easy—it's not. A ton of work sometimes goes into making songs, not to mention the incredible work that has gone into the models themselves.

The key is to never mention AI. Just say your crazy Uncle Larry and his musician friend made it for you, or that you hired one of those "song on demand" services to record it for you. Yes, it's not the truth, but you will see the difference in people's reactions to how "great" it is.

[kindred minnow]

with this one i managed to keep the quality for 4 parts i dont know how

https://www.youtube.com/watch?v=jnlMTu2s-CA

[quiet mountain]

@pliant cave I just hate that though. I am such an advocate for AI so I'm open about it - but you are right. Do you or anyone know a good use case for these besides a youtube channel? Seems to have opened a world of possibilities.l

[kindred minnow]

yes.. im doing 3d movies with blender so i can get unique music for my movie

[pliant cave]

It's only limited by your imagination in what it can be used for; there are many ways you or even maybe Suno haven't thought of yet.

[quiet mountain]

Imma ask William - my chat. LOL

[kindred minnow]

you do publicity you can get a jingle for your commercial

[pliant cave]

That's a good use case. It's great at Jingles.

[kindred minnow]

there is so much ways youc an use music lol

[thin flint]

Hmm. I can't get the create page to load now.

[kindred minnow]

suno got shot down by the industry!!! lol

the page not available

hehehe

[pliant cave]

Look how effective the "JG Wentworth" jingle is; everyone knows that one.

[kindred minnow]

probably server overload

[thin flint]

I got a Vercel security checkpoint and it's not redirecting.

[kindred minnow]

Also, next i still try to get the best 3:30 soounding song i can, when i get it i will try to do a Ai video clip, using midjourney for generating image animated with runway gen2

i try to log in to see if i can access the create page

[thin flint]

It's stuck on loading...can't access library either.

[kindred minnow]

create still work for me

[quiet mountain]

Works for me too

[pliant cave]

Looks like it was a mini outage

[final hazel]

https://tenor.com/view/bruce-almighty-comedy-jim-carrey-frustrated-come-on-gif-3406367

[static wigeon]

hey guys, can i bother you to ask some idea of prompt to have this type of drop 😱🤯🥺✨✨✨ if you have some prompt i takin all x) (at 0:19)

[pliant cave]

I see Tony is on; maybe they are looking into what happened with the servers.

[quiet mountain]

what genre do you all thing is the best sounding on Suno? Do you have a most realistic favorite?

[wooden stag]

Suno excels in EDM/Electronic/dance music. Does great with Pop and also metal. R&B can be great. It does poorly with acoustic singer-songwriter songs. Rap is hit and miss. Blues timing and scales are pretty good. I find it impossible to get a truly gritty and soulful voice. 50's, Americana Roots, and Funk are pretty impressive. Live performance sounds are also outstanding. I haven't tried it yet, but I bet you can get some really powerful gospel songs with using a blend of live performance, organ, uplifting, and soul.

[quiet mountain]

Oh wow - did ya'll know you can get your own custom vinyl pressed with these? What a cool bonus and idea as a gift!

[gilded mist]

I would love to have the ability to generate Lyrics to a song.
I. E. generate an Instrumental song, and then have a button like "Extend Song" that's called "Generate Voices" to be able to input the Lyrics in there and it generates the voices for the song. Currently, I sometimes have an awesome track but the AI just doesn't sing like I want it to or it gets toned out in noise and the next batch of songs already plays out very differently.

If we'd be able to generate Track and THEN the Voices, that be amazing. If we would be able to download the lyrics only, we could use another AI and replace the vocals if we like the text-flow but not the voice in itself.

[rose ore]

Okay

I've solved it

When it starts over-harmonizing, I switch to v2, render a small section, and then continue with v3 which resets it

[gilded mist]

Just a question. How long did it take from V2 release to V3 release? (The Full release, not the Alpha)

[pliant cave]

Approximately 5 months ago, v2 came out in November with a soft launch where only those who were observant happened to notice it was the new model on the web app, lol.

[kindred minnow]

should be great that suni give a.wav

i after used @mastered to remaster the track but i dont know iif its sound better

[gilded mist]

I see. I'd assume roughly the same time till V4. I just found out this website exists on Friday and I already cannot wait the ability to generate 4m long songs (if that is in V4). 2 Minutes is already extremely awesome, but if we'd be able to single-prompt full songs from start to finish OR get rid of the issues the system has now with continuing songs, I don't know if I would have enough money to put into this with how quickly I already now burn through credits.

[pliant cave]

Well, I've been here since the first day or two the Discord opened, and the progress has been exponential since they clued into the musical capabilities of Bark and focused on that. So this is the best it's been; you showed up at a good time. My guess is they roughly already know when V4 will be tested since I'm pretty sure they plan their training runs well in advance and have to rent cluster time, etc. I'd say focus on prompting for now and make the best of what you can.

[gilded mist]

Oh definately! I'm here to stay! I am already more than amazed on the leaps AI in general has made in the past years. Cannot start to think where we're going once NVidia starts to tap out the new GPUs and sell them to datacenters which companies train on. I think they talked about an up to 30x increase. That is just insane if true and applicable.

[quiet mountain]

guys I just love this app - had to say - I've been working on what I'm going to use it for and ideas keep coming to my mind!

[charred lake]

Took forever but I have mastered and uploaded my 27 min Suno house / dubstep / grime / dark wave song. Hope you love!! https://on.soundcloud.com/ZmmCtZdqWxtLEwYb8

[flat field]

I only wish they can provide us ability to generate songs with seeds so we can keep the same 'singer' and 'bands'.

[pliant cave]

It's being worked on, or at least they intend to. In the meantime, I use the 'continue from anywhere' feature to recreate the same effect/vocals/band with a new topic and new lyrics.

[lyric sonnet]

I notice in some suno songs that certain frequencies seem to reverberate overly loud. I can correct that with a notch filter, but is that a known issue?

[untold briar]

There's been some feedback along these lines, yes. And a long standing issue where sometimes errors or degradation kind of feedback and get worse over time, a problem common to most LLM models. (But better than it used to be already.)

[ashen moon]

Hi all, I have a question.

I got a 2-minute, partial completed song (let's call it part 1), I like the partial song but only from start to 1:26.

Would it be possible to use the 00:00-01:26 of part one, then continue to generate the part 2? Thanks.

[limpid spade]

Use "continue from this song" using the three dots menu... Under Title on the left you will see a timer to change to 01:26

[ashen moon]

I tried. Just not sure that the performance is not good or it does not work this way

[limpid spade]

This will generate 2 x Part 2 songs. If you like one of those you can Continue Part 3 etc. Once happy, click the same menu on the final Part and Get Whole Song which costs 0 credits

Are you in Custom Mode or simple mode? You need to switch to Custom Mode to Continue

[ashen moon]

I am using custom mode, when I "continue from" 01:26 the Part 2 does not sing my additional lyrics

[limpid spade]

Also. Update the lyrics to start Part 2 from whatever is at 1:26 otherwise it will sing verse 1 etc

See above. Clear your lyrics and only put in lyrics you need from 1:26 onwards

[untold briar]

Just put in the new lyrics, whatever is not actually sung in the audio of Part 1, start there. If it still doesn't sing you may need to backup earlier than 1:26, could be some random bad luck about that time.

[ashen moon]

I am guessing if it wants to deliver some better melody, but my lyrics may not fit. So I will try continuing from other timestamp. Thanks!

[last vale]

/[Verse]
夜深人静心不眠
梦里追光逐影转
泪落成诗，心向明天
风雨过后见彩虹

[Verse 2]
挑战来临勇敢面
坎坷路上更显英勇
每一次跌，每一次爬
铸就今日不屈的我

[Chorus]
别怕失去，别怕痛
每一道伤，都是勋章
笑对人生，笑对风
你我终将风华再现

[kindred minnow]

i noticed that the most prohiminent problem with suno to be solved will be the phase cancellation of similar sounds frequencies, and try to separate them more.. Specialy with powerfull song with vocals and also the calibration of effects.. example whenthere is reverb and echo on a vocal its seems to sometimes create not a feedback but a lopping of the reverb that cancell either they synths or the vocals cleanest

sometimes its even seems to create somekind of a crouwd sounds in the back that resonate

[flat field]

I wish if v4 going able to generate entire song instead of 2 minutes max. I keep finding that the continuation song tend to be different as if the AI forget

[rapid bluff]

try to pull the full song after every continuation and continue from the full song

[flat field]

It cut off mid-chorus and i have burning tokens for nothing since it still won't pick up the chorus and continue from it. I had even tried returning several seconds behind and even if it resumed the 'chorus' tune is markedly different. I wish we were able to 'lock' the seed.

[rapid bluff]

are you creating the first part of your song still? or have you already started doing continuations?

[pastel shard]

hey guys ,is it possible to gen only one song at a time

[rapid bluff]

no its not, sometimes wish it was

[flat field]

The continuation.

[rapid bluff]

seems one of the songs always gets the short end of the stick

[flat field]

V4 perhaps. Hopefully so.

[rapid bluff]

continuations only last 1 minute after the initial song

[flat field]

Yes I know, I complained that the continuation didn't pick up the chorus properly

[rapid bluff]

can you share the link of the song you are continuing from?

[flat field]

Wait a moment

https://app.suno.ai/song/b38d14ff-7607-4296-9d48-9a114e43adda
First part

[rapid bluff]

ok i see where you could have your issue, you should structure the song to end on the chorus for each part

[flat field]

Hmm... I guess going to cut midway

[rapid bluff]

since you already have a chorus, I would start the verse directly after the chorus as a continuation

[serene creek]

Is there a good way to control tempo of the music?

[night pelican]

Wondering if there is a good way to invoke a British accent

[flat field]

Well, I tried to add BPM in the title. So it might help somewhat.

[flat field]

You can add 'Britain' in the style. That's how I made this Arabic Latin abomination
https://app.suno.ai/song/bd23573b-f8f5-4ff9-82ef-607d90e6d540

[exotic fiber]

try with

[pre-chorus, build up]

ur pre chorus (try to make it just 2 lines)

[Chorus, drop] or [drop]

ur chorus (4 or 8 lines)

Usually this is the structure i use to write my lyrics:

[INTRO]
[VERSE]
[PRE-CHORUS, Build up]
[CHORUS, Drop]/[HOOK]
[BRIDGE]/[INTERLUDE]
[PRE-CHORUS, Build up]
[CHORUS, Drop]/[HOOK]
[BRIDGE]/[INTERLUDE]
[OUTRO]
[END SONG]/[END]

Besides () and ("Instrument" solo) or (no kick) etc etc

What other things v3 accepts?

[clever karma]

/CHIRP

[flint haven]

The /chirp command is not available in this channel

Go to the #🐣┃chirp-beta-1 channel to use the /chirp command, and check out the #🐣┃how-to-chirp channel for step-by-step instructions for making music using Chirp on Discord.

[smoky pine]

Hello! I tested two songs already, and mentioned "Mozart, chorus, four-part, cello, flute, hymn, classical, strings, trombone, harp, bass drum," but noticed that cello, flute, strings, and drums are not included. Additionally, the style feels very modern rather than classical.
I'm curious, if I subscribe, can I access a more extensive sound library?

[limpid spade]

BPM is not recognised. You would have better luck with slow tempo, mid tempo, fast tempo or musical terms for tempo like adagio, allegretto, vivace

[limpid spade]

I believe the training data applies exactly the same for free users as Subscribers. Adjusting prompts is the best way to get closer to what you want, not all tags will be used in your generation but they might in the next one. Ai music generation is not a perfect science.

[quasi bridge]

Can I use suno to generate natural sounds? Such as cicadas, waves, etc., I always feel the official music melody

[untold briar]

sometimes but usually as part of song. For waves try "field recording" and maybe "ocean" or "waves" if I had to guess (not tested)

[dusky sphinx]

I have tried the music style of cinematic (elsewhere with tthe bot not here), it made a Metal style. Is Cinematic not an option?

[limpid spade]

I've made plenty of cinematic songs with V3. You can even specify things like legato cello, staccato strings, trumpet fanfare etc to try and direct Suno closer to what you want.

[dusky sphinx]

Ok so I need to try and be more specific? I only used "cinematic" but I assume I should include instrumental, epic and so on ?

The website is better for my options I think.

[untold briar]

Each version reacts different to different prompts. The discord bot is V1. The website is V2 or V3. Some style words they will have in common but some words might be very different. In V3 especially some words have almost no effect and other words very strong, and it is hard to predict. words like "cinematic, epic, dramatic, musical" probably work in all 3 versions though. You can mix V2 and V3 clips in the same song and this can be a useful workflow .

Oh if a word has some effect but is not very strong you can repeat it in the prompt to make it stronger.

[dusky sphinx]

Ok thank you. The website has the option of instrumental which is great but when I used the words "I am vengeance! I am the night! I am Batman!" it sung it in heavy metal style, I did but cinematic as the style though.

Not to worry, I have created something with v3 which is closer to what I was after.

[digital talon]

You may find it useful to try and breakdown down the style to create a more detailed prompt. I've found ChatGPT and Google Bard handy for that.

[dusky sphinx]

This is part one: click and part 2 here which sounds quite different despite using the same keywords that it auto gave for part one in the "style of music" on the left hand side of the webpage. Its less of a continuation and more like a new track entirely.

[real turret]

Hello guys! Do you know how to make a song longer than 2 min?

[broken lotus]

Cinematic is not a genre, it is a style description
[cinematic ambient]

What you probably want is called ‘orchestral’
[cinematic orchestral villain theme]

[final hazel]

click on the 3 dots menu, and choose "continue from this song".

[real turret]

Get! Thanks

fyi: click on the 3 dots of the second part and click on "Get Whole Song". You'll find a full song in your library

[graceful condor]

I was about to ask. Thanks man!

How can I control the lengths of a 2nd part? Even I am adding plenty of text, there are only a few seconds added. What am I missing?

[final hazel]

I think the AI feels like your song is ending then? If the first part has some silence or fading out, it does that. Try using the timestamp to continue a few seconds earlier.

[static wigeon]

thank a lot, i gonna try this, what is Build up ?

[vast ledge]

Does the continuation ever get to be 2 minutes long as well or we are stuck at a minute each so if we want a 5 minute song, we have to use up 50 credits a day or pay for premium?

[final hazel]

continuations are 1 minutes max...
As a free users you get 50 credits per day
If you pay you can choose between 2 plans: pro (2500 credits) and premium (10.000 credits) which can be used the whole month

[pliant cave]

I went back and used v1 after not having touched it since v2 came out. Now it feels like a novelty compared to v3. Back at the time, I thought it was a modern miracle (it is), but it's amazing how perspectives change.

Technical observation, since this is the technical channel:

I've noticed, and this is an observation not a complaint, that the time from submitting a song generation request to waiting in the generation queue to the start of streaming has increased significantly since last week. My account is on the priority queue plan, but I assume this is due to an influx of new users and increased load? It used to be a matter of seconds. Again, not a criticism, I'm grateful for the streaming option instead of waiting for a full 2-minute generation, but just noting what I'm seeing as a user.

[exotic fiber]

Intro: The introduction sets the mood and tone of the song. It's usually instrumental and occurs at the beginning of the song before the vocals come in.

Verse: The verse is where the story or message of the song is typically conveyed. It usually consists of the same or similar melody and chord progression each time it's repeated, though the lyrics may change.

Chorus: The chorus is the catchiest part of the song, often containing the main message or hook. It typically has a different melody and chord progression from the verse and is repeated several times throughout the song.

Bridge: The bridge provides contrast to the verse and chorus. It often has different chords, melody, and lyrics from the rest of the song, serving as a transitional section that leads back into the chorus or next verse.

Pre-chorus: Not present in all songs, the pre-chorus acts as a buildup to the chorus. It typically appears before the chorus and serves to heighten anticipation for the main hook.

Buildup: The buildup is a section of the song where tension and excitement gradually increase, usually leading up to a climactic moment such as the drop in electronic dance music (EDM) or the final chorus in pop music. It often features rising intensity in instrumentation, dynamics, and/or rhythm.

Drop: Commonly found in EDM, the drop is the climax of the song where the energy peaks and the beat kicks in at its fullest. It follows the buildup and is characterized by a sudden release of tension, often accompanied by a powerful melody or rhythm.

Outro: The outro occurs at the end of the song and serves as a conclusion or a way to fade out the music. It may feature elements from earlier sections of the song or introduce new elements for a final resolution.

[static wigeon]

big thank to you for this,I understand better

[stark sky]

I wonder if there is already some list of tips for successful continuation? For example it seems that I get more lucky when I set the time rather into the middle of a phrase/verse rather then trying to put it into a gap between verses. Also I still haven't really figured out if I should leave some previous parts of the lyrics in there (even though they are already in the part to be continued) or if it is better to remove them from the continue-prompt.

[tepid wedge]

soooo... did the devs delete the ability to download chirps?

[grim nexus]

I was trying to generate a song with several names of my daughter friends from school (in Czech) but it is flagged for moderatiom - why is that?

[ivory needle]

How to change lyrics of generated song without changing the mood, instruments and tempo overall? I tried via continue (part 2), but getting silenced/bugged song. Remix option changes everything. I need to fix 1-2 rows

[exotic fiber]

maybe one of the names is also an artist name... they r trying to fix this.. also happens with, for example, Latino on the style.. it gets flagged cause theres an artist called Latino...

[exotic fiber]

sometimes, changing the lyrics a little bit works

[broken lotus]

some tips for successful continuation..., but you are correct: it is more an 'art' than a science
https://www.suno.wiki/faq/making-music/my-words-are-wrong-after-continue/

[grim nexus]

I see… thanks

[exotic fiber]

and, take into consideration of at least, letting 4 or 5 characters left on the Lyrics space.. if u use all the characters space, it always will be flagged

[fickle hull]

can you specify the key and bpm in the song generation of suno? and if yes is there a format i can use for it? ty i´d appreciate

[exotic fiber]

u can, bit works like 25% of the times.. I force it a little more with bpm + Speed

Like,

Slow Trap. 90BPM,

[fickle hull]

okay cool ty

i wonder if anyone publish some documentation what works best, whats best format to get what i want etc. when i just google a song and copy that genre its not even close most of the times xd

[exotic fiber]

use this at the beginning of the style

{Voice type, unprocessed, key, bpm, era} Style/Styles

{male, unprocessed, c#, 90bpm, 2020s} Hip Hop, urban

[fickle hull]

oh nice 🙂 what does the brackets mean? is it like in stable diffusion where it pays more attention to it?

[exotic fiber]

theres no perfect formula... just trying to write the lyrics with cotrrect metrics, rhimes, song structure, etc etc... suno AI doesnt make miracles :p

[fickle hull]

tthat alreadys helps a lot tho ty :3

[exotic fiber]

we still dotn have confiormation from the devs, but for me works..

and for the lyrics space,,

[] for song parts structures and () for eschoes and/or second voices

[exotic fiber]

https://www.suno.wiki/

[fickle hull]

i just figured out earlier you can use ( ) for repeat of a word in as echos/reverb

[fickle hull]

nice ty i´ll check it out

[exotic fiber]

Usually this is the structure i use to write my lyrics:

[INTRO]
[VERSE]
[PRE-CHORUS, Build up]
[CHORUS, Drop]/[HOOK]
[BRIDGE]/[INTERLUDE]
[PRE-CHORUS, Build up]
[CHORUS, Drop]/[HOOK]
[BRIDGE]/[INTERLUDE]
[OUTRO]
[END SONG]/[END]

[fickle hull]

I assume it wont fit all one clip ^^ seem to only allow certain amount of characters for each generation

[exotic fiber]

yep, more than 1 generation.. i usually finish a track after the second continue (3 parts)

[fickle hull]

im trying to learn how to remake a song so i can get better understanding how the AI for music works. still working on to getting it work locally but seems still too knew to have smth like suno

what im surpised at is it can handle english/korean combinations

[exotic fiber]

handles almost all languages, even dialects

[fickle hull]

also funny thing is sometimes the Ai sings whats in the bracket XD. ive tried smth like
Verse 1
I wanna be a good man [Female 1/Male 1] Just for you (as background vocal)

[exotic fiber]

i made an afrobeat with an african dialect

)or what i think it is) xD

[fickle hull]

^^

[exotic fiber]

i just translated some words to that dialect and thenn used it.. worked as a charm

[fickle hull]

defo gonna do some experimentation later when i buy more credits. dont have any left anymore ^^

[exotic fiber]

Enjoy the ride!

[fickle hull]

Ty!

joined yesterday cuz i saw it on a live stream, he got a lot free coins 😭 😄

[exotic fiber]

https://tenor.com/view/i-know-the-feeling-well-marc-wheeler-twilight-zone-i-know-how-it-feels-i-can-relate-gif-21893952

[glass spear]

Hey dunno if this is the right channel for this question, but anyway: is it possible to generate songs of specific length? I'm just learning to make prompts and like 20 seconds would be enough for that purpose

[toxic harness]

why unprocessed ?

[scarlet mica]

OH MY GOD! They did it! They let you cut out cringe endings from an otherwise good generation!

Shucks. I think I'm going to buy another month to support the innovation. I really like that.

[final hazel]

lol.. they had it already for a month.. But glad you could join us 😉

[scarlet mica]

Eh. I was pretty burned when I lost a few thousand tokens to the month ticking over.

[light oyster]

now we just need it in the other direction

[scarlet mica]

Grafting a newly generated beginning to a good end?

[final hazel]

paint-in can.. they are working on that

[light oyster]

oh can it will? 🤭

yes yes i know. i am looking forward to it. in the meantime i would still die for a hard cut option rn

[cinder panther]

Not sure if this is the right channel, so sorry if it isn't. Do the credits from subscriptions reset by interval (x days) or calendrical per month?

[final hazel]

they are monthly, so they do not roll over.

[scarlet mica]

Yeah don't over-buy tokens.

Or make sure to generate as much as possible at the end.

[cinder panther]

Therefore i'm asking when they reset aka expire, end of month or end of 30 days subscription?

[scarlet mica]

I think it's 30 days. I could be wrong.

[final hazel]

you can check that in your account. Click on subscriptions. There you can see when your nenew date is.

[cinder panther]

Thank you 🙂

[exotic fiber]

someone commented (not confirmed by devs) that vocals get a cleaner sound with that.. and, to be honest, i also feels that, but dnt know if its just fantasy xD

[wooden stag]

hmm idk, I just tested and honestly the "unprocessed" sounds like the results are more like levels not mixed in post production, but the autotune was just as heavy

[weary wharf]

If I use the [Hook] meta in a song, suno tends to repeat it and passed the new lines in part 2,3, ...
Any ideas how I can force to go to the chorus? Tried, breaks, interludes, cuts... It always hallucinates rendition of the hook part.

[wooden stag]

Did you try [Chorus]?

[weary wharf]

Yep. I even just added gibberish text, new lines... always repeats the hook 🤷‍♂️

[wooden stag]

You can try [Hook x1], I've never tried with 1 but with x2 or x3 it will repeat.

[weary wharf]

I'm never using hook again 😆

[wooden stag]

It might be the timing structure though too. If it thinks it needs a filler to finish the verse at the right time it will add stuff.

[weary wharf]

I'm going back second by second right now... hoping I can find a nice starting point where the hook is not prioritized.

[neat nexus]

Have you considered using it in the style prompt not the lyrics prompt?

[wooden stag]

How would you do that

[weary wharf]

I'm on the "part 3", and all I get is the hook in different versions... nothing breaks the loop, even changing the style of the music

[neat nexus]

Ouch. That has happened to me before, but last time was in v2

[weary wharf]

This worked, but now the music ended of course 😄

[Outro]
Aah... Ooh...

[thin flint]

Well the AI fails jokes hard, but v3 final seems to have the delivery of comedy down. Just can't use AI to write good jokes. https://app.suno.ai/song/73c50dbc-17f3-47b4-9fa3-2469b73d3aa1

[fallow gull]

I had a song where the Bridge started with the same words as the Chorus "Why do you". Suno almost never included the bridge in a generation and when it did the bridge was often scrambled. I left the song unfinished for now.

[weary wharf]

If only it just returns back to a chorus or verse... I'll cut the mid part in Audition.

[thin flint]

I didn't feel ike continuying this, but it's cool that it can do it. https://app.suno.ai/song/c72798c5-bbf6-4d46-bd64-4a288c4bdf5d

[wooden stag]

That image though

[torpid elbow]

Any idea to keep music base code in suno?

[thin flint]

https://app.suno.ai/song/1656aebe-1745-4960-80cc-4ab6dcfc1ac5

[wooden stag]

Anyone knows what causes skipping/glitching?

[errant lotus]

how can I move a downloaded song to thumb drive to play in my car?

[broken lotus]

under the [...] menu find Download -> Audio
if you are on mobile, try switching your browser to *desktop view *to download rather than 'share'

[exotic fiber]

that not only happens with Hook, happens with every empty parameter.. if u also left CHORUS empty Suno will try to fill it with some words from other parts. But usually it produces good "lyric mixes"

Keep in mind that happens when you have assigned lyrics to things but not to others, like

[chorus]
hjshgkhgkdfhgf
dfgdfgdfgdgdg
DFGsdffgdgdf
dfgdgfdgdg

[bridge]

[verse]
fdghdfgsfgdfg
dfgdfgfdgdfgdf
dfgdfgdfg
sdfgdfgdg

In order to get control again over the lyrics but mantaining the "remix of words", do it like this:

[chorus]
hjshgkhgkdfhgf
dfgdfgdfgdgdg
DFGsdffgdgdf
dfgdgfdgdg

[bridge]

[break]
[verse]
fdghdfgsfgdfg
dfgdfgfdgdfgdf
dfgdfgdfg
sdfgdfgdg

After the Bridge steps are completed, it will start singing the next verse...

(works for me most of the times)

[exotic fiber]

also this!

[eager topaz]

I like using the [Break]. Also the [Pause] can work quite well, although with Suno you never know! Sometimes Suno's fillers and arrangements are fire 🙂

[neon pollen]

Is it possible to make song with same start and end so i can play it repeatable?🤔

I'm trying it but wondering.

[untold briar]

No. Well there's no feature to do that. But hypothetically there might be a style prompt which the gets the AI itself to do it. Doesn't seem very likely but not impossible.

I'll try "loop" just in case the easiest possible thing actually works, you never know. Low chance but takes 1 minute to check.

[neon pollen]

Well if it's impossible then muted start end might be also good.

so thanks to let me know

[untold briar]

"loop" seems to make the music itself very repetitive, like stylistically. It might be repetetive enough to almost be loopable but is not techincaly looping.'

[neon pollen]

Yup i got result what you expected 😁

[untold briar]

you know... one sec

might be close

it's like hold music though lol

it's like hotel tv background music vibe https://app.suno.ai/song/745ca516-de1b-4fe3-bef3-d2376f51e8aa/

It changes so little you could make it loop but it isn't exaclty

[neon pollen]

I'm interested in making game of programming part so..

using music tools looks cool

https://tenor.com/view/shocked-surprised-gasp-what-cat-shock-gif-635629308990545194

[neon bramble]

/CHIRP

[flint haven]

The /chirp command is not available in this channel

Go to the #🐣┃chirp-beta-1 channel to use the /chirp command, and check out the #🐣┃how-to-chirp channel for step-by-step instructions for making music using Chirp on Discord.

[flat field]

Is there any way for the melody of each verse to be same? I keep finding the verse between Part 1 and Part 2 have different melody even if the wording stays the same...

[alpine venture]

In my experience, you will have the most success with continuation based of prior prompt adherence within a given progression. Personally, I don't think its worth trying to force, ultimately you have to evaluate the final composition at any given moment on its own merits. Almost always you can lose multiple lines without losing as much weight as if you had forced the model in a direction it is not inclined towards whatsoever, usually because lyrical flow is trash or it is just not catchy, since a certain amount of weight pushes it in that direction in the form of, presumably, human preference optimization.

but prior adherence is the single biggest factor

However, there's another useful trick: you can often recover "lost" lyrics using instrumental cues, or a blank stanza after a prior that was cut off

ie it will actually read them off the lyric sheet despite not having said them previously. This is an example of how I was able to recover the cabbage line in this song:

Here it is recovered:

https://app.suno.ai/song/a64c490e-3817-4352-85e2-137ffb819705/

and that is a continuation from this, which has no adherence:
https://app.suno.ai/song/9ff46ccb-19eb-4984-ae20-e0bef7968cc9

.

which was eventually rolled into the full song

[alpine venture]

can you post an example? This likely has to do with style tags, if you have something in there that has a lot of dynamism you will have melodic shifts. If you want something very repetitive, goth, house (presumably), dance, etc are likely to preserve a repetitive riff

[flat field]

https://app.suno.ai/song/610e5f0e-08c3-43a2-a10d-d1d863cc8c1f
Part 1
Vs
https://app.suno.ai/song/c6c9ba2a-c13b-41ab-9f66-ff5e4f58873d
Part 2

Even the chorus isn't consistent

[alpine venture]

https://app.suno.ai/song/3fa95d63-ebb9-4601-8345-1d667211ee7f

the most repetitive thing I ever made. I don't know why I did this.

[flat field]

This is one of the reasons why I kinda want to wait until Suno is able to reach 4:00 mark before continuing

[alpine venture]

it might be autism.

[flat field]

I had better result with fast paced song

Even then it's hit or miss

Oh well, I guess I can think of this as the limitation of the current model and just play around it...

[alpine venture]

Wait, are these continuations? it doesn't show "continued from"

or is that just on mine

like, maybe I can't see your continuation

[alpine venture]

ita definitely not, I am able to keep melodies in general, but most songs don't repeat the same thing over and over, so that's where style tags matter

[flat field]

It's fine on mine

[alpine venture]

if Indonesian music has progression in some fashion, of the melodies

then that will do it

[flat field]

I guess I'm going to keep on the usual fast J-pop style of song

I think the slow pace of the song is hard on the algorithm

Which remind me, how can we make a song faster but not that fast (BPM ~100)

I read that Suno ignore BPM tags for now

[alpine venture]

examine what styles of music have a lot of that type. Then make sure your lyrics are amenable to it via experimentation.

you can then blend those styles with your preferred style until you get a sound you like

it's all experimentation at this point. Pink has good pacing, specifically dance-punk

but your music sounds much lighter

[flat field]

I am actually experimenting since someone asked me if Suno can make emotional song

[alpine venture]

oh it definitely can

you just won't see it on explore

[flat field]

I guess they would need to satisfied with the short version

I keep failing continuing the slow songs, even the chorus keeps changing. I guess the current algorithm can't really deal with it I guess

[alpine venture]

here, an example of both the way the model deals with slow songs, as well as emotional:
https://app.suno.ai/song/b9305985-fb17-4ccc-9993-4b7d29089180

I just realized this is technical, thus likely only for reporting bugs, so I am going to stop commenting here, I think we're flooding this chat

I hope that helps 🙏

[flat field]

Yeah, sorry for that

[untold briar]

No technical is supposed to be discussing prompting in detail, this is exactly on target

[alpine venture]

ah OK cool

[untold briar]

#🪲┃bugs-web is for bugs

[alpine venture]

duh haha

bruh that Vyvanse takes too long

OK

so slowcore virtuoso, for example

will usually give you something dynamic, at a reasonable pace

[flat field]

Okay then, I usually put the first part as this:

[Intro]
[Verse]
[Chorus]

then the next part would be

[Verse]
[Chorus]

Should I place [Bridge] in part 2 or part 3?

[Bridge]
[Chorus]
[Outro]

[alpine venture]

this is because virtuoso is FAST and DYNAMIC, and I put those in caps because it's insane. Whereas slowcore is SLOW and REPETITIVE

now, it also has sadness, so that will pollute a happy song

[alpine venture]

this I can't help with, someone else will be better with style tags. I don't use them much, and just go with experimentation with line breaks and let the model decide whats what

but I don't use a lot of rhyme, mostly just meter

sorry, not style tags, I mean the "[]" bracket tags

I only use them to force certain instrumental interludes

[flat field]

It's kinda confusing... I mean one of the songs I made... Is kinda messy but the continuation is fine and works well:

https://app.suno.ai/song/f66ca9cd-2af2-44f7-8932-0d06147567c4

So I guess for the moment, it would be hit and miss... And burn tokens until something sticks...

The chorus stayed the same all the time. So I was confused why the chorus changed in other songs. Must be genre related I guess

[tall zodiac]

320 bitrate is the holy grail, if v4 achieves that with 3 min tracks, its gonna be wildddd

[cold zenith]

i'm wondering if they are working on 320

[toxic harness]

i hear nothin...

[exotic fiber]

no feature but also not hard to achieve,

as u said, we need a loop.. but why to ask Suno to create the loop if we have all the control we need to do it?

a loop its something that "ends" "where starts", so

Just end the song with the same thing u started with, not using the [END]

something like:

[VERSE 1]
Line 1
Line 2
Line 3
Line 4

[PRE-CHORUS, build up]
Line 1
Line 2

[CHORUS, drop]
Line 1
Line 2
Line 3
Line 4

[INSTRUMENTAL SOLO]

[VERSE 2]
Line 1
Line 2
Line 3
Line 4

[PRE-CHORUS, build up]
Line 1
Line 2

[CHORUS, drop]
Line 1
Line 2
Line 3
Line 4

[Bridge]
Line 1
Line 2
Line 3
Line 4

[VERSE 1]
Line 1
Line 2
Line 3
Line 4

As you have started and ended with :

[VERSE 1]
Line 1
Line 2
Line 3
Line 4

without going to the [OUTRO]/[END] part, you just need to remove one of the VERSE and thats it, you are going to have a complete long smooth loop with coherence, Because, Technically, the song ends at the [BRIDGE], (Connects to the VERSE, same as the one u started with) so you re going to have 2 Pre chorus changes, 1 intrumental solo to avoid having vocals all over the loop and 1 Bridge change. And maybe we can Add more Dynamism if we add [OUTRO] and then try to make it get back to the [VERSE]

Loop done!

Also, maybe with a simple Script the Loop functionality can be added! 🙂

[exotic fiber]

https://tenor.com/view/i-dont-know-gif-18174494

[exotic fiber]

The [Bridge] in music is just a "connection" between to parts to create dynamism in the track, u can put it wherever u want, theres no formula... thres a lot of different song structures... just be creative... 🙂

[versed valley]

Welcome! For help, check out Suno Help

[hexed pumice]

you could try just [intro] blhblahblablabhlabhal [repeat intro]

[strange viper]

help :c

[hexed pumice]

or generate and then remove intro and add [repeat intro]

[haughty flare]

Is there anyway to change the written down lyrics after I'm done with the song? I had to fudge a bunch of generations together and now it sounds great but the lyrics are screwed up

[upbeat flame]

how would I use a prompt for Suno to use sidechain effects? (as in filter house)

[haughty flare]

Cause I had to keep feeding it lyrics from the previous generation as it kept losing it's place

[alpine venture]

ey yall, to those who find this stuff interesting, I've been using suno music creators in the style tags for funsies

https://app.suno.ai/song/2a7ea5e0-2271-4989-83c4-5ff47198f9ec

I'm p sure that song titles have in some fashion been used off our liked gens, which would explain why 🌳 and - Yolkhead are effective for certain sounds

but it could be confirmation bias. but I mean, it's a bit too coincidental imo. Either way, if it's true it gives a great avenue to basically easily get your own style

as a baseline shortcut

for what amounts to "allowed" artist style tags

example 2 (more approachable, altho i love the former one):
https://app.suno.ai/song/e998c170-eec4-4144-bc08-9b7690ca9bd8

[sour flare]

Hello, why all the /chirp last 40 sec ?

[final hazel]

Try using the website then. Discord is V1, and website has V2 and V3. Just go to https://suno.ai
There you can make the first part up to 2 minutes

[sour flare]

thx

[proper hemlock]

I wouldn't know all that

But I know good Music when I hear one

[marble token]

haha

Good judge

Where do you listen to your music?

Spotify?

[proper hemlock]

Lol

No

I use gala music streaming platform

Bobby Millette songs are my best most recently

[marble token]

I don't even know him

[proper hemlock]

He's an upcoming artist.

You can invest in the token too. $MUSIC
it's a worthy investment

[marble token]

I would look into it

Tnks

[cyan holly]

Wow, impressive stuff you have there Suno. I just discovered it. I've been using multiple different AI music stuff previously, such as Soundful & co, but the quality of your tracks is next level. I also tried to read Terms & conditions, but I didn't receive the answer I was looking for - if I generate a song while holding the Premier Plan membership, then will I keep the copyright/right for commercial use even after my Premier Plan gets cancelled at one point, as it was generated while I had Premium membership, or not?

[upbeat folio]

Yes but it's not retroactive. Anything you made on the free plan don't change copyright when you sign up. Only what is made during the subscription time.

[river lodge]

Anyone knows how to make Suno make a duet? i tryed a lot of different cofigurations to have verse 1 MAle, verse 2 female, chorus male and never does it well. Even writes [Verse 1: Male] and the singers of that verse is female...

[final hazel]

i've got some study material for you.. the song is not finished, but you can check the lyrics etc: https://app.suno.ai/song/5045e5f8-d213-4cac-9d87-99b370e89adb

[river lodge]

Did you start the song from custom mode?

[final hazel]

no instrumental to get the intro

[river lodge]

ok thanks! ill try it hehe

[final hazel]

np.. here you can find a video as well: #💬┃general-chat message

[fallow gull]

I was getting some duets just by adding "duet" to the style, but it was more random when the voices changed.
I was also including the names inside the section tags. That didn't work out well.

[stray narwhal]

Um,why is it at the end of my song there is a thunderstorm and then police sirens that sounds real?? I didnt ask for this lol https://app.suno.ai/song/62f893d3-54e9-4c21-8552-638eb29f7c9c

Time is about 1:30 when you can hear it

Dont really know what goes on behind the scenes but either way i still love Suno!

[fallow wave]

Is there a way to selection a different mp3 bitrate? Or any option to select a lossless format for audio download?

[eager topaz]

No, sadly!

[fallow wave]

That is not good

[final hazel]

That's is the current state of development. The model will become better over time.

[fallow wave]

yeah!

😁

Please consider flac or other audio formats

[final hazel]

I am sure the Suno team will.. it's just a matter of time

[flat field]

I think that one of the song used to train one of the tags: (classic rock upbeat) is using the music video version, and that MV has police siren at the end of their song.

[obsidian tendon]

Hallucinations

[charred lake]

Question for the Suno team - do we expect to have .WAV file downloads for "Full Songs"? I'm resaving these as wav's when mastering, but I feel like starting with a .WAV instead of .MP3 would bump the quality for DSPs

[fallow wave]

This is interesting on first try. This is just style of music prompt
https://app.suno.ai/song/f5eea442-df74-4272-9c88-8b018ce347cd

This is style of music with reverb and expander compression prompt
https://app.suno.ai/song/31dcffdd-7adc-4dd5-918c-c27c7f541a53

Maybe it affected it. Or I am just thinking it is changing the audio

[limpid spade]

It's been said a few times (search discord chats) that in the current version a .WAV would simply be a larger file of the exact same audio quality as the mp3. #💬┃general-chat message

[midnight kernel]

اغاني عربية

حان حان العيد
هيا هيا للفرحة كن سعيد
اهلا اهلا بالعيد
فالعيد يلين القلب من جديد
هيا هيا هيا
افتح قلبك لاتكون عنيد
هيا هيا هيا نمرح
فالعيد في جدار سنتر لن تكون وحيد
ياعيد ياعيد ياعيد

[alpine venture]

Eid Mubarak (I didn't think it was Eid yet but what do I know)

Anyway, more experimentation with prompting using previous Suno song titles

https://app.suno.ai/song/3d5e05e0-2d1d-4589-8899-f9fa94b23010

so this is an exact title I've used before, and of course using my " - Yolkhead" appears to confer a definite inpact

it is clear that the title does impact the resultant music

what is unclear (but I think is the case) is that the semantic data on word meaning is impacted in their fine-tune or even base training by title information on Suno produced songs, which it seems likely are used to improve the model with something akin to dpo

for sdxl

meaning, 🌳 for example is semantically distinct from its previous meaning, which was likely null 3 months ago or so, but now creates a lot of oscillating sound when included in title ie 🌳 's actual style

and " - Yolkhead" produces dynamic sad chord music

with a poor recording quality or distortion

but idk; in any case, it has improved my ability to produce music I enjoy since it makes it way easier to get the sound I want.

to show more clearly the difference, here is small, representative sample of that prompt without my name in the title.

https://app.suno.ai/song/d2b7295a-c4d8-4165-8c95-77d1ca046cbc

https://app.suno.ai/song/e6fc2322-cf96-41e4-bdd9-8b794e397f0e

https://app.suno.ai/song/711b0a87-9226-4970-ab4b-97997f63eb57

now let's examine with " - Yolkhead"

the whole sound changes

https://app.suno.ai/song/e61d1e1e-63b8-4982-809c-d0d12b6c8a50

https://app.suno.ai/song/de8b40c4-962a-4972-9179-de1127d45868

https://app.suno.ai/song/bf8dea73-d151-40c4-905e-d2ae43971c4f

the last one is p extreme example of how much it changes

Anyway, normally I'd just assume this is normal variance and/or the tokens of course influence the model, but it isn't dependent on our prior outputs

buuuut

you'll hear fairly often in songs with mine a certain clicking noise/sound fx. It's the sound of a cassette tape. This pops up a lot in some of my songs, especially in v3a.

I suspect thus this is indeed using our title information in their direct preference optimization ie if you use consistent title tags, you may be able to, in the future, preserve a certain style

unless I'm wrong, which I probably am

or they change their method

in any case, it DOES definitely suggest an intriguing avenue in the form of improved semantic descriptions through the use of community tagging, since music styles are not easily "boxed" and having superior shortcuts to provide that information accurately to the model would be awesome. If you could, for example, use a particular playlist as what amounts to an equivalent to a LoRa, this would be much more powerful in terms of its descriptive power than natural language.

but minimally tags would be helpful, since you could just use a string of numbers for certain features you like

and then locate those later on

[exotic fiber]

the format doesnt matter if the bitrate is the same.. wav, flac, mp3, etc etc.. the information the file has is the importasant thing, higher bitrates, more information, better quality.. at the moment suno exports at 192 kbps.. so, no matter if the file is a wav or an mp3, the information will be the same, the quality will be the same.

[shy sand]

Guys, a question, how would you create a boy band? back street boys style for example, with 5 different singers, I'm curious.

[alpine venture]

wanted to share a technical prompt:

https://app.suno.ai/song/8695213d-9f1f-4d82-bb5c-1cc5333c2d35

as the last part of a song

Since it can sometimes be hard to get it to end in a more "engaging"/dynamic melody

this was a p big style switch in this case

but fairly seamless

[graceful condor]

Hey guys! How do I unpublish? I accidentally clicked the "Public" button in the wrong version of the track.

[broken lotus]

A dev said there is no 'spill' from the title...
(I have seen the image-gen DOES pull from the title)

Since we can rename songs after they are generated, you could leave the branding for later and then it can't influence the song at all.

[broken lotus]

use the [...] menu and set the song to LINK ONLY

[quiet mountain]

/bark

[flint haven]

The /bark command is not available in this channel

Go to the #🐶┃bark-beta channel to use the /bark command, and check out our docs for step-by-step instructions for using Bark on Discord:

https://suno.ai/docs/discord/bark

[quiet mountain]

Can we use our songs for commercial purposes? Who owns them?

[final hazel]

Yes if you pay for a subscription.. 🙂

[final hazel]

As a Suno Pro or Premier subscriber, you have the right to use your generated songs for commercial purposes, as you own the songs created during your subscription period. This allows for a wide range of commercial uses, including uploading them to platforms like Spotify. However, for users on the free version, the songs are for non-commercial use only, as Suno retains ownership.

[quiet mountain]

Thanks! How much do I love Suno? THIS MUCH!

Wait - so I can license them as my own? Correct?

[final hazel]

Yep, that's right! As a Pro or Premier subscriber, you can license them as your own. Your creations, your rules!

[quiet mountain]

Not so much a question for Suno I guess. But very cool - thanks I'm so egg-sighted

[alpine venture]

hmm interesting

[alpine venture]

oh I'm not complaining, I'm saying it's useful if you want to use "band names" that are allowed by tos

however, if it has no influence on the song then it's irrelevant

in which case its likely just my own bias

but I'll take a look when I have more focus

[broken lotus]

I do tend to believe my ears over things people tell me... 😉

but just repeating what they said...

[alpine venture]

for sure, I mean, they have no incentive to lie

However, for example, if GPT-3.5 or another LLM creates hidden "instrumental" prompts when that's checked, there may be bleed-over that is unanticipated. In general, if an agent is overseeing production behind the scenes there can be a lot of unanticipated behavior with careful adversarial prompting

so it could still be possible, but unlikely

nah, just tested with the usual stuff I use for this kinda thing (you can fake out most LLMs just by making something seem like it would work)

https://app.suno.ai/song/fd6e8e8d-2d93-4e56-a008-b5f0b061b20f

it's just baroque, thus titles have 0 influence

style, however does

including 🌳 , etc in attle

style

but I'm much less certain now that the titles have any impact

that being said, you could stick a certain set of emojis in your style tags

and then see in the next version increment if it has a particular impact

[clear osprey]

https://app.suno.ai/song/012f4ed0-615f-41c2-9d38-0a5eaaaa782f

[versed valley]

Welcome! For help, check out Suno Help

[boreal fern]

hey, i have a few questions. 1. why can i generate lyrics, see, reroll and change them without using credits for all parts except the first?
2. Is there a way to reuse/remember a melody ?
3. I struggle to create a 3th part that sounds like the first one. is there a trick?

[vapid minnow]

was literally about to talk about this

[boreal fern]

nice^^

[vapid minnow]

made three songs with it already and it's really nice but they never take the same melody from the chorus - after the 2nd part, and you get to the third, it starts to change and fall apart

I've just been splitting the song up into stems and editing it accordingly to have the same melody as needed, but having to do that is kinda annoying. why does it not remember the melody after 2nd part?

[boreal fern]

what do you mean by editing the song? is the rea cutting tool?

[vapid minnow]

even after just having the chorus for the third part continued from the song, the melody changes

[vapid minnow]

no. I do it myself.

[final hazel]

it does, if you do it right.. maybe my video helps: https://youtu.be/5uJw5kb05as?si=gGiiztZWEX9rFmV-

[vapid minnow]

at what timestamp do you explain the fix for this

[final hazel]

I explain to make full songs in between so it has a longer memory

If you watch he full video you see my chorus is consistent.

[vapid minnow]

so, this is fine, but this doesn't help for songs that are already in progress.

I can adapt this workflow for the next song, but I already have a song going that I like, with a chorus I like

[final hazel]

Why not.. you can already make a full song in between?

[cunning mica]

Hi folks. Do you know a way to avoid / remove over-saturated reverbs / delays? I get that tinn-y echo of cranking the wet knob to high way to often and dunno why.

[boreal fern]

@final hazel did you write the lyrics yourself?

[final hazel]

yes why?

[vapid minnow]

oh, I did not see the other part, okay

[boreal fern]

because i try to create one with genered ones and have this 3th part problem. but i will try again tommorow

[vapid minnow]

it's a 10 minute video so I skipped around, timestamps/chapter markers would be best

[final hazel]

I alread tried to make it as short as I can.. if it's too lang so be it.. just don't expect me then to type it all out for you 😛

[vapid minnow]

did not ask but thank you all the same

your video was helpful nonetheless

[final hazel]

Well you did ask how to get chorus melody the same.. my videos explains 😄

Have fun 🙂

[boreal fern]

i guess you can only subscribe right now by credit card?

[vapid minnow]

idle question: why does it sometime just ignore a whole lyric?

[broken lotus]

Yes... Devs have mentioned getting more payment options in the future, but I do not know an eta on that...

Some people suggest to buy a pre-paid card (if you are buying an annual sub).

[boreal fern]

thanks

[cunning mica]

I think that's related to the rhythmic structure of the underlying audio. I had this too and found out that when you remove a syllable or two it will acknowledge the rest of the line in a new iteration

https://app.suno.ai/playlist/fbc8baaa-666b-4c55-b258-301484efc286/

[vapid minnow]

question, I've reviewed the whole video, it's helpful indeed but, my issue is that my song where I want to continue ends during the chorus, so I'm not sure how to continue it with the same melody

[cunning mica]

In this track the third line of verse 2 first only had 4 syllables. It got ignored from number 2 and moshed into line 4. Then I alligned the count to lines 1 and 2 and in the next iteration it was recognized completly

[vapid minnow]

ah, I see

[cunning mica]

A thesaurus can do wonders regarding this topic 😆

feature idea: pop-up thesaurus in the lyrics box 😉

[final hazel]

For me, if a chorus cuts off, I just start the next continue before the chorus so I can create a new one. I often feel they even turn out better then..

[vapid minnow]

hrmmm..

[cunning mica]

Is there a possibility to break the links between parts?

[boreal fern]

but the individual parts don't vanish?

[cunning mica]

Aye. I have a track which switched styles in the middle, but I made two completely separate follow ups out of it. But now when I want to finish the full song, In one of them theres still the, now not matching anymore, part for the other branch. I still need that, but not in THIS specific track

https://app.suno.ai/song/34df8a71-2a7e-4768-aa10-cba8875322eb/
specifically this one. The switch happens at 03:33. I have a finished follow up for the first half, so that's fine, but because of the link, I can't get rid of it in this track here.

[boreal fern]

7:39 👀

y, i understand your problem

[cunning mica]

😖

I feel like a 40k Techpriest here: I have no idea what I'm doing, I don't know how this thing works, I'm just saying my praye... ah, I mean generation phrases to the Omnisi... I mean Suno, dangit, in the hopes it gives me remotely what I want 🤣 #nerddom

[kindred minnow]

on the website, the new song page does not update anymore... same song on there for the past 20 hours or so

[uneven egret]

/chirp

[flint haven]

The /chirp command is not available in this channel

Go to the #🐣┃chirp-beta-1 channel to use the /chirp command, and check out the #🐣┃how-to-chirp channel for step-by-step instructions for making music using Chirp on Discord.

[thin flint]

This video is helpful.

[rose ore]

So if anyone's interested. Here's an example of how I cut between takes to get a better performance out of the machine. Don't worry, this one's upbeat an happy, even if unfinished
https://app.suno.ai/song/0e1f27ab-78b9-434e-bf2d-8d08498e93aa

[quiet mountain]

HOw are you going to rerun and finish that? The problem I have is getting the same version of a song - but adjusted. Is there a way to do that after you hit create? It seems every creation is a brand new sound?

[rose ore]

I sgement the song into clumps using the whole song feature. This keeps things compartmentalized in the neural net, so it'll scale out and finish the song a little more reliably. If I was going to re-render the whole thing, I would thumbs up all my takes that were good, which is going to keep the reading and voice mor or less intact.

Like, for example, provided I keep the words the same in the second chorus, I would get a reading based on this one, with a lot of the same quirks, which is what I'm shooting for

I just need to write a second verse and spoken word passage that make sense structurally.

[quiet mountain]

I had no idea the thumbs up kept the ready and voice intact -

[rose ore]

If you do enough of it, yes

But in general, it'll try to balance out the comp towards the takes you like

[quiet mountain]

I just thumbs up on all the things I like - my library is huge - I am deleting stuff now

[rose ore]

Oh yeah

Last time I got to the end, like two months ago, it was like 1000 pages

don't even want to think about how big it is now

lol

[river lodge]

Happening to anyone who wants to continue a song in a X minute (v3) and first thing you write is your chorus (used previously so it repeats it) but Suno jumps to the next thing after that first propmt and then mixes everything you prompted? Would be so easy to JUST REPEAT my chorus! idk, am i doing anything wrong?

[safe ruin]

It happened to me as well. Or it did play the chorus but in a totally different way making the song inconsistent

[urban dagger]

my lyrics keep getting flagged im not sure why

we evade their bullets. Griddy on! up Mount Olympus
fizzed out heros shredding, full squads are on us
downed three chuds gave one a fatal injury, on our way to get a crowned victory
lets destroy these chuds, prove how game winners get epic dubs

[cunning mica]

chuds maybe?

[wooden stag]

From what I heard, Suno uses top rated tracks to refine training. But the "autotune" resides in most songs noticeably. Curious if there's some kind of filter for this or if we're going to be seeing an increase over time

[quiet mountain]

thinking it could be the fatal injury - or 'fatal' and concept of three chuds?

[cunning mica]

I noticed that after just two days of using. If you use certain key words in certain genres you can bet on soon hearing very sameish blocks over again

very noticeable in the EDM genre

[urban dagger]

no i made a whole song about chuds

it was called chuddie buddies

[cunning mica]

hmm 🫤

dunno, that "fatal" approach above seems reasonable, try to replace that

[urban dagger]

hmm okay thank you i will try it

[cunning mica]

If you need inspiration, watch Monthy Phythhons Parrot 😉

[urban dagger]

i think it was sthe fatal

changed it to fateful

[wooden burrow]

Hİ

[jagged hatch]

Hi

Is Chirp what is used on suno.ai? i.e. the "v3 model"?

And is it, or will it ever be open source?

Also, Google uses the name Chirp for a speech-to-text AI model.

[neat nexus]

Yes, the latest model which is available on the website is v3.

[neat nexus]

Unknown at this time; ask again later.

[neat nexus]

https://tenor.com/view/conspiracy-conspiracy-time-tin-foil-hat-gif-20017112

[jagged hatch]

https://lablab.ai/t/chirp-tutorial-how-to-use-chirp-model-on-google-cloud

(whoopsie on the branding?)

[broken lotus]

you'll notice it only says 'Suno' on the website...

[jagged hatch]

On which website? I'm saying that Google has their own AI model called Chirp. But Suno's is also called Chirp. Both are for different purposes. Google's is for speech-to-text while the Suno one is for text-to-music. Seems like this could lead to consumer confusion. Try to Google "chirp ai model" and see what I mean.

[neat nexus]

Yes...and you may notice that if you search the conversations here we went from calling it Chirp, to Suno Chirp to now mostly Suno and occasionally just 'the AI' which I suspect gets really confusing... hehehehe...

But yeah it's an issue that has been noted and that is why most of us have been changing how we refer to it. Habits can be hard to break though, we've been calling it chirp for a while here...

[exotic fiber]

next time, try with Fei-tal.. ;p

[urban dagger]

ohh thanks

i just changed it to fateful it was close enough

[chrome jasper]

https://app.suno.ai/song/0b47fcb2-5a53-4262-949b-09d782528be1

[chrome jasper]

https://app.suno.ai/song/8ce68f27-4f96-496f-b9b5-c927c3c2debe

[night pelican]

Hey... Dutch accent? 😆

[leaden crown]

Sorry if this has been asked already - Getting a vague error on the site when attempting to remix some songs. Is this a known issue? Is there a fix?

[fallow wave]

I try to switch between male and female singers in the same song and it is not possible. Seems like it is only one singer voice type is possible.
https://app.suno.ai/song/5cefe9de-a413-439f-aa53-6bf301f4d2a4

I tried these meta tags:
[Man Voice Verse]
[Woman Voice Verse 2]
[Group Voice]
[Chorus]

And the style of music: Caribbean steel drums, tropical, ocean waves, wall of sound, spector sound, Man voice, Woman Voice

Is it possible to switch vocalist in the middle of a song?

[untold briar]

Yes, it's a bug caused by the Remix feature copying all the lyrics from a Full Song, which is too long for a single clip lyrics field and causes error. The fix is simply to shorten the lyrics manually for now.

[tight zephyr]

hahaa very much so

[broken swift]

https://suno-ai.notion.site/Using-Metatags-34944efe09ec41d693e314f13af44695
Divide a song into segments and experiment with metatags like [boy]、[girl]、[woman]、[man]

[quartz heron]

Hi people! i am from Russia and I can't pay the service. somebody can help me?

[shut hollow]

use googlepay. Thats available in russia also.

[quartz heron]

i cant add russian card to google pay

[shut hollow]

well. There is no way to help then thou? you wouldnt be able to transfer money to any one either.

[quartz heron]

You are from support?

1. I can transfer someone in russia who can pay
2. my friend in china can transfer someone in china who can pay
3. can you give me access before payment and in 2 months I will move to Bali and make a card there and pay on the suno site?

[untold briar]

Actual Suno support is behind email [e-mail]. I've seen some users here post about using prepaid credit cards or gift cards or other methods, when they didn't have access to a working payment method, but I don't know the details myself.

[quartz heron]

thx

[flat field]

https://app.suno.ai/song/6df4bbee-637d-47c1-9d08-ba3b30b098f7

Got an unintentional duet... So funny. I guess it's because of the tag: Indian movie

[undone kelp]

Does anyone know how to Remix? I click the Remix button, select two "songs" but no button forward to launch the process. (Get Whole Song, got it! Nevermind! 😊 )

[tepid wedge]

@open magnet I just had a quick question. I wan to buy Suno for a month, but I also want to know if my credits will roll over after the month is over. I was thinking of buying the $30 tier just so I have a crapload of credits and I can work them down over the next month or two (or three, idk how much I'll be using it?)

So yeah. Do they roll over and stay in my credits permanently until fully used?

(also sorry for the ping, but I'd rather ask someone who knows for sure than the community)

[untold briar]

The monthly $10 or $30 tier credits don't roll over. (The extra credits do, but not the monthly ones.)

[tepid wedge]

wait, what 'extra credits'?

I was only going to buy it for one month

[untold briar]

You can always upgrade from the $10 to $30 tier mid month, if you decide you are going. to use all the credits. Extra credits are on a 'buy more credits' menu that appears when you are on any pay plan. They are more expensive than monthly credits.

[tepid wedge]

ok

so what are 'extra credits'

[tepid wedge]

oh

that sucks the energy out of my desire to buy a plan

so if I buy a 10,000 credit plan any unused credits just disappear after 30 days? :/

[upbeat python]

/chirp

[flint haven]

The /chirp command is not available in this channel

Go to the #🐣┃chirp-beta-1 channel to use the /chirp command, and check out the #🐣┃how-to-chirp channel for step-by-step instructions for making music using Chirp on Discord.

[tepid wedge]

@upbeat python the website is easier to use IMO https://app.suno.ai/create/

[broken lotus]

Mine disappear before 30 days most months....

[tepid wedge]

gdi

I bought it anyway

but that is ... I don't want to say anti-consumer, but it feels anti-consumer

if that happens to me, this will be the first and last month I pay for Suno :/

[scarlet mica]

Yeah. Unfortunately, I learned that the hard way.

[sweet mauve]

I sure hope that the rules for forbidden words in lyrics get revamped. When I want to make a serious metal song, it's bound to contain some dark topics like death or violence. :/

[scarlet mica]

After nearly a month of trialing it, I wanted to support them. And the bonus credits was gravy.
But then the credits disappeared.

And it really left a bad taste.

[fierce fjord]

what if you generated a song stem during the free period but then upgrade to subscription and then extend the song stem generated from the free period? Is this extended version that is extended during subscription commercially usable for the user?

[scarlet mica]

There is quite literally no meaningful way to enforce that policy, unless you are just out and out bragging about commercially using the free version.

[broken lotus]

Seriously... there was A song that convinced me to buy subscription because I could hear the music was good enough for my needs – but once I had a lot more credits, and experimented a little more... That song was forgotten. lol.

Agree 100% with ShadowMender above. Trying to split legal hairs over an old generation becomes moot when you can make 50 new ones in an hour...

[runic dock]

I dont have experience with DAWs. What can we do in cunjunction with suno ai ? Except trivial things like removing the end of the song when suno goes to hallucination after a wonderfull start ?
I would like to get FLStudio.

[hollow dove]

There are a lot of things you can do with a DAW. You can enhance Suno tracks with things like studio compression and reverb, tape emulation mastering and more, depending on the plugins you add. You can use stem splitter plugins to separate vocals and instruments. You can also add your own tracks to the mix, such as vocals, sound effects, samples or your own instrument playing.

[runic dock]

Ok thanks for the answer. When you use the stem splitter can you separate the different instruments or just get two tracks : vocal and instruments ?

[hollow dove]

I'm using the OpenVINO AI Music Separation plugin that can give me drums, bass, other instruments and vocals.

[neat field]

https://github.com/suno-ai did they take down chirp repo, and does anyone have a copy laying around ?

i have 64gb ram, don't care how slow cpu gen is, i just want as close to the excellent quality of suno.ai v3 truly open source locally generated

my ears have heard the glory of suno.ai, and it cannot be unheard, i will eagerly await other GPT music gen workflows to get close to it

[radiant heart]

hello there! Ok let's say I can remake whatever song Suno generates.. what's the real deal with the lyrics and the melody of the lyrics? can I copyright this song as mine?

I know I can change lyrics and remake the chorus and do all kinds of production gymnastics to get something completely different from the original, but if the chorus is in the money, what's the legal part of this?

[meager lodge]

Dumb question but how do we download the song? It’s only downloading as a video which wont open in audio apps. On iphone using both app and desktop versions.

[final hazel]

You are presented 2 option when you click on download: Audio and Video

[meager lodge]

Unfortunately on iphone there is only a save option which saves as a video. On PC, the download link only offers mp4 video, no audio option which is odd for an audio program. Is this a paywall issue? I was going to subscribe ($$$) but not if I can’t download audio from the audio creator.

[final hazel]

Hey Epic. Only Bark was on github, and still is 😄

[broken lotus]

switch your browser to 'desktop view'
you will see the normal downloads that way.

IOS and Android intercept with a SHARE menu.

[quiet mountain]

That's my only issue. If I do a song I like, but I want to change words or adjust - I can never get it back again - : (

[radiant heart]

what do you mean?

[quiet mountain]

When I do a song, if I like it, there's no way to say - I want this same song, but change these things about it - make sense? I thought that's what you were talking about. It's like every one is an original and if I do one for someone and they want to change it - to bad so sad can't

[radiant heart]

ahhh, I see what you mean. What I would probably do in that case is just get the main idea from suno and then re-record it with my voice with the changes I want and then use something like Audimee to replace my vocals. My question is more in the legal aspect. Because whatever song I get to generate with suno I want to have other people sing it, but sometimes the song that suno generates is so good I want to remake the music and keep everything lyrics wise, then get the artist to re-record it whit his/her voice and then release it to the label

[neat field]

#🐶┃bark-technical message i made a bark specific post, however chirp Apache 2.0 or MIT license is what i want (as close to v3 as open source can produce with todays SOTA)

[final hazel]

Chirp is not opensource, and never was either according to the Devs 🤷‍♀️

[neat field]

please consider this some sort of feature request, even if it's delayed while you spin up revenue (devs gotta eat!)

[final hazel]

I am a volunteer guide, I have no say in that 🙂

[neat field]

thanks for your answers!!

[cerulean pagoda]

Is there a way to input inflections, accents, and chord progressions?

[kindred minnow]

#been using Suno v3 for the past 2 weeks.. i think i have enough for now... i done extensive test. Key point to focus on for the next version: 1- Keep the quality consistency from part 1 to x number of part. The quality is diminishing often from part 2-3-4 etc. 2- focussing on equilizing and mastering sounds. Often when there an epic chorus with synth or orchestal instruments with a powerfull signing voice there is a lot of noise and phase cancellation preventing to get a good audio no mather the key words you put you can put clean hi fi audio, mastered, good mastering.. high sound quality... you can even give bitrates... it dosent change this noise. I think suno could create some amazing original song ready to use for a project in the futur, but for now the quality is not good enough.. i noticed that for music like rap or some country music or any king of music where the voice is not pushed and intrumentals the app is great, but as soon as you go into emotional build up with a few layers of sound, the phase cancellation prevent of hearing the sound clearly as soon as the signer is to powerfull it break the quality.. i think with those 2 things worked on this would be awsome.... I re rendered a lots of times same songs sometimes i re rendered 30-40 times the same part and in each time when there is powerfull sound this happend...

For now i will take a break and comeback in a few month

its also often the higher frequencies that have to most problem

[solid plaza]

Hello, I have an issue with a song I would like to continue. The AI managed to create a perfect continuation the first time. I am now trying to continue the second part and add the final lyrics but the style changes. How can I achieve a perfect song keeping the same style, same voice and so on?

[kindred minnow]

@solid plaza do you use =the web site or you use the discord app

[solid plaza]

I am using the web site

[kindred minnow]

and each time you render a part it changes style when continuing the song??

[solid plaza]

Precisely. It changes the style, the beat, the instruments, the voice. It is driving me mad.

[kindred minnow]

Strange and you dont change the style description??

[solid plaza]

It is frustrating because I am following the instructions and I know the first time worked perfectly.

I have tried leaving it blank and also using the same tags, but the result are always different for some reason.

*results

[kindred minnow]

can you post the example here

i will listent o it look at your prompt

[solid plaza]

Of course. Give me a second.

Note that at the time there was only the v2 available, so I continued with the v2

[kindred minnow]

ok

[solid plaza]

Original prompt : https://app.suno.ai/song/3a22b19b-59f5-490b-96f1-4dd4ac10bec0
First continuation : https://app.suno.ai/song/34a26978-2177-4805-ad01-31f4ce594450
Third (fckd up) : https://app.suno.ai/song/18f50346-ada2-4458-8d8f-3f945bf962b6

If you listen to the first two they are the same beat, the third and all subsequent ones are different

[kindred minnow]

i listen to it, give me a few minutes and do you mind if i try your prompt to just to see what its doing in mine i will also use the v2 for the test

[solid plaza]

Take your time, I am making dinner so I am not in a hurry 😉

Thank you for your time

[kindred minnow]

no problem im just an enthousiat to this tehcnology i dont mind helping to seek he problems

[signal temple]

Hi, i need help i hope this is the right channel.
last month i subscribed to Suno (from the site) and got credits (i subscribed just for 1 month to play with the credits)
i did create some songs and i left many credit, now this month i entered to spend some more credits and i see it reset my credits bank.. why??
i did pay for these credits, i spended maybe 10% of them.. why its gone?

thanks for helping

[final ravine]

Ugh I just used this for the song description
Atmospheric, Dark, Sinister, Darkwave, Ethereal, Mysterious, Gregorian male choir, Male singer (+ what the song is about)
And it completely ignored everything except the lyrics 😭
What am I doing wrong? How can I get the right style?

[kindred minnow]

ok, listening to it i dont see a real problem, the first 2 part the beat is on, the 3rd part the beat goes away, but instrument are similar and the voice is the same.. its mostly a beat drop that suno created.. in music there is a variation of beat, with drops whenre sometimes you dont ehar the bass drum for a moment or either to create a break in the song to either finish the song or to put the beat back on later that give more impact to the music if the beat is not constant... but if you really want you beat to be constant maybe try to render it a few times.. i often render lets say part 2 of continuation 10 times even 20 times, i listen to all of them and chose the one i prefer to continue with suno will always create something different.. but in your case its not really a problem i think maybe just that the 3rd continuation is not fitting well with the rest but try to render it a few times you might get one that suits better for the ending of the song or the 3rd part (dont know if it was the ending )

this is one i made, the sound quality is awfull but the song have multiple variation giving maybe the feeling suno change things but once conected it make a powerfull emotional vibe

https://app.suno.ai/song/0c32495f-2293-4266-8f1c-a9c1082b406b

[quasi canyon]

Ownership & Copyright
https://suno-ai.notion.site/FAQs-b72601b96de44e5cacd2cd6baa985448#47fc7b6ad1114a1597903b11c73ade77

[solid plaza]

I will listen to it as soon as possible 😉 Thank you

[kindred minnow]

@solid plaza im making a version of yours now, but i think even if you started with v2 you canc ontinue with v3

https://app.suno.ai/song/89db0fff-24c9-443c-9bd3-7df36bc02aed

i continued your song.. i tryed to put vocals.. but yeah seems like you were right maybe a bit of inconsistencies but.. i dont know if suno do this for artistic reason to build up a song or if it a bug??

in this version it did not even put vocals just a synth and beat continuation

[fossil ore]

Just started using Suno. Seems promising. Question: When I upload my lyrics, do I need to indicate verse, chorus, bridge, instrumental by having those words in parenthesis prior to the section?

[solid plaza]

It is too bad because I generated the text with chat gpt and then I changed some parts to make it more original. I really hoped to achieve a full song...

[solid plaza]

This is more consistent than mine, I wonder if you wait too long to continue the song the AI kinds of forgets it, even if it should have memory of the prompt

[solid plaza]

Yes you have to write [verse] , [chorus], [pre-chorus] and so on

[winter inlet]

Any ways to download a song that shows the lyrics (like previous ver)?

[final hazel]

No.. that's gone. They are working on a new design.

[jagged hatch]

I've heard Suno. It's good but it's not perfect. When it starts layering on sound, it tends to sound digitally compressed.

Does Chirp do parallel processing of instrument tracks and merge them after completing full tracks or does the model inference only create pre-blended audio?

[worldly juniper]

I have a question: if I buy the subscription, can I own the comercial copyrights of the musics I created in free version? Or can I buy them separately?

Remix doesn't always generate the same version of the song, and I didn't found the answer in the FAQ...

[final hazel]

You can only get the full rights of songs you created under your paid subscription.

[worldly juniper]

If I just continue my own song, that doesn't count, right?

[final hazel]

@worldly juniper from the faq:

[worldly juniper]

Oh, that's why I need new glasses lol
Thank you for clarifying that! But isn't there any way to get the rights for a particular music even if it means paying by it directly?

[final hazel]

For things like that it would be wise to contact [e-mail]. I don't have the answer on that.

[worldly juniper]

Nice! Thank you for the guidance! Really appreciate it :)

[tulip granite]

How do I cut the end of a song to not have an abrupt ending?

[fossil ore]

When I click on 'remix' it doesn't seem to do anything.

[final hazel]

The word remix is kinda confusing. It just copies the lyric and music style from a song and paste it in a new generation.

[neat field]

Manually do lyrics and add [outro] tag to the lyrics at the end

[tulip granite]

I just finished my first song. Suno is really good. It's impressive.

[fossil ore]

Oh bloddy hell that IS confusing. Thank you.

[sudden parrot]

hi y'all guys i have a question how can i add female chorus after certain phrases

[azure crane]

So does it also generate two versions when you Create/Continue with the subscription and that will cost you 10 credits?

[teal raft]

YOu always get two generations

No matter what

That's the core of how it works

Right now you will never get 'efficency' with credits like that

For example, I, and many other people, will have many credits burned for every complete song

A single 3 minute song might have 500-750 credits behind it, if not more

You aren't going to 'use' every generation. You get the audio, it's there, but if you are working toward a song you probably aren't going to use it

[azure crane]

Hmm, can’t understand it. You waist 250 songs then if you only use one of the two generations made. They should change their info then.

[teal raft]

I agree about how they sell the subscriptions but it's technically all true

You do get the audio

It is generated and uses the GPU cycles

[exotic fiber]

They call generations -> Songs

[brazen path]

Hello. I'd like to make Suno sing a song in an incommon language (a dead indigenous language from Brazil). Since Suno doesn't recognize the language and its writing, is there a way I can tell Suno the right PHONEMES to speak, and the proper syllables to stress?

[untold briar]

Your best bet is to phonetically write it using a language that is closely related and shares a lot of the sounds. You can't use IPA or anything.

Regular accents can work in many languages for stress.

[brazen path]

Unfortunately, as an indigenous language (old Tupí), I bet extant related languages (like Guaraní) are also unknown to Suno.

[untold briar]

Well it could be any language even if completely not related, as long as it's capable of producing similar sounds, Suno can deal with made up words to some degree.

Before assuming it 100% doesn't work try Guaraní though, you never know...

(I wouldn't be able to tell whether it was working very easily.)

[orchid whale]

Everytime I try to continue a song it only generates a 3 second song and I get my credits refunded. Am I doing something wrong?

[untold briar]

That can happen with a particular clip. Try using the custom continue time and choosing a time that is further from the end of the clip instead.

[orchid whale]

I'll try that and see how it goes, thanks!

[untold briar]

I tried anyway https://app.suno.ai/song/90c2ca3f-fe92-4cc4-a14f-dfc43229a5b9/ vs https://www.youtube.com/watch?v=FAcNH-rzG84

National Anthems are sung in such a specific way it is kind of a terrible test song lol

[brazen path]

As I feared, it's not pronouncing correctly even standard Guaraní (where "y" has a sound different from "i", apostrophe stands for a glottal stop, etc.) 😢

[untold briar]

You could probably trial and error find some ways of rewriting it to sound correct, but it could take a lot of patience and credits.

[orchid whale]

So when I continue does it actually continue the song? Or is it gonna be a bit different? First one got the same voice but not the bgm/cadence

[untold briar]

Can you post (or DM) the clip? It could be bad luck, where the clip sounds like the end of song, so what's happening is omore like getting a whole new song.

[orchid whale]

I just noticed that the end of the original clip is complete silence so that might explain it

[untold briar]

Yes that is a common result from silence. So in the "Continue From" field, find a time in the clip before the silence starts and you may have more success.

[brazen path]

Thanks anyway for your kind attention. 🙇‍♂️

[orchid whale]

Also for the "style of music" field should I keep it blank or make it exactly like the original work?

Oh it's working now

thanks jonathan

[untold briar]

It may be worth searching to check if anyone already created something like a 'romanization' way of writing it (like writing Japanese using English characters, but not just English) that uses one language to reproduce the sounds as close as possible.

[untold briar]

There is no clear answer. A lot of times the style field does not have a strong effect in a continue either way, unless the clip is very short, so it doesn't matter that much. But sometimes it works better with or without it. Or you may want to try and change the style to get the end of a song, for example.

[untold briar]

Correction: IPA can work, sometimes. I don't know why it works and why it doesn't, but at least once in while it can: https://app.suno.ai/song/340b0566-bafa-43a6-a2e1-1b2625179d0a/ That's supposed to be three blind mice and the first bit works okay.

[azure crane]

Is there a way to not having the instructions included in the lyrics when you are in simple mode? It seems that the use of [ ] doesn’t help.

[rose ore]

So I made a breakthrough today. Did an entire song in four takes, and got a perfect reading. It was the 12th time I recorded it, but what made the difference, I think was that I kept upvoting the good takes and the final song for every draft. And the machine just got better at doing it. It's also one of the more interesting country pieces I've done.

[finite dew]

@solid plaza Did you tried to “get whole song” before continuing? Making a whole song after each new part will result in more consistency, because stitching all parts together will give the ai a longer memory.

[tight zephyr]

Using ChatGPT with something like "Write this out phonetically in the xx language" works quite well especially with some few shot samples.

[nimble quiver]

I’m so curious about the details behind the new ability to continue songs from any point. I’m wondering if the ability was always there but just unfeasible to release (code improvements, hardware upgrades, etc). The new feature also hints that it might be possible to merge different parts of a single song - or even go back and fix previous parts of a song while retaining later parts. I’m curious about the challenges getting these new features into place. It’s super fascinating

I assumed at some point the model becomes sort of a black box, so I’m wondering how much of the process is offloaded, etc. Are there white papers I can start with to see how suno works? Or even non-suno papers using the same methods?

[solid plaza]

I will try it, thanks.

[neat field]

Start with Suno GitHub and Bark

I suspect chirp is bark, with some intuitive but incredible prompt/embedding workflow

Watch Microsoft Vall-E on the open source side, and musicgem from Meta on open source too

They will be watching Suno and I suspect they have similar improvements for open source within 1.5 years

Suno is just ahead, just better, than SOTA open source right now

My stance is suno should contribute something to open source around this as they establish revenue from their subs - they have first mover advantage

[serene jewel]

How can I make it show the lyrics in the final product?
It used to do that out of the box, but now it doesn't show it anymore.

[final hazel]

That's gone.. they are working on a new one.

[prime crypt]

Can anyone help me or guide me to the proper discussion? I have created one song, via browser, and now I want to implement diferent lyrics but to maintain the same singer and style of production. How do I do that? Eventually I would like to create 4, 5 songs in the same style and voice, just different lyrics

[mossy granite]

1st post! A question: is there really no way to render the tracks in higher quality than 192 MP3?

[final hazel]

The only thing you could do is use a DAW to montage it outside of Suno. In Suno keeping the same style, is not 100% possible because it never uses the "same" seeds (like it does with making AI with images).

[lyric sonnet]

I assume there is no way to change one phrase in the middle of an otherwise perfect song? Is there a plan to add such a region-edit feature?

i can re-render from the beginning of that section, but then the end doesn't fit

[final hazel]

No.. atm moment the Devs are still developing Suno and hope to be able to over higher quality in future versions.

[mossy granite]

You want a DAW like Auria. The potential for what you can do by cutting and pasting is pretty intense. Alternately, you could do it with something really cheap like AudioShare, but it would be much more difficult…

[final hazel]

Right now, Suno can't change just one part of a song, but they're exploring a feature for editing specific sections (knows as "paint-in"). So hopefully in the future we can.

[prime crypt]

So basically, this only comes down to luck, what kind of track it is going to make? 😄

I mean, you give the instructions and that's fine, but the outcome can be in the millions of different results

[mossy granite]

Thanks. Even 320 would be a huge improvement, the low end esp. gets lossy at 192. And the reverb loses air… maybe just as an option for finished songs, as opposed to every clip? I would definitely pay for that!

[prime crypt]

I would say that adding a function which says "copy" this track/result you've created, but put this lyrics, would be the best thing to have

[final hazel]

Absolutely, Darcela! It’s a mix of luck and your creative inputs. Each instruction can lead to a myriad of unique outcomes – that’s the exciting part!

[final hazel]

Try the button "remix"

[prime crypt]

But what happens with that button, not sure I understand how it works and what I need to do?

[final hazel]

It copies your lyrics and style of music (doesn't cost credits)

[prime crypt]

@final hazel if you are avaiable for a quick call, we can dicuss it privately for a few minutes

[prime crypt]

Yeah, but I want different lyrics to add to one of the results it has created

[final hazel]

https://www.suno.wiki/ this might help

[lyric sonnet]

I am using a DAW to master the Suno tracks I make but in this particular case it's not so easy; it's the end of the verse blending into the chorus and if i re-render that, Suno generates a chorus with a different melody

[final hazel]

I am sorry.. that goes a bridge too far for me as volunteer

[mossy granite]

Use “continue song”, keep the same style prompt, erase the old lyrics and add new ones.

[prime crypt]

Understand, but remix is not explained 😄

[final hazel]

Hit is.. and you will see

[prime crypt]

Tried, but doesn't really maintain the track 😄

[mossy granite]

You mean you want the exact same track?

[final hazel]

keep same song with new create.. is just not existing in Suno

[prime crypt]

Yes, same production in terms of bass, voice of the singer etc, but change the lyrics/update them slightly with new words for exampole

[lyric sonnet]

one thing you can do is keep the beginning of the song up to a certain time point and extend it with new lyrics (modify the lyrics before rendering); this preserves much of the song character

[mossy granite]

Maybe you need to try more variations?

[prime crypt]

Okay, and any idea/explanation on how remix works and what does it do?

[final hazel]

#📚┃suno-school message

[prime crypt]

This "It copies your lyrics and style of music (doesn't cost credits)" really isn't an explanation, through which I can really understand what does it to.

Does it create another variation on the same track/lyrics for free, is that the answer?

So If I input one final lyrics and prompt for style, I can create unlimited versions of it (for free/without credits usage) with the remix function?

[final hazel]

Funny enough this is how I explain it often. But I will say it in a different way "it copies the prompt". I just pastes the lyric and style of music from one song to a new fresh one, so you only have to hit "create".. that is it

Again.. what you want.. Suno cannot do. You cannot keep the same style and voice in a fully new song. Just doesn't work that way.

[prime crypt]

Okay.. I've now tried remix, but it did use my credits. I thought you've said it doesn't use them?