📚┃suno-school

[frozen pivot]

For me... Location is USA, Maryland

[charred merlin]

https://tenor.com/view/breakingbad-science-yeah-science-gif-5954775

[serene viper]

Ok

[charred merlin]

👉 If you are unable to "Get Whole Song", share whether your song starts with uploaded content. 👈

[rose jasper]

omfg, I got a woman in pain screaming and growing and making grudge noises, you'll love this Tree man!

[graceful condor]

let's take that out of context now

[rose jasper]

HAhaha, I'm talking about horror!

[wet minnow]

ok. get whole song from a extend of a upload doesnt work for me too

[rose jasper]

https://suno.com/song/07b783f6-73b0-4507-a9c3-03f3372fc5a8

[graceful condor]

it was working yesterday for sure

[rose jasper]

This might be a satanic summoning lyric hahahha

[frozen pivot]

It was even working for me a few hours ago

[serene viper]

YEs I was able to get ful song at 12:50 PM Eastern Time Zone

ANd when I try like an hour ago, starting having this issue.

[rose jasper]

Here's the one I mentioned earlier, thanks guys for the assistance, I got my desired result, "enjoy" wouldn't be a word I would say, to listen to this, but.... Listen too it, it's not pleasant hahha:

https://suno.com/song/06443f3f-98c6-4783-acf5-ca5364b35d8a

This is, not nice

But, wow, that Suno could make it

[wet minnow]

i think thats really really nice. super athmosphere

[rose jasper]

Yeah, it's wow... Yea

It makes you think!

[hollow hare]

Yeah, think "Oh man, how? Why? Why am I still listening to this? Huh, that is pretty cool, wonder if I can use it in my tune." 😛

[rose jasper]

I'm the same hahahah

You can see infront of you, just a woman in torment

[unique stone]

Recent upload -> Extend -> Get Whole Song = broken
Old upload -> Extend -> Get Whole Song = works
Old extension of an upload -> Get Whole Song = works

[serene viper]

Interesting

Well that sucks for me

[onyx linden]

Samesies.

Hopefully this isn't indicative of a change. I saw a new popup when uploading forcing me to further agree that I certify that I own all rights to the uploaded material.

[unique stone]

That's always existed since the audio upload feature was released, though I did have to agree to it a second time recently.

[onyx linden]

Well I am old and broken so... I could have missed it the first time.

[serene viper]

Yes, that old news.

[graceful condor]

also had to accept it rencently too

[onyx linden]

I blame Crowdstrike.

[hard mango]

This is the case for me. An upload from last night is "whole song"able but not any new uploads.

[charred merlin]

Thanks for following up with that, hamster 💙

[charred merlin]

...is that still ongoing? I know the CrowdStrike outage messed up a lot of things, but Suno didn't seem impacted at the time, and that was close to a month ago.

[hard mango]

here's an example of what's going on when you click "get whole song"

[onyx linden]

Nah man, it's long over. I work in cybersecurity so I blame them for everything now.

[onyx linden]

It's the cybersecurity version of "Thanks Obama."

I am still unable to get whole song.

Not my job. 🙂 I know it's not me so that's the end of the troubleshooting unless Suno's willing to get billed.

[hard mango]

when did you upload the audio that you are extending from?

i can whole song from files that i uploaded last night, but nothing from today

well the heart runs on electricity, so of course it's heart moving!

[graceful condor]

considering how much they want to lock us more and more, justified

[unique stone]

Get Whole Song gives a HTTP 500 error if you are using it on something that was extended from a recent upload.

Devs have been notified.

[austere dragon]

Any ETA?

[graceful condor]

they try to find out who's dog pissed on the server cables

nah really, that is a big issue, so they should post it in announcements once fixed

[austere dragon]

Suno Rats chewing through all the wires

[hard mango]

i've been refreshing every 5 milliseconds, still no luck

i don't understand what's taking so long, devs! just hit the undo button!

[austere dragon]

I'm afraid I've wasted a hundred credits before I realized it wasn't a problem on my end.

[untold briar]

Get Whole Song doesn't cost credits. Or it shouldn't...

[austere dragon]

oh? had no idea.

[untold briar]

If you did lose 100 credits though, that would be another bug too, let me know!

[quasi sonnet]

Hi, I am having an issue with my song I just made. it won't let me select "Get Whole Song" I press it and nothing happens

[untold briar]

If your song was created from an "Upload" then this is a known bug. Suno is aware. Hopefully they can fix soon.

[quasi sonnet]

gotcha, thanks

[untold briar]

Almost certainly just random chance, or at least the random chance that [something_blah_something] might work. I guess not completely impossible to imagine Suno left random Bark voice names in random music related code, so still worth a quick experiment, for sure though.

[hard mango]

its working again

thanks suno

@untold briar

[charred merlin]

https://tenor.com/view/ohhh-shit-a-rat-dance-oh-shit-a-rat-oh-shit-a-rat-gif-14187081

[uneven locust]

Hello I have a recomendation: when the song is generated I guess the video also is generated along the mp3 or wav file, the video is taking long time to generate, lets live it un-generated unless the thumb up is activated

3.5 model has no clue or loose the latino or latin styles (Cumbia, Bachata, Vallenato, Mariachi, Ranchero, etc} is a luck of the draw if the style is understood by IA and not to mention the difficulty for the IA to set the gender of the voice vocals

you can add an extra space to select the voice or voices like (male [with diferents tones and accents] female [with diferents tones and accents], male and female like duo singing, choir [male, female} or both

[serene viper]

Just got back and it worked

Thanks for escalating the scenario @wet minnow

[onyx linden]

Can confirm, the "Get whole song" issue appears to be resolved.

[astral flame]

im having payment problems on the website. how can i get help?

[hollow hare]

You will need to contact [e-mail]

[astral flame]

thanks

[astral flame]

lyric generator will not process lyrics ive only used 1497 / 3000 characters. ive had this in the past when too many words will make the lyric generator not work. but why make a limit if its not ture?

[charred merlin]

https://discord.com/channels/1069381916492562582/1259902708765167709 This is a known-bug -- if you use more than 1248 characters & click "generate" it will fail like this -- the current workaround is to create the lyrics on another platform like https://www.chatgpt.com or https://copilot.microsoft.com and paste them into the lyrics box afterward 💪

[astral flame]

thats what i am doing. i pasted lyrics in

[charred merlin]

Are you clicking "generate lyrics", or "create"?

"generate lyrics" takes anything you type into there & reimagines it as lyrics. If you just paste your lyrics in the box and press "create" it will sing them 🙏

[astral flame]

i dont see generate anymore i am clicking here

[charred merlin]

It’s the same thing as what I said before, just a different label now instead of “generate lyrics”

[astral flame]

wait. are you saying generate lyrics about anything and then paste in the ones i want instead and then use the create button?

[charred merlin]

#1270928841614823546 message

You only need to press "write about" / "surprise me" / "generate lyrics" if you want to take the words you've entered & ask Suno to come up with entirely new lyrics based on those.

For a SUPER clear example, skip to the last demonstration in the video I just linked 🙏

So if you're asking Suno to generate the lyrics for you, you're not able to use more than 1248 characters before pressing "write about" -- BUT, Suno will attempt to write a full song's worth of lyrics based on ANYTHING you put into there if you press "write about"

[astral flame]

oooh! so if i add my own lyrics i dont need to hit write about!!!! omg! ok! thank you

but i did want suno to do it since it sings best to the lyrics it makes. but now that i know this about the other lyrics its super helpful

thank you!!!!

[charred merlin]

🙂 Happy to help -- I would wager that the "generate lyrics" function is tied to the 1250 character limit that v3 is limited to, as v3.5 can use 3000 characters, and I wasn't able to replicate the same error if I used <1248 characters. If you upvote this post in this link for visibility, that way the devs will know that it's effecting people still 💪 https://discord.com/channels/1069381916492562582/1259902708765167709

[astral flame]

it wont let me upvote it

[charred merlin]

😅 My bad, I didn't check that the post was still open, you know what they say about old habits. 🤦‍♂️ 😛

I think it's at least on their radar regardless, and since there's a non-intrusive workaround, it might be lower priority at the moment.

Either way, I hope your <t:1724219280:t> is going beautifully & best of luck with the song -- I'm sure Laura will love it 🙂

[astral flame]

thank you kindly! and i hope your well whatever time it is for you too

[foggy wing]

Is there a way to hide the prompts when sharing the Suno link? I am all for us all developing and sharing with each other, but there are instances where I would not want it to be seen.

[graceful condor]

edit the "lyrics" to remove tags?

[sand moss]

Is there list of all tags possible?

I need tag for poetry lyrics to be read instead of sung

[wintry saddle]

[Spoken Word]

or [Spoken]

generally works before it

and use less musical styles in genre list

put Spoken, Poetry etc...

[naive quail]

"Spoken Word" and "Poetry" are a good start. I've also used "Audiobook", which seems more inclined to pull in some background sounds/dramatic music, and even "Disclaimer". https://suno.com/song/173ac910-bc00-4c67-8834-ca70eac7de79

[jade mist]

It's over for you, Suno! Stop claiming you don't clone existing songs. And let us choose an artist style without saying oh, it's copyrighted. Is this why they're taking time with V4?

https://suno.com/song/d82d28d8-d61b-4a2 … aca9bd3f03

[bronze helm]

The way generative AI works is it is trained on a large dataset and it then produces original content based on its training.

The reason why we cannot choose artists is that as long as AI is transformative, its outputs fall under 'fair use'. The moment an artist's style is imitated, it becomes more complicated.

[jade mist]

It made this: https://suno.com/song/d82d28d8-d61b-4a2f-86a1-e7aca9bd3f03

[fallen flare]

I don't know which channel to write to, I'll write here. It's extremely difficult to get suno to make folk music, it constantly rolls it into pop. For requests like "nordic folk" and others on my text, I managed to get only a couple of real folk music from a large number of attempts, not pop or something with a guitar

[hard ruin]

The tempo or BPM often gets messed up automatically, jumping from the original tempo (e.g., 90) to 160 without any explanation. This ends up wasting unnecessary credits, leaving users with no option but to discard it.

Audio playback repeating option missing 🔂

IOS APP

My entire Playlist collection not doesn’t load in the SUNO App

[naive quail]

They've known about this one for a while, so it's likely somewhere in their todo list. Give it a few months... https://discord.com/channels/1069381916492562582/1257470377211465750

[gusty acorn]

I would like tips on stems separator. The stems from Suno are still making really poor quality.

[violet tundra]

how to clean up suno vocals / AI background his for 3.5? I usually keep generating until I get the cleanest possible "recording," but I'm trying to get actual "human" vocals. Tried cleaning it up with Kits.ai and Controlla as well as Ace Studio. No luck.

[violet tundra]

that's because they are not stems ... their not even instruments... it's just one long AI-generated interpretation of music with different frequencies and patters labelled "drums" vocals, etc. Even if you use Lalal.ai. the stems need to be generated on the server side, individually by the AI rather than by stripping them from a composite.

[raw marsh]

Hello! Does anyone has a list of commands that can be used to make custom music with our lyrics? For example [chorus] works great. but I don't see anywhere an explanation of how to use these commands, how many are there

[coral plover]

is there a better one by now? 😅 @naive quail @neat nexus

[neat nexus]

https://suno.wiki is the unofficial community wiki that I am familiar with. I do not know who is behind that other one.

[coral plover]

aaahhh i see. i was looking for this one but my google spitted out the other link

[coral plover]

https://github.com/panyanyany/sunoaiwiki.com

thats the user of the other one. doesn't seem to be problematic or fake but haven't checked further

[neat nexus]

It's overly complicated. And I'm not convinced that all that does anything. Or at least it's not doing what it says it's doing.

[naive quail]

I haven't looked through that one to any great length, but the the bits that I have seen in regards to style prompting... meh. I haven't been able to reproduce that stuff with ANY level of consistency and I spent a TON of credits playing around with it. This alone makes me cringe:

[silent saffron]

/chirp

[flint haven]

The /chirp command is no longer working

Chirp is retired, and no longer on discord. You can make music now on our website: https://suno.com

[gusty acorn]

Yes, I know. I should have wrote "stems".
Still though. There are applications that can extract stems from s song using AI. Even get VST for the instruments.
I'm interested in recommendations.

[alpine venture]

it's a link to a tutorial

I'm just using Instagram because I can't keep up with 5 different social media streams, and YouTube is my weakest

[crimson widget]

Hi guys, just wondering what's your plans on API access?

[upbeat idol]

Hello guys, I often have a problem with the first word of the song when the track starts directly with a lyric. It is not often the case (like 1/3). I can add an instrumental intro to avoid it but I don't want to make a longer song. Is there any way to solve this issue ? Ty

[mystic flume]

The lyrics it makes are a good guide, but I wouldn't say Suno sings "best" to them. Use them as inspiration if you find a particular song you really like but want to use your own lyrics. Often I'll rewrite my lyrics to fit the song Suno has generated from them, just to help smooth that interpretation. It usually results in a quality result (eventually). This route can be expensive, but if you have the 10,000 credits plan you can afford a little trial and error.

[austere dragon]

Anyone getting a preponderance of fingersnaps in 3.5 generations?

[bleak gust]

Hey suno on my computer isnt working properly, Its not loading up my page

[frigid sapphire]

Posted this in general but maybe I should've posted it here instead. I got my song flagged for its "content" so it can't be made public. I primed the song with me saying "one, two" and mainly used ChatGPT for the lyrics, so I have no idea why the song was flagged. No help available on site so I thought I'd check here.

[naive quail]

It's related to that "one, two". If Suno detects voices in the uploads (or anything that it THINKS is a voice), it will prevent any resulting tracks from being made public. 'Not sure what the logic is, but that's what we're stuck with right now.

[formal tree]

hey how do i make it sing with an accent again ?, i want to make another Lullaby but need an accent to perfect it.

[violet tundra]

Lalal is the best. https://www.lalal.ai/

[languid junco]

Im using UVR5

[cunning mulch]

so, what's the absolute earliest time you can Extend from in a track?

I'd like to at least keep the initial 'seed', so 0:00 is just fine for my purposes

[coral plover]

hahaha

i will definitely test some too and give feedback. maybe there are some that do the trick

[fiery phoenix]

Hello. does Suno provide access to own API?

[mystic flume]

No

[mystic flume]

0:01, but it doesn't work like you're probably hoping. Each extend has its own seed as well. So you'll get something completely new.

[cunning mulch]

.... :/

yeah, I'm not a fan of having to throw out bangers that just need small fixes

[eager topaz]

I use RipX DAW Pro, split to stems and clean the vocal. It's a great app, probably THE best for the job. They have a 21 day free trial so you can give it a try.

[alpine venture]

https://suno.com/song/fd40fec5-3bd5-4907-928f-92f08baa63f3

this is another in my recent few where I've been re-exploring the noise growth in AI gen music, now that its way easier to control

I personally think its one of the most unique sounds in AI music, and the most musical nuance we currently have in it, so I thought technical might find it interesting

I used to play with this a lot in v3

but now its much easier to just shoot for clarity, but with this, I get the best of both worlds, and noise growth outros are really fun

and like, its just so unsure of itself, I like it

its kind of like you get it into a position where it has noplace to go but up, it wants to go five directions at once and also none

[shut light]

Is there a way to hide, clean up the prompts on what is displayed?

[charred merlin]

#1270928841614823546 message

Anyone can edit lyrics, but only paid accounts can "regenerate video"

[shut light]

I am a subscriber:)

I mean this

[neat nexus]

Not the style prompts. Paid plans offer the ability to edit lyrics. (I think this is a paid feature like setting the thumbnails. Check and see if you have the option or not to edit the lyrics prompt at the end of the lyrics to see if you have access.)

[shut light]

Edit lyrics was a free feature I think and pictures are paid.

[vast bobcat]

So how long will the "flagged for moderation" continue. It comes and goes and it feels like it's on overdrive again?

[hollow hare]

Does it occur when you try to Generate Lyrics? (or as "Write About" nowadsys)

If so, that is ChatGPT — the moderator for the lyrics — saying it did not like something in the lyrics (whether it be certain words or an imagery that GPT did not like).

Try modifying the problematic portions. If you do not know what caused the issue, you can ask ChatGPT directly to look over the lyrics and list the issues

[rough dune]

hello ,nice. who knows what time or how can i use the https://sonoteller.ai/ , it always results "Our total free daily limit for new tracks have been reached" .. what can i do to use it while i need it ? thanks

[charred merlin]

Out of curiosity, have you attempted to contact sonoteller's own support service?

At a glance, it looks like the website is just a playground-style front-end demo for people who want to use their API services? Maybe they only offer a limited number of free uses through the website (across all users, so people can be greedy and hog the public usage limits, but since it's a simple demo, and their paid services are their bread and butter, they might not care about paying for a fix/ui redesign)

Either way, it looks like [e-mail] is their contact, or there might be more info at https://x.com/sonoteller

SONOTELLER (@sonoteller) on X

AI engine to analyze music, including song lyrics summarization and musical features such as genres, instruments or moods. Part of @chordalcurated

Twitter

[narrow warren]

So I followed the local instructions on https://huggingface.co/suno/bark and it says no gpu being used. How would I select my gpu?

[rough dune]

Yes, I have also taken these into consideration. But I am also willing to pay when there is a real need to use it. Just can't find the entrance, and there's no reminder on its public webpage to download the API application. so i am confusion for that.

[charred merlin]

If you click the 3-line dropdown menu on their website, they have an announcement about how they're joining "Chordal", which looks like they're still working out the logistics thereof

[charred merlin]

Oh, looks like my reply from earlier still stands -- you need to reach out to [e-mail] to let them know what you want to use it for, otherwise it looks like they're more interested in providing the feature as a service through this "chordal" platform -- looks to be free to join, but I'm already way off topic by not talking about Suno stuff

[rough dune]

thank you for my doubts . you are so nice

[charred merlin]

💙🫵

[warm rune]

Hi folks, if you are working on audio models (ASR, TTS, or multimodel LMs) and trying to model audio using tokens, do checkout our library audiotoken on GitHub here: https://github.com/cmeraki/audiotoken. You can use this library to create audio tokens (think of this as tiktoken, but for audio). We have spent a lot of time in making the tokenizers fast and easy to use. Please feel free to try and share suggestions if any!

You can install it via PyPi (pip install audiotoken).

[narrow warren]

I can't seem to find how to use a gpu

[prime prairie]

Getting really fed up with having to extend a billion effing times before it will FADE OUT!!!!

[narrow warren]

Also, it sometimes completely ignores input text

[prime prairie]

This too! 👆🏼

I just LOVE when it will go through the first couple verses and get ready for the bridge, and then jumps back to Verse 2 all over again, then skips the bridge altogether!

[untold briar]

Can you link to the exact command or script you are running? Although maybe Bark discussion is better in Off-topic, even though it's technically still Suno, it's not common in this channel any longer.

[mild rain]

Why not just export the WAV file and drop it into Audacity (free) and fade it out using that, it will save you a bunch of credits and you can select the fade-out point and length.

[prime prairie]

Thank you! I'm pretty new to all this stuff. I don't know how to do most things. 😂

[mild rain]

Audacity has a really good help system built in and there a tonne of videos on YT. It would literally take you less than 10 seconds to do.

[prime prairie]

Cool, thank you! 👍🏼

[mild rain]

Anytime, if you get stuck hit me up.

[prime prairie]

May I friend request you?

[mild rain]

I was experiencing this too but one thing I found is everytime I had a verse written as [Verse 2] it repeated that verse twice once I changed every verse to just be [Verse] no matter if it was verse 1, 2 or3 then it didn’t repeat and went to the bridge correctly or to the next part correctly without repeating. It was so frustrating.

[prime prairie]

I will try that. Thank you!

[torn remnant]

Guys, can you give me advice on creating 15-20 second videos?
I always end up with videos around 2 minutes long. Maybe there's a specific tag for this?
This translation maintains the informal tone and key points of your original Russian message, asking for advice on creating shorter videos and inquiring about possible tags or techniques to achieve this.

[untold briar]

What is the reason you want only 15-20 seconds? Are you trying to end the song?

If you are trying to end the song, try using model V3 or V2. And style "end" and in lyrics "[end]".

[torn remnant]

Thank you, yes I tried to end the song, but in version 3.5 it continues even after the end tag.
I haven't tried it on other versions.

[cunning mulch]

That's weird. I never had a problem with it doing that, but my problem was repeating lines from the first verse before even first chorus

wound up solving it by structuring as 6-line verse, 2-line pre-chorus

[untold briar]

When Suno is repeating verses, don't just 'retry' over and over if it isn't working. Change the tags: [verse], [verse X], or even no verse tags. And change the Extend From times as well. Sometimes even 1 second difference can change things. And if you need to, backup further.

[cunning mulch]

At first I wasn't even Extending

it was doing right from 0, and no amount of tweaking solved it

it just absolutely detested the idea of an 8-line verse, followed by a chorus

[untold briar]

Adding slight variants into the lyrics themselves, an extra (Ah) or something, can help too.

Make sure to try without any verse or chorus tags. If your song is less typical in structure, that can be good.

[cunning mulch]

okay...

Here is the original Part 1 I wound up using. It wraps up the verse just fine. What on earth would compel it to repeat lines 7 and 8 again? https://suno.com/song/d3e1dda8-4539-497b-91a3-a9a0a71334c0

at 44 seconds in (ignoring the flub at 35 seconds, trying to read the word 'our' and mixing syllable counts)

[untold briar]

So, that's a pretty complicated prompt in general, so maybe it's just Suno getting confused overall. But...

Have you tried without this? ** [To Chorus]** ?

The threads of hour heritage, that weave our hearts as one,
Would be torn apart, and left to be un-done! [To Chorus]

That's a structure tag in an unusual location. If I had to guess one thing, it'd be that.

[cunning mulch]

that was me trying to get it to stop repeating lines 7 and 8

I was throwing things at the wall and seeing what stuck by that point

even Extending, I couldn't get it to stop doing so, until a point where it choked the end of the verse into gibberish

here's the Merged result: https://suno.com/song/95818bde-cb5e-4706-933c-41fe8677ec20 (edited Displayed)

any other time point resulted in it going on repeating

burned through a good 1300 credits on monthly, AND a re-pop trying to get this to work

I've already considered dropping back to basic in sheer frustration

and yes, using the word "hour" in its place was me saying "Okay, lean in to the two-syllable version used"

I mean, it's stuff like that that really makes me want some ability to simply re-generate vocals from lyrics. It already has some idea of the contents of lines, so there has to be some way...

hell... you wanna know HOW desperate I was to try and get it to re-do what I had before? I was sticking a tag like this into it: [Melody And Vocals: 95818BDE-CB5E-4706-933C-41FE8677EC20] (to try and say "Hey, do that again!")

[charred merlin]

Seeing as how UUIDs are intended to be used as unique identifiers, I wonder how Suno would look at this string and try to associate it with any patterns it learned during training.

That said, using unique identifiers for repeatable voices should be very easy to test the efficacy thereof, though it feels more akin to "coding attempts" -- attaching a picture of the UUID run through ChatGPT's tokenizer to see how Suno would be "reading" the UUID, but more sharing as a curiosity or visual aid than a point in itself 🙏

[cunning mulch]

like I said, I was into the 'madness' phase where I was grasping at straws and trying anything I could think of

[charred merlin]

Sorry, I wasn't sure it was intended to discuss different methods you've attempted so that others could collaboratively build on the ideas, or if you were casually sharing a funny anecdote about the lengths you've gone to in the "madness phrase". ☕ 🙏

https://youtu.be/KwIe_sjKeAY If you'll excuse me, I have my own "Madness Phase" to get back into 💪

[cunning mulch]

mostly, I was just praying it might somehow manage to retrieve anything from that generation

[indigo smelt]

The only way to chop off part of a song is to export it and edit it in audacity or similar, correct? I want to chop off the extra instrumental bit suno added at the end of this: https://suno.com/song/1b46e587-f722-4c82-9d3b-694c6093d29b

[vestal rapids]

I use ffmpeg on the command line on my Mac.

ffmpeg -i input.wav -t 00:02:17 -c copy output.wav

[indigo smelt]

I'm more comfortable with audacity myself, just was hoping there was a way to get the corrected song in suno.

It's not difficult to fade out in audacity, I just was hoping I could have the full song in suno.

I may do it anyway, just so I can add some fading crackling flames at the end.

[obsidian siren]

/chirp

[flint haven]

The /chirp command is no longer working

Chirp is retired, and no longer on discord. You can make music now on our website: https://suno.com

[simple osprey]

is it possible to create a background choir that sings like "ooooh ohhhh" in the background while a singer sings the normal foreground chrous?

[cold rapids]

Hi, is there a way to make a crowd of people chanting for a chorus?

[jagged narwhal]

Suno was perfect but now in period I guess XD too robotic

[ivory adder]

I used the male vocals tag in the style, but it still generates a female chorus. How can I make a song with only one male voice?

[lapis rune]

I have some quest I write "Female vocal" and in stye of music I write "Female vocal". But when music are generated the sound of man are singing

[indigo smelt]

I saw this trick the other day. Maybe check if it still works: https://www.reddit.com/r/SunoAI/comments/1d5bmja/duetmulti_voice_in_35/

[hoary cobalt]

HI ! Is the “flow slow” prompt taken into account by the AI?

[eager knot]

Anyone else having trouble to play a song? Since yesterday I have troubles to let Suno play songs

today it worked fine until now, reloading etc. nothing helps

[jade mist]

Is there still no news about V4? They said it's in the works already in March but still nothing? Or is there a thing again in the .json or whatever that says model name V4?

[graceful condor]

Hey guys, I wanted to ask a question. If I had a song made by Suno (I wrote the lyrics), can I cover it, use similar instruments for the instrumental, and publish it using the free plan?

[charred merlin]

No news as of yet, it'll be in #📢┃announcements or #🚀┃updates when it comes out, so make sure you have notification settings ready!

[fringe steppe]

I am not realy happy whit the song I missing whit the vocies for the song so will sound more like a borderlands 2 song need help whit what type of comands there are and effects comands

https://suno.com/song/794bc7fa-8128-4369-98b4-47e5f88daa3d

[charred merlin]

If you're simply looking to share the song, I think attributing the ownership of the song to Suno covers your bases, but if you're looking to monetize the content, you'd be better off contacting a copyright lawyer or reaching out to [e-mail] to figure out the best way to handle the matter.

Considering best case, worst case & most likely cases, releasing your own version of a song that Suno has ownership over is a legal grey area to begin with, further considering the biased source of where you're asking. I'd imagine that the best case scenario is that you need to pay some form of cover fee to a licensing agency like Harry Fox Agency, similar to publishing any other cover song.

Worst case is you get taken down for infringing copyright, or worse still, legally Supplex'ed for attempting to subvert or circumvent "legal ownership".

for the record, not a lawyer, none of this is legal advice -- you do you 💪

[graceful condor]

Thanks sm!

[wide ore]

I need assistance. I have had a "Pro" subscription for a week, but I can't download "WAV" files. I was able to download one song on the first day of the subscription, but since then, I constantly get the error "Download failed, please try again."

My Specifications:
• System: Windows 10
• Browsers: Chrome, Edge (cleared cache), mobile browser Chrome
• I do not use a VPN

I’ve tried downloading at different times: in the morning, afternoon, evening, and night, but I always get the error almost immediately after the notification about the WAV file generation disappears from the bottom of the screen.

I’ve encountered similar issues, but there has been no response or solution, at least none that I’ve found. From what I’ve read, it seems to be selectively related to profiles, independent of the user's hardware. In other words, there might be a server-side error tied to specific profiles, but that’s just my assumption.

Is there any way to adjust the WAV download settings that I can try? I feel like I’ve tried everything. My profile is "@nick_fx," and I haven't yet posted any songs publicly. I plan to upgrade to "Premier" if I can resolve this issue. Thank you.

[charred merlin]

https://discord.com/channels/1069381916492562582/1252656600770940968 It's a known bug, I'm not sure what's been discussed internally regarding the matter, so probably best to email [e-mail] to let them know what the issue is and what steps you've taken to try to resolve the matter so they can give the most informed response.

[hollow oar]

Hello I need to speak with someone with an issue within the billing section.

[hollow hare]

You will need to contact [e-mail]

[vapid nymph]

Do things like "solo" and "shredding" work as style prompts or better used as meta prompts?

[charred merlin]

Style prompts effect the entire generation, while metatags are localized to whenever they'd be interpreted as lyrics 💪

So it really depends on what you're trying to accomplish.

[vapid nymph]

Well I've been trying to insert bass guitar solos and the issue I'm running into is getting just guitar solos, no solos, or like 3 second solos. I've also kind of noticed it has trouble with "bass guitar" as well. Wondering if hyphenating it might change the way it's interpreted by the a.i.?

[charred merlin]

Hyphenating absolutely changes the way that Suno reads the inputs -- lots of previous discussions regarding tokenization if you search this thread's history. @naive quail has tried to tackle bass solos, and 5 days ago came to me hat in hand, blubbering like a fool, saying "Suno ignores all bass prompts".

Despite his peculiarities, I respect the hell out of the dude, and if he says that it's difficult, I'm comfortable siding with him, knowing the ridiculous amounts of work he puts into his songs & experiments.

[vapid nymph]

Anyone who fights for the bass is a hero lol

Man that's kind of a bummer

[charred merlin]

That said, I'm sharing secondhand observations without knowing intimate details about how he tried to approach the matter (given his username, it's fair to assume he focuses more on guitar)

I'm confident that you can make ALMOST ANYTHING within Suno, given enough time and effort, and creative & experimentative use of style prompts & trying to grok how their respective weights effect one another.

For instance, you'll go blue in the face trying to directly prompt "theremin", but it didn't take long before people noticed that "spooky halloween" or whatnot routinely includes thememin because of the association.

Similarly, when I make Indie Math, it almost goes SUPER complex, and I need to pair it with thematic descriptors like "sleepy", "winter", "considerate", etc to "calm it down" and get more of the Delicate Steve vibe I'm going for

[vapid nymph]

@naive quail care to give me a tl;Dr of what you discovered regarding bass prompts?

[scenic sequoia]

that explains so much. i thought it was just me having issues with bass

[vapid nymph]

Same lol why I finally brought it here

[scenic sequoia]

right, i was trying to make a dubtep song and ended up with nothingggg close. but it sounded like a pretty good edm song

[charred merlin]

@vapid nymph I was sharing my experiments and encouraging likeminded-individuals to collaboratively contribute over in the pool party up until recently > https://discord.com/channels/1069381916492562582/1256301881966198805 -- you can't post there anymore, but you might be able to find something that helps get an understanding of how I'd break songs into smaller chunks

come to think of it, I never got around to posting the basic tutorial workflow of how to use extensions -- I should probably get around to finally making that

😂

[scenic sequoia]

cheers, ill take a look, i just found out about that u can add tags to influence songs, im addicted to this site so im always down to learn somthing new.

[naive quail]

I've discovered that Suno hates bass... and is also not particullatly fond of solo drum stuff... and is also vindictive. I waaaaaaay over-prompted for bass and Suno gave me ALL percussion. lol https://suno.com/song/a0881d67-ddbf-48de-b0bc-891eeb492f11

[vapid nymph]

🤣🤣🤣🤣🤣

Did you try "bass-guitar" or cello?

Maybe slap cello?

[naive quail]

Slap Suno might be more rewarding 😄

[vapid nymph]

🤣🤣

[naive quail]

I sat on the other one that was paired with that render for a couple weeks before using it for this track (after maaaany extensions): https://suno.com/song/7fa600c0-2f75-4d32-aefb-e1ec1dac442f

[charred merlin]

Legit one of the more enjoyable "musical conversations" I've heard in some time.

(close second to The Duel from Metalocalypse https://youtu.be/gBYQAR8fG1w)

[charred merlin]

Getting way too theatrical and colourful for a moment; following the musical-slaughter he just put on in "Some Time Ago...", I absolutely love how Skwisgar starts off at 1:00 with a "low-effort" shred, almost waving Toki off dismissively; Toki panics and tries to counter by turning it into a call & response (1:17, raised eyebrows -- oh shit, this might actually be something)

1:20 it turns into a quick paced aerial dogfight as Skwisgar tries to outpace Toki by throwing more and more layers of complexity, (love the imagery as Skwisgar becomes a Thunderhorse and Toki becomes a Flying V-esque eagle)

1:38 - 1:50 , the gloves come off and Skwisgar is out for blood, barely able to compose himself, and Toki doesn't back down for a second, giving just as well as he gets

1:50 > 1:55 they actually start realizing how evenly matched they are and develop a shared respect, completely unrecognizeable to those "outside the conversation", and even though it looks like they're trying to tear out eachothers throats, they've finally found someone to challenge them in a way that could make them improve.

2:10 -- "we might actually be onto something. Are you feeling what I'm feeling?"

2:13 -- "🤘"

2:15 -- snapping back to reality, they're no longer depicted as bourgeoisies, but rising above it all, followed by the imagery of the Thunderhorse and Flying V

better together 💪

ffs, I thought this was #🗣️┃chill-chat 🤦‍♂️

[silent viper]

A custom GPT to extract tags for the "style of music" field.

Indicate the song/artist (you can paste the YouTube link if you want) and ask GPT to analyze and extract the tags to copy into the "style of music" field.

Suno Prompt Master v2:
https://chatgpt.com/g/g-SY5QcDEVS-suno-prompt-master-v-2

[nimble iron]

When I think of guitar converstion, my mind goes to this clip
https://youtu.be/XT8KS4J0gLg?si=WtitgJRXvpWcuFyI&t=36

*also a show by Brendan Small!

[charred merlin]

He - Is - Franz - Kaf - Ka

Franz Kafka

[vapid nymph]

Hot take: without Dwayne we would have no dethklok

[hallow pollen]

I need help

Someone who can help me with something?

This is a Suno interface related inquiry

Moderators, guides, team?

@charred merlin

I require thy knowledge oh wise one

[nimble iron]

Everyone always returns to the harbour harber

[hallow pollen]

Yes, need some help by the techie peeps

Moreover, ask a favour

Thing is that they would probably have to change something

So perhaps it is a delicate thing to ask

Anyway, will leave this here just in case someone from the team reads it

I need more characters for the lyrics

That's it

You might ask: "7500 is not enough?"

Well... Not in my case

Why, do you ask?

I made a 22 minute song

With 6 parts

So yeah

[hollow hare]

Er, is there an issue with Extending? @hallow pollen

[mental ridge]

site is really slow atm

[hallow pollen]

No

You see, I made a 22 song, the song is ready, but I want to put the lyrics to the song, the complete ones

7500 characters are not enough, sadly

In the lyrics field

[hollow hare]

Oh, no way to do that in Suno. Ideally, you should have been adding the lyrics as you kept Extending

[hallow pollen]

Yes, but hit the limit

Bummer

[hollow hare]

Ah, I was not aware that there was an upper limit to character limits (aside from the 1250/3000 for each part you Extend)

[hallow pollen]

The limit to each song is, apparently, 7500 characters

I would need uh...

At least 1500 more I think...

I made a monster of a song

I could probably send you the whole thing here

But they would probably ban me so

[hollow hare]

Well, nothing much we can do to help you here, though I am reaching out to some people to ask them about that. I suppose you can ask in suggestions or bugs but the best method would likely be to reach out to [e-mail]

[hallow pollen]

Yeah I might do that lol

Thanks for the help though!

Sent an email

Hope they respond kek

[maiden sparrow]

maybe a point and click ability to tip people by the cent for their song or tip in credits. like micro tips or even nano tips. some kind of emoji like currency

[hallow pollen]

Idk who you are, but yes

Give this peep a seat in the board

[glacial scarab]

You are only talking about the displayed lyrics right?
If you made multiple parts you should be able to work around this. Add most of Lyrics to part 1 add rest to part 2 get whole song. Now you should have above max.

[hallow pollen]

Yes, displayed lyrics, but I reached the limit

Song is just too damn long

Lol

[glacial scarab]

Ah yea that can happen. But my way should work. Each part can have 7500 characters and the get whole song combinations ignore the limit. You might need to add a silent part at the end if you already merged everything though

[hallow pollen]

Hmmm

Makes sense

Might try that

Thanks pal

[charred merlin]

https://cdn.discordapp.com/attachments/1266927804751151117/1276331824045363344/staffguidesmodstheharber.png?ex=66c92423&is=66c7d2a3&hm=6d81bdffa300a1c165840477f96754ee638d17304644e6ec2c8f3299798db7ed&

[hallow pollen]

Amazing

[naive quail]

no feet tho...

[charred merlin]

Sorry for the delay, finally got around to watching UFC 305 -- absolutely wild.

I'm scrolling back through chat now, but I've not heard anything about a 7500 character limit -- as Paimon noted, the characters are only counted/considered when generating the clip, afaik.

For example, if you upload a clip with a rapper singing really quick lyrics, the character field was never used, and it's just personal conjecture, but I wouldn't imagine the Suno model itself is listening to the lyrics, transcribing & placing token limitations on them. (there's likely a separate service working in conjunction with the model to make sure that people aren't using Suno in bad-faith.

That said, I don't doubt that you've run into SOMETHING weird regarding how Suno is outputting content if you've been extending for a bit trying to get a specific outcome -- I'd be curious if you're able to consistently demonstrate the behaviour, especially if it's presented in a way that others can easily understand and attempt to reproduce.

[hallow pollen]

Awesome Mr. Harber I will def try to extend and see if that solves the issue

If not, I have sent an email to our Suno overlords

For them to instill their wisdom on me, a mere mortal

[delicate idol]

For my eyes only dude. And ears too…

@charred merlin (“Let me see dem feets!” 🎶)

[devout rune]

Heya, I encountered an unnormal behaviour regarding generating a song. I prompted this: film-noir dark jazz
and I got someting like Rock/Metal thing.
https://suno.com/song/d73dfa76-c49d-47da-8513-6eeef1c178a4
This sounds great, no questions, but it is not jazz and even nothing like bladerunner-ish sountrack

[charred merlin]

v3.5 veers toward "generic" without effective use of prompting or a strong musical context to "hook onto" when extending from a previous clip.

Have you tried v3? As the model dropdown suggests, it's a lot more versatile and broad in its outputs, while v3.5 REALLY excels at solidying the song structure and lyrical cohesion once you've created the song.

[rain spear]

Hey Guys you just have to get rid off the hissing in the back ground. I ripped the files to tracks but the synth is terrible exteemly bad when lound. Strings are even worse. You just cant clear the noise. It ruins most songs completley useless.

[leaden lark]

Can say, going from 3.5 to v3 and back seems to be kinda of interesting

[leaden pawn]

Is anyone else having trouble with adding songs to playlists from the create screen?

I get the pop-up window, but its transparent and I can't select anything. I've reloaded the page, but it didn't fix it. First noticed today - was working yesterday (EST time zone).

I also cannot close the window, I need to refresh to remove it.

[loud gazelle]

After I sent another mail like yesterday the Suno-Team finally made my account come into use again and the "Create" Button does work again

[wind venture]

I don't have membership at the moment, but I remember the other day I was able to generate or upload a cover image. Today I don't really see the option anymore and it seems to be locked behind members. Did something get changed recently?

[charred merlin]

#📢┃announcements message Only been available to paid users for as long as I can recall.

[wind venture]

ah okay i must be misremembering then, thanks for clarifying

[charred merlin]

Not a worry 💪

I just checked on a free account, it looks like you can still see the section, but it's subscription-gated

[normal warren]

been getting this when trying to export .wav anyone else?

[frozen pivot]

Have you tried again?

[normal warren]

its every song unfortunately, i think it because i have created over 300 songs in my library

Ive probably made and deleted over 1500 songs but only have kept 300 ish. Im sure the average account holder doesnt do that many transactions.

[frozen pivot]

I do have more, unfortunately... Try clearing your browser's cache or try a different browser

[trail fulcrum]

It's a shame that they still can't manage to improve the audio quality. And that you don't get a single midi track so that you can rearrange it in your DAW.
Maybe also consider offering all tracks not just instruments and vocals.

[frozen pivot]

Midi tracks has been a hot topic over the few moths in the suggestions groups. Let's see what the next versions hold

[alpine venture]

https://belladoreai.github.io/llama-tokenizer-js/example-demo/build/

wanted to point out that there are LLama tokenizers as well

so likely we could do some tests with words that are tokenized differently

to determine which way the model is tokenized

I would think LLama, just bc it is easier to fine tune, but the deal with msft a while ago made me assume gpt

in any case, it's often very similar

"The LLaMA tokenizer is a BPE model based on sentencepiece. One quirk of sentencepiece is that when decoding a sequence, if the first token is the start of the word (e.g. “Banana”), the tokenizer does not prepend the prefix space to the string."

[untold briar]

Suno Bark TTS uses "bert-base-multilingual-cased" but a quick search can't find a playground or visualizer for that exactly. This has one Bert: https://huggingface.co/spaces/Xenova/the-tokenizer-playground or maybe you can past the name into 'custom' and it works, not sure if it is, or the UI is using a default.

[runic stream]

The playlist cannot be run today, and I am unable to add music.

[untold briar]

Are you on a desktop browser (Chrome, Firefox) or on a mobile device?

[untold briar]

I see the bug. The "Add to Playlist" menu option is missing on the right side of the screen. But the menu is still here:

[tulip pasture]

is it possible to create songs with separate stems already?

[untold briar]

It isn't possible to create songs with stems, from the beginning. But for paid users there is a "Get Stems" menu option in the "..." menu. On a finished song, it will seperate vocals and instruments (though not perfectly).

[jagged tinsel]

Здаввствуйте, я случайно загрузил трек, на который нет моих авторских прав, что делать ?

[runic stream]

yes, can not work.

[untold briar]

This looks like a bug. I see some other users with the same problem already reported it here #🪲┃bugs-web message Let me see if i can find temporary workaround...

[untold briar]

Try forcing a 'Full Refresh' in your browser. In Chrome: Ctrl + Shift + R. Or maybe: Ctrl + F5.

[maiden sparrow]

Thanks for the comment. Wrote a song on it but needs a rewrite. If you have heard the expression "nickel and dime you to death" well, it'd be nice to have ability to tip an artist 5 or ten cents and flip that and maybe Nickel and Dime you to Life https://suno.com/song/aa68f09b-f29a-4efa-b500-1791db799e1f

[lusty basin]

hello! what is the quality of mp3 / wav after downloading from suno with subscription?? thanks for the answer!

[vestal rapids]

the same

[eager knot]

It's very annoying that the music player just stops randomly or won't start playing my songs. I have a hard time to work with it rn...

[charred merlin]

https://discord.com/channels/1069381916492562582/1252656600770940968 It's a known bug, I'm not sure what's been discussed internally regarding the matter, so probably best to email [e-mail] to let them know what the issue is and what steps you've taken to try to resolve the matter so they can give the most informed response.

[charred merlin]

You're only able to download .wav with a paid plan.

[crisp scroll]

my wav download disappeared...

[mortal pendant]

hello, I need help changing my e-mail related to my payment. The website doesn't give me this option. Is there any support channel?

[charred merlin]

Do you still have a paid account? Go to https://www.suno.com/account to see if you're prompted to update payment info.

[crisp scroll]

I don´t, but a week ago I still was able to download .wav

[charred merlin]

The "official support" is still in fledgling stages on Discord, the first-most-effective method is to email [e-mail] to request this.

[charred merlin]

Sounds like you might have had access while your account was still paid, but if it's expired now, you're back to free limitations -- no .wav download, no custom photo uploads, no video regeneration

[crisp scroll]

I never had the pro plan, thats the problem

Why did it worked before then

[charred merlin]

Sounds like you got a cool free bonus, I'd be happy that I got something normally restricted by a paid subscription for free.

[mortal pendant]

thanks, theharber. I will. Before, let me better describe what is going on, so maybe you can help me: I created a Suno account using a secondary google account. I tested it and I liked some songs I've created in this account. Now I want to subscribe to suno, but I changed my logging method account to another google account. However, when I go to subscribe, the previous google email is still related to my payment info and there is no option to change it.

[hollow sapphire]

So, since I'm always curious about everything, and basically, I only believe what I see with my own eyes, I tried uploading a short audio clip to Suno... The verdict is clear! 'I love to love' won't get through^^ 🔥 😂

[charred merlin]

Happy to help with a huge grain of salt (since I'm not a suno guide, a discord moderator, or a staff member, please take ANY of their answers over mine)

If I'm following, the easiest workaround should be to log out of Suno, log in from your preferred Google account, and make the subscription from there (this way there's no "redirects" or anything, just a simple route-one account subscription)

This might require a bit of "tidy-up" if you have songs from your other account that you want to continue from the paid account -- perhaps making a private playlist on the original account, including all of your unfinished songs, then using that playlist from your official paid account to finish making them.

obligatory TOS stuff about not making multiple paid accounts for the purpose of abusing free credits, additional stuff about how I'm not a lawyer and encourage you to look into copyright and ownership laws in your country

[charred merlin]

Bonjour Netor; ca-va? 🙂 ☕

Can you talk a bit more about what you were attempting with the audio upload? Is there a section that sings "I love to love" that was cut off?

[mortal pendant]

thanks for the answer, Theharber. I will contact them and see if I can get a quick solution. Have a nice day. 👍

[hollow sapphire]

Ha ha ha!!! No, no, it’s a joke: I knew that this little bit from the song “I Love to Love” by Tina Charles wouldn’t get through! But... I had to see for myself! In the end, I saw! Suno's copyright wall is very effective! For those who still have doubts^^👌

[charred merlin]

🤦‍♂️ Sometimes I am 100% in "troubleshooting mode", and fail at basic human interaction 🤣

I agree though, the copyright detection is very powerful, but since it's an automated system, there is always room for "false positives", in which Suno accidentally flags an innocuous song, or "false negatives", in which Suno fails to flag copyrighted content.

I'm very much the same, and learn best from doing, so we're on the exact same page in terms of respectful experimentation 💪

[stray slate]

Hello y'all I was thinking of the possibility of a Discord bot that can play suno songs like jockey music for other platforms? Is it technically viable?
Regards,

[vestal rapids]

can't be done as there is no API

[stray slate]

There's a bot for download though 👀

[hoary otter]

I have observed on rare occasions creators specifying a vocal value, assuming that works it gives creators a bit more control over which vocals they use in their creations. An idea I would gladly welcome. Does anyone know of a resource exists in which you can review the AI vocal styles by description/gender to aid in the tag process?

[wide ore]

After subscription, the MP3 quality remains [192 kb/s], but the ability to download WAV quality [1536 kbps / 16 bit / 48000Hz ] is added. While not everyone has WAV downloads working, there may be a download error, mine is in Chrome. However, I installed Firefox and the download works wonderfully there. Also, after subscription, the quality of playing music on the site will be much better, maybe 320 kbps, but it feels like 1536 kbps!

[untold briar]

**V3.5 can occasionally lock in and hallucinate/write lyrics with real language (words, phrases, lines, even verses) instead of word-salad/SimLish. **But other times it's word salad as normal. Anyone happen to notice which prompts hallucinate more coherently?

This song lyrics prompt https://suno.com/song/6c1ddee3-eb6c-44a1-b8d3-77afd4fb3a23 looked like [field-recording, chorus-acoustic-cello, copy-copy-copy, verse, verse-exotica, exotica-vocals, exotica-renaissance, renaissance-harmonics, --] - I was curious about structure tags like "chorus" and "verse" in bracket prompts. But I don't think is the cause.

[wide ore]

If you need to download WAVs like me, try switching to Firefox, it helped.

[unique stone]

A little confused by this. Suno's MP3s are 192kbps regardless of whether you have a subscription or not. Which you mention at the start of your message but then state that after subscribing the quality will be much better.

[hollow sapphire]

👍 My English teacher (I don't know if he's still around) always used to tell me, if you don't try (he was talking about practicing English), you'll never know if you're good or bad at it! I think this can apply to anything! You have to give it a try!

[wide ore]

The quality of the file that you download in MP3 format will not change, it will be 192. But the quality of playback on the site itself will be much better, it will also be possible to download WAV, and it can already be processed and mastered. Or simply convert to Flac or MP3 320 kbps.

[unique stone]

Ah. Yeah that might be dependent on how the browser is decoding the MP3. The actual MP3 you download is the same as the one that plays in your browser.

[charred merlin]

Out of curiosity, is the workflows section intended for users sharing detailed explanations of how they do what they do, or is it okay for users to solicit suggestions?

I've been trying to redirect some focus in there, but just realizing how close my assistance is coming toward moderation, so I'll step back from that channel, and just wanted to see if there was an intended direction for it

[charred merlin]

I can't express how much I enjoy seeing "Simlish" being used effectively in a non-Sims troubleshooting chat -- absolutely heartwarming.

[charred merlin]

lol, there's also companies claiming to offer a suno API -- not official, and anyone cashing in on it is blatantly scamming you.

[stray slate]

I've seen them, I'm going to test one that has almost the same UI also 🤭

[charred merlin]

You do you -- just be careful providing anything to shady-scam-devs, try to protect your internet identity, and since using unauthorized methods of accessing the service is a ToS violation, maybe consider how use of the scam-apps could impact your access to the service, since Suno reserves all rights in the matter.

"heh heh heh, ToS doesn't matter if I choose to not read it, I'll just play dumb."

"hey, I have active credits available, but the site isn't working for me. I'm actively paying for a plan. Please help because I use this service every day"

[unique stone]

Yeah I feel this is how it'll end up playing out.

[charred merlin]

"wtf do you mean, tos violation? I'm a paying customer. Give me access to your service, or I'll just make a separate account and you can't stop me."

[unique stone]

also stating an intention to do something that breaks ToS, in the discord for the service whose ToS you intend to break, is not a good look.

[charred merlin]

for the record, this is most likelyexactly how it has played out many times. I just took the liberty of voicing their quiet parts out loud.

Hope you're having a wonderful day, hamster 💙

[stray slate]

The web that I found is called remusic (snitch mode on)

[charred merlin]

(kuak, my "one-robot stage play" wasn't intended to be directed toward you, you genuinely seem like an excited tech-enthusiast, so I feel like we're on the same page , just wanted to share an anecdote of how they could play out -- the stuff about protecting your identity is super serious, but an equally important part (that I left out because I didn't want to come across like a shill) is that these services aren't being set up honestly, and likely abuse the HELL out of free accounts -- meaning not only will every song generated from it potentially vest in a DMCA claim or legal issues, but the scam-devs are actively abusing the system and making it worse for everyone else, both in the immediate-term for server load, and in the mid-to-longterm, because the legit-devs are regular people who just so happened to have worked on a cool project that skyrocketed -- any time they spend bot-swatting is time that they actively aren't spending researching how to make the next models better)

[stray slate]

Don't worry bro, I haven't feel like a personal attack nothing that you have said, on the contrary, I've felt it like a warning and a positive one. I'm glad that you are clarifying everything for me and others reading 🙂

[vestal rapids]

I have noticed over the past few weeks that generation times have been longer than they used to be

[nimble iron]

Question,
how long does it take for a video's regeneration to kick in?
If you wait a while after regenerating, it still will show the old one when posted to Discord.
Is it, like, a day long wait?

[hollow hare]

It shouldn't be more than 10 minutes at most (and that is being far too generous).
The issue with Discord is that it caches the link. Just adding a / at the end of the URL should be enough for it to treat it as new content

[nimble iron]

It's been an hour and a half 🙃

[leaden beacon]

Hello Guys, does Suno have an API which we can access to generate the musics? where can I find info about it?

[hollow hare]

🤷‍♂️ Try downloading the video from the website. If it shows your recent changes, that indicates an issue on Discord side (likely because of the link being cached as I mentioned)

[nimble iron]

Yea I downloaded n checked, it's a discord thing. Good to know, thank you 👍

[charred merlin]

discord caching + suno is legitmately weird.

@naive quail mentioned that he tries to make whatever video edits before sharing the link, that way his preferred version is the only one cached.

I tried to time how long it takes for things to be updated, but I lost track and noticed that it at least updated the following day

[hollow hare]

That is because "numetal" does not exist, they are two words. Try "nu metal" or "nu-metal". 3.5 definitely should know this style

Side note: There are some genres that 3.5 had a hard time recognizing, let alone knows it (like phonk), in which case you should start the generation in 3.0 (better at knowing more diverse genre) - likely as an instrumental piece - then extend that into 3.5 for the song structure

[nimble iron]

It's sad that most songs' chance to be seen is the day it's created, so you have to get everything perfectly right before releasing...

[hardy lynx]

Where can I find the most updated list of prompts and tips and tricks for suno?

[solar lynx]

Hi everyone 🙂 Is it possible to manually mark the stress in words somehow?

[alpine venture]

https://suno.com/song/4f2ecf0f-4a74-4744-80e0-e1388808b764

more experiments with noise

and I liiiiiike

I wish I could end it properly, but also, I think it's kind of fitting

but it's a part 2

and part 1 is not actually a part of it

[little cargo]

is there any meta tags i can use to improve the quality or fidelity of the songs im making?

[charred merlin]

After extensive testing and failing to notice ANY discernable result, let alone with any degree of consistency, I'm relatively confident saying "no", but I'm not going to attempt to argue with anyone's confirmation bias if someone includes related metatags or prompts by rote or genuine belief.

#1256301881966198805 message if it helps, here's a prompt experiment playlist that Jon Fly and I contributed to, with DAViD lending his exceptional ears to confirm that they all just sound like standard Suno generations.

[leaden lark]

I see, I’ll have to look into when I can

Soon clocking into work so can’t look at much

[charred merlin]

There isn't a "standout" in the playlist, it's more to demonstrate how many separate attempts we made to try to get something like "dolby atmos" or "wide spectrum" to work.

That said, it doesn't mean that they're "wasted prompts", it just means that it doesn't give the result that we'd misinterpreted Suno to understand likewise. I think Jon Fly noticed that the "dolby" prompts seemed to give a more "movie quality" feeling to the country-style prompts he was using as a test. ( it made cinematic, movie-score country or something)

I kept you in mind specifically when I added the section about not arguing with anyone's prompting choices -- I've seen "dolby" in some of your songs, and they turned out great regardless of how that prompt was interpreted or expressed ❤️

[lusty bear]

how to make the 1st 4 lines in lyrics singed by chorus and the rest by artist voice ? i mean the 1st 4 lines are repeat 2 times in the song i need it song by chorus and the rest with the singer?

[leaden lark]

Thanks I mean I just use Dolby because why not, but I have no idea if it even is doing anything with the genres and styles I do

[charred merlin]

why not
right? Nobody gets hurt by you using it, and while they aren't exactly hurting for publicity, I'm sure Dolby Labs wouldn't be upset that we're trying 😂

I see it as a "thematic descriptor" rather than a "command" -- it's like carrying a lucky charm to work to prevent elephant stampedes -- like, I'm probably fine either way, but if there's even the slightest chance...

[untold briar]

Trying for another song with semi-coherent hallucinated lyrics, with no singable words prompted, required more aggressive creative hearing, and more creative interpretation of what the song might be about. first song was a lucky gen. But i do still think V3.5 hallucinates more coherently, overall, this was much harder with earlier models. https://suno.com/song/ea438822-5ec4-4512-aa01-9009bed2761f

This isn't doing the very useful Suno workflow where you hear a hallucinated vague verse, then prompt a plausible real line of lyrics that matches the syllables, and use the new version where the words snap into focus and sound clear. all these lyrics are unprompted, all the time.

[faint crypt]

How to stop ruining my songs with 4 minutes songs?

Why does it keep going???

So ass

[hollow hare]

Can you be more specific as to how it is ruining your song? What you had mentioned is very vague.

If you are talking about songs unable to end in 3.5, that has been known since like the beginning and alternatives has been mentioned many times.

Basically: 3.5 is not great at ending songs. If you want to terminate a song at a certain time point without using an external program, you will need to [End] it with an extension set to 2.0 or 3.0

• extend the piece at the point you want the song to stop.
• Set this termination extension as 2.0 or 3.0
• Set the style as "End, Short" or just simply End. Up to you if you want [End] in the lyrics

Here is a graphical version of the instructions that was pinned in general: #💬┃general-chat message

[faint crypt]

Ok

[muted timber]

after 3 month of Pro i saw now that the video function works. But i only get this waiting/progress symbol. mp3 and wave i can download. Is there something i have to do to get it also that i can download the video?

[glacial scarab]

yea i have made the same experience V3.5 Hallucinations are sometimes very coherent. Like here for example also unprompted.. well except for the meta tag and some arrows. But no lyrics
https://suno.com/song/afb7853b-8d39-431d-8fdc-496f882f2f25

[hollow hare]

You should be able to download the video as soon as it has generated, like at the same time the audio has generated.

If a loading icon remains after the audio has generated, try refreshing the page. It could be a visual glitch

If the download itself is being prevented, try using incognito mode or another browser in case there is an extension interfering with the download

[muted timber]

ty very much for your fast reply - i will try it out 🙂

[austere dragon]

Anyone have any suggestions for percussion intros? I haven't been able to generate them in Suno. I've been building them in either another AI generator or in Audacity with loops. Would love to know if anyone has had success with this. I've searched the website, and I don't see any good examples.

[charred merlin]

Oh man.

One of my favourite tracks has a complex drum intro, and getting what I wanted was a journey, but it's probably my favourite song.

Focus only on the drums at first, don't include ANY lyrics, or other style prompts or metatags -- anything else will overpower how specific of a track you're looking to create in isolation. Instead of trying to guide it through metatags, try experimenting with different style prompts that specifically reference drums -- use v3 to force Suno to be a bit more versatile & broad, so that it might be accepting of spitting out a result that's just a long drum solo, where v3.5 would try to turn it into something more musically coherent.

These songs "live outside of Suno" since there's no collaboration toggle, so I just link to them externally, but I can look up what style prompts I'd used for the drum intros

[sweet wigeon]

I've been experimenting with arpeggio technique for gumbe kora and guitar https://suno.com/playlist/17ad1856-3bfb-4e83-94f9-9e2b2d884744

The best are derived from AMaj root, interval 1-3-5.

[charred merlin]

Turns out getting the intro to "sidewalk surfboard" was a heck of a lot simpler to get started -- the trick was using just the first couple seconds, experimenting with where to extend from, and what different directions I could take the extensions in once I'd "locked in" the intro to "hook onto"

I use a random word generator to pick project names for me, to act as impromptu "workspaces" since the swapping through playlists is an absolute travesty for productivity -- pleasant coincidence that my favourite drum track was called "bobbing"

[honest abyss]

https://youtu.be/wsAqokt_5uY?si=uqAzeCuF713zQCRT

How do I get my song to sound like this

[glacial scarab]

ive played with the thought of using hashes or guids and keeping them external in a list.. random words are a nice idea too.. but might have some overlaps in search.. well depending on what words it generates.

[charred merlin]

Instead of using a predefined list, I use a free api that pulls random words from the dictionary, I would be genuinely surprised if you make enough projects for it to be an issue 😛

It's incredibly simplistic, but I built it collaboratively with ChatGPT and felt proud of how quickly it came together 🙂 Even the .ico file was a random color noise image generated in chatgpt using python.

I ran into an issue where the text insertion didn't work on a different website and I learned about how that site handled their text fields differently (and got a glimpse into why @vestal rapids gets so pissed at every website trying to handle things differently 😛 ), but with a single iteration I was back up and running 🙂

[glacial scarab]

i was more thinking the words might appear in some lyrics or styles as well as the search searches everything. But i guess if you don't use super common words it's a none issue. I'll probably still use hashes of some words just for fun

Can't have my library not look like a collection of a mad man 😂

[charred merlin]

Oh, it's just a right click option to insert a random word, nothing special

[glacial scarab]

Yea, still handy. I first went for the putting it all in playlists but as you can not add while it's generating i always forget

I don't want to know how many possible bangers are lost to that 😂

[simple osprey]

is it only me or is the sound quality dropping at night?

[charred merlin]

I'm not comfortable speaking on behalf of everyone who isn't you, but the sound quality seems consistent on my end regardless of the time of day 🌞 🌜

[simple osprey]

[leaden lark]

[mental forge]

Hi, I'm curious whether Suno can create songs with traditional genres like dangdut or genres from a country's culture?

[leaden pawn]

Is it just me or is the website super slow at the moment?

[frozen pivot]

The easiest way to check is to use those genres in the "Style of Music" section of the prompt. List the genre, instruments, etc... up to 120 characters. Use similar listings within the Lyric section between []

[frozen pivot]

It's normal from my end; east coast USA

[leaden pawn]

Seems to be OK now.

[frozen pivot]

👍🏻

[past drift]

Guys help me please. Chirp Bot is not responding to me (The application did not respond error)
What could be wrong with me? )

I created a server, added a bot, authorized it with all rights, but it didn't work

[frozen pivot]

This is no longer in service. You can now create your music at suno.com

[frigid sapphire]

Hey, guys, if I have 5 credits remaining from my premium subscription at the end of the day, will I still have a refill of 50 credits waiting for me tomorrow or do I have to use them all up? Because using up 5 credits is impossible. Just wanted to know if I need to try and manually create a 6 second song to get my free credits.

Oh, and is there any actual point in reporting bugs? I've generated the same song 10+ times now, but it always deviates sharply from the lyrical structure, and so I've been reporting them for this.

[ocean flicker]

How do i make these music videos for my songs without that SUNO logo at the bottom of the video?

[flint palm]

Hi there! Question to the SUNO technicians.

Is there any prompt / trick / a way to generate song w/o music at all ? Just singing voice w/o any music in accompaniment.

Is there any solution how to achieve that? We tried many ways, but always music being involved. NO music, without music description in prompt simply not working.

Any help / advice will be highly appreciated.

Thanks.

[rugged torrent]

Hey, guys. I have question about rights.
I have free account. I create song (without any publish), then buy Pro/Premier account and extends this free-generated song.
I haven't rights to free version, but what is my rights to extended version?

Suno help page don't answer to this question.
Thanks.

Or who from Suno team i must ask about this?

[charred merlin]

You edit them externally — videos directly output from Suno will always have the Suno logo, at least currently.

[charred merlin]

[e-mail]

There is no formal instructions because this is a separate request from what they offer users in the ToS.

I’ve heard many instances of people being able to acquire the rights to one of these songs after contacting support, but I don’t have any other details like how much it would cost.

[charred merlin]

Haven’t gotten to my desk yet, but you can ctrl + f in #1256301881966198805 for my prompt experiment playlist on the differences between “a cappella” and the anglicized “acapella “

[charred merlin]

Unless you buy more credits, wait for the refresh, or generate another <7 second song, you’re right that 5 credits isn’t enough to generate another song.

Whatever credits you have left over are reset at…

[flint haven]

Free Credits Refresh

Refreshes at 0000 UTC+0 (<t:1713225600:t> in your localtime)

[charred merlin]

…>Report>Bugs>… won’t immediately impact your results, and you won’t receive a response from Suno on them (you can try the bugs channels here #🪲┃bugs-web etc if there’s something people haven’t already mentioned in #1252292428258480209 , but that hasn’t been updated in a while )

Any reports you make help improve future models 💪

[charred merlin]

There is no official Suno bot.

If there was a simple mistake, no worries!

If someone represented themselves as a member of Suno, or told you to install a bot for your server please do a security check.

(There is no official Suno bot, and the API has not been made public for others to create official bots. Nobody here is upset at you, but I am concerned that someone might have attempted to scam you, or run unfamiliar code if you’ve allowed any sensitive permissions to the bot 😨 )

[flint palm]

@charred merlin Thanks! Will try "a cappella". Thanks, once again. 👍

[sweet wigeon]

I did some experiments with Flamenco style for so called malagueñas. Included guitar, tambourine. I wanted to have very fast finger picking which is included in the prompt. The last version includes kind of modern style, which I do not like. I wanted solo, but the prompt was included in the song. Never mind I got at least two amazing melodies. https://suno.com/playlist/400a138f-246c-4c26-b95c-61953dd62e2d

[severe solar]

hi, I'm fairly new. Does anyone know why i cant set my songs to public? the toggle button wont work.

[charred merlin]

If you've uploaded audio content that has been flagged as containing audio, you can't set it to public -- it can still be shared externally, and it will appear in public playlists that it's included in, but it won't be directly searchable within the social-media side of Suno.com.

[polar spade]

Re: A CAPELLA -- One simple check to see if Suno can do, or to find a prompt, just search for what might work and find a list of other's songs. Review for the sound you want and try their prompt for your own.

You might use "a capella" as a prompt, but may need to add something like "chapel" or "choir" to nudge suno to sing without music. To get less of a "church" sound, I've tried "choir rock" that brightens it up a little, but might throw in percussion or , of course, some music.

[polar spade]

For instance, this starts out with no music... eventually including a minimal of sound and the lead vocal remains a single person and that "a capella" group sound we might want. https://suno.com/song/951c2e32-ca25-46d5-b6fa-60c64fd2c4a5

Oops... I didn't want the visual. Sorry.

[polar spade]

Adding "barbershop quartet" will definitely (maybe) give that 4 part harmony with no music, but to nudge it in other direction with still no music might take some flair.

By putting "choir" up front but moving "barbershop" towards the end, with other prompt in between, I got a 60s style rock sound, like Beach Boys, or a do wop group, but without much do wop. Again, percussion was added, so that might need post production work to just remove out all but vocals. https://suno.com/song/a5a3bb25-e96c-4d2d-9f4c-a8393db2c9b3

Ok. Good luck.

[robust surge]

are suno generated songs dolby atmos compatible?

[charred merlin]

dolby atmos compatible
To clarify, are you asking whether you're able to edit the audio externally so that it can have dynamic spacing, or are you asking if you can tell Suno "dolby atmos" and have it know what you mean?

In either case, 1) yes, Suno songs can be edited externally, so whatever audio engineering methods go into the application of Dolby Sound, you can do the same process to Suno songs.
2) No, using "dolby atmos", "wide spectrum", "wet mix", "tight compression" etc, don't force Suno into applying these EQ filters, but might help act as thematic descriptors when used as prompts to try to push a song in a specific diection (jon fly notes that "dolby atmos" combined with "country" gave a more theatrical movie-soundtrack style country song, but the audio quality doesn't differ from any other generations)

#1256301881966198805 message > a playlist demonstrating a large number of examples, none of which resulted in any discernable "spatial audio"

[thin flint]

I'm curious, I saw some prompts while browsing songs with elements of the prompt in brackets, E.G. [Artcore]. What effect does that have on the prompt? Will it follow the prompt better or something?

[polar spade]

Pro tip: If you want to nudge something into an indie-voiced sound, you can certainly try "indie" [indie pop, indie rock, etc] but also try CCM (stands for Chistian Contemporary Music). As much of CCM includes the indie style, with addition of harmonies and a "broadway" feel, it can nudge a (non CCM) song into a more folksy sound. Great for adding to my prompts "mellow gold, yacht rock") to move it into a earthier Doobies, or 70s style sound.

[charred merlin]

"thematic descriptors", or "genre-adjectives" -- doesn't always have a direct sound, but it can infuence a specific style.

Think of "east coast" -- by itself it might imply Acadian kitchen-party style music, fiddles, guitar, singing, hand-percussion

But "rap, east coast", and suddenly you're talking Biggy

[hoary otter]

Hope everyone is having a great morning. I scoured through workflows last night and found some great tips but realized that I really have no real workflow for album art. Meanwhile I see some great creations from many creators on here. Would anyone care to share their creative process for Album Artwork? I know this new creator would love to learn from it. Thank you!

[pastel gyro]

Suno API when

[polar spade]

I'm curious: If you change lyrics on the original gen, then extend it (adding lyrics to the extended) , when clicking "whole song" it will sing the whole song with the original lyrics, not any edited words. That seems obvious.

But if I go to original and edit lyrics then delete everything, THEN do an extension, suno most often will give me 2 to 7 seconds of blank, or a couple of times just a fews seconds of a song ending. I get many 'credit return" errors.

So changing lyrics and saving doesn't affect the audio playback in the extend or whole song, but deleting the lyrics of original , and then saving, does affect extensions and whole songs? It's interesting at least.

[stark ibex]

Hi. I bought the pro plan edition and i don't recieve 10 working songs at once. I still recieve only 2...why? 🤔

[jade osprey]

Sr. Sunors: I don't know how you're generating your vocals, but on the off chance you're using white noise and then some diffusion-like method, maybe start with brown noise instead of white noise?

[hoary otter]

hmmm it gens 10? Really wish this was configurable. Sometimes i think I would prefer just 1 so i can be more frugal with credits for when suno gets it right.

[thin flint]

Why are people getting the idea that buying a subscription plan means more generations? It does not; you only get more credits to do with as you please. It's not this massive feature upgrade people are imagining.

[stark ibex]

Hmmm i think you are right

[stark ibex]

Thanks for your information

[brave gull]

pro plan has more simultaneous generations (meaning you can start more jobs without waiting for previous ones to finish)

generating 10 at once right of the bat would be very silly indeed

[hoary otter]

suspect also very taxing on the AI infrastructure

[brave gull]

also; my question being: is "regenerate video" broken atm? been trying to change an image and lyrics and nothing's happening

well I found a workaround, just redo "Get whole song" I guess :D

[charred merlin]

https://discord.com/channels/1069381916492562582/1253766525458841610

Nothing has been announced, but give this post an upvote to the let the devs know people are still interested.

[charred merlin]

There might be some mistake -- each "create" generates 2 songs, this will always be the case.

What you're referring to are the "concurrent jobs", which are how many separate times you can click "create" before it tells you to wait.

• Free Users can only have 2 concurrent jobs (meaning 4 songs created & processing)
• Paid Users can have 10 concurrent jobs (meaning up to 20 songs created and processing)

https://discord.com/channels/1069381916492562582/1254418510885687366 This is a known-suggestion post where users have requested the ability to create different amounts of songs with a single "create" click.

[charred merlin]

https://tenor.com/view/spoilers-doctor-who-gif-8623491

[stark ibex]

If someone knows.. Can i edit a songs lyrics, keeping the same voice and remake the song? If yes What is the way?

[charred merlin]

Because you're allowed to have 5x of the concurrent generations, listed clearly on the subscription page

[charred merlin]

When you "extend", it tries to use the existing song as context to continue from -- if there's existing vocals in the origin segment (like the vocalist singing the first verse), then you'll have a MUCH easier time continuing from that voice.

If you extend/regenerate from before the voice started, you're just hoping that your style prompts and metatags are enough to influence it the way you want, but it can be a roll of the dice.

And trying to get consistent voices between songs is an ongoing challenge, but there's a lot of awesome Sunovators out there doing stuff like it, for instance @mystic flume has multiple albums released under the same AI band, Celtic Moonshine, and there's a degree of consistency between the vocals (though I'm sure there's a degree of confirmation bias on my end)

That said, there's a known-suggestions thread for reusable vocalists, if you wanted to upvote it to let the devs know people are still interested. https://discord.com/channels/1069381916492562582/1252789683176345641

@jade arrow, I keep reading your name as "mallrat" and I'm genuinely questioning if that's the intended meaning, or if I'm experiencing late onset dyslexia 😂

[jade arrow]

Hi! I made a good song with suno but i used not my main gmail account for that. I want pay for premium plan with my main gmail account where i do these payments. But how can i use my well made song which i created with not my main email where from i would like pay for. Can i use that song on youtube after i payed with 1 account or how can i get that excat song for my main paid registered suno account?

[jade arrow]

:D:D yes you are right this is a late dyslexia i think too. 🍮

https://tenor.com/view/coffee-coffee-cup-tea-caffe-caffeine-gif-13667954778907566683

[polar spade]

I thought it was 10 simultaneously, as in, click create at lease 5 times without waiting.

On free, you have to wait for your 2 to finish generating? I've forgotten if that's so.

[stark ibex]

I really thank you so much!!! 🙏

[polar spade]

Yes. I'm right. 10 running at the same time. Right. Click 5 times.

[polar spade]

I'd prefer the "1 at a time" as well. Got it right on the first try, then go on to extended or changes.

But as humans we want up to three as the best options to choose from feel like we're not missing out on something better.

But three at once would give us that "wasted 2" feeling. Only 1, that "could there be something better" feeling. So 2, is the answer, I guess. 😊

[jade arrow]

help pls ||
||

[charred merlin]

(ownership / authorship / copyright are sometimes incorrectly used interchangeably , but the differences are very important. I'm going to use "own" colloquially, but advise you to do your own research into the matter, as my advice isn't intended as legal advice)

If the song was created on a free account, it is "owned by Suno".

If you make a song with a free account, and then sign up directly afterward, Suno STILL "owns the song."

If Suno owns the song, you ARE allowed to;

• share it with friends
• call it "your song" colloquially between friends or on suno.com
• upload it to your social media, provided you state that it's owned by Suno

If Suno owns the song, you ARE NOT allowed to;

• do ANYTHING with the song that would potentially lead to you receiving money for it
• posting the song anywhere that you would need to claim ownership (like publishing it with a distribution partner, or otherwise distributing it to streaming services such as spotify or apple music)

If you have a song created on a free account, and you'd like to purchase Suno's rights, titles, and interest in the song, you'll need to email them at [e-mail] to request this.

[craggy delta]

what's wrong with SUNO this month? no matter how much i say 80s, it can't stop giving me a 2000s singing style. SUNO downgraded for 80s music since july sadly 😦 no more the beast it was.

i'm wasting and wasting tons of credits trying to make a 80s hard rock song and even if the intro sounds 80, it pop ups with that crappy 2000s singer

[jade arrow]

Thanks for the info and your help T! Funny part i made another song with my main gmail account and i want that too 😄 i want all of them 😛 So i need write email for that song (which created with not my main gmail account) before i purchase premium plan on suno with my main gmail account? or after i purchase premium?

[charred merlin]

The models aren't changed in the way you're suggesting.

Curious, are you approaching the songs you'd like to create from different angles and creative approaches, or have you just spent credits re-rolling "80s hard rock" on v3.5?

[vestal mason]

How I can buy subscribe if I from Russia?

[charred merlin]

I would advise to only make songs on your paid account -- that way you are the author of any new songs you make. You'll need to let suno know if there are multiple songs made on a free account that you want to buy ❤️

[charred merlin]

[e-mail] are the only people able to answer this properly.

https://discord.com/channels/1069381916492562582/1254232844767854665 other payment methods have been requested, but nothing has been implemented yet.

[craggy delta]

1980s hard rock, true 80s voices, genuine 1980s hard rock band. 80s sing style this is my style using, last month worked, now not

https://suno.com/song/c3077959-03c2-4b4f-9ecc-e28e22f04d33

Do you think this is a genuine 80s rock song style?

This is completly Crap

that was a prompt it worked last month, Why isn't working now?

[jade arrow]

Thanks a lot T!

[craggy delta]

what prompt do i have to put for v3.5 to give me a realistic 80s singins tyle?

Image for ticket to the moon
ticket to the moon
genuine 80s hard rock song, with a 80s hard rock singer,
this is what i put it now!
Listen to the piece of crap it come up with? Why in the world i get this crap singings tyle? I hate that way of sing, it sound true synthetic, with a lack of emotions,. 2000s killed rock and roll and now SUNO is doing the same!!
https://suno.com/song/c21c9573-d553-4d80-9ea5-ec66951879ee

https://suno.com/song/0e0535e8-4936-4199-b78a-9b5b019e8745

Listen to this crap, It have nothing to do with 80s!!

thanks Suno for making me throw to the trash 10 dollars this month!!! ):@

at least last month SUNO had good days and bad days, now only have bad days, what a downgrade

[charred merlin]

Usually I try to help people navigate these situations, but if you spent $10 and felt like it was a waste, and weren't able to learn anything from the experimentations, I worry that the expectation will be for me to "tell you how to do it", instead of working alongside each other to create what you want.

There's a LOT of really bright minds here in tech-chat, many of which have gone onto other things, but look through the pinned comments, search for previous similar discussions, look for names like SixStringToker, Tinda Lee, Furrybeans -- lots of other names that aren't coming to mind at the moment.

In other news, listen to this PHENOMENAL track that I was able to make with the help of @olive drift -- sure seems like Suno is getting better by the day, eh?

[violet tundra]

I actually click "generate" about 10 times very quickly when I get the lyrics right. Sometimes 20 or 30 times.

[brave gull]

yeah me too, but usually you want to generate a few first just to see how thing go.
I meant it would be silly if you didnt have the option and it just burnt all credits

[final hazel]

I noticed it helped to spell it different sometimes like: eighties or 1980s

[craggy delta]

i also tried 1980s, sadly, no longer works

it worked before, not anymore 😦 suno is giving me headache today

[final hazel]

eighties worked the best for me

[craggy delta]

hmmm, i'll try this way

[charred merlin]

Hi @final hazel, thanks for locking #1256301881966198805 when that bot account was going wild.

I reached out in private, but you might not have seen the message -- are you able to unlock the post now? If not, I'm happy to continue assisting users here in the technical chat 🙂

[final hazel]

And last tip.. double down. Say it twice

[brave gull]

eh, almost reached 1000 pages in my library
now the number really doesn't fit in the field hehe

[final hazel]

Done! 🌼

[charred merlin]

Thanks!

[craggy delta]

it barely worked, but it still giving me a bit of modern style 😦 i hope they can fix this bug soon

[violet tundra]

Suno is the best thing to happen to lyricists since -- forever. I've written, produced, and sung and played keyboard in my songs and have collaborated with people -- very expensive - much more than $10. And not very productive, considering. Suno has allowed me to collaborate with poets, writers, and also to put my own writing to songs. Collaborated with Ukranian poet Svetlana Litvinchuk on "Heart Cup." Very pleased with the result. Suno is never a waste.
https://www.youtube.com/watch?v=Qvu1IUyiUjw

[craggy delta]

yeah. If you love modern music, but not in my case, music died in the mid 90s

[violet tundra]

Suno rocks 70s, 80s, 90s.... not sure what Suno you're using, but I have a couple classic rock songs as well as a 90s hair band song, and some grunge -- it does everything.

[craggy delta]

i'm using v3.5

[violet tundra]

Well, there' are tons of talented people on this thread, and if you need hints for prompts and processes, just browse the showcase lyrics -- the prompts, the styles -- it's all right there... and then there's the wiki.

[craggy delta]

i think is more a bug than a prompts problem, as i used the prompts it worked for me in the past

[violet tundra]

A "good song" for me (I know that's a subjective term), costs between 300 and 2,500 credits. Average is about 500 to 800. That's approximately $2.40 per song generated.

Ah, well, yeah. Always frustrating... good luck!

[craggy delta]

thansk buddy

https://suno.com/song/1ee31d6f-52cf-4a42-93da-8771d82d5222

Finally!

[violet tundra]

Best lines...

Living on the dark side of shattered dreams.

[Verse 2]
They built a rocket, just for me,
Said, "Wave goodbye to gravity!"

I'd extend an instrumental and piece it together to the beginning, then I'd slowly build the song from these three lines ... Could be pretty catchy... Gotta write your own lyrics and get rid of those Suno ones...

[craggy delta]

cool, check out the lyrics i'm writing 🙂
[early 80s montley crüe style]

[Verse 1]

I have a problem
i jus't can deny
i'm adicted to the danger
no matter where i am
i drive within seatbelt
i open the parachute
way near of earth
God is with me
anyway

[Verse 2]

I love feeling my heart
working like a v12 engine
moving fast and wild
challanging my life
i don't care
if i'm too high
or to low
my harness is half tied
the rock looks solid
what do i care

[Pre-Chorus]

I got an adiction to

[Chorus]

Top Adrenaline
my vains feel the danger
but my heart need more
keep me pushing
to limits i know i can try
Top Adrenaline
i love when the danger
is near by

i just put early 80s motley crue style just to try if it can give me a cool early 80s hard rock song

[violet tundra]

Some good stuff there. The off rhyme between "deny" and "am" - this is what you wanna stick to, Suno does "moon June" crap. Just make sure you find an iteration that enunciates it well. I would plug all these lyrics into custom mode and enter your style, and see what lines are sung at pace / tempo you like. Then, continuing to modify your style until you got the sound you want, I'd keep adjusting the lyrics to match the desired tempo -- then I'd click generate over and over until you find something that blows you away.

[craggy delta]

awsome!! good luck on that BD

[violet tundra]

If you want, I'll work on it and toss it back at you and see what you think...

but otherwise, that's what I do...

the process is 1) generate a sample you like 2) find lyrics sung to the tempo you like 3) adjust lyrics and style until you love it 4) rewrite lyrics repeatedly until you got everything done 5) generate multiple iterations until you have either one great piece or multiple pieces 6) splice muliple pieces together in your favorite DAW 7) add a beginning by extending an instrumental 8) piece the instrumental into the beginning to create a nice opening 9) don't worry about crappy one-off mistakes in the singing because 4.0 is coming out - just wait for inpainting to fix That about covers it...more or less.

(or use credits to fix one-off mistakes if you really want to release the song soon) Finally 10) copyright your custom lyrics. You can copyright 20 sets of lyrics for $65, which is reasonable.

[craggy delta]

https://suno.com/song/7f76dd3d-6ac0-4263-baa2-c76a2020bd56

now this is epic

[craggy delta]

coool!! 🙂 go ahead

[violet tundra]

Nice. Right up there with Triumph & Night Ranger (hope that's not taken as an insult). I like both.

[violet tundra]

gimme a sec...

[violet tundra]

almost got something...

Okay, I am starting to like this...
https://suno.com/song/c171c7ff-71b0-472d-ae6f-fc798b3ddc01

Original prompt: [80s glam band, metal, Style of Motley Crue, percussive, guitar [problem just can't deny gotta problem can't deny, addicted to ] [had to drop Motely Crue prompt due to warning]
[chorus: top adrenaline]

Resulted in this song:
(https://suno.com/song/4792fe8a-c501-49c9-a036-d5c32f850a9d)

I think the link should show the original generated lyrics.

Below are the modified lyrics, which I would continue modifying.

[repeat intro]
[Verse 1]

Bruised and battered
Doesn't matter,
Nerves on fire
Wheels spark the wire
Voice screams out my rage
blood-shot eyes shining
One on one wherever you wanna meet
Can't destroy me
just try to destroy me
And I never cheat
OHH!

[instrument interlude]

[Verse 2]
Body's frame cracks
doesn't break
I can go the distance
I can take the strain
One on one wherever you wanna meet
Can't destroy me
just try to destroy me, ---
....
And I never .... cheat

[Chorus]
Top adrenaline
Gonna feel the crush
Screaming souls
them versus us
Can't deny
that In the rush
Feel the need
courting chaos
Always feeding
And I never cheat

I would take the intro instrumental off that original song and splice it to the extension. I love the beginning of the original...

And, of course, I broke my rule of going with off rhymes.... (but you have to start somewhere).

And I would keep working it until it was structured well and had more elements that really stood out.

It was 160 credits. So ... depending on one's taste in 80's glam rock - it could be well spent or a complete waste 😄 (But I like the overall trajectory of it) this idea of top adrenaline, living on the chaos ... addicted to the battle, take all comers --. and I never cheat. so, ...

??

Of course, I had to muscle the tempo with commas, hyphens, and ellipses ... And I stuck in an [interlude] to pace the music between stanzas.

I'd stretch this out into a 3-verse song with an instrument and lyrical bridge with a slam-cut ending. The arc ... I'm not sure of the narrative arc yet... but chaos, always play it straight (raising the stakes (another off rhyme I'd use) .... I think the narrative arc I'd choose is... take all comers ... and I never cheat ... stanza 1 .. by verse 3... it would be meet wherever you will... take all comers... and I never kill. No need to kill OH! can't destroy me, just try to destroy me... but you never will... [repeat chorus] Because if this guy's so tough, he doesn't need to kill or fear reprisals. He just eats a sandwich.

I updated the original version... somehow I messed up the link. but the intro instrumental is on the original link now.

When I find something useful for extending, I always edit my files to "SOURCE with notes" so I can identify them later

And that crazy speed-core drumming in Verse 2 is best saved for Verse 3 or the bridge -- so I would spend some credits getting that fixed.

And I spent a few credits getting the AI to sing "shining" to rhyme with "rage." which turned out kinda okay.

[noble raptor]

so i'm trying to create an jazz instrumental track. it's coming out nice but the tracks are 55 sec to 1min 30 and most of the songs sounds like it's starting from the middle of a song and this is happening a lot and unfortuntely wasting credits to try to prompt it not to. is this happening to others too? anyone have a solution?

[polar spade]

If you haven't yet, just putting [intro] as a tag at the beginning can help sway suno to start somewhere than just jumping in.

[noble raptor]

in style of music? hmmm ... once we hit the instrumental button, we lose the lyrics block.

[true night]

making some structure in the lyric tab often dont make it sing it many of my longer works thats instrumental has structure in the lyric tab. it is ofc a bit hit and miss if it starts singing the structure but as @polar spade said add a simple [intro] may help it giving you a nice sway intro that fits youre styles 🙂

[signal girder]

Anybody here with the tools and know-how willing to assist me? I have a generation here and the Music is REALLY GOOD but the quality is beyond mid. I'd like to know if someone a little more experienced with outside tools would be willing to assist in a remaster, and share with me a 59 second sample 🙂

[violet tundra]

You'd need to convert 1) wav to midi (which always turns out like crap), but it would give you some initial ballpark ranges with which to work 2) with the messy midi imported into your online or offline DAW, manually trial-and-error match the messy converted midi notes (for each track) until they matched your Suno generation ... very fiddly work 3) you'd need the basic instrument VST plugins to match the Suno instruments and 4) an arpeggiator plugin to save some hassle turning the beats and chords into progressions, which would lead to some happy surprises.

I have yet to find an online wav to midi conversion tool that automates this in a high-quality way. so ... very fiddly work.

[signal girder]

I figured, outside suno my Musical and production knowledge begins and ends at autiory learning on a piano xD. Hence my asking, even if someone does it for me. (I am also not disillusioned, I don't expect free work, 100% willing to pay someone for their time if I have to)

[violet tundra]

Thing is, you'd find someone who can do it easily 1) paying for their expertise 2) find some (like me) who has done it with effort (paying for their effort / learning curve). I have lots of songs that I'd love to "fix," but 3.5 is good enough . I would do this first before spending any money. 1) separate the music into stems, 2) import stems into a user-friendly program such as Premier Pro. Amplify the best stems ... ususally drums, bass, and maybe guitar. Forget backup vocals, strings, and "other" because Suno seems to use these channels as hiss-collection tracks. Create a backing track using your original iteration + amplified tracks of your choice on top of the backing track. Then - most important - upload the adjusted version to an online mastering program -- take your pick. The mastering program will get rid of (some) of the hiss and poor quality, running it through an EQ session and adjust for LUFs, max db output, etc. You might find that for around $30 you can have a version that sounds pretty good -- while having 30 days worth of subscription to do even more stemming and mastering.

[signal girder]

Ya lost me lol but I can definitely take some time to do the research 🙂 I started uploading to music platforms and such so I guess getting the experience could help me streamline the process and prevent subpar audio coming through in the future, thanks for the info!

[charred merlin]

❤️ Send me a DM -- I'm happy to offer my services for free since you seem like a pleasant musical collaborator 😄

[signal girder]

Appreciate the consideration, I will touch base with ya soon, I do want to take some time to research it on my own as well 🙂

[charred merlin]

I do want to take some time to research it on my own
I want to hug you right now, internet-stranger/potential-friend.

[leaden lark]

[charred merlin]

Really good timing, actually! I'm shifting toward collaborative learning projects lately, especially regarding music production at the moment 🙂

Not only could we work on improving the quality of your song, but we could document the process & create amazing practical tutorial content for people who are in the same situation, but might not have the foresight or follow-through to reach out to likeminded individuals.

If you're interested, you get your track remastered, we both gain from the hands-on experience, and I get to create content that could help others. Even if it's just to touch base about what software or workflows within have worked for me through previous experimentation, I'm happy to offer what assistance I can to those who want to learn. 💪

[charred merlin]

🇾 🇴

[leaden lark]

How are you today? Been testing a bunch of style recently

[charred merlin]

Everything is trending upward 💜

Oooh, I love your NAI stuff, make sure to send and standouts 😄

[leaden lark]

Will do, and how has 🐝 been as well? I’m still working on that song

[charred merlin]

How're things on your end? Did they finally catch that Robin Hood of a saboteur who kept breaking things at your job ever so slightly that you didn't need to be sent home, but were able to be paid for hanging around?

Whoever they are -- protect them.

edit - I'll carry on in DMs so "the phone lines for tech-chat stay open", but @delicate idol is doing well, and you're a peach for asking 🙂

[leaden lark]

It’s fine, a lot of things in this building are old and they won’t upgrade stuff. It’s the usual stuff. But I’m well, and hanging in there

[solar lynx]

Hi everyone ! I have generated something i like, but I would like to change there few words - is it possible to regenerate keeping the style ?

[charred merlin]

The only way to handle this at the moment is to place an "extension point" before the lyrics you'd like to change -- everything before this point will be kept the same, and it will attempt to continue the song from that point, using whatever input you give it (different music style, different lyrics, etc).

If there's a strong enough musical context for the extension to "hook onto", the extension will continue the same style, vocal timbre, motifs, etc

[solar lynx]

got it, thanks!

[safe hazel]

The rumor is that version 4 of Suno will allow partial regeneration within a song to fix things like this

[low vigil]

hi all in anyone experiencing suno ai just ignoring prompts ? even if you just type in 80s and gives you randome songs ??? i never exsprinaced this before burned 1000 credits lol
started today

[violet tundra]

How many "voices" are there in Suno? There seem to be about 12 to 20 distinct voices, more or less.

[violet tundra]

Assuming you're typing in "80s" by itself, it's kind of a broad prompt, which is probably the reason. You could add some other descriptions to it that describe the type of 80's music you want, and that might make things more precise.

[jade osprey]

The GIF doesn't make sense to me. I don't know who that actress is.

[charred merlin]

I don't know the actresses name -- the gif shows a clip from the popular television series Doctor Who, it features a character who is aware of privileged information / things-to-come.

When this character is asked questions about thing-to-come, they often respond with a cryptic and sly "spoilers" to playfully hint that the answers might come to light eventually.

pro tip, you're probably not going to get trade secrets about their novel music gen methods, but it might be worth looking into https://github.com/suno-ai/bark , the open-sourced text-to-speech software developed by suno.ai, if you wanted some light weekend reading

[jade osprey]

Thanks for the protip, but...

[charred merlin]

open-source

You realize when Suno was developed right?

[jade osprey]

I think it's just too meta for me. I don't know what you're trying to say. Anyway. I hope they fix the noise. ¯_(ツ)_/¯

[charred merlin]

Apologies if I'm misunderstanding; it's actively unhelpful of me to try to explain without making sure we're talking about the same thing 😅 legitimately my bad 🙏

Sr. Sunors: I don't know how you're generating your vocals, but on the off chance you're using white noise and then some diffusion-like method, maybe start with brown noise instead of white noise?

I understood this as you asking specifically what type of audio noise that Suno uses to generate songs, similar to how image-gen uses different types of color noise diffused into bespoke images-- or am I mistaken?

If that's the case, I was trying to say that this information is a strongly protected trade secret about the internal workings of the platform, and we're not privy to this.

Suno/Chirp (text-to-music) and Bark (text-to-speech) were developed at the same time, presumably using similar methods. Suno is closed sourced, but Bark has been open-sourced so you're able to pore over every line of code if this is something that you actively want to know.

[jade osprey]

It's just a guess on my part. The vocals sound like they're swimming in a pool of white noise. Maybe it's happening at training time. Maybe it's happening at inference time. I figured if it's happening at inference, and no-one knows why the vocals are so noisy, an easy test would be to start with brown noise as the seed "latent" (does that even apply in audio?) instead of white noise, if they are indeed doing that.

I'm not terribly interested in learning how it's done or getting a sneak peak into trade secrets, just want to generate vocals without a nasty white noise hiss overlaid.

[charred merlin]

I gotcha -- completely legitimate question, unfortunate part is that we just don't know because the magic of how it works is their bread & butter.

I didn't think that your question was based in bad-faith, or I wouldn't have put so much effort into explaining ❤️

That said, the routine response I see to questions like this that may be too technically-specific for the devs to directly comment on is "your choice of input is the biggest determining factor on Suno's output".

[jade osprey]

I guess I wasn't really trying to ask a question so much as suggest a quick/simple thing to try. I imagine that for a dev to try out this suggestion would be a change to a f() somewhere in a dev/test env. And if it doesn't work... <shrug> I guess I'll never know. But at least I threw it out there in case that's the source of the pain.

And thanks for the discussion. It's good times.

[solar lynx]

if I click here it will bind part 3 with part 2 ? will I be able to revert this binding ?

[charred merlin]

#💬┃general-chat message

If you click ... > Get Whole Song on part 2, it will stitch together the origin segment with part 2 taking over from the timestamp you chose to extend from, and you get the full song in your library (for free).

Clicking ... > Get Whole Song on part 3 does the same, except now it stitches together all 3. (note that the original tracks are left alone, and only a new one is created, so this method is non-destructive 🙏 🙂 )

They always stay as separate tracks, until you "get whole song", and even afterward, part 2, part 3, etc stay separate, and only the "stitched version" shows all of the creative decisions together.

[solar lynx]

thanks a lot again 🙂

[charred merlin]

You're very welcome 🙏 (ramble incoming!)

#1256301881966198805 message here is a link to technique I call "terminator extensions" to use multi-part extensions to great effect.

Each extension can only "see" a certain distance previously into the clip it's extending from, so a cool guitar riff at the very beginning of a 20 minute song might be too far away for an extension to "see/hear" and try to repeat.

On the inverse, by using multiple extensions (part 3 and beyond) you can very powerfully limit how much of the song's context you wish to show to Suno -- Terminator Extensions demonstrate cutting the extension's "context" or "song sight" down to as short as a single second -- that way the extension "doesn't know" what it is extending from, and is a lot more willing to "turn on a dime" and go from hardcore metal to opera 😂

So if you notice that creating extensions from extensions from extensions creates bizarre results, try using "get whole song" to "lock in" segments that you love and want to persist in that song-project.

[safe hazel]

It seems like the more I extensions I use, the worse the audio quality of those extensions get. I try not to extend more than once for that reason

[little cargo]

what else can i do to enforce the fact i want a female or male vocalist other then [female vocals] [male vocals]?

[safe hazel]

put it in the style section

I usually put it in twice and I rarely ever get the wrong vocals

[little cargo]

im trying to alternate between male and female and the AI is fighting me

i got it to do it exactly once.. but by the end of the song.. the fideleity gets trashed..

[safe hazel]

yeah, there is no way to ensure that, probably the only way to force it to do that is by using the extensions feature and doing what theharber was talking about

[little cargo]

thats the problem it too several extensions to tweek it to do it.

check this out.. i havent finished the lyric or anything.. but it did exacly what i wanted.. two singers. but damm it eats the quality in the end. the song is great.. but i dont think ican make it a live.. We really need a way to fix the fidelity after a song has been made... or hold the quality at least

https://suno.com/song/6fa53c5e-1d3a-49f2-9b95-ab960b95759f

[violet tundra]

A lot of the fidelity issues, it seems come from Suno's understanding of song -- it seems to think that more is better and that song's must build, and in the building comes volume and accompaniment and complexity, which introduces some noise -- if you fiddle with your prompts and iterate enough, the ending will stay clean.

[oak jungle]

Hello, is it possible somehow to ask suno to sing the lyrics of a song with a male voice and then switch to a female voice for a specific line ?

[violet tundra]

yes, you can actually talk to suno . I often write [these lyrics should be sung in such and such way] It works 10 percent of the time--often resulting in the instructions becoming part of the lyrics But 10% of the time will get you there. More or less.

Most of Suno's disappointments come from the competing desire to save credits,. Enough credits and fiddling will allow you to make a truly custom song.

[little cargo]

take a peak at the song i just posted a few lines up i did that very thing

[violet tundra]

I'll run it...,

[sterile yacht]

• • • female vocal * * before the line works for me 7 out of 9 times

[oak jungle]

hmm i see, well yeah, would be better if we could just have a specific format to ask this. like this for example: [Voice: Male]
(Verse 1)
Lyrics in a male voice...

(Bridge)
Character says,
[Voice: Female]
Lyrics in a female voice...

[Voice: Male]
Back to the male voice for the rest of the song...

[sterile yacht]

adding duet to style also helps push it to better switching it up

[little cargo]

I came here asking that same question about 5 mins ago... it took many spins and extensions to get it to do that

[little cargo]

yup i added that it seemed to help i thought

[violet tundra]

yeah, that's going to be a credit sink

you could extend another version in male vocals, using the same drum beat and instruments and splice them together.

[unkempt cloud]

With my 7th suggestion in "suggestions-browser" could help with that over time, so it's not impossible to improve it, so we just have to hope that gets implemented.

[tough helm]

I still can't get over it how I managed to get a duet work once, and never again
https://suno.com/song/796bb5e8-97b3-45af-a5a0-ec0f24a61ec9

They actually followed the "Male verse 1" "Female verse 2" instructions, but i never got it to work on any song again xD

Anyone any idea how to make this work more often?

[unkempt cloud]

Haven't played around much with that, I DID however have a little luck with it by giving the "singers" a name instead of using "Male/Female"

It at least seemed to recognize Ahren Gray's style of "rapping" or whatever you call that from this song (I looked up the lyrics for it and it contained Jonathan's and Ahren's names in the lyrics so i just went with it): https://www.youtube.com/watch?v=a5Pm2bQhpdQ

[tough helm]

Hmm doesn't really seem to work either for me

[unkempt cloud]

Weird, worked fine here (even though it didn't recognize Jonathan's style, probably because the AI isn't trained on him): https://suno.com/song/9d93e82f-a83c-4ce9-9fa1-d69156303825

[tough helm]

Even with the same prompts (like (Verse 1: Jonathan Young), (Verse 2: Ahren Gray)) I get a female voice only on both verses

It's so weird how it sometimes works and usually doesn't xD

[unkempt cloud]

Wish I could help you more, but we'll just have to hope my 7th suggestion in "#suggestions-browser" gets implemented, then EVENTUALLY it could learn from that and probably be more accurate.

[tough helm]

Yeah would be nice in the future if you can give instructions like this

[unkempt cloud]

Quote from my suggestion: "could be added as an option in custom where you can choose up to for example 5 singers and for example have one melancholy singer male, one high-energy rapper female and so on" that would help CONSIDERABLY in of itself.

At least in this case

Would also help if the AI can look at those choices and refer to them as for example: "Chorus Voice 1 & Voice 3" as them doing a duet and so on.

[charred merlin]

Do you "get whole song", or are you extending an extension? It's like photocopying a photocopy of a photocopy. etc.

If you grasp the "song context" idea from the Terminator Extensions, it might explain why extending off extensions gives weird results, because it keeps changing how much of the song it's able to use as context -- probably explains why it feels so weird -- try "get whole song" on part 2/3 before extending, and notice how the next "part 2" off of "full song " sounds WAY different, and likely closer to what you were expecting in the first place 🙂

[charred merlin]

https://discord.com/channels/1069381916492562582/1252789683176345641

Being able to "re-use vocalists" between songs is a known-suggestion -- if you wanted to throw an upvote on the post I linked, it'll let the devs know this is still a feature people want to see added 💪

[signal girder]

Sent

[jade sapphire]

Hi everyone, I have a (probably) stupid question but it irritates me ... I am on my second day at Suno, got my 50 credits, made 1 song, it went down to 30, clicked on my profile, it went up again to 40 as it should ... is this normal? I know its minor but... thanks

[unique stone]

Sometimes the credits don't display properly when you make a song. Not sure what causes it but refreshing the page should show the correct amount.

[jade sapphire]

thanks so it is a known bug of sorts? Okay, thanks. I am pretty clueless so far about anything , and tech is not my favor either way lol apologize also if my question doesnt really belong here but still making my way through it all

[unique stone]

Yeah just a cosmetic/visual bug. It's a little annoying but harmless. Have seen it reported a few times.

[safe hazel]

I don't really do the "terminator extension" technique at all, since it's kind of a credit sink. But when I do extensions I only use them after "get whole song", the alternative isn't weird, just bad quality. Noisy, fuzzy, muffled vocals, just bad audio quality.

[amber vigil]

does suno have a place to report copyright violations?

basically i want to issue a takedown notice

[charred merlin]

ToS states it, gimme a sec

[e-mail] (might be suno.com? Not sure if their email client still reroutes .ai for them. probably)

Make sure that you're following this to a TEE, because filing improper DMCA isn't good for ANYONE involved, least yourself.

[amber vigil]

thanks @charred merlin - you are the man.

Some bruh swiped my lyrics- including the (c) Foggy and genned two songs and made them public.

[charred merlin]

Also, make sure that you've made a genuine effort to contact the infringing party, as it can speed up the process by showing you've made goodfaith efforts throughout the process 🙏 best of luck, dude

[amber vigil]

How can I contact a suno user?

there are no ways on the site

[charred merlin]

Doesn't seem like there is -- not a hard & fast requirement, but beyond the possibility of a quick & clean resolution, it can't hurt to show that you tried to resolve it without resorting to third parties.

[charred merlin]

Everyone is free to use Suno however they find it works best for them, and I'm not pressed for whether or not people find success with methods I've found success with, but from the description of being "a credit sink", I genuinely wonder if you're understanding the concept the same as myself and others do.

Either way, best of luck with the audio quality 💪

[charred merlin]

Bear in mind this is all armchair advice, but unless you've filed a registered copyright, expressing an interest in copyrighting can give you a stronger case than if you didn't have anything "stamped", but it's important to note that it might not be an entirely open and shut case, especially if you might need to prove that the lyrics were written without the use of generative AI, depending on your legal jurisdiction.

[oak jungle]

just fyi, writing "strong male voice, duet with girl" in style of music seems to help a lot for this too, sometimes i get the opposite though, but mostly it's good

damn this ai is so goddamn addictive lol

[amber vigil]

pretty sure copyrights are granted automatically to writer. At least in USA

[safe hazel]

Yes, I understand the concept. It requires extensions on an extension to erase the memory of the original song. That's two extensions each time you want to add on.

[untold briar]

Specifically losing memory is the mechanism by which you can improve Suno quality on long extensions. It's trading musical consistency/coherency for quality. Though you can Get Whole Song again later, and recover some long term consistency.

coherentness?

coherence

wow

I must have traded some brain cells today mysefl

[oak jungle]

i've noticed that when generating a song, mostly, it's the second version that is the best, is this also like this for you guys?

[untold briar]

At the extreme, with 0 memory, Suno has no long term memory degradation of course. But also it's completely unrelated music. I still wish we could set 0 second extensions, I'd use it.

[charred merlin]

each time you want to add on
Terminator Extensions are only used in very specific instances, like when someone is having troubles swapping genre. Again, it's not something that's just used "for the sake of it", it's a powerful method to use when people have already "wasted credits" trying the same things over again to no avail, like trying to make the song forget a specific style or motif.

I was referencing them to explain why extending from an extension gives worse quality.

[charred merlin]

Is the single second from a terminator not enough for your usecases?

[untold briar]

The thing users most often report is the first generation is the best (one of the first 2 clips). I'm 95% sure this is placebo effect, but the 5% is because I could imagine some UI or software bugs that mgiht produce some weird behavior, so it's not out of the question.

[untold briar]

If you want to improve quality in the same song, like keep the voice, you probably need to use a series of 2 to 4 second clips of just the voice.

But 1 second is good enough if you want to mostly restart, yeah. Not always though, like if the song has a sound is 100% present, then even 1 second might not clear it always.

[charred merlin]

I agree, I guess I'm confused about what a zero second extension would do, if not what a terminator does but with brutal destruction for previous musical cohesion, like whiplashing to a different song without a single second of transitory lead

Mind the violent imagery 😛

[untold briar]

A lot of things in the past wanted to use 0 second extensions to do, I can now do with Upload. They are often experiments. What happens if Suno tries to 'continue' from audio that is literally two completely unrelated (not even by 1 second) songs just smashed side by side? Up until we had upload, we couldn't know.

But Upload only lets us do this for 60s, as the first clip. There are still some things 0 seconds could allow in longer songs. It's not a high priority feature at all, but it's the kind of thing like, "Is this just changing a variable from 1 to 0? Then if the value is > 0, why not?"

[charred merlin]

@untold briar, we almost never bump shoulders except for brief conversations where we start to peel back some really cool layers of the separate projects we're working on in our own little corners, and it usually happens right as I'm heading to bed.

You're a cool one 🍻

[bronze helm]

Sorry to interrupt, but is this the right place to share an interesting trick I discovered in Suno when it comes to Bulgarian lyrics but may extend to other languages?

[untold briar]

Yup

[bronze helm]

#📚┃suno-school strikes me as the type of channel to share model tricks

[bronze helm]

Thanks

[untold briar]

What the community needs most is a trick to get Suno to switch into Mandarin or Cantonese, specifically. (read: what I get the most DMs about)

My guess is there is probably an easy trick - something you do in prompts, reliably. I spend a whole evening trying to find one myself, but it was too excruciating to try and listen to the samples and judge if the language was correct, having 0 familiarity with either language.

[bronze helm]

I've tried this on Suno v3.5 only. I've found that Suno often likes not to care about what I want to hear and instead imposes what it thinks is right for my songs, which can get annoying but that's just a model limitation. This can range from incorrectly stressed words to just plain wrong pronunciation (i.e. weird Russian-Bulgarian hybrid pronunciation).

A very unconventional trick I tried is to force the stresses, using Latin stressed letters. Instead of "Началник" for example I could write "Начàлник" and this forces the model to stress the correct syllable across the board from my experience.

When I was making a song, this made it sound from weird rap-like garbled nonsense to actual Bulgarian vocals. Perhaps it extends to other languages?

[untold briar]

That's funny, I was about to post the 'one weird trick' I do know about is from a Russian user who said exactly the same thing. But it's very useful to have confirmation the same trick works in Bulgarian. But it worked much better in Suno than capital letters.

The Russian user said "Instead of capital letters, use a latin style accent character". He said you only find such accents in Russian in a children's book, or something like that, so it's not typical. But accents worked much better in Suno than capital letters.

[bronze helm]

I think I have examples of when I tried this vs. earlier attempts.

Also regarding the accents, for Bulgarian they are available in dictionaries, but in written text we don't use accents.

It's useful to know which syllable is stressed

[untold briar]

In Bulgarian, is there something like the Russian use of capital letters in the middle of word, to indicate stressed syllables?

[bronze helm]

I did try it and results are inconsistent. For example, in one song the first verse always sounded Russian no matter what, but capitalizing the "О" finally steered it towards Bulgarian faster. But it doesn't always work and I don't think it works for stresses.

Latin accented letters seem to work all the time.

Also, using the Latin accents made the vocals sound entirely Bulgarian too, which is interesting.

[untold briar]

I'm clueless on languages. There are two different meanings 'stress', I am just now realizing. One use, like the Russian users, is stress that differentiates between different words. So you find this stress in regular written Russian text.

The other use of stress is to emphasize how a word should be pronounced - to help Suno - but you would not find stress marked like this normally. Is this the type of stress you mean in Bulgarian?

[bronze helm]

Yes, I mean the latter use. If you speak French by chance, you may know that "regarde" and "regardé" would change the stress i.e. which syllable is more pronounced than the rest

In the IPA this is often marked with an apostrophe before the stressed syllable

[untold briar]

As a general trick for Suno, you can also think about adding more obvious signs of the right language, as part of your prompt. For example add some very obvious Bulgarian phrases, that would hard to mistake for any other language, after all your real lyrics. Then when you Extend the song, choose a time before the fake lyrics are sung. So you just add them to the prompt so Suno sees the whole song as more obvious Bulgarian.

I'm going to post in suggestions that Suno specifically train the model to sing in IPA. Why not? You can find IPA songs on YouTube right now! It would be useful!

[bronze helm]

I'll keep that in mind! A dead giveaway is probably the way we translate the English W. Many Slavic languages if not all opt for "V", for example... I guess this is the first thing that comes to mind but I hear other Slavs say Vinks compared to Winx, whereas Bulgarians say Winx as well (Уинкс).

[crimson terrace]

I do the same for arabic, we have something called tashkeel that suno actually knows how to pronounce .. and i use the google translate speak function to hear the word before i paste into suno

I already suggested they add this feature long time ago but they didn't care

[bronze helm]

Similar experience in Hebrew. Sometimes Suno struggles especially with names and gendered pronouns so Niqqud steers it in the right direction all the time.

[bronze helm]

Or for example I just found out that Russians say "Vindovs 10" instead of Windows 10. In Bulgarian we say "Uindous 10" which is basically Windows 10 again. Obviously 10 is different for each language 🤣

[untold briar]

I just looked up "tashkeel" seems like its used a decent amount (not only in children's books or something) so makes sense song lyrics might also use it, and Suno would know it. Non native speakers of a language looking up song lyrics, must be one of the more common things non speakers lookup actually.

[bronze helm]

Unsure as to why Bulgarian decided to be more integral to English compared to other Slavic language but I'd rather say Uikipediya than Vikipediya 😅

[untold briar]

I have heard there are no good Hebrew TTS models, so strangely Suno music might actually be the best quality Hebrew "text to speech" model, but of course you have to fight with Suno to keep it from bursting into singing 😂

[bronze helm]

Despite Niqqud being available only in the Bible and children's books, Suno does understand it perfectly. For example when I wrote "Alex" in Hebrew without Niqqud it sounded different all the time, like "Alkekes" for example. Added Niqqud and it was just "Alex"

[untold briar]

Bulgarian, French, Hebrew, English... impressive spread of languages to know there.

[bronze helm]

Thanks 🙂

[untold briar]

Especially since they are not super close languages.

[bronze helm]

They are all from different families ✌️

[bronze helm]

This is likely because AI is involved in Suno. TTS systems aren't really what we'd call AI today. They don't reason as to what should be what and why.

For Hebrew this is because vowels are never written, only consonants. So TTS systems need to figure out which vowel is where based on literally everything in the context.

Niqqud is that - vowel markings which directly show the vowel placements. They're in children's books and the Bible.

Everything else is written without niqqud, including song lyrics in Modern Hebrew at least

Point is that Niqqud helps wonderfully with Hebrew pronunciation in Suno, and Latin accented letters with Bulgarian pronunciation

[bronze helm]

Maybe Bark could be a good TTS model for Hebrew speech?

[untold briar]

I think any modern TTS could do it, it's just a matter of priorities, for sure Bark could do it. Bark loves context.

I think even a fairly simple TTS could do it, just not something as simple as like, I don't know, Speak and Spell toys from the 1980s

[bronze helm]

This reminded me of "my roflcopter goes soi"

Though I've noticed with Suno v3 and v3.5 that English and French vocals tend to be pronounced without any egregious mistakes

Likely because of the greater number of samples available

The music industry is quite saturated with English and Spanish vocals especially

I can't say for other languages outside of the 4 I speak

I can just say that Portuguese can be messed up because of the Brazilian and European dialects

I can tell them apart

But it's always interesting to see how different languages are optimized on Suno. Latin characters, Niqqud, tashkeel, [PT-PT] I remember is decent for "Portuguese of Portugal"

[untold briar]

Portuguese is a common complaint, but not nearly as common as Mandarin and Cantonese, probably just because of more users rather than Suno being worse

You can use the same Suno prompting methods/tricks for accents, like British or Australian English

[bronze helm]

I tend to notice plenty of Chinese folk in AI communities, especially the Stable Diffusion Discord

[bronze helm]

Had a song switch from British to American English the moment Verse 2 started 🤣

[untold briar]

I haven't tested this in Suno in months, so not 100% in latest versions, but if you want a foreign language accent, you can sometimes prompt it by starting in one language, then continuing in another language. So for example Clip 1 is 100% french song. Then Clip 2 is 100% English, you may get English with a french accent, which is fun.

The accent can weirdly fade away though

[bronze helm]

Exactly what I noticed as well, without doing this too

Interestingly, when I was making Mizrahi-styled EDM (lyrics contained Hebrew), the first one or two lines that were written in English had a slight Hebrew accent

(In English) Edited

[untold briar]

On voice-to-voice singing models (RVC) most of the big projects are from Chinese authors (or papers from Chinese companies). So not starting with text like Suno, but converting one singer's voice into another. I was always wondering what the reason for that is, probably just a popular thing on social media I guess.

[bronze helm]

I do know about RVC. I actually intended to use it locally to make an abridged dub of a series 🤣

Local inference is very fast with 12 VRAM at least

[bronze helm]

About Stable Diffusion

When I was more active on the Discord I often noticed messages like these. Like at least a few times a week, which led me to believe it was definitely popular with Chinese folk

So it may apply to Suno when you say that many people complain about Mandarin and Cantonese as opposed to PT-BR and PT-PT

[gaunt barn]

Hi

[undone marsh]

Hello, I have two questions that I hope someone can help with:

1. How do I set the prompt words to speak the text directly instead of singing
2. How do I make the generated music use my own timbre instead of the AI ​​timbre
   Thank you, guys

[untold briar]

1. Speaking can be difficult. You can try style prompts like "spoken, spoken-word". "live comedy" also works, but of course, it sounds like a comedy show. Also in the lyrics prompt, if the words look more like spoken text, than song lyrics, this helps. For example, a paragraph of complete sentences looks like spoken word. Instead of a verse of 4 short lines of text - that looks like a song.

[untold briar]

2. For timber, you can describe it in the style prompt with words, For example: "bright, dark, warm, cold, breathy, velvety" This sometimes works, not 100%.

Sometimes the words work, but are weak. Can try repeating it "country, banjo, female, breathy, breathy, velvety, velvety".

I haven't tested those specific words in V3.5. They are examples of the type of words that can work though. "velvety" was useful in previous versions for sure.

[untold briar]

2. Another option is to use the "Upload" feature with an audio sample of the type of musical sound, the timbre, you desire.

[undone marsh]

Thanks for your reply, Jonathan, i will try it

[untold briar]

Remember you can also Upload pieces of Suno songs. You only need 6 seconds. So if you have a bad song, that has the right timbre or voice, you can make a new song from a piece of that song.

[remote crown]

is it a bug or a feature that you cant generate a spanish song with word "fuego" meaning fire, because its someone's artist name?

my prompt doesnt event have the word but AI is adding it there it would be nice to have a command like: -fuego to cancel something the AI wants to add there

[untold briar]

It's a bug. But the bug is only in Simple mode, or in the "Write About" button in Custom mode. If you write your own lyrics, you can use the word fuego. I just tested: https://suno.com/song/1787bdb1-ba0f-4bbb-a466-9725ddaac66b

[remote crown]

okey I'll have to better my spanish

[untold briar]

I used regular ChatGPT for that

[remote crown]

Okey I'll use ChatGPT untill fire is fixed, does ChatGPT make better lyrics?

[untold briar]

ChatGPT makes worse lyrics if you just say "Write a song" but if you give more specific instructions it is about as good. Also remember to ask it to make it better "This is kind of generic, and you use the word fuego too much, can you try again?"

[remote crown]

thanks where would I post lyrics, that get flagged for moderating that dont have a single bad word? so they could debug whats frong with the AI?

[untold briar]

We used to have a Google Form, but that was when lyrics moderation was even stricter than now, and there isn't a specific place right now. But if it's a very bad false positive, like you can't even guess why it was flagged, can post it in the bugs channel here.

[remote crown]

En haluaisi satuttaa sua, Sydämen lyönnit yksinäisyyden.
Ei täällä tähtiä tarvita, kaiken voi sanoiksi pukea,
Ilman totuutta, Kuiskaa mitä haluat, niin saat sen

[jade sapphire]

Interesting language discussion, I noticed yesterday when writing lyrics in German that it has a direct impact, unknowingly, I have used flüstern (to whisper) and you can guess what happened. that was a surprise 😀

[undone marsh]

Hi, I just tried the"Upload Audio" feature, but it didn't work, because the music expanded based on the uploaded audio still uses the AI ​​timbre, not my own timbre.

[remote crown]

and here the language translated: En haluaisi satuttaa sua (I dont want to hurt you), Sydämen lyönnit yksinäisyyden(heart beats of loneliness)
Ei täällä tähtiä tarvita (we dont need no stars), kaiken voi sanoiksi pukea (everything can be dressed up as words),
Ilman totuutta (without truth), Kuiskaa mitä haluat (wisper what you want), niin saat sen (and you will have it)

[undone marsh]

The AI ​​timbre sounds like a robot, I want the generated music to use my own timbre😂

[untold briar]

Can you link to a Suno song as an example?

[weak bobcat]

hi:). is there a way to give direction on chord progression for verse and/or chorus?

@undone marsh I was hoping to upload a 60 second voice memo and build on/around it... is this possible?

@untold briar Hi. So, is it possible to use a recording from my phone and use upload audio option to give direction for song creation??

[untold briar]

Yes. Using Upload is not always simple. You usually have to try different options: try different audio samples, try different audio durations (10s is very different from 60s), and so on. But many things are possible with Upload.

Generally:

1. Use long uploads if you want Suno to continue the audio you uploaded. Keep many things the same, voices, melodies, instruments, etc.
2. Use short uploads if you want Suno to take inspiration from the audio you uploaded. Maybe just the voice, or the tempo, or the drums. And make a whole new song, that only some smaller part of it similar to the upload.

If you want Suno to keep things very similiar to the upload, sometimes a blank style prompt - no words at all - works best.

this was one of my first upload tests, maybe worth reposting

[subtle python]

I'm trying to make a song that alternates between Male, Female and Duets. This is what I attempted but it was all female. Help would be appreciated. [Verse 1 - Male and Female Vocals]

Underneath the starry sky so wide,
A blue jay flies, a unicorn by his side,
Different worlds, but their hearts collide,
In a love that no one could’ve denied.

They look up, the night is still,
Hoping that the stars will grant their will,
“Can we pretend that airplanes in the sky,
Are like shooting stars as they pass us by?”

[Pre-Chorus - Male Vocals]

“I could really use a wish right now,
To make this moment last somehow,
To bridge the gap between our lands,
And hold each other with our hands.”

[Chorus - Male and Female Vocals]

Mordetwi, Mordetwi, it’s more than just a meme,
A crossover dream, love so supreme,
Mordetwi, Mordetwi, against all the odds,
We’re breaking down the walls, ignoring the flaws.

[Verse 2 - Male and Female Vocals]

Twilight’s magic, Mordecai’s flight,
Together they bring the dark to light,
In a world where love seems out of place,
They find their way, in time and space.

And though they come from different shows,
Their bond is stronger, and it grows,
“Can we pretend that airplanes in the sky,
Are like shooting stars, just you and I?”

[Pre-Chorus - Female Vocals]

“I could really use a wish right now,
To keep you close, I don’t care how,
For in your wings, I find my grace,
And in your heart, I’ve found my place.”

[Chorus - Male and Female Vocals]

Mordetwi, Mordetwi, it’s more than just a meme,
A crossover dream, love so supreme,
Mordetwi, Mordetwi, against all the odds,
We’re breaking down the walls, ignoring the flaws.

[Bridge - Female Vocals]

Through the pixels and the gifs,
Through the jokes and the tiffs,
We find a story that’s pure and true,
A love that’s strong, and breaks right through.

More, but size limit

[untold briar]

There's no simple answer for this. I have the best chance of duets using prompts that looks like scripts. I quickly tried a few with your song, using ChatGPT to format it (it added some extra stuff that is being sung that shouldn't) but you can get duets at least sometimes like this.
https://suno.com/song/ee775020-673f-4df0-b4a2-35f273207aba
https://suno.com/song/c40be6cc-e454-414e-87bd-70d064acbf73

I used no style prompt to maximize the chance the duet formatting works. With some music styles, it is harder to get

[stiff talon]

Hi, I have a text that is flagged for moderation but i don't see any reasons, words are quite inoffensive and it's not copyrighted text. Can you check please? Here's the text :

[Verse 1]
Весна приходит, тает лёд
Душа поёт, мечта зовёт
Ветер шепчет, день встаёт
Дорога вдаль ведёт

[Chorus]
Жизнь - это битва и полёт
Любовь и боль, огонь и лёд
Не бойся сделать шаг вперёд
Пусть сердце путь найдёт

[Echo Chorus]
Жизнь - это взлёт и паденье
Радость и грусть, мечта и сомненье
Смело иди, не зная преград
Сердце укажет, где твой клад

[Verse 2]
Гром гремит, и дождь идёт
Но солнце тучи разорвёт
Время мчится, мир растёт
Надежда не умрёт

[Chorus]
Жизнь - это битва и полёт
Любовь и боль, огонь и лёд
Не бойся сделать шаг вперёд
Пусть сердце путь найдёт

[Echo Chorus]
Жизнь - это взлёт и паденье
Радость и грусть, мечта и сомненье
Смело иди, не зная преград
Сердце укажет, где твой клад

[Bridge]
От зимы до лета, от ночи до рассвета
Через море к звёздам, сквозь память и забвение
Друг мой, верь в победу, ищи свою свободу
В каждом вдохновении, в каждом пробуждении

[Chorus]
Жизнь - это битва и полёт
Любовь и боль, огонь и лёд
Не бойся сделать шаг вперёд
Пусть сердце путь найдёт

[Echo Chorus]
Жизнь - это взлёт и паденье
Радость и грусть, мечта и сомненье
Смело иди, не зная преград
Сердце укажет, где твой клад

[Outro]
Пусть сердце путь найдёт

[Chorus]
Жизнь - это битва и полёт
Любовь и боль, огонь и лёд
Не бойся сделать шаг вперёд
Пусть сердце путь найдёт

[charred merlin]

I’d recommend feeding your lyrics into https://www.chatgpt.com or https://copilot.Microsoft.com to ask what lyrics are being flagged as inappropriate, and you can even ask it to rewrite that section.

[sudden bough]

Hi there, not a big time Discord user but have it around when i need it. I have heavy problems with designating the styles lately, and just can't figure out why the AI sometimes completely ignore the whole instructions at all. It is really good to waste thousands of thousands of points, but it's beyond annoying and frustrating to be honest. There is a exe file or any table where i can pick those what it recognizes because obviously it didn't have a single idea when i giving Official genre and music types, instruments or general instructions about what i like to get. Thanks for the answer in advance.

Side Note: This is keep happening the last couple of days.

[mystic flume]

Maybe your prompt is a bit much. I'm not really familiar with Skid Row but what about something like this? https://suno.com/song/209064ee-6d33-4ffc-9add-f5c2e255bbf2

[mystic flume]

Give us an example of what you mean when you say you're giving official genre and styles.

[sudden bough]

Official genre of music styles. I don't really know how to say that otherwise. The Official names of different genres of music styles around the world. I am into music for years, but nobody ever asked that question ever before. Sorry, i dunno what you mean. 😆

[sudden bough]

Maybe this helps. Genre and Style of musics. If not this what you meant?
https://en.m.wikipedia.org/wiki/List_of_music_genres_and_styles

[mystic flume]

I was asking for examples of what you're prompting Suno. Sorry, my effort to be precise was probably unhelpful.

[nocturne maple]

hi all. how do I get the "seed" of a song I created to use that in a subsequent song?

[sudden bough]

Finally i found what i looking for, it was taken a little time, but i got the list of genres what Suno Ai recognizes. I was precise enough, asked about the genre what the AI recognize. And found the list. Genre of the music is the very basic start when creating music, can be whatever style genre is, but without it, there's no music at all. Like biology. Without genes and codes, there is no life at all.
Genres like: Hip-Hop, Rap, Soulfunk, Rock, Metal, Classical, Symphony, Opera, Country, Folk... etc etc. You know, the Genre or Style. But thanks for your helpful answers anyway, i got it. 😄🫡 Have a good day.

[stiff talon]

Thanks, I actually did with sonnet, there's nothing, it's the most inoffensive text, translate and you'll see how bland it is

[charred merlin]

Don't use Sonnet, use chatgpt because that's what Suno uses for recognition.

I didn't check the lyrics because they're in Cyrillic, but still, ask ChatGPT what possibly could have been flagged as inappropriate, ask it to rewrite, and if you feel like there's seriously nothing wrong, send an email to [e-mail] to let them know the hypersensitive content moderation is preventing you from using your desired lyrics (include an english translation of the lyrics in your email to them to ensure a faster response, as Suno is an American company)

[cunning mulch]

...is English the least-moderated language? Seeing how far people can go, I've got plenty of room for lyrics with my type of spice, so it's weird to see other languages getting dinged for stuff like that

[stiff talon]

like I said, bland stuff

most probably a false positive, that's why I reported it here

(first time I got a text flag moderated even in Cyrillic)

[cunning mulch]

yeah... I know some songs that make go "HTF did they allow that?"

[charred merlin]

I gotcha -- we can try to give community advice, but the "official support presence" is incredibly fledgling on discord at the moment, and seems more directed to intentional follow-ups than first-contact issue reporting.

That said, I'm just a regular user like you, so I'm not trying to tell you how to use the discord, or how to talk to Suno, just trying to help give actionable advice 💪

[stiff talon]

I see, thanks a lot for your help and telling me what to expect from official support

[charred merlin]

You're welcome! [e-mail] has been helpful resolving the vast majority of issues, and the Support Lead they appointed last month has been increasingly active in the community trying to address any edge cases.

Like @cunning mulch alluded to, there COULD be some kind of bug where otherwise innocuous lyrics are being misinterpreted and lost in translation, so bringing these types of instances up to support, especially in a collaborative and friendly manner, will help the team make things better for everyone.

[stiff talon]

yeah ok

[charred merlin]

tbf, this was all a lighthearted project to make music through a simple discord bot, and now it's the fastest growing AI company across multiple markets.

It's very easy to expect Apple-level standards from them, but they're regular folk like you and me who are needing to wear a LOT of different hats (at least until they can properly find respective heads to wear the hats that fit least comfortably). Like almost every startup, a lot of learning as they go, as best directed as they can hope for.

Takeaway -- if you want to make change, it's more productive to get involved than to get upset

(last part not directed at you, but a general motto in my life)

[stiff talon]

yeah and there's still a lot that is not working as expected so they must be flooded with feedback

[subtle python]

Thank you, this is very helpful, I'll try editing the verses in this way.

[ashen oracle]

Hello everybody.
I unsubscribed from Suno, and Suno keeps trying to debit my card... What else I need to do? If I'm unsubscribed why are you guys still trying to charge my card??

[vestal rapids]

email [e-mail] and they will fix it for you

[charred merlin]

Sounds like you didn't remove your payment details? There's certainly some confusion that would need to be addressed, so definitely seconding robert's suggestion to reach out to the billing team because they handle all account-related issues.

[violet tundra]

"they are not regular folk like you and me." Unless there's a huge difference between you and me (making them more like you than me).

[charred merlin]

I'm caught off-guard by the approach, so to clarify, are you suggesting that the devs are different than us somehow?

they're regular folk like you and me meaning that it's easy to forget "The people" behind "the company", but it's still a small startup (regardless how quickly things are growing)

Apologies if I'm off base, but it seems like you might have misinterpreted my earlier comment as saying "they are different than us", when I tried to express the exact opposite.

[violet tundra]

1. Skills. Mad skills. 2) economic class -- they were millionaires pre Suno. 3). which makes them titans. I'm not a titan. You might be. I'm not.

[charred merlin]

To attempt to clear the air -- the devs aren't magical geniuses, they're really good at the specific things they've been educated in, and beyond the fact that we REALLY admire the work that they do, it doesn't mean that the work is ALL that they are.

Just the same as everyone else, they have their own complex lives -- some people get very upset because the product or company isn't the way they envision it, and somehow feel justified attacking Suno.

[violet tundra]

Well, to be fair they 1) never answer emails. 2) Development is slow for a company raking in $12M per month. To be fair, I suspect the reason is limited talent pool. To be fair. $12 per month is a conservative estimate.

[rose jasper]

@cunning mulch I don't know man, it's not really working with the structure anymore, keeps on repeating Verse 1 over and over, you've had any luck with your structure?

[charred merlin]

You're completely entitled to this opinion, but I don't really feel like it's conducive to any productive conversation, and it actively removes a human element from the interaction, so I'm going to drop it now and get back to work.

I hope you have a wonderful day 🍻

[violet tundra]

Suno is great. Which is why we stick around - sluggishness and minor frustrations aside. I don't think any of us give a sh*t about their lives.

[violet tundra]

I certainly didn't subscribe to their lives. I subscribed to their awesome music-making bot.

[charred merlin]

I don't know you, and I don't have an personal invested interest in you, so I'm certainly not going to follow up on the matter to ensure it,

but as a general statement, looking for small chances to look for "good", I genuinely would rather that you have a better day than not.

[charred merlin]

By making assumptions about their personal motivators, you're absolutely subscribing to their personal lives, you would just rather subscribe to the version that fits your preconceived notions of them.

again, sounds like you've had some colorful history with their support email, so I'm not going to say there isn't anything to improve, and I can tell you care a lot about Suno as a platform. I'm excited for things to get better as well 😄 🍻

[vestal rapids]

you seem to be a very rude and confrontational person, I would appreciate it if you would tone it down a bit

[hollow sapphire]

👋 Hi! While chatting with a few friends in the French channel, we started wondering why there isn't a multilingual channel. It's true that it might be annoying to manage at worst, but perhaps very constructive for the community at best? The reason I'm saying this is that despite the undeniable number of open channels, there's still one missing for multilingual discussions! However, many users would like (here, I'm speaking for myself) to be able to express themselves with everyone, don't you think?😎

[rose jasper]

If I'm not mistaken there is one already?
https://discord.com/channels/1069381916492562582/1154772025940922438

Or you mean you want multiple languages spoken in the same channel?

[hollow sapphire]

Yes! that's true, but not one mutltiligual, in the same^^

[untold briar]

Could be a nice option. Worth considering. 🤔

[hollow sapphire]

Yes, I think so too, because we don't really know each other, so maybe a bit of synergy from the best of everyone could create a nice atmosphere. 👌

[violet tundra]

Apologies.

[charred merlin]

No worries, thanks for taking such an active role in tech-chat and helping others who ask for best-practices 🙂

Sorry we got off on the wrong foot 🍻

[solar lynx]

got a problem, that when getting full song from multiple extended parts - the final lyrics info is totaly wrong - is there a way to fix it ?
because if video is generated it is printing the wrong lyrics there

[hollow sapphire]

Yes, I’ve had the same thought several times. When you change the lyrics and everything fits well, and the song is finished, and you decide to modify the text in the editor, save and confirm it, then you think, great, you download the audio and the MP4 video, and voilà! The lyrics are saved with the old text from before! It’s quite a bummer. 😎

[solar lynx]

yeap, ok, so no solution yet, I guess^^ thanks for confirmation!

[untold briar]

"..." menu, "Regenerate Video"

[hollow sapphire]

Yes, with a paid plan?

[solar lynx]

the problem that the lyrics are already saved wrongly - it regenerates video where lyrics are still wrong. Need some way to manualy modify the lyrics part where song is already complete

[untold briar]

It may be a paid feature, I can't remember off the top of my head.

[untold briar]

It might just be cached, and will be updated in a little bit? It should regenerate with the new lyrics, if not, is a bug.

[cunning mulch]

wish we had the ability to edit lyrics in the song audio after it's 'finished'...

[solar lynx]

ok, will report it then, thanks!
What it did - it kept the lyrics from all my extensions, so that lyrics are duplicated for each new extension

[craggy delta]

hey, not bad at all!!!! 🙂

i can try this style

[untold briar]

You can edit the lyrics before regenerating, for the 'Full Song' which should fix this. if you did that, then maybe it is a bug.

[solar lynx]

hum, triing to understand. I have:
part 1:
verse1
verse2
verse3
creating part 2 using timeframe before verse2 is starting with lyrics:
verse2
verse3
creating part 3 using timeframe before verse3 is starting with lyrics:
verse3

Then I get new song and in lyrics I see:
verse1
verse2
verse3

verse2
verse3

verse3

but expectation is:
verse1 from part 1
verse2 from part 2
verse 3 from part 3

And I don't see any button where I could force set the correct lyrics for the final song. Bug or am I missing something ?

[charred merlin]

Hey @solar lynx 🙂

One of the challenges I often find in helping people use Suno the same way I've learned to is that my teaching methods don't always "click" with people's preferred learning styles -- (suddenly wanting to go back and write thank-you letters to a few gradeschool teachers who made a positive impact on my love of learning -- some of the teachers were clearly there just for a paycheck, but some sincerely cared about making a difference)

To help get an idea of what kind of content people resonate the most with, are you the type who prefers ordered, technical instructions, how-to videos, etc? Do you find it helpful to see generic examples demonstrated (so you can gain a general understanding of the ideas and try to adapt them to what you're trying to achieve with Suno), or can it feel too abstract unless it's using a song that you're actively working on?

Lots of people willing to help, but taking a beat to make sure that we're all playing at the same tempo and key makes sure that any wild improvisational guitar solos feel constructive instead of simply the guitarist's ego not being able to shut up 😅 😛

[ruby garnet]

Hi! I didn't renew my subscription, but I have 2500 Credits again. Why?

[charred merlin]

sounds like your subscription renewed -- you might not have cancelled in time? Contact [e-mail] for account related questions, but you can contact your bank to issue a chargeback if it was unintentional (keep in mind this could negatively impact your suno account)

[untold briar]

it's this button on song page. But make sure you edit the lyrics on the same song you click 'regenrate video'

[safe hazel]

scroll all the way down

there may be a bunch of blank space between the final lyrics and the button

[solar lynx]

Hi @charred merlin ! Nice to see you ! Thanks again for the tips from the past 🙂 It helped and I finalized my first song today.
I started with some videos from docs-guides-resourses, and now comming back to it I see, that the information it actualy contains is very good, I just failed to understand it by first trials.
What I think is missing is how to lead the sond in preffered direction using [] tags, if there is anywhere some predefined list of tags that one can use, or if you can put there any word and it will try to interpret it. Or maybe one can put there even multiple words inside or even put sentences with descriptions.

I.e. I tried small with [Verse1][Fe,ale Vocals]
but most of generations were still with male vocals.
another problem for me - tried to force word spelling with ` like fláme insteed of flame but still could get result with singing it flamē

[solar lynx]

thats awesome, was searching exactly for this, thanks so much! searched on wrong place apparently ^^

[ruby garnet]

How do I cancel my subscription? I logged in on August 24th, had 50 Credits, and today it's 2500 Credits.

[charred merlin]

There's no one able to assist with that on the discord server at this point -- for now, you'll need to email [e-mail] 🙏

[craggy delta]

i just notice somethign interesting!
V3 is more prone to follow genere style while v3.5 is more prone to give you more clear sound clearance and more minutes per song.
Check out this two examples:
this one is with v3.5
https://suno.com/song/d54e648e-12f5-457e-8827-6c6349d7e6ad
and this with the good old and reliable v3
https://suno.com/song/18218b77-9790-4de3-b900-58ff6d1bce21

which version fits the style of music? 😄

hope this kind of bug gets solved in v4 ... if there is one in a future (but with this Sue thing ... i doubt someday we'll have a v4)

[alpine venture]

for those who find it interesting

I'm p sure the tokenizer is llamas

so my previous style sheet is partially wrong, but mostly still right

as there's a lot of overlap

but I need to find more examples to be sure

this is based off one prompt

slacker

ir rather ", slacker"

which in got is sl acker

and in llamas is s lack er

and basically through a few combinations, it seems acker is much more heavily tied to electronic sounds

while lacker is veery close

just harder

than normal slacker rock

or rather -lacker vs -acker

since the tokens don't have a lewding white space

so you need to start with a dash

[rose jasper]

Can you please correct how you're typing your messages, instead of posting one line at a time? @alpine venture ?

[alpine venture]

no

I'm kidding, sure, I'm just accustomed to typing this way

my thoughts don't really come to me all at once

[alpine venture]

v3.5 though is much better at following meta tags

[craggy delta]

hehe, not really:
https://suno.com/song/0ff2188a-1ecf-4b1e-ae07-d8e5c1f16982

even if i ask for an 80s vocalist, it gives me a 2000s vocalists

[alpine venture]

I mean, they both speak a different language, but 3 has a tendency to read the meta tags

[charred merlin]

Congrats on releasing your first song! It all snowballs from here 👀

Ah... metatags & style prompts are certainly tricky because they're all "weighted" differently, and it's less of "a guarantee" that something will work.

(there are no actual numbers assigned, and this is meant solely for explanatory purposes, and likely WILDLY oversimplifies how Suno actually works)

An example of a "strongly weighted" style prompt might be "a cappella" -- this prompt is derived from Suno's training data, so if MANY songs have this "tag", then it's able to associate what "a cappella" is meant to be solely from pattern recognition of songs labelled "a cappella" -- let's imagine this strong-prompt has a weight of "10".

Some style prompts might be more vaguely associated, or might cover a large range of genres, making it harder to Suno to clearly associate what specific patterns are supposed to be associated -- for instance, "rock" or "80s" covers SUCH a wide range of songs. I'd consider "rock" to be a weak-prompt, because it influences the style, but it's not descriptive enough for Suno to know clearly what you want -- let's imagine "rock" has a weight of 4, still recognizable by Suno, but vague.

Some prompts are so specific that there might be very little-to-no tag association in the training data, like when people ask for invented subgenres, regionally-specific sounds like "Washington DC Rap" -- even "phonk" is a newer genre that might not have a broad presence in the training data. While Suno can understand what these are supposed to mean, they're so underrepresented in the training data that they get an incredibly weak weighting score, like 1~2.

---

Since they aren't all respected equally, while there are prompts that are more strongly weighted (again, no official list) I worry that people will assume that any gaps between "what they wanted" and "what they get" from Suno won't be viewed as an opportunity to learn and experiment by making small changes to what you tried previously, but will be viewed as "them not using the right prompts", and lead people to blindly copy others by rote, instead of learning to understand the "Suno-Instrument" from a more personal viewpoint.

As a former guitar teacher, I've seen WAY too many times people trying to play guitar solos because that's the fun part, but without really understanding the techniques that make those possible, it seems more like "magic" than just technical expression of learned patterns.

[alpine venture]

but broadly yes, 3 was better at generalizing

although the gap has closed

substantially imo, either due to our better understanding of 3.5 prompting, or lora

[craggy delta]

yeah, but depends on what, if it means vocal style or decade way of sing, suno just ignores it and keep it with the post-90s style of sing, which in my opinion, is the worst of the worst

[alpine venture]

in 3.5 you can actually dictate weights

[charred merlin]

(forgive the wall of text 😅)

[alpine venture]

in meta yags

[craggy delta]

what? O_o

[alpine venture]

it's kinda wild

[Meta: your-text-here, weight:X]

put several of those in succession

it worked for me several times

[charred merlin]

Do you have a demonstration playlist for comparison & contrast?

[alpine venture]

but it also didn't several times

it's part of what got me thinking the model is a llama one, bc I just tested that one day as a "oh, the meta of the song is..."

[safe hazel]

I tried out the style known as "popera" and Suno certainly does not know what popera is (just my reflection in a discussion of niche styles)

[alpine venture]

but I realized this might be an accidental trigger as llama is a meta product, and so it might be just doing soothing that gets a certain type of attention. to be fair though, all the models train on one another's datasets, so often they confuse themselves in some sense with one another, since "identity" is sort of fluid to them

[alpine venture]

I do not haha

but I recall it, I just didnt think much of it at the time

as the meta tags are increasingly powerful and complex in general

[charred merlin]

That's curious, and I'm intersted in looking into this a bit more -- I'd only focused on the inclusion or exclusion of metatags and styles, but haven't really looked into any user-directed weighting, like how you can wrap stable diffusion prompts in parentheses or square brackets to increase or decrease attention to specific phrases.

All of my experiments into "command-style prompting" gave negative results, and no one had proposed any method of testing reproducibility, so admittedly I haven't considered or explored all of what metatags can do.

I just find it really hard to nail down anything with certainty regarding how Suno works internally the more I'm accepting of my own cognitive biases, and I'm desperate to learn more to get a firm grasp.

[solar lynx]

the weightning explanation helps 🙂

[charred merlin]

I feel like I can find my way through "indie math-rock" in the dark at reasonable pace, only occasionally smacking a shin off a style prompt I expected would have a bit more "give" 😛

[alpine venture]

OK, so here's some examples

had to modify it slightly but it works as proof of concept

as usual, it's not precise, but thus illustrates it; the piano here overpowers the harpischord prompt bc its biased by the model, but this isn't relevant, as it's illustrating the violin switch

hang on, here's the baseline pair

https://suno.com/song/c4ea3334-162e-43c2-ac55-d6d4d9c57d81
https://suno.com/song/22f7d812-b203-4bfd-8e22-d92ccc54fd5a

OK, and here are the same prompts, but with violin cranked to 20

https://suno.com/song/cdc5ec26-f8e0-4298-8d18-0d7850dfc81a

https://suno.com/song/20135af4-ca6a-4d30-81ac-f2eb1542728d

it's not actually modifying weights

it's simply inferring

and basicslly simulating a primitive

version

[charred merlin]

I'm glad you could find something useful in the wall of text 😅

I know it's a pain in the ass because people understandably don't want to be given homework, they just want to know what they're doing wrong, or what to do better -- but since the songs we hear in our minds ear are so subjective, and because Suno can interpret things however it chooses, it's tough to say for certain "indie math-rock" will to XYZ if you include it, because I have NO idea what you're stacking beside it, so if I say with certainty how it'll work, you might "play the instrument" in a totally different way, and I seem like a musical-idiot for not understanding the genres you're suggesting.

Have you already shared your first song in #🐣┃suno-showcase? If not, please DM me the link when you do, I'd love to hear 😄

[alpine venture]

so you need to coax the model with three parts minimally

to get it in the mindset

to get the idea

thus the harpischord

Anyway, long story short, you can get the model to do all kinds of stuff if you have enough practice with LLMs in general. It's harder when it's a black box, but I like it strangely. If it was known what it was, I would have it stripped for meat by now haha

(metaphorically, what i mean is it feels more "magical" when you still have some doubt)

still wish I knew though 😜

example of a failure
https://suno.com/song/63147abc-8f97-4afc-b7bd-e13ffb397fa7

although this is with a space with the colon. Small things like that can sometimes be overfit. Most ai generations involving weight have no whitespace there, so this tracks that's the inference might be weaker.

[ashen onyx]

The auto scroll. Please Suno! Just remove it! It's driving me crazy!!!

[charred merlin]

Hey @ashen onyx, the tech-chat is more meant for users sharing ideas with eachother, and the devs don't really poke their head in here to look for advice on how to build the website, BUT they set up #🚧┃suggestions-layout to encourage this kind of feedback 😄

Yolkhead also pointed out #💡┃suggestions-web and #📱┃suggestions-ios for ideas about the actual music-generation or the iOS app ❤️

[alpine venture]

they're not reliable, but they're MUCH more reliable than they used to be

I also don't think there really is a way to test bc black box

I've thought about this, bc basicslly, it could very well be thay temperature increases with successive gens

on the same prompt

or something of that nature

no way to know

so adherence could go down over time unrelated; it always does seem to me that the first gen is the most "by the book"

but again, I'm subjective and very very prone like all of us to finding patterns that are just the result of bias and limited data

[charred merlin]

Exactly -- perhaps using an extension as "a seed", but still ridiculously hard when we're not sure of how best to create control tests.

I know you really throw yourself into your experimentations, so whehter or not anything ends up coming out of the investigations into user-guided prompt weighting, I'm curious either way 😄

It would be super neat to have a novel method to suggest instead of something like "violin, violin, violin, violin" 😂 🍻

[alpine venture]

yea, I doubt I'll pursue it much, I don't think it's worth teaching the model it as it's prob better at natural language prompts in comparison

and ultimately, you can get the same results by just changing the order, plus if we do it and it trains on it, then it will become highly biased towards the styles of the user's who specifically use it. Basically, it's a long way to improve that skill compared with based stuff like basically feeding style tags into the meta data

sorry, I'm saying Metadata, I mean lyrics

I usually literally just thrown in [sh--- music, annoying] and call it a day

[charred merlin]

when you specifically request "annoying" and it gives you a song you genuinely enjoy

https://tenor.com/view/the-office-dwight-what-is-going-on-what-are-you-doing-identity-theft-gif-21922300

[polar spade]

Add to that that the shape of the lyrics, syllables in each line, adjectives/nouns used, punctuation and so on, I feel from experience so far that suno will change inflections , combine lines and so on to match more the style that the words seem to fit.

With each generation, certainly , it will try variations of a theme or genre, but , for example, have the word "christmas" in the lyrics and more often than non-christmas lyrics, suno might throw in some jingly bells that aren't prompted for....for example.

If you don't properly metatag for a buzzer or laugh, suno should by protocol sing those words as lyrics, but as AI I think there is some logic in your lyrics that it can figure out that you may want a sound effect. Similarly, because of the shape and metering of your lyrics suno logically guess you may be trying for folk music instead of hardcore and subtly lean the song in a folk direction underneath any prompts you've given.

Does anyone else get that spidey-sense?