📚┃suno-school

[broken lotus]

we are all using the website now
https://www.suno.com/create

same deal: free daily credits, but MUCH better audio.

[shrewd light]

Hi, in version 3.5 you can't use the "Extend" function, it always ends up with an error.

for example
https://suno.com/song/09c1d5a0-0d23-4fb6-9137-03361ad7c412

[broken lotus]

the Devs have seen a few reports on this... They are working on a fix.
It seems to effect a small number of v3.5 clips – they can't be extended from other models....
Unfortunately there is no workaround yet...

[shrewd light]

Thank you

[cerulean flower]

Hi, how can I know what the hell is wrong with my flagged lyrics ? XD
Trying to do a joke song with quotes from ganados in RE4 and "¡Detrás de ti, imbécil!" is already flagged
I mean, i get they don't want us to use bad words, but "imbecile" isn't that bad 😂
There's really no way to know what is flagged ?

[undone star]

you can run it through chatgpt and ask what might be an issue with the lyrics.
as an alternative - remove the line or word you think might be wrong and replace it - if that works you know the issue. ||you can also spell the word differently and try to weasel yourself around it||

[cerulean flower]

yeah I'll try that thanks

[charred merlin]

Hope I picked appropriate reactions trying to plug up the reaction spam in announcements 😅

[feral coral]

Is there a name yet for that ubiquitous 'noise wall' that gradually encroaches? It's most obvious in quiet tracks.

[charred merlin]

As a huge fan of Control (2019), I nominate "The Hiss" if a name hasn't been chosen yet

[feral coral]

Not sure of the reference - it seems like a bigger range of sound than Hiss implies, but it's on the right track

[brittle bolt]

Hey everyone, interested in the technical details - how could i train a bark model from scratch on new language or fine tuning existing one

[broken lotus]

it's called THE SUNO RING

[feral coral]

lol

i like it

[broken lotus]

Not many Bark Experts here anymore. 😞
Probably the best place to find like minds is on the github where Suno/Bark can be downloaded.

[brittle bolt]

Thank you

[glacial scarab]

Last time there was no bark training code available anywhere i looked though. If you find anything let me know 😅

[brittle bolt]

Nothing so far

[grim karma]

Are the images that are generated with Suno also usable for commercial use like the songs with a subscription?

[broken lotus]

yes, the same deal.

[neat field]

Bark is fully generative text-to-audio model devolved for research and demo purposes. It follows a GPT style architecture similar to AudioLM and Vall-E and a quantized Audio representation from EnCodec. It is not a conventional TTS model, but instead a fully generative text-to-audio model capable of deviating in unexpected ways from any given script. Different to previous approaches, the input text prompt is converted directly to audio without the intermediate use of phonemes. It can therefore generalize to arbitrary instructions beyond speech such as music lyrics, sound effects or other non-speech sounds.
https://github.com/suno-ai/bark?tab=readme-ov-file#️-details

one of the only open source repos i found with fine tuning for audio task is from Meta (facebook)

https://github.com/facebookresearch/seamless_communication/tree/main/src/seamless_communication/cli/m4t/finetune

[manic pelican]

hello i am new in here. i got error says "messages cannot be loaded" how can i make a music?,

[runic perch]

Hi I was wondering do we have "render" function in the Suno? as for like I gave it a rough sketchy melody and they can sort of "render"it to make it sounds like a complete song?

[glacial scarab]

like this?
#💡┃suggestions-web message

[simple osprey]

is it a known problem that after the song "ends" it takes a few seconds and then the Ki starts with random shit again? V3.5

[weary wharf]

that's coming soon 😄

[broken lotus]

it's this... #📢┃announcements message
coming VERY soon.

[green walrus]

Having this same issue! Anyone know how to fix??

[runic dock]

I think that i am quiet ready than

[woeful sluice]

hello

I was wondering is there a way to make subtle changes to lyrics in custom mode of an already generated song without having to regenerate a track which will be completely different ?

[runic dock]

Not yet

[woeful sluice]

ok thanks !

[weary wharf]

Short answer no.
Long answer, you can try creating a new part from where you want, and hope continuation stays the same, you can merge both parts in a daw or an editor like audition.

[wheat rose]

working with 3.5 full release today seems a bit improved from the beta.. but the new problem is getting the song to actually end at some point instead of it continually repeating

[tacit hawk]

Hello. This is my first message here. Is there a way to indicate that different parts is the song should be sung by different voices? Like (Male, Female, Soprano, Alto,Tenor, Bass, Chorus) ?

[fluid robin]

[Rockabilly intro riff]
[Upbeat Rockabilly Vocals]

[Instrumental Breakdown]

[Rockabilly Guitar Solo]
[Driving Bassline]

[Rockabilly Drum Fill]
[Raw Vocals with a hint of gravel]

[Boogie Woogie Piano Riffs]

[Rockabilly Guitar Solo]

[Drum Break with energetic Snare rolls]

[Wild Saxophone Solo]

[Boogie Woogie Piano Riffs]

[Instrumental Breakdown]

Handy tags if u want to make rockabilly music

[broken lotus]

https://www.suno.wiki/faq/style-and-lyrics/duets/

[simple osprey]

Currently Suno wont create any sound?
got 8 empty songs in a row?
help!

anyone from team? is this known? 😄

[vast surge]

Is it possible to get a song to end naturally before 4 mins? Also is it possible to not get that rising background noise in every song? Not sure how to describe it

[simple osprey]

try [End] or [Outro] at the end

[vast surge]

Thanks!

Does this work for instrumentals too?

[simple osprey]

i think

[vast surge]

👍

[weary wharf]

Whoo... [Rockabilly Guitar Solo] works really good 😄

[karmic tartan]

hey anyone got a better way to make it stop than [end]? it used to work on v3 but on 3.5 it just seems to draw it out for a full 2 mins of absolute madness

actually nm i just switched it back to v3 for the last 7 seconds of fadeout

that worked

[karmic tartan]

ok is there a way to make the vocals louder ? or a seperate track so i can mix it myself? the mix is off so often its actually quite bothersome

[hollow hare]

Great to hear that method worked out. IIRC, you can use {end} to directly call the terminate procedure, though I do not know how effective that is in 3.5

In case you did not know, here are the other ending tags: https://www.suno.wiki/faq/making-music/how-do-i-end-the-song/

[low coral]

Is there a specific, generally-reliable way to prompt in order to get it to generate a specific chord progression?

[karmic tartan]

if you work this out mate tag me in it. i NEED this

[low coral]

I'm not having any such success so far, unfortunately! I tried style, maybe it'll work in metatags

[weary wharf]

Chords do not work. You can only specify major and minor chords in style as name, maybe a little more with details, but saying [Cm] or [E#], those do not work.

[light ermine]

question, when I try to add a musical or melodic vocal intro it always skips. am I missing something?

[gray swallow]

I know this is not a simple ask but sometimes the voice used is perfect then when it gets later in the song it becomes more metallic. Is there a way to get the voices better?

[snow maple]

i didnt know it automatically renewed, i didn't use any credits can i refund

[fallen pawn]

I have noticed that as well, especially with long songs.

[south kindle]

I’ve just been re-rolling the dice if a particular extension doesn’t maintain the audio quality. If you aren’t careful you can lock in that unwanted sound. I just keep rolling until you find one with good audio, good beats, and good vocal rhythm. The holy trifecta.

[frozen badger]

The voice quality is really autotune like at the moment. Hope the team can sort this because ElevenLabs incoming and Udio are just so much vocally better than Suno at the mo in my opinion. Also half of my soul stuff sounds country. Ugh frustrating.

[valid flax]

I think i found easter egg (~~ bugg~~) ... when i use hurdy gurdy as Style for music it forces to generate random style of music with quite an interesting melody variation. It was also able to generate few songs with vocal humming which was quite suprising as well...

it also kinda edges the hallucinations

here is one of the songs https://suno.com/song/29dd746f-f27f-4b3c-a98c-1a06e210b297

[south kindle]

I think piano is just the fall back when it is like "I dunno wtf you mean"

[valid flax]

This is just one example from like twenty songs

[modern belfry]

I have a dumb question. I want a video that will beat to my music. But I can't for the life of me figure out what its called lol
something like this for example
https://www.youtube.com/watch?v=e6nXweiNaeI

[versed nymph]

why do suno 3.5 songs sound so generic, regardless of genre?

[valid flax]

wdym?

Looks like skill issue tbh

[final hazel]

Can you drum the beat? with your fingers or pots and pans?

[versed nymph]

me and my friends like to do songs in suno

[modern belfry]

lol I mean I could hahahah

[versed nymph]

but when 3.5 released for all

we notice that all the songs are generic

and very common

[valid flax]

send me example

[unique stone]

Would you describe this as generic?
https://suno.com/song/dc8529a8-ef93-45d8-bb7d-9b0d28b30a1b

[versed nymph]

all songs sounds like clothes shop songs 😭

[final hazel]

Then the solution is coming soon: https://x.com/suno_ai_/status/1794367911408353349

Suno (@suno_ai_) on X

Make a song from any sound. Coming soon 🎧

VOL-1: A watering can, but make it psychedelic rock

Twitter · May 25, 2024, 2:00 PM

[hollow hare]

Use dashes instead of spaces for a compounded terms will help get the correct genre (e.g. hard-rock)

See Jonathan Fly's post about 3.5: #📚┃suno-school message

[valid flax]

i found 3.5 to respect genres more alongside labels in lyrics

[valid flax]

reminds me of those old flash games ... they ofteh had this kind of sountrack

[unique stone]

It's not mine, it's by Mr.Tree but it seems unusual

[modern belfry]

No no nothing like that lol. Just watch the video I put on here. You'll see what I mean

[valid flax]

people need to understand that RNG gods has major word in suno's music gen

[south kindle]

i would be interesting to see a contest where there is a reference track and you need to try and generate something as close as possible.

[final hazel]

well.. it actually can also work with audio as far as I understood.. so if it's copyright free, you can probably use this. I saw another example of someone playing paino in one of Suno's teasers.. so I think it actually can, in the nearby future 🤷‍♀️

[karmic tartan]

so i found a work around for the poor mixing . theres a site called vocal remover that lets you seperate the music and the vocals you can change the balance on there and it works really well. afai can tell its just one dude and some ai you might wanna hire him..

[south kindle]

Try Ultimate Vocal Remover 5. It is free and open source.

[modern belfry]

It took a bit. But I found out what it is I'm looking for. It's called a beat visualizer video. Thanks for trying to help out tho! I appreciate it!

[south kindle]

you can split all sorts of stems, as well as match the eq of other audio, etc.

[karmic tartan]

cheers io'll give that a google now

[south kindle]

You need to download the various models that will do the heavy lifting once you install it. I dont think it does a whole lot by itself out of the box.

[lone oak]

I didn't expect that the rock melody produced by 3.5V is worse than that of 3V. Although the sound quality of 3V is like an outdated tape, it often speaks nonsense and even can't distinguish the type of rock. But now it's good. 3.5V always reads lyrics to you and sings mixed songs of men and women. It seems that 3.5V can only pick up the slack for songs generated by 3V.

[unique stone]

I think I have a better idea now of what the scratchy/grainy quality technically might be - sample aliasing/dithering. Here's a short snippet of hallucinated vocals. The first half is the original, the second half has been saved as 8-bit WAV. The fuzziness you hear in the quieter parts would also be affecting the louder parts, which sound about the same.

This is obviously a very primitive test and I have no idea how the AI actually generates its samples, I just remember some audio stuff I looked at years ago where there were different ways of "smoothing" the sample amplitudes (typically when resampling/mixing) so I wonder if something similar would help here, when generating the audio. I don't think it's something that can be done easily as a post-processing thing but I'll see what I can find.

[charred merlin]

@broken lotus, just wanted to share-- the revelation that Soundcloud doesn't allow the distribution of AI generated music bugged me enough to sign up for a Musician+ account on Distrokid, after having seen it mentioned by yourself and others.

I'm sort of glad Soundcloud has a narrowminded definition of what constitutes as "music", because that was the kick in the pants that I needed to take my fun projects more seriously and take steps to more professionally release it than with the occasional YouTube drop.

[lone oak]

Don’t know how Soundcloud determines whether your music belongs to AI?

[charred merlin]

Honour system at this point, but it's stated in their terms & conditions that you're not allowed.

Doesn't mean people still won't, or that they're able to necessarily tell in every instance, but that's why Instagram added a "made with AI" toggle they want you to use, policed largely (if not solely) by the honour system.

[lone oak]

https://www.youtube.com/watch?v=V7FCEr0keUU&t=431s

[lone oak]

OK, but now there are AI splitting tools that can separate the accompaniment and vocals, and the vocals can be trained by AI and replaced with your own voice. I have seen people process AI songs in this way.

[charred merlin]

#📚┃suno-school message

[lone oak]

Although Distrokid has a low starting price, it actually requires continuous renewal and you will have to pay high fees for other services in the future. CDbaby is more conscientious.

[charred merlin]

I appreciate that you're trying to help, but I'm not really looking for unsolicited advice on that matter at the moment-- thanks though, and I hope you have a wonderful night otherwise 🙂

[south kindle]

to be honest, where do you draw the line with adding an "AI" tag? If you use a keyboard that generates drum beats to play along with, would you need to declare that as well? I feel like it could be a debatably slippery slope.

[rapid bluff]

I am so confused at how this guy is pricing things, most people are just trying to get their music on Spotify or whatever music streaming platform, with Distrokid, its a minimum of $22.95/year to achieve that unlimited music uploads as often as you want which is great for people like us who can create a full coherent song in under 15 minutes. CD Baby is 9.99 per upload AND a 9% commission.

Me personally have 61 releases only 7 months in to my subscription with United Masters which is $59.99/year. If I released mine through CD Baby thats $610 give or take, that would have cost me. If I didnt release anything else for the next 10 years it would equal out.

[lone oak]

Because some of Distrokid's distribution channels are paid, there are hidden fee services compared to CD Baby. For example, if you cancel your subscription, the songs will be removed from the shelves, and it will be a huge expense to keep the songs.

[lone oak]

His calculation method is not very reasonable.

[delicate orchid]

suno have a "was this audio generated with bark" classifier model, so potentially they either have a similar model for music or the bark one also works

[rapid bluff]

if you only release music a couple times a year, sure it makes sense. I can promise people creating AI music will release probably 10-15 times more than the average artist not using AI every year, which $23 is not bad at all if thats all I have to cough up per year and not lose any commision

[cunning vector]

@modern belfry I use FL Studio's ZGameEditor plugin. Not sure if it comes with the free "unlimited trial" version or not; or if you'd be licensed to publish work while under trial. But IMO it's worth paying for because it's my creative outlet most days. Here's an example (no idea why the timer is wrong - didn't notice that until I published it on YouTube whoops!):

https://www.youtube.com/watch?v=i4ae5Td8v4A

Plenty of visual bases to choose from (like a mountain flyover or something more Matrix-style) plus visual effects you can layer on. I barely know what I'm doing with ZGE so I just pick premade ones.

[young frost]

Is there a size limit for the cover art? The only reason why i can think that some of the images are not sticking is that the image is too large. Strange that it appears to attach though and disappears only after refreshng the tab

I just got told there is a size limit so will try reducing image size and reporting back

[icy valley]

Guys, can i replace the lyrics on the second part of my song using extend while keeping the exact melody and rhythm of the first part?

[young frost]

setting the image size to 1024 pixels worked

[gray swallow]

Agreed however from a creative standpoint I can Hit my ideas so much closer with suno. Over all I like it better just when the voice starts out perfect then the modulation gets all tiny and a little shrill.

[young frost]

The images are gone again. Is there a maximum number that can be uploaded?

[rigid jolt]

i'm here croza

[half void]

@rigid jolt Alright, so to make a colab work without diminishing quality, my method is to space it out minute by minute. Triplet lyrics work well in rap, sometimes it'll take to longer sentences if they're spaced and placed beneath each other in the lyric prompt too. If you start at lets say, 0:12 and drop some lyrics in the prompt, Suno will take what you make, and usually speed rap it if your lyrics are placed right. The shorter the bars, the faster the lyrics go normally, sometimes you'll get fun little switch ups too, but with v3.5 you're going to get a lot of consistency with the original melody so don't worry too much about that.

[clear stag]

Overall, the improvements in the 3.5 update are pretty good, but I have the feeling that "Style of Music" no longer has as much influence on the generated song. For example, the genre "phonk" is no longer covered at all. I've probably tried to create 20-25 songs in the phonk style, but it just doesn't fit at all anymore.

[remote dome]

I have similar experience - I found that generating style and genre with v3 and then extend with 3.5 is a good "workaround"

[untold briar]

I think that's right. Try phonk, phonk, phonk, phonk maybe. Some other suggestions: #💬┃general-chat message

[untold briar]

Also a useful method and not one I mentioned on my list.

[sharp kraken]

Try using bigger genres and narrow in tooo - like if you want oldschool memphis phonk start with dirty south rap and then add memphis phonk

I have to do that to get to psycore through darkpsy

[stray hull]

If you are on paid subscription should the Suno watermark still appear on the lyric video ?

[graceful condor]

Not necessary

[fluid robin]

[Eeire Music Box] Scary Haunting Female voice DARK - Good for the darker type of songs

[runic dock]

Yeah. I got the same issue with roots reggae. Putting multiple times in the style generates now some reggae songs. But unfortunately not able to have what i was having consistently with V3. I mean for the specific type of reggae. I will try putting à dash (like Roots-reggae) to see. Definitively, seems broken for this style.

[cold zenith]

they should put the 4min on V3 too.

[twilit patio]

How exactly does the trending songs list work? Is it only based on how often people listen to a new song? Or how much it is liked? What is the mixture? And how and when does the ranking decrease when the song gets older?

[broken lotus]

it is a combination of all of these, but there is also a stricter content filter.

[twilit patio]

It seems that if you make songs available they don't appear in the new-section.? But sometimes there are two or more songs of the same user in that list? Is it completely random which songs are listed there?

[pulsar grotto]

working on a suno reproduction now - got some killer vocals that make me happy

[ripe shard]

Hi everyone

I have question, when i create a song with lyrics and i like the instrumental, i cannot seem to generate the instrumental alone, is there away to do it? I tried to extend the song but when i try to generate it with instrumental it does not work, any help?

[lone oak]

V3.5 is very much like the song sung by the translation device. The language is more accurate than V3, but there is no good melody. On the surface, it can generate 4 minutes at a time. The number of defects is more than V3. V3 can generate good results in dozens of times, but now V3.5 needs to double the number of times. When this situation is serious, I don’t even believe that V3.5 can generate moving melodies. Now it seems that suno cannot take into account both accurate language and moving melody. Perhaps it is a better solution to accurately locate the scope of Style of Music and strictly limit it to the type of music, while increasing the scope and weight of lyrics punctuation and marking.

suno can take more points from me, but I don't want to spend more time generating flawed music

Song generation and management is taking more and more time. If there is a way, I suggest that the official website UI and file classification method be improved.

[jolly gale]

How do I get a song to end? When I generate a song with model v3.5, it wants to continue. Is there a certain prompt I use for the song to finish?

[lone oak]

This is a bug, and the official may not be aware of it yet.

[jolly gale]

Oh okay. Thanks.

[lone oak]

Although the official gave a generation time of 4 minutes, they gave a location of 3000 bytes, which has not been reasonable since V3.

[jolly gale]

I see. Is there a way to contact the official to make them aware?

[lone oak]

[e-mail]

[jolly gale]

Thank you.

[lone oak]

The official maximum range is about 150 words per minute. It is hard to imagine what the official is thinking. I have to send an email to remind the official.

[steady owl]

"My credits have disappeared, and the subscription points are not adding up. I had 2500, now I have 550. This is called professional swindling."

[vestal rapids]

did your sub end?

[lone oak]

This unreliable design is reflected in the volume of the lyrics in the V3 version. If there is too much amount of lyrics, the speed of speech will be accelerated. It is reflected in the V3.5 version that will cause the song to continue

[lone oak]

The new month starts on the 2nd. Have you recharged until the next month?

[broken lotus]

I just switched to v3 for my [fade out to end] worked perfectly

3.5 wouldn't get off da stage...

[jolly gale]

Oh, that's perfect. I'll have to try that. Thanks!

[unique stone]

The subscription month doesn't start on the same day for everyone - it starts on the day of the month that the subscription was started. For example, mine is on the 6th:

[vale flume]

I assume it's random or does not appear due to the number of created songs. There are probably maaaaany people using Suno already.

[little tinsel]

Premium needs Stems

[rain mist]

any good way to force a song to end? in this case, cos there's a good 20 seconds of silence at the end i want to cut off~

[charred merlin]

External audio editing software, like audacity

[little tinsel]

I always put [outro ] then [end] - seems to work well most of the time

[rain mist]

i had [Big Finish] then [End] - it ended nicely, but yeah a lotta nothing afterwards up to the 4m mark

[charred merlin]

If songs don't want to cleanly resolve, another trick is to extend a solo / chorus / etc at the end, and use a studio fadeout as it loops in the background

[undone star]

It can't create perfect stems though as the training data is full songs imho. So the chances of the stems being better than uvr5 (which is free) are low.

[unique stone]

Stems would be incredibly useful. They're suggested quite a lot and I expect are probably a long-term goal. Suno would need a model trained with separated stems which there are likely far less of (compared to fully mixed songs) and also generate multiple audio clips per song (more compute power.) Unsure how feasible this is at the moment. I'll definitely be celebrating when it's possible though - I'm often basically prompting Suno in a way to tell it "yeah that was nice but could I have it again as a vocal solo please?" 😄

[broken lotus]

Also..., all apps where stems are useful, now have a built-in stem feature.... and you might only need to demux a section, not the full song.

even my DJ software has internal stems (even live stems 🙀 )

so ... I see a workflow issue downloading multiple files from Suno to load in software that can already create its own stems from the single download.

Suno's stems would need to be special somehow to justify the extra file management.

[bleak bobcat]

I prefer Cakewalk by BandLab, but it seems is gone to be by payment the new versions. I like Audacity, but i do'nt know is for me, but editing a song change tempos, velocity, and corrupted the project repeatedly. Cakewalk has multi-tracks and many possiblities, but only i have to edit some clips and crossfades.

[vestal rapids]

I use GarageBand to edit audio, it's good and it's free!

[bleak bobcat]

Thanks. I see is for Apple I tried to search for Windows if exists 🙂

[vestal rapids]

it's only for Mac

[unique stone]

The stems offered by Suno would need to be actual true stems to be worthwhile - so we could pick an individual instrument to silence or adjust the volume level of, or apply effects to etc. It would be a huge milestone.

Existing stem separation software does a pretty good job, but I notice obvious "holes" in the sound where percussion has been removed, and/or I have trouble with vocals where there's a vocal harmony I want to get rid of, or the vocal blends partly in with an instrument so it ends up with a portion of the vocal in an instrument stem, or a portion of an instrument in the vocal stem. I can sometimes work around this by the above method of convincing Suno to output an isolated part - typically this is for vocals, as I can at least re-create the instrumental parts manually.

[vestal rapids]

Aiva does stems, which is good but it's more mature than Suno

[fallen pawn]

Does that work on 3.5? I am having a very hard time finishing a song in 3.5. Nothing seems to work. I changed to 3 and all the tags work.

[charred merlin]

Studio fadeout using external tools as mentioned in the other nearby comment ❤️

[fallen pawn]

its seems that on 3.5 the song extends until the fills 2min no mater what I do at the end.

Yes I do get that, but it should work. If it does not needs to be considered a BUG.

[charred merlin]

I've found it can be tricky specifically asking the AI to end a song if that's primarily what the extension is trying to create. It's designed to make music, and we're asking it to create a "new song" that's largely composed of a musical conclusion.

I don't know it's a bug so much as a misunderstanding of how tags work. They're not explicit instructions, they're suggestions

If the AI isn't creating specifically what you're envisioning, it's not 100% a failure of the model, you might need to be more adaptive in the way you approach the issue; try choosing an earlier extension point, remove all existing style tags and replace with requests to fade out or resolve the music, try experimenting with lyrical metatags, use external software to isolate tracks then mix & master the old fashioned way, etc

[fallen pawn]

True, I think it should at least follow the suggestion but it seems that it is not even trying to do that.

[charred merlin]

Not to say that there aren't improvements that can be made on the model, we're on such a bleeding edge that even 'experts' with this particular model are learning new things daily from users sharing their feedback and discoveries. Within the last week they've pinned new content to this channel because they found helpful experimentation tips

I think more can be gleamed from this discussion than just "why doesn't it listen to me?"-- presumably the devs would like it to 100% adherance to prompts and style requests, but generative AI just isn't there in general.

[broken lotus]

then report it in the https://discord.com/channels/1069381916492562582/1082775596767117312 channel...? 🙏

[charred merlin]

Fixing the problem isn't exactly as easy as "increasing a slider"-- for example, CFG (classifier-free guidance scale) for image generation boils down to "how accurately the AI should stick to your prompt" vs how much it should "interpret" things.

Setting a higher CFG doesn't mean it understands better and can produce a better result, it often gives a really wonky result

[fallen pawn]

Yeah sure, but one thing is when the song is kinda on it's peak lyric wise and when you ask it to extend it goes all the way to 2 min, another thing is there is no lyrics at all, there is a prompt to "end" the song and it still goes all the way to 2min

[charred merlin]

In order to get better, more specifically accurate results from prompts, advancements like dalle 3 need to be made, which are near-exponential steps in the parameters and training required to create them.

It WILL get better, but try not to be so critical of the robot, it's learning and trying its hardest for us.

[wheat rose]

had a really great generation with a tiny goof early on... perfect opportunity for inpainting if we had it available 😦

[fluid robin]

[Eeire Music Box] Scary Male voice DARK Happy [Old static] [Old voice] Clear voice, energetic, upbeat, rock, metal

👆

[fallen pawn]

Sure. Not critical at all. We are just having a conversation, and I am expressing my opinion. Not dunking on anything or anyone. It is super fun how it is, and like all AI related things needs time to improve.

[lone oak]

I gave those sad songs that I fed to 3V to 3.5V and they all became songs full of youth without exception.🤣

3.5V turns the sad label into cheerful. 3.5V is really an emotional AI.

[mint spade]

Hey guys, great work, not sure if repeated notes, but my initial feedback after the first hours.

• Easy to use, lyrics generator is great that it takes in whatever input you add before
• Musics end abruptly (not sure if happens all the time, but happened with the ones I generated) - should have a breather for the notes to fade out
• Love the extend feature - really adds to the creative possibilities
  All in all great fun!

[broken lotus]

lots of tips here, including how to get it to end in different ways
https://www.suno.wiki/faq/metatags/

[charred palm]

1.5 days in Suno. 400 Credits burnt - No regret!

Loads of smile on friend's face! Much good laugh!

[sharp kraken]

Yeah suno Roks guys, thank you

🙏 bless da dev

[torn needle]

Stems are needed for those of us who take suno songs and "play them" with other VST's in DAWs. Been on a couple of months and now subscribed, I do mainly scores. Version 3.5 is killer IMHO. The thing with current ways to produce stems (I use Melodyne) is the time it takes to seperate the notes. If you have a knowledge of note ranges the cut and past is not too bad, but it would be very nice to have the ability to down load the stems for each instrument. The time saved would be great. For those who need to produce scores in a hurry, Suno is a great tool, but stems would make it a tool for professionals that would be a "must" have IMHO.

[vestal rapids]

Stems in Logic Pro work well, it separates them very well.

https://support.apple.com/en-au/guide/logicpro-ipad/lpip1b60ada3/ipados

[raven turtle]

Hi all. Is there a tips sheet anywhere that can help get different vocalists than the usual male/female drawl that it churns out? I've tried many suggestions like "German accent", "baritone", "dry Australian female" etc. nothing seems to work.dammit

[broken lotus]

each model has quirks and prompts a little differently

https://www.suno.wiki/faq/metatags/voice-tags/

[lament cove]

[Spoken] tag works really well now if you want a [Spoken Intro]

It works with reggae and raggamuffin style at least (that's what i have experimented at the moment)

[raven turtle]

You star! Many thanks!

[small crypt]

How do we download the song with all the extensions/parts as a whole?

[graceful condor]

Click on "get the whole song".

[small crypt]

Thanks

[graceful condor]

By the way, you must click on "get the whole song" on the LATEST part you've made so far. Then, it will start on second 0:00 and end on whatever minute it is in the part you chose. If you chose an early version, let's say, the intro only, even if you have created more parts, they won't be taken into account.

I suppose that makes sense.

[small crypt]

When we publish them how do we publish the whole song?

[graceful condor]

First, you must "get the whole song" and once it's made a single file, you publish it, and that's it.

if you make only a short part public, only that part will be heard, not the whole song.

[small crypt]

Cool, thanks!

Can we remove the parts of the song? especially at the end. When extending, some times they song starts again after ending.

[graceful condor]

I know what you mean. That's a bug. That cannot be done on Suno directly but you can download the song, remove the glitchy part with Audacity, Adobe Audition or any other audio software, and then export it again and upload it on Youtube.

[small crypt]

Right

[graceful condor]

But then you won't be able to share that fixed version on Suno directly, of course.

[small crypt]

Hopefully that feature is added soon

[graceful condor]

Yes, and also some "Audio Inpainting" like Udio has. Have you tried it? I like Suno way better than Udio, but the "inpainting" tool is amazing.

But the overall experience on Udio cannot match Suno in my opinion, specially because it's only 30 seconds per segment. Too short.

[small crypt]

Yes I have

Yeah Suno has better structure

[fluid robin]

if u want to delete alot of your bad parts in large quantity hold ctrl and left click Which allow you to select a large number of generations on the page u are on

[graceful condor]

Great tip!

[charred palm]

Hi Sunoites, I just blew the biggest amount of credit ever trying to create a Norah Jones AI knock off. I got a song that was good but it ended abruptly. The style of music I put was "Jazz, minimalistic, slow". Somehow the AI will always create something too over the top. The more styles I add, the more of a trainwreck it becomes. Any tips?

[sleek jay]

Please give me a prompt to specify the tempo of the song. For example, when I create a smooth jazz song, it has a slow tempo. I would like to make it a fast tempo.

[graceful condor]

As far as I know, you cannot imitate a particular artist, but just emulate the genre.

[graceful condor]

Try [Fast tempo] before the verse starts. That's how it always works, commands go in brackets.

[charred palm]

So I am doing a slot machine here...

[graceful condor]

For now, that's how it is. They said they will allow you to upload music as guidance to generate AI music based on the sample, like image AI does.

But that will be in version 4.0

[charred palm]

Got it... so I should manually trim the semi-baked product's ending for now. 🤣 Will be patient meanwhile!

Appreicate the tip! Let me try that!

[graceful condor]

You're welcome!

[indigo imp]

I can change the cover image of my song, but when I export as video, the original auto generated image shows. is there a way to export with my new image?

[graceful condor]

Sounds like a bug. You should report it.

[charred palm]

Ditto. This happens to me too.

[indigo imp]

will do

[hollow hare]

This is because the video generates when the track is initially generated hence why they get stuck with the title & picture it had when it was made.

Updating a video each time someone would change their image would be too costly to do so (esp. when someone is testing out several images).

You can suggest to have an option for a regeneration of the video at #💡┃suggestions-web

[sleek jay]

Thank you! I got a nice instrumental piece with the [middle tempo][intro] prompt.

[graceful condor]

Excellent. That's how it works for most cases. Just make sure the command makes sense within the music context. There's actually nothing written about it. It's just common sense, trial and error, lol.

And always remember the brackets.

[charred palm]

Works for me too. Tempo is much slower now!

[sour delta]

Is it just me and my bias of understanding these more than other genres; but is chirp more proficient with Jazz/classical than other music types? I find it much better and close to indistinguishable in those genres - almost like it feels theres a training or method limitation with other forms of music that they do not have? or is it simply based on how much it was trained on of each (jazz and classical are more commonly public domain so..)

Noob with the model side of this but looking to learn by being curious about these kinds of observations,

cheers

[oak merlin]

Anyone know how I add an instrument to an existing song?
Context: I got a really good song, but I'd like it's ending to be even more intense, adding instruments, like a guitar and whatnot.
I could get higher energy choruses by extending the song and rolling the dice until something good came out, but I can't seem to get a "[piano solo]" to happen at all. My idea would be to extend the song with a piano solo, the use audacity to insert the piano into the chorus to make it "richer? deeper?" not sure the term for it, but to give it an extra umph.
Example: https://www.youtube.com/watch?v=7b9DWZ1OxwI at 4:15 for the last chorus they add an extra (guitar? bass?) string instrument to intensify it even further for a big finish. I wanted to do the same but I can't seem to figure out how

[normal patio]

Brand new Suno-ist, just started last night. I saw a meme about telling it when to end the song. I had that issue where half I made concluded but the other half just kept going even when I tried extending it. Is there any best way to make it happen?
Also, is it possible to have alternating male and female singers in verses or no? I tried that and it just seemed to default to whichever one I said first.

[broken lotus]

lots of tips on the community wiki
https://www.suno.wiki/faq/style-and-lyrics/duets/

[thin girder]

Is there a way to organize/hide tracks other than "Like/Dislike"?

[normal patio]

You mean other than playlists?

[worthy ridge]

Getting a lot repeating verses in 3.5 even though they're labeled [verse 1] [verse 2], etc. Getting duplicate lyrics in [verse 1] or [verse 2] mostly. This was fixed in v3 I thought.

Keep to Delete ratio in v3.5 is about 1 to 30 at the moment. Burning trough a ton of credits because of the verses issues.

[charred merlin]

If I create a song with Pro/Premium, does anyone know the legalities behind "gifting ownership of a song" to someone? I'm about to take a dive into that, but I wanted to see if anyone had already taken a look

[charred merlin]

I've started using playlists for organization instead of sharing or playback purposes.

[charred merlin]

https://discord.com/channels/1069381916492562582/1082775596767117312 ?

[worthy ridge]

You can do whatever you want with your copyright. You can gift/sell/whatever. As long as you sign off on it (for their sake incase you want it back). I'm a producer with BMI. You could also get them to register it as well

[charred merlin]

Okay, I'd read here https://sunoaiwiki.com/en/tips/2024-05-05-how-to-resolve-copyright-claims-on-suno-ai-generated-music/https://sunoaiwiki.com/en/tips/2024-05-05-how-to-resolve-copyright-claims-on-suno-ai-generated-music/ someone had suggested the means for handling copyright disputes for whatever reason would involve showing an active suno membership (or having had an active membership when the song was created), etc. I wasn't sure how that would work in the instance that I gave ownership, or how my proof of original ownership would be proven or "blockchained" along with that.

I'm sure I'm overthinking the process, just new to the legal ownership side of things.

[worthy ridge]

Typically, regardless of what venue, and or software, and or method, and or instruments, etc. The minute you "Write" a song, you are the Songwriter. It's your copyright. The Time/Date signature on the .WAV file master is your proof of ownership. To make that Copyright "Ironclad" the song should be Registered with BMI "Broadcast Music Industry" or 'other' production corporation by an Industry Producer. That will ensure NO ONE, not even a legitimate Corporation or Songwriter can ever make a claim without you releasing the songwrites. BMI keeps track of all that.

This is why everyone in the industry freaks out about "Masters"...

[charred merlin]

True-- I grew up with stories of people mailing themselves cassettes to prove they had the idea first should they ever need to show the stamped & dated envelope and open it in court 🤣 paranoia grows out from a healthy sense of caution

[worthy ridge]

lol. OMG.. I actually used to do that. Yup, exactly

Especially today with kids liek Jojo Siwa, snipping good songs

[charred merlin]

I suppose everyone is just playing along with the honour system until it gets challenged (legitimately or otherwise), that's when proving ownership comes into play.

I doubt that anything I end up gifting will put someone in that sort of situation, but I at least wanted to think of the contingencies before throwing songs around to people & getting caught with my pants down because I misunderstood some legal disclaimer.

[worthy ridge]

There are mega stars in the UK who are fighting Suno already. They want to own music ai in the music Corps and not allow the general masses have access. Cray part is.... This aint new. I've been using AI for my music for 40 years... So.... There is that.

[sharp kraken]

can anyone tell me more about Unicode? https://suno.com/song/54a5232f-2b49-4bc3-8ff5-0539caed0b5b

[charred merlin]

"Jagger & Gallagher's Secret Plot to Destroy AI Music?! Suno's Plan to Fight Back will Shock You. More on pg 13"

[worthy ridge]

Thom York has also been doing a giant play on owning all the music rights for every song on the planet. AI is messing up that plan yo. 🙂

[charred merlin]

Unicode is a system that assigns a numeric value to every variation of every letter, symbol & emoji, ensuring that the same text is represented when you exchange data worldwide.

In terms of the usage in the song, it sounds like you've overloaded the system's interpretation or context & it's spitting out stuttered hallucinations.

[sharp kraken]

it changes the language too

i know what it is i was just wondering how it relates to suno more or lesss

[worthy ridge]

There probably aren't many neural nodes for it to interpret what your looking for, so it's trying to give you something. The way Suno works, if you keep working on it, it will 'eventually' learn what you are looking to get out of the prompts

[sharp kraken]

I mean, you can read the prompts and it stays pretty on course until the vocal breaks come in

its not like the whole song is some crazy hallucination - and all the tags seem to be working fine like, the vocals are pretty cool

[eager frost]

Would all agree the vocal quality is worse with v3.5

It definitely has more issues with hallucinations and not following custom lyrical prompts accurately.

When it works, I do find an improvement in song structure, with having a beginning and end to the song, instead of a abrupt starting or ending as happens regularly with v3.

[worthy ridge]

I can get the same flow from 2 B, D b, do, da, do, da. The intro flow is just converting the unicode to a text prompt. When it gets to your equation, its probably just doing a conversion to whatever alphacode it can interpret and spitting it out.

[sharp kraken]

I disagree - i havent gone back to 3 yet, for me and my music it is going great

[sharp kraken]

but this looks cooler "♬-♫♬--♩¸¸♪♫♬♩¸¸♪-♫♬♩--¸¸♪♫--♬-♩¸¸♪

[worthy ridge]

Song structuring in 3.5 is a better, but there are some interpretation problems with the lyrics still. Would like to be able to keep some of the instrumentals and ditch the lyrics on the messed up verses. Maybe we'll get stem splits in 4.

[edgy token]

why do some people put things like dots or morse in the start of their songs?

[charred merlin]

Because they saw someone else do it-- it's like tiktok videos, or decorations for social media names

[sharp kraken]

um i do it to control the timing

of instruments and effects

[severe jewel]

Fr

[eager frost]

Stem splits would be great. An in app voice model modifier would be awesome, that can select part of the vocals that sound good and improve the entire song. Or the modifier could clean up noise.

[sharp kraken]

suno knows what this means :𝅘𝅥𝅯

and it knows what this means: 𝅘𝅥𝅮 and if you use instead of the other it will adjust accordingly

but maybe if you don't know what that means it just looks like gibberih to you

[thin girder]

with the 3.5 release, is there a better understanding of tags and keywords?

if so, is the knowledge base better updated with that info?

[broken lotus]

It’s more like 3.5 is a return to metatags, that use to be more responsive before v3.
Metatags that were less reliable before, are much more reliable now.

The meaning of the words haven’t changed, but they may have stronger or lesser effect, they may combine with other words differently and because of 3.5’s aesthetic voice training 3.5 may ‘lean’ towards lyrical genres over weird mash-up genres v3 could do.

It’s a little early to have a completely rewritten wiki.

[worthy ridge]

tag placements most def respond better. If you put [verse] or [chorus] tags into the structure without words, it still produces a full length song in the structure you want. It used to just end the song at 45-1 minute, no matter how many blank tags you put in.

[thin girder]

if you make an extension to a song in a playlist, does it automatically add it to that playlist?

[thin girder]

it there an explicit tag list as well?

[broken lotus]

extensions don't get added to playlists automatically, we have to add them.

here is the 'primer' for metatags, and you can find a couple of style word lists on the site too
https://www.suno.wiki/faq/metatags/

[thin girder]

much thanks

[cerulean steeple]

My songs keep disappearing after a while. Is it just a setting to keep servers not full? I'm just feeling a bit afraid that Suno moderators suddenly finds something wrong in songs i generate

[thorn nimbus]

Sorry if this. question is repeated, I want to read about the implementation (high level overview) on how suno works, also can i use the songs, maybe in tiktok, shorts etc without worrying about copyrights ?

also if there is an onboarding and cool tricks guide 👀

[cerulean steeple]

If you're not a paid member, you should credit to Suno, as far as i know

[thin girder]

are they diapering from your library? I find mine leave the "create" area often which is annoying, but still exist in library

[cerulean steeple]

Yes, they dissappear from Library, although songs that i set public are still in my profile

right now, the difference between my "create" and "library" areas is only one song. Any other songs just gone.

[paper root]

Hello. Using the new 3.5 version, everytime I create a song, despite using the [outro] input, the bridge and the final chorus repeat twice.

[subtle otter]

hi!!!! first of all, 3.5 sounds SOOOOOOOOOOO GOOOD, so much better than before, i love you all @Gim @Helena

prompt question:

the only thing is, with v3 i would always get a really nice high pitched singer when i made emo or pop punk songs on there, and with v3.5 it keeps using a low pitched singer which doesn't match the genre. any ideas how i can get that old high pitched singer back? i tried using "high pitched singer" and "tenor singer" in the musical style but it doesnt do any thing :(

[fluid robin]

Swing.50s. Strong vocalist., heartfelt.Fast paced. Drum Snares. Guitar Riff.Trumpets.Vocal male in style of music

for a classic swing song

[umbral grotto]

chirp

[fluid robin]

no chirp bot sadly

[fluid robin]

https://suno.com/

[umbral grotto]

thank you..

[covert anvil]

Hello, in Sumo v3 i was able to use style in lyrics to define what I wanted in my songs - for example I used [saxophone solo] to get honestly great sounding parts, but now they are completely ignored in 3.5. That's a bummer since the new model sounds so much better. Am I doing something wrong? Did the keywords change? or am I just stupid (which is something fair, ngl)

[broken lotus]

please look to top right of screen, you will find the pagination links to view further back in your library

[prisma wraith]

I can't even begin explaining my jealousy at the amount of credits you have

[broken lotus]

sorry, I meant to edit that out... they are suno's credits to test upcoming features...

[cerulean steeple]

thank you. but are they supposed to dissappear from "create" page in a while as well?

[broken lotus]

I think each page holds a maximum of 20 clips...

the create page will hold more, until it is refreshed.

[night bluff]

Hi, when I try to create only instrumental music it keeps putting vocals. This is the 3,5 version. Is there any way to eliminate this?

[broken lotus]

3.5 seems to be trained to sing... like it REALLY wants to sing....
for instrumentals, I am using v3... especially to make an instrumental opening before switching to 3.5 for better voice.

[upper dock]

is sono same as bark voice?

[night bluff]

Thanks for the response. I kind of figured it would be.

[broken lotus]

Bark and Chirp were products'made by Suno the company.
Bark is a text-to-speech model, and Chirp was the original name for the music bot here on Discord.

ever since they started the website, they've rebranded the service as 'Suno'.

Bark can still be downloaded on Suno's github, and used free on HuggingFace.

the current Suno music bot evolved from Bark, I guess...

[upper dock]

is there no Chirp now?

is everything bark now

where i can find the music bot?

[broken lotus]

https://www.suno.com/create

[upper dock]

I thought it was a bot in voice chat

is it not used in this discord server

[broken lotus]

there are no more discord bots.

[cerulean steeple]

Btw, is there any info about whether devs have plans to release a mobile application? It would be much more comfortable than the website

[broken lotus]

they had hiring notices for ios and android developers...

I haven't heard an eta, but I think they are starting alpha testing soon. (Suno doesn't always keep us in the loop until they are ready to start testing something)

[cerulean steeple]

Okay, i really hope it will be released sooner, thanks for response

[upper dock]

since it is called suno is this just the music bot chirp or also bark too there?

confused

[broken lotus]

Bark is not on the website.
it is open source and free to download on github.

the current music bot online is capable of spoken speech... although it will try to start singing often.

https://suno.com/song/fc49e3a9-b671-461e-95f0-dd6b3ebbda53

[upper dock]

thank you so much. Is this music bot not opensource?

bark could produce music too i was wondering what the difference between the two would be for music generation

[broken lotus]

the music bot is a subscription..., like Midjourney

it has evolved many generations from Bark...

[upper dock]

Is Bark abandoned in terms of upgrade in favor of this music bot which is the upgraded version of Bark for music?

[coral plover]

i tried bark few days ago in the colab demo very "interesting" results

does anyone know is there a suno or chirp paper or something?

https://colab.research.google.com/drive/1eJfA2XUa-mXwdMy7DoYKVYHI1iTd9Vkt?usp=sharing

can check out bark here btw

[broken lotus]

they wanted to write one but we all yelled at them to get back in the laboratory and finish making us v4.

j/k...
I think there is a paper under 'Chirp by Suno'...
maybe more like an interview, but tech oriented.
someone showed me, but I have lost track of it.

[broken lotus]

it looks like Bark is abandoned... aka 'open sourced'

[serene viper]

Is anybody experiencing this error when trying to extend?

WOndering if it just me

[weary wharf]

No, I have no problems currently. Just extended a song. Try a refresh.

[serene viper]

Ok thanks

That;s odd still same

THat means I need a break

I try again later on today

[broken lotus]

are you trying to extend in a different model (not v3.5)...?

[sharp shard]

Okay...

How does the "Full Song" Button work?

Cause, I had a five minute song, I added a third part which was 35 seconds...

And the full song from that is 5 minutes and ten seconds.

...

Where did my 25 seconds go!?

[graceful condor]

maybe you extended from a timestamp 25 seconds too early? or does the song cut off abruptly at the end?

[sharp shard]

So, this is from pressing "Full Song" from a Part 3 that was from a 5 minute song. Part 3 said it was 35 seconds long. I got a song that was 5 minutes and 10 seconds.

I was wondering if it can shave parts to make it connect seemelessly?

[graceful condor]

no, it doesn't do that, the "full song" joins the two segments but doesn't cut anything off

[sharp shard]

huh...

[graceful condor]

if there's a part missing, chances are the 3rd part was extended from a time 25 seconds too soon

it can happen, it's a bit finicky

[sharp shard]

Well, it still works anyhow.

It just confused me is all.

[graceful condor]

it's good practice to double check (or even triple check lol) before extending a song to make sure the timestamp is correct 💜

[south kindle]

I would recommend always getting the "ful song" after an extension and making sure there arent any issues before extending further. I've made what I thought was a masterpiece only for it to end up in the garbage after realizing the very first extension didnt work like I thought it did.

[neat field]

can one extend v3 after first starting v3.5?

[south kindle]

yes

[charred merlin]

You can use any of the models for extensions-- while 3.5 has better song structure, lyrical adhesion & flow, etc, I sometimes get a closer match to my envisioned sound through v3 (probably why they have it described as "broad & versatile"). So often I'll use v3 to get the ball rolling on a song, then extend quite early into the song to force 3.5 to use that specific style and apply a more cohesive structure to it.

By the way, I hope you enjoyed those tracks the other day, there were a few great results in it!

[upper dock]

will it ever get new updates? the last update in the readme is from 2023.05.01

[broken lotus]

It is open source. If anyone is going to update it, it is now up to the Bark community.

[wheat rose]

any tips for getting a more natural/less "autotuned" sounding vocal?

[weary wharf]

"clear vocals" mostly work for style or as a tag.

... just in 3.5, not in 3

[wheat rose]

i'll give that a try

[vestal rapids]

You can only use songs commercially if you subscribe. Songs made when you have no subscription cannot be monetised.

[charred merlin]

If the song was created for free, you need to attribute Creation by Suno, and subscribing can't retroactively change the licensing for the song. You can subscribe and try to recreate it if you want exclusive rights over it

Can someone share insight on how copyright works if someone extends a song that you make public on Suno?

I understand the company is putting $1,000,000 in prize money for becoming 'popular' in terms of using the site people use Youtube or something, but the idea of becoming a social media tryhard for the sake of making prize money is even further soured for me by the thought that someone could make an extension from one of my songs and then claim it as their own, especially if they're able to bypass any form of creator attribution because I chose to share the song on Suno.

[untold briar]

I can't comment on copyright generally, but specifically for qualifying for the Suno contest, I asked Suno and the thinking is all clips in a song need to be created in June and all clip need to be from a single user.

[charred merlin]

Okay, good to have clarification for the contest, but I'm more concerned about my content moving forward & outside of the Suno platform. I appreciate you having reached out regarding that, but is there an appropriate contact email I should use if I wanted further clarification regarding the copyright status of extended songs? (even if it's as simple as" 50%+ of a songs content counts as ownership" / "whoever created the first clip has ultimate copyright" / "uhhh... no one has asked us this yet")

If there isn't a lot of consideration for that, I'm not going to jump through trending hoops for a chance at competing with community "celebrities" for a chance at making money if it means I'm handing my copyrights out like candy (in a very liberally exaggerative sense to further emphasize a point)

[south kindle]

Pretty sure Suno has the underlying info the be able to connect the dots to know if someone extended your song and where it originated from.

[broken lotus]

I am using DistroLock before I make public and post here.

Another Sunoid can build on my song, but my song's 'thumbprint' is already distributed to most (all legit?) content streamer/hosts.

I pay a fee to DistroKid, not to distribute to those stores, but to lock my song out of those stores (and yt).

People are going to download as 'collectors' – can't do much about that.... If a Sunoid wants to be creative, I am ok with that honestly – I will judge them on effort and creativity, but that is just my ego...

but the scammers who are just lazy crooks, are not going to bother to add on to the song to make a legal loophole. They are just going to steal and upload the song as is -- this is what DistroLock blocks.

[south kindle]

I mean, it lights up the sub parts when you play a song, so it knows.

If you are that worried, then don't post it publicly. But I like to think of posting your songs publicly is a way to help others learn and be inspired to make their own creations.

[charred merlin]

I'm not so much concerned about what happens on the Suno platform, but if I make a song, someone takes & extends it, I'm wondering where the copyright falls.

Do they need to generate the majority of a song in order to dominate copyright? Does Suno even take a stance on the matter, or have they accidentally promised the song to both parties through lack of definite wording?

There's a middle ground between releasing all songs and not releasing any, I'm just trying to figure out with a relative degree of certainty where we stand if not siding with either extreme.

[south kindle]

i think it is probably safe to say that we are cutting a somewhat new path through the legal jungle, in a way.

[charred merlin]

(for the record, by no means am I trying to compare my work to The Beatles, I just find that to be a common example everyone can understand)

I might have to look into copyright law in general to get a better idea how this is handled (I recognize y'all are just trying to help, sorry if I made you feel like you needed to give an official statement on Suno's behalf, I'm just enthusiastic 🙂 ), but do you have a feeling it would be handled the same way if someone were to try to extend or modify "Let it Be"? After having thumbprinted the song, unless significant changes are made to the underlying work, it's not legally distinct enough to count as an original work?

This is where my head starts to swim regarding the venn diagram overlaps of "ownership" and "copyright", so I'll just limit my question to the above 😛

In that scenario, even if I were to take the 3m50s of Beatles, generate 3m51s of additional extended content afterward, or if I were to cut the intro off and finish the rest of the song myself, there might be an argument made for my "ownership" of that new version, but it wouldn't be distinct enough to file a copyright?

[rough pelican]

This just helped me out, thanks. i had to use search to find this. Maybe this should be pinned? i don't know

[bronze helm]

Has anyone else noticed a significant decline in prompt adherence for v3.5 ever since the public release? While some claim the model is not changed when released publicly, I find it strange that so many of my pre-release tests were of incredible quality, and now none of the generations are even passable. Female voice gives me male-sounding vocals and sometimes there's this weird "Is this voice trying to be female or male?" sound to vocals. Like they sound female and they slowly transition to slightly more masculine vocals in the traditional sense, and so on.

Reposting from #🪲┃bugs-web

[alpine frigate]

I am receiving this error:Application error: a client-side exception has occurred (see the browser console for more information), when trying to log into Suno. I have cleared Browser History and Cache as suggested on general channel but am still erroring out. Any ideas why this is the case? NOTE: I ended up clear cache for a few weeks and it worked...back on Suno

[cursive crater]

50 credits every 24 hours?

[charred merlin]

For free accounts, yeah

[broken lotus]

So... Suno is a company. They can't give you a copyright, that's a government thing. They have to verify your lyrics, etc

A lot of projects are unpublishable because the copyrights are scattered among too many people. Recordings of live shows, early broadcasts, were done before licensing became the standard – or demos that involved non-union musicians. It's actually very common for projects to be 'poisoned' by copyright – that's how it's meant to work.

However, we live in the future, and where as music recording to distribution use to take 8 or 9 months..., it now takes minutes. Someone can publish my song to yt in 30 seconds, literally before I finish filling in the distributor forms.

The idea that copyright will save us, or impact us in the least... By the time we got a courtdate 2-6yrs from now... that music is irrelevant. Probably doesn't even have sentimental value at that point.

Really, splitting hairs over copyright is not going to prevent someone stealing your music in the real world. If you are curious, the US CopyRight Office has a website, where these things are answered in plain English, without sensationalism or AI paranoia.

None of this is new, it's just so much faster that if you are not planning a very large commercial project with a corporate lawyer budget, it's silly to keep you up at night. Go to the actual govt website (whichever country) they are public resources, this doesn't need to be a mystery or a drama.

No legitimate publisher would take a poisoned project, you're talking about a bootleg. Again, we're not talking about a stolen bicycle and you call the cops and get it back. This is digital data, once it's out in the wild, it's not under control anymore.

[charred merlin]

Didn't think I was applying sensationalism or 'AI paranoia', just genuine curiosity. I didn't mean to have struck a chord or gave you the impression I was staying up at night worrying about this, but for all of the reasons you've provided that differentiate songwriting through traditional means vs using the Extend feature in Suno, it's not unreasonable to be curious about the legal differences.

I'm certainly going to follow up looking into copyright law, as I had suggested I would in my second message, but we'll have to agree to disagree that North American copyright law is written "in plain English".

Thanks for your efforts to help, even if you came across pretty condescending. :/

[grave cradle]

Good evening, I have a question, I'm about to subscribe and I don't really understand. For a month is given 2500 credits, and do I understand correctly that those 50 credits that are accrued every day for free will not be accrued if I subscribe. And that is, if, for example, I spend all 2500 credits in 10 days, then until the end of the month i will sit without credits?

[charred merlin]

They can't give you a copyright
But they promise general commercial terms, and I was asking about those.

Do you have the email I was asking about if I wanted to contact Suno about this, by the way?

[broken lotus]

[e-mail]

[charred merlin]

Thanks 🙂

[broken lotus]

You get the whole month of credits at once. It's up to you to manage them. Use them all at once, or spread them out.

There are top ups if you want to buy more, which will rollover, and only get used after all the monthly are gone.
(I have bought them. they feel expensive, but one topup padded me for a several months)

If you make a lot of music (like every day) go for the Premium subscription, it is worth it. If you just have a few afternoons a week, Pro is maybe better, just space them out.

But it's not a race, you don't win free credits on top of the subscriber credits just because you used them all up quickly..., lol.

[grave cradle]

thanks a lot! just wanted to be sure about this one! have a great day

[elder oyster]

Anyone having the generate lyrics button not showing up on the custom create a song?

[elder oyster]

It's usually between the lyrics and song style. But today it's been disappearing most the day.

[soft gyro]

I was wondering how many different "human" voices Suno uses or generates....

[broken lotus]

if you can scroll the panel...?

Also thos text boxes can be squeezed shorter, maybe enough to reveal the button?

there is a little angular corner bottom right on the lyric and style boxes. Click and drag those up...

[serene viper]

Sorry just saw this, I guess is that instrumental the problem maker.

[scenic swift]

Is suno working on quality? I have noticed they are coming out with new features but would love to get suno to the point where the tracks sound like a fully mixed and mastered release

[untold briar]

Quality improved quite a bit in the last week (between the release of 3.5, and today)

[scenic swift]

Oh okay I generated stuff last night but it's not perfect like this

[charred palm]

Can I convert myself to Premium from Pro once it runs out? I feel like whatever credit I have won't even last me for one more week...

[untold briar]

If you are on Premium, you can select Pro for the next month. Or even select Cancel for the next month, and then when you need credits, upgrade to Pro at that time. This may be the best option if you sometimes run out of days in the month to use all your credits, because you will always restart your cycle on a day where you are prompting.

[charred palm]

Cheers to that @untold briar ! 😸

[old quarry]

Hello. Will it be possible in the future to create a song with lyrics / voices / vocals and then have the option to download the song with and without vocals?

[charred palm]

Try stem splitting using other software?

[old quarry]

I did but it did not work very well. I would be nice if suno.ai would offer a solution. Not sure if the is possible.

[charred palm]

It would be awesome of they do TBH.

[old quarry]

should I put this in the suggestions channel?

[sharp kraken]

one of the things I love about suno is you can do some really strange musical things so easily - something i always dreamed of was 7/4 drum and bass - now here it is!
https://suno.com/song/d7c0f708-04cd-4490-9a73-1ab0db43f7ed

[grizzled slate]

I have only just noticed the Character Count on both the Lyrics and the Style of Music. Great, little addition!

Also, there are 3000 characters for the lyrics now, too!

Many thanks! Keep up the great work.👏

I can't wait to see what you come up with for your first Birthday! 🎉

[cerulean steeple]

Just want to thank everybody who is related to Suno developing and testing! v3.5 songs sounds so natural (even breath-in sound can be heard) and voice flows through lyrics so well and almost doesn't mess with different lyrics parts. Can't even imagine how cool v4 is expected to be

[cerulean steeple]

i think yes. Although it is gonna be too hard to realise, because Suno generates the whole song. If they decide to implement your suggestion, then they most likely will have to train suno to generate music first, and lyrics then, right "from scratch". Just my opinion tho, not sure how it actually works, but it sounds logically for me

[elder oyster]

Using mobile, the resize option isn't there. But in the mobile browser settings, if I choose "Desktop Site", the button reappears.

[spring sail]

when i "edit details", it is only updated online. it's not reflected in the video download. how can i update the download?

[lament cove]

If you use the combination of [Bridge] and [Final Chorus] it seems (sometimes) to give a nice octaves/chords change for the ending of the song.

[Bridge]
...last verse lyrics...

[Final Chorus]
...last chorus lyrics...```
Tested this for an epic metal song.

[charred palm]

We talked about this yesterday. This is a bug. Reported but feel free to report again.

[halcyon folio]

Are there any plans for an official API? I'd love to integrate Suno into a new /music function of my large AI Discord bot. https://geneplore.com/bot

[weary wharf]

If you like Woodkid give this a try. Depends on the lyrics as well, but I see that the outcome is pretty good most of the times:
symphonic-acoustic, orchestral, symphonic, symphonic, orchestral, intense, desperate, betrayal, male singer, clear vocal

[graceful condor]

can someone help me out make a full ai song, i got the lyrics and verses i js need like help with suno and ideas

[trail badger]

I want to tell the difference between v3 and v3.5.
In v3, the music is more natural and it handles cultural music better. It reflects emotions better as music. We get better results. In v3.5, the music is always processed in the same rhythm. Vocal sense is better, quality is better. v3 is obviously better now. Vocal improvements need to be made while keeping the music quality and naturalness of v3 the same.

[hollow hare]

Sure, here is the rundown: head over to https://suno.com and sign up/in if you did not do so.

Go to Create > enable custom mode > enter your lyrics, style keywords, and title > create.

Some helpful starting links: https://suno.wiki
https://youtu.be/lj8utBd_IUM

The wiki page is very useful. Try reading all the entries you can at first to get an idea.

You usually can ask an LLM for the genre/style of a song if you need a starting point (though remember they can lie convincingly). There are other sites as well to identify song genres like https://sonoteller.ai and Chosic's music genre finder (Spotify only, <
https://www.chosic.com/music-genre-finder/>)

[cerulean steeple]

Btw, I've seen the telegram bot that says that it has Suno in it. Didn't tested tho, since it has no trial attempts, but if it Suno, i can assume that the API is probably available for organizations

[hollow hare]

There is no official API currently available but it is on the devs plan for something since several companies did reach out to them about it.

The methods that most of these bots and unofficial apps on the app/playstore used definitely violated Suno ToS in how they went about it

[cerulean steeple]

If there is any way to report such fraudulent action, i will definitely report

[knotty socket]

I`m wondering about, will the tinny sound and artificial robot voice get uptdated as the v4 come out _

[halcyon folio]

I could use an unofficial one, but I would like to support the developers of Suno and make my features stable.

[hollow hare]

Unfortunately not really AFAIK. Suno are aware of these apps but not much they can do (or are doing) about it right now. It is relatively simple to find out what accounts they are using to create music with (if they are one of the "good" ones) - assuming they are not simple redirects apps.

The app/playstore will likely do nothing about it unless Suno contacts them with a lawyer. Which may happen when the official app comes out and only for those that impersonates Suno

[knotty socket]

i know its possible cause udio is so clean at vocals, but i think suno is much better at melodies and cross genres ]

[warped furnace]

Does anyone know if it is possible to add sound effects, e. g. gun shots? How would I prompt that? Nobody is being killed but love is compared to a bullet to the heart, so a gunshot sound in the background would be nice. 😁

[naive quail]

https://suno.com/song/2ad4ff91-a1ae-4bda-a05c-3e103d37488b

It didn't want to end. lol

[trail badger]

Hahaha 😄 very nice

[warped furnace]

Ah! Thanks for the prompt. I sound like Vincent Price. 🤣

[cerulean steeple]

What the... It sounds too good lol

[weary wharf]

Hahaha. So good 😂

[misty leaf]

So [Final Chorus] then is the last part and then it'll end the song, correct?

[lament cove]

enigmatic <whatever style> in the style prompt trigger interesting rhythms, like syncopated rhythms

random changes of drum kicks

[south kindle]

Have you tried Ultimate Vocal Remover 5? I've had some decent success using this model and cranking up the segment and overlap sizes to what your computer can handle. If I crank it up it can take like 5-10 minutes to split, but the quality can be much nicer. Each track is different, however, and some other models might work better for a particular song.

free and open source. Worth at least a look.

[meager inlet]

Is it meant to be that when searching in our library that we can only see one page of results and not all the songs that have the words or styles that are in the search bar?

[carmine torrent]

(how to turn off percussion) THIS IS NOT A BUG, I'm just wanting to know how to create song where there is no drums or any percussion, no matter what I do with or without prompts Suno always includes drums in the song, How do i get Suno to NOT include percussion?

[south kindle]

go to the next page...?

[broken lotus]

you should see pagination links top right of the screen, near the trashcan

[hollow hare]

(Blue circle)

[south kindle]

Worst case scenario you could use UVR like I mentioned to split out the drums to get a non-drums version and see if that works for you.

[meager inlet]

😑 I know how to search through pages online..... See how when I search it is grayed out?

[carmine torrent]

'ultimate vocal remover'? I don't know how to do that and I don't know how to see UVR comment you mention.

[meager inlet]

why does my setup look different??

[south kindle]

Just google it so I don’t have to give you any links via discord. It’s good stuff tho.

[unique stone]

Ohh UVR can split drums out? I'm a complete novice with it and only tried it on vocals

[south kindle]

you need to download different various model, as I have learned, before you can do that, but yeah. it is kind of insane how powerful it is being free and open source.

[jagged mango]

Have you tried Fadr?

[carmine torrent]

Not yet, but I will, thanks

[unique stone]

https://tenor.com/view/shocked-owl-gif-10746536

I had no idea it could do that and assumed just from its name that it was for vocal/instrumental splits. Gonna try this later. Thanks!

[south kindle]

I havent tried Fadr yet either, but have tried RipX DAW, and noticed that since not all suno songs will split stems in quite the same way due to how they are generated I do like having the option to choose a different model and try re-splitting them because I seem to get varying levels of results using the different models.

I am also not convinced that once you split them up, even if you track them back together it is 100% not the same as the original mix. stuff will get lost, so I've been just layering them back over the original mix to make tweaks.

I'm a professional. I've been doing this all month. 🤣

[carmine torrent]

Wow! that worked for me. Impressive! But I'd still like to do that in Suno.

[jagged mango]

It would be AMAZING if Suno could generate vocals seperately from the tracks, and in time stems for all tracks!

[unique stone]

I expected that to be the case. As I understand it, it works like a very complex audio filter, so to simplify: If you used audio filters to split stuff into different frequency bands, you'll always have some degree of overlap, or some dropout, it'd never be 100% the same.

Generally what I've tried to do so far is use an online stem splitter to get vocal/bass/percussion/instrumental parts and then I start layering soft synths or other samples on top of the different parts, with the aim of either reducing or eliminating the original audio. So for drums I just try to reinforce what's already there, bass tends to be replaced entirely, vocals I intend to transplant with other AI vocals and the instrumental parts I mostly replace but I'll keep a little of the original in the mix because there's often just a nice ambience to it.

A lot of this is theory for me though at the moment, I haven't actually completed any songs doing all of this stuff, just having fun learning and trying things

[meager inlet]

Soooo no answers now to the fact that my page turning arrows are grayed out when I'm searching in library?

[south kindle]

well, if you split a stem and then invert the wave and layer it back into the mix you can essentially "delete" parts of it, at will, as the inverted wave will essentially cancel out in the main mix.

I've used it on occasion to remove the odd errant sound that seemed to randomly appear in a part of a song.

[unique stone]

I wasn't sure how well that'd work / how much it would manipulate the sound but that's interesting to know. That's a classic old technique for doing vocal removals where the voice is mono (flip one channel, mix it over the other) which had the unfortunate side-effect of making the resulting audio mono.

[south kindle]

it was tricky, but I think it is best to make sure to isolate it as much as possible before inverting it. Sometimes what might sounds like a 'noise' to remove when isolated could really be the complementary part to something else that just got caught up in the stem cut.

[neat field]

new failed task: whisper spoken word dialogue/duet (any gender, just 2 speakers)

even if the song starts with a soft spoken word, 2 speakers will not whisper together, ever - i don't think this case exists in the data

[south kindle]

From what I understand duets are near impossible at the moment and require some crazy luck to pull off.

[neat field]

v3.5 has higher success rate than v3, and yet yes very rare

[jagged mango]

With Suno? I do have one, but the dude only sings second verse, and half of second prechorus

A duet, that is

[south kindle]

I think they mean where they sing together, at the same time. right?

[jagged mango]

Way I quick fixed that was extracting the vocal with Fadr, then uploading the vocal to another online AI software that changed voice into a male voice. Note that you should have a part with very dry vocals for a good result. The part you need must be dry

[fierce glacier]

something has gone wrong 😦 cant veiw my profile

[final hazel]

Click on edit profile and check your handle name

[lament cove]

[Voice End] before the [End] can give a nice blend of the last chorus voices with a full all-in instruments of the song

...Chorus lyrics...

[End]```

[fierce glacier]

its the same as its always been Suno.com/@jamesmartin

[final hazel]

Can you provide a screenshot?

[final hazel]

Thank you. I will report it to a dev for you.

[atomic nexus]

Does Suno have to 'understand' the lyrics when generating the song? I'm trying to generate a song with lyrics in Bulgarian. In v3 - it didn't work well with lyrics in cyrillic letters - sometimes it would just generate its own lyrics totally ignoring the provided ones - so I just typed how the words should be pronounced in English - with accents etc. I'm trying to generate Salsa and Bachata songs. But with v3.5 I have zero success. It seems to read cyrillic words now. But it usually just generates some pop song - ignoring the style prompt. I even have some results that just sound like an audio book. (With v3 this wasn't always consistent too - sometimes it would change the rhythm from bachata to kizomba in the middle of the song or something like this). When dancing to these styles - you usually 'count the beats/steps' (or at least I do - I'm still new to this). Is there a way to improve my results - to get a song that you can actually dance to? Is the language the problem (for example if it should be able to subdivide the words to syllables or something like this when generating the vocals).

[south kindle]

nope

See here: https://suno.com/song/82c38649-353f-40cf-9c48-1108042e5128

there is a 'drum solo' section for the vocals.

Just gotta mess around and find out.

I've had a song pronounce a word two different ways in a song before. it was weird.

[atomic nexus]

I've seen this too - with words containing 'ch' - sometimes it reads it as 'ch', sometimes as 'sh' and sometimes as 'kh'.

[runic dock]

It seems really dependent on the music style.

[elder oyster]

Was there any reply on the vanishing "Generate Lyrics " button on mobile after last update? It only reappears now if it change my mobile browser to view as "Desktop site".

[elder oyster]

If I change it* I meant

[coral marlin]

Hello, does anyone know, or has any experience, putting a SUNO song into another software to cover the lyrics with a specific AI voice? Im trying to get my songs (or most of them) to appear to be sung by a single person.

[lament cove]

What are your tricks to avoid this: sometimes happens that in a 4 verses long song, it repeats verse 3 a lot more than using verse 4

[cerulean steeple]

Is there any limitation for how many songs can we have in our accounts? I mean, the servers' capacity is not limitless, so i wonder whether i have to sometimes completely delete songs from my profile or not

[cerulean steeple]

4 verses? That's actually too much. You can try to extend from the timecode where verse 4 is expected to start

[cerulean steeple]

#🪲┃bugs-web it seems to be a bug that some users reported already

[viral lotus]

Yes, I have no generate lyrics button either

I cannot generate a rock ballad. No matter what I enter, every ballad is a country song, except for a couple of generations with heavy metal ballad I just got a very hard heavy metal song that was not close to a ballad. But whenever I enter for example any variation of a Rock ballad it always comes out country. Inversion 3, I enjoyed power ballad as a genre but now it's always coming out country.

[lament cove]

No 4 verses are a perfect length in almost all genres, i get songs ending before the 4:00 minute; there's even space for a last chorus repetition.
The problem that happens sometimes is that it starts repeating verse 3 (after verse 3+chorus ended) and not the fourth verse (or a final bridge).

Not always, but sometimes

Especially in rock/metal songs

[cerulean steeple]

That's the reason because 4 verses is actually too much (i mean almost no a bit popular songs have that number of verses, usually 2 or mostly 3). And that's might also be the reason why Suno repeats the only verse instead of end a song.
Just try to extend with 4th verse lyrics

[old quarry]

Thank you for the hint. Never seen this program. Will try it out.

[lament cove]

I think you didn't understand. If i extend the song, i'll get a Verse 5

All my song have space for 4 verses and they end before the 4:00 minutes or in 3:58

The problem is the AI not grabbing the actual verse, but repeating one verse behind

Verse 2

Verse 3

Verse 3

4:00```

Or

Verse 1

Verse 2

Verse 2

Verse 3```

Sometimes it works flawlessly

Verse 2

Verse 3

Verse 4 (or Bridge)```

[cerulean steeple]

You can try to extend from the timecode where verse 4 is expected to start

[cerulean steeple]

Verse 2

Verse 3
<EXTEND FROM HERE>
Verse 3

4:00```

[lament cove]

Ah yes, if i'm ok with the song i do this usually

But it's "annoying" when it happens

[cerulean steeple]

Indeed. AIs are not perfect 😅 Hopefully in v4 it will also be resolved

[glacial rampart]

Yeah, it's gone for me too. Probably a bug I assume?

[warped night]

how do i tell suno that "amaro" is not an artist, it's an adjective

[cerulean steeple]

Use English instead i guess

[broken lotus]

maybe
-amaro

[warped night]

for now a synonym will do, im not sure about mixing languages in the prompt

[cerulean steeple]

Oh, i meant to make the whole prompt in English

[past cliff]

have you tried to change it to something else, and then back?

[warped night]

eh but i dont trust it for now telling it in english to make a song in another language

[fierce glacier]

No, just tried to view and it didn't work. But devs fixed it a few hrs ago

[modern belfry]

It did that to me as well when I wrote Zippo

[molten smelt]

can you prompt for a unik vocal? i mean, if i kinda want the vocal to be similar track after track?

[soft gyro]

is there any place where there is a user manual for newbies where all the tags, the genres and the behavior of the "music monster" called Suno is described?
these are the basics actually to be able to produce expectable results. The help page does not offer such a thing actually!

[south kindle]

Best thing I can suggest is to look up tunes you like and see what prompts they used.

the faqs are a good guide, but there is always strange nuance that people discover.

[slender trellis]

Anyone know how to force end the song properly? I can't get it to fade out or end on a proper chord or anything. I've tried [End] and [Fade Out End]. AI just seems to like singing more...

[eternal nymph]

Hello all! I love Suno! ... I have heard some really awesome songs by others. I'm new but have watched tutorials and understand the meta tags etc. and played around with my first two weeks of my subscription. My question is how much is skill and how much is luck? Will I get to where I can consistently make good songs or is each song going to be a try until it suddenly sounds good?

[vestal rapids]

Prompting is an art and leveraging tags is skill

[wheat quail]

is it "meant to be" than things between () in style (not lyrics but the style) they are used in the song??

like in the lyrics actually singed

(like i put (metal) it begin the song with "METAAAAAL")... you get the idea?)

[ocean salmon]

I'm a pro member and I've always written my own lyrics but if I generate them with suno, do I still own them? I'm sure this has been asked multiple times sorry

[low remnant]

Your lyrics are your own.

[faint spruce]

I'm experiencing this same issue. It's not following the lyrics and it's suddenly all over the place. Recently, a song sang the first two lines of verse 3 after the chorus and then sang the last two lines of verse 2 before the chorus.

Been happening a lot today.

[long remnant]

Not sure how to properly search for this- but where is Suno sourcing it's music? Like what's the database it's working from?

[fluid robin]

in style of music = [Dark], [Rhythmical beat], [Vocal old style], [whistle], [Humming], [Big band] [Laughter], Mystery

for the darker songs

[vestal rapids]

it's not working with a database of music, it has learned how to create original new music in the same way as once you learned to count to 10 or learned to draw a picture

[long remnant]

But it's working off of a knowledge base, right?
If so, if it's working off a base of out of copyright/creative commons ok, but is it using just access to the net to find music?

[vestal rapids]

no, it's a stand alone AI model, it doesn't have a knowledge base, it has learned how to create new music. It's a difficult concept to grasp without knowing some AI fundamentals

[rose bridge]

Hi, what is the command or structure to customize your song cover?

[past cliff]

click the dots. edit song. you can change the art there

[long remnant]

That's great!
I'm really having an issue with the AI/trying to subvert actual artists thing. Especially bc so many of these AI things are actually stealing from living artists.

[past cliff]

give specific examples if you claim nonesense

[vestal rapids]

who is your favourite music artist?

[long remnant]

LOL. I mean; that's a loaded question.
I'm a Bowie fan, an Eiffel 65 fan, Mozart, Eisbrecher, Mono Inc, Loreena McKennit, SYML, etc,

I'm fully aware that almost all musicians and artists draw on older artists, but I think there's a difference in a human being building off of what previous artists have done and a computer literally storing and correlating data and a large corporation using an algorithm to avoid paying actual artists for their work.

[long remnant]

Uhhh... ok. we can talk about the recent Marvel movie literally using AI to avoid having to pay a human being for their work?

[past cliff]

This is Suno.com buddy

not marvel or a movie

[long remnant]

You asked?

These are my concerns.

[vestal rapids]

ok, that's the thing, AI models don't store their training data. They have learned how to create from scratch. It's a hard concept to grasp but learning something is different to storing it.

[sharp kraken]

legally the ftt is considered transformative

just ftt it and you own it

[long remnant]

BING and similar are literally using copyrighted and original content as their basis for their generatied content.

[past cliff]

you write with literally no proof lol. maybe you are not lying. but prove it lol

prove it

thats all you do and i will believe you and back you up

right now you are just yappin

[long remnant]

I'm literally just asking how Suno gets its content because I'm A) interested in how the algorithm works and B) don't want to be taking away from artists who already have a difficult road.

[past cliff]

it makes it

[long remnant]

Dude, there's no reason to be an ass when I'm literally just asking a question?

[past cliff]

they mentioned it 3 times

[long remnant]

I'm looking for peoples' experience and official documentation.

[past cliff]

this is what you stated. on top of your original question

that's what i responded to

fair enough either way. not trying to start crap with you. you are making claims.

it's generated music though

it trains from probably many types of music and uses transformative LLM

[long remnant]

I'm literally not making claims, but go off?
I'm just trying to get info about Suno, how it generates content in relation to what it uses to train from.

If it's one of the rare AI sources that trains from it's own db, fantastic! I'm getting amazing, spine tingling results from my prompts.

But I don't want to be working off of content that is taking away from some poor independent artists who are already underpaid and unrecognised.

I'm sorry if that triggers something for you. : (

[past cliff]

i'm good to go. go re-read everything you wrote and you will be less confused lol. have a good one buddy. Suno.com is the website that has all your answers as well as emails to places that can answer the rest. anything that isn't answered is a good chance that it's proprietary and wouldn't be shared purposefully

[long remnant]

Caveat: I have already contacted Suno and haven't received an answer in over 48 hrs. So I was hoping that this server, given that all of you have been here and have been sharing your experiences, would have input or information that might answer the questions I had.

[vestal rapids]

Can you bear with me and follow a simple thought exercise with me? I'll explain how training and inference works

[long remnant]

Not here to start fights; not here to battle about AI, just curious. Have a LOT of people in the film and entertainment industry that I love and value and are seeing their jobs disappear because the Big Wigs would rather steal than pay a human.

[long remnant]

Certainly,

My concern is what Suno draws its information from. CC0 and stuff out of copyright OR Suno's own copyrighted material: Fantastic!

I think things like Suno are an amazing resource for people like me who have an "idea" but can't properly translate it into literal sheet music

[vestal rapids]

ok, assume you don't know how to draw a barn, castle or house. I will show you 20 pictures of each and ask you to reproduce similar pictures. Then after going through this for a while I take back my pictures leaving you without access to them. Then I ask you to draw me a house. As you have seen and reproduced 20 pictures of a house you can do this with ease, each time they are new and original pictures. This is what training and inference is.

[long remnant]

: nod : Understandable.

[vestal rapids]

ok, then it's the same with training AI. You actually teach the neural network how to do something. Because a neural network runs on a computer it can do things at scale

[long remnant]

But is there not a difference between a human looking at something like 20 different pictures of a house and learning the difference between angles, lighting, shapes, etc and a computer storing and using data from 100’s or 1000’s of artists, a traditionally underpaid job in the first place, who either are not aware of or are actively being exploited?

I just am interested in Suno sources bc I fully recognise my own limitations.

As I’ve been told: “you’re creative, but not artistic.”

I don’t wish to be taking bread from the mouths of actual artists.

[vestal rapids]

that's the thing though, a model does not store it's training data

[long remnant]

I think suno is fascinating and my results have been exactly what I would want

But me plugging a prompt in doesn’t make me a musician

What it can do is give me something to build off

At least that’s how I feel

: )

[unique stone]

When I first got software for making music, I would try to (very poorly) re-create existing songs. I got notes wrong a lot of the time, but eventually got better at it. Then I started doing things like re-making a drum pattern or chord sequence I'd heard. Essentially the end result of what I make is combination of different structures and sounds I've become familiar with over the years, bringing them together in a mixture of what I find sounds "nice".

AI is basically doing that process but much faster and 100% concentrating on learning. It doesn't have to eat or sleep or anything.

What I do now is ask Suno to make things, and then I try to re-create the music it generates and I can then customise it as I wish. I prefer to use it as a foundation to build on, but it's also nice to sometimes just ask it to make something and enjoy how it sounds.

There's never been any information publicly available on what music Suno is trained on, but I would expect that they would be careful, to avoid legal troubles. They deliberately stop you from naming existing musicians/bands to avoid people making imitations.

Ultimately it's producing songs that - in whole - wouldn't otherwise exist. So it's not stealing songs (and the earnings from those songs) because those songs didn't exist in the first place. But, the existence of music-generating AI does mean that you can create good music, quickly and cheaply... when it can make music that is CD-quality or above, well it could affect the music industry overall so then it impacts all musicians equally...

Anyway, Suno might create something with a similar sound to an existing piece of music, within each genre. But it will still be unique. There are already pieces of music that may partly sound like something else (without meaning to, i.e. not a cover.) After all, there are only so many notes and chord sequences, types of percussion and rhythm etc. 🙂

[past cliff]

ai will replace easy jobs. just always remember that

if it can replace jobs and noone minds whatsoever. not really much to say about that

[long remnant]

If you think that art and music is a “monkey job” that’s extraordinarily sad.

[past cliff]

noooo. only mediocre art and music is a monkey job

great art and music is not duplicatable

[vestal rapids]

I would be less worried about artists and more worried about manufacturers of musical instruments, especially synthesizers and drum machines. The advent of this technology is going to move a lot of bedroom producers from hardware.

[final hazel]

Monkey jobs sounds extremely unrespectful.. and let's stick to the topic of this channel please

[long remnant]

I’m a former “orch dorch” ie: played violin for 10+ years but was never particularly talented. I could play by ear, but was exceptionally pathetic when trying to sight read .

I think ai could be phenomenal for people who can “feel” music but can’t properly translate it to the staves.

[unique stone]

Live performances will always be a thing at least.
Strangely I'm actually going from 100% software synths to using AI to make music whilst also building hardware synths. AI takes away a lot of direct control and part of the reason I'm making hardware stuff is so I can physically interact with it because moving controls on a screen isn't as satisfying.
(but anyway, that's me going a bit offtopic now 😄 )

[long remnant]

I legitimately didn’t want to cause a kerfuffle; I just have been playing with the app and I’d like to know something about the technology before I continue further

I don’t want to be stealing tracks from anyone

I don’t want someone being stolen from who’s unaware they’re being sourced

[vestal rapids]

you won't be because Suno's AI model doesn't store tracks

[long remnant]

I value all my musician and artist friends too much to actively steal from them

Suno has come up with some great base things I can work from

I recognise my limitations.

[unique stone]

There's also the upcoming new feature in Suno where you can put sound or music into it and it'll make things from it. Seems like it might be possible to just hum or whistle a tune (one demo shows a piano piece being played and then Suno turns it into an accordion.) So you'll be able to give it some direction and it uses its "knowledge" to build on it.

[long remnant]

That’s one of the reasons Suno appealed to me; it didn’t seem like it was a literal “help we’re here to literally suck Spotify dry and make our own shit out of it while avoiding paying artists”

But for the reason I have been playing with Suno

It’s extremely important to me that it’s not just taking from artists who historically have a hard time even being recognised, much less paid.

As I said; I’m “creative, not artistic” . I know how incredibly banal and pathetic my own brain is.

I need the assistance of other people, or in this situation, a computer programme, to help me workshop things

[vestal rapids]

I used Suno to make a kids song about the Zoo, I typed the prompt, got the fully formed track first generation, downloaded it, mastered it then distributed it the the music platforms. The whole process took 15 minutes.

[long remnant]

That’s awesome!

[unique stone]

The thing is, generative AI means that anyone can create content cheaply and quickly. And people do. Some people will create a load of pictures or pieces of music with it, and then upload it somewhere and/or try to sell it. I think some distributors are even not allowing AI content, probably because there is so much of it.

But there is already so much art and music anyway. There's just now a lot more uninteresting content, too 😅 (as well as good stuff of course...)

The technology isn't going to go away though. Eventually most people will end up using it, to some extent - I think it's very reasonable to use it as an assistant. If you find it morally wrong to generate and then sell the output directly, you can't stop anyone from doing it but you don't have to do that as well 🙂

[long remnant]

What I don’t want to see is multi million or billion dollar companies using things like Suno as a way to generate subpar, inferior soundtracks as a way to avoid having to actually pay artists for doing actual work.

There’s only so many people Walmart will hire

[long remnant]

I agree 100%.

As I said:

I recognise my own limitations 100%. I know that I have a passion for things but also I have a limitation on what my autistic brain can do, so things like Suno help me to work through and towards things

[vestal rapids]

I have been releasing AI compositions since 2019 when Aiva was made public. I am a classically educated musician.

[long remnant]

Ai is not going anywhere

I think it’s an amazing tool

[long remnant]

What instruments do you play?

[unique stone]

I think I need to stop typing so much when all I meant to say was "this is a good way to use it" 😄

[vestal rapids]

piano, french horn, clarinet, saxophone

[long remnant]

That’s amazing! I was terrible at wind instruments

And saxophone is insane to learn

[vestal rapids]

I prefer brass to woodwind but I am pretty good at the wind instruments. My main instruments these days are electronic though, I have a room full of synthesizers 🙂

[long remnant]

They banned me from flute in middle school

[nimble dirge]

I LOVE SUNO! I want to make something viral, maybe I'll try it on a bilingual channel. Although my main language is Spanish but I have Rockabilly /EDM Tecnho Dubstep / Alternative Rap playlists :3
Is any advice possible? I'm going to spend the night editing videos for my channel to see if I can promote somewhere I want to make an income, any ideas? I'm open to advice :3

[long remnant]

As you can see Ive sent a friend request. If you have the spoons I’d love your input

[vestal rapids]

I just approved it 🙂

[long remnant]

My mother played piano, guitar, and trumpet, but unfortunately I didn’t inherit any of her skills lol

[vestal rapids]

In my opinion piano is a base instrument, if you are competent at piano you could theorectically master any instrument

[normal patio]

Is there a remix option to edit the lyrics of an already generated song? I saw one article mention that but I don't see that as an option anywhere and can't find any other reference to it

[vestal rapids]

no, such a thing does not exist

[normal patio]

They even included a screen shot that had "remix" as an option... so it's not even locked behind a paywall? Now I'm just confused

[warm parrot]

So I changed the Album art for a song, anyway to get the video to show the new album art?

[vestal rapids]

I've never seen that option

[vestal rapids]

videos are generated once on initial download, you can't change it

[warm parrot]

😢

[unique stone]

See my reply in the general channel where you also asked this
(best to just ask in one channel 😉 )

[normal patio]

I mean I asked it here first and someone else said they'd never seen it so...
I guess I won't thank you in two channels then

[warped furnace]

Somebody know how to prompt a faster beat? I'm into Merengue and Reggaeton. First one always comes out way too slow. Of all my tries I had only two which were the right tempo but I could never duplicate them. 🤓
I get Merengue but it's always too slow.

[eager topaz]

AI now is receiving exactly the same criticisms that synthesizers and drum machines received when they were invented 🙂

[vestal rapids]

correct 🙂

[edgy roost]

Is there a way to have suno stop a song early? Instead of extending? Add a fade to silence for instance when extending?

[nimble dirge]

For the moment I think you can do that by downloading and editing with FLSUDIO /Capcup or similar video /music editors, but in the future I trust that we can make real -edition of edition in this magic web: 3

[edgy roost]

This is what I've done so far. But I can't upload the edited video/song to my Suno account again, so the "wrong" song will still be displayed there. 🙂

[bronze helm]

Someone might've answered this, but I remember the "Remix" button simply being a "Reuse Prompt" - it has been promptly renamed.

[normal patio]

Yes, but thank you

[magic basin]

i just "extend" and then pick the time you want it to end and type END in the lyric box. it should create new parts with just nothing. then you can use the "get whole song" option.

[rough pelican]

anyone know how many different voices can be added to a song? i am trying to do

1. german - deep distorted male
2. english - average female
3. english - average male

i like to experiment, but didn't achieve it yet. i thought i might ask here

Edit: Suno seems to understand JSON data in the lyrics field, which is actually SUPER interesting to know 😄 I just need to find out how to utilize this to give more specific instructions

[iron wind]

Good evening question I've got a few songs I've changed covers to and I LOVE the synced lyrics videos but how do I get the video to change the cover & Song title even when I edit it suno it doesnt apply

ignore previous I've seen no way to change thx

[warped furnace]

Good luck with that. 😅 You can try to put in as a prompt but not even duets are possible. I only get them by chance when I don't need them.

[rough pelican]

maybe define it as a single voice line... once i figure out how to define voice lines xD i think there is a chance to do this with json data format

[warped furnace]

Well, let us know if you succeed. I've tried to get two voices. It didn't work with "duet" nor with [female voice]/[male voice].

[rough pelican]

ty good to know 😄

[rough pelican]

well, as of right now i can't say if i am getting anywhere with my approach. i'm a web developer, but a noob when it comes to understanding AI xD

So, don't expect any updates soon

[ebon crater]

Hi. Can somebody help me in genera/styles ? I needed list of genera list as the website doesn't have one.

[faint spruce]

https://en.wikipedia.org/wiki/List_of_music_genres_and_styles

[rough pelican]

The solution might be simpler, lol. I should have searched before experimenting.

I haven't tried it yet.
https://www.reddit.com/r/SunoAI/comments/1d5bmja/duetmulti_voice_in_35/?utm_source=share&utm_medium=mweb3x&utm_name=mweb3xcss&utm_term=1&utm_content=share_button

[teal turret]

Hi how to make ai music?

[past cliff]

goto website. type words in boxes. win

suno.com

[cerulean steeple]

Hello. Do v3.5 songs include 20000th frequencies?

[sharp kraken]

so whats best practice i your suno is getting all nasty sounding ater a few gens

[dawn remnant]

how can I add video as backgrond of the music i generated

[eager topaz]

Go back to the best gen and try extending again. Other alternative is to change the genre, or modify it until you get good sounding results.

[untold briar]

If you don't mind changing things up you can do a reset with a very short clip, just like a genre shift. even if you do mind changing things you can Get Whole Song after the 'shift/noise-reset' and try to put more of the song back together

[dawn remnant]

I am using pro now I love it but need to play around and I think is best tool just if we can generateusing AI for videos also with the music and generate it will be very good and great software

[sharp kraken]

How exactly DO YOU do genre reset, bc im unfamiliar with that noise-reset? sounds like exactly what i need

[sharp kraken]

tooo many sunos to do that

[untold briar]

Extend From a time where you want to reset, then Extend From that new clip at time 1s or 2s. The song will be mostly 'fresh' since it has memory of either the music or the actual song you are making. Retry until you get something nice and clear. Then Get Whole Song on that clip. Finally, Extend From the new Get Whole Song Clip, after a reasonable sized chunk of the fresh/clear audio, and pray that Suno can now sing parts of the actual song (before the reset) with less noise. It works for the noise, but the music or melody may not exactly be the same

[sharp kraken]

I have been doing it with just 1 ext at a natural transition -

so your kinda forcing a transition i guess

I have to play all this stuff later tho, so does tempo and timing remain consistent?

I guess you could doubletap the drop - suno is pretty easy to get too

[untold briar]

You can use longer chunks than 1 or 2 and it can work, and preserve more timing, but may have to reroll many times

[unreal zinc]

If I remember correctly, the remix option once exist.

[untold briar]

right, the extra clip is if you want to choose a time. If you don't care you can ust any time near the start of a clip. But often that doesn't line up with where people actually want their song to take a little diversion, which is usually what happens if you do a reset, at the least

[unique stone]

"Remix" is now labelled as "Reuse Prompt"

[unreal zinc]

quite funny to be honest

[unique stone]

Why?

[unreal zinc]

I did a little bit of experiment when "Reuse Prompt" does not implemented, i made a song as a base then remix it with slightly tweaks, even though it is different, the tone is similar, but "Reuse Prompt" is not.

[unique stone]

It's exactly the same feature, they just changed the name because all it did was reuse your prompt so "remix" made people think it did something else

[peak topaz]

Before, putting Female Voice worked, but now in 3.5 it just doesn't seem to anymore. How do you force a voice?

[eager rampart]

Do we know if they’re planning on separating vocal and instrumental for download in the near future?

[eager rampart]

“Female vocals”

[unreal zinc]

shouldn't it have a function called bootleg?

anyway

[eager rampart]

Imagine if it could actually remix the song (club remix or acoustic version, etc)

[dawn remnant]

can anyone guide me how to change the image in the video as their is preblem in Language so can upload personnal image on the video tried doing enable can you please explain ?

[unreal zinc]

wish it can but not so soon i guess

[unreal zinc]

perhaps back in the time you can manipulate the style of the song to make it sounds similar to the base.

[ivory tartan]

Hey all, are you able to edit small parts of songs? I have just one word that Suno is finding hard to pronounce. Can I edit a word or line or do I have to generate all over?

[sharp kraken]

I write all my scripts in a circle, i usually like the variety in takes and I d a lot of lyrics then commands and try to break it up so that all 3000 words are relevant

but can you overload suno? that might be afactor in my case. I have to beg suno for an end

[cursive jetty]

Suno is actually supposed to sing (and I am a backup singer) in a background (or second) voice eg:
Hello I am a singer (and I am a backup singer) had success as utilising this as well as alt meta tagging with style, ints, sex changes eg: [Rap Verse Male - Bass, Piano]. x. [Chorus Female - Guitar, Drums, Bass, Piano] hope this helps...

[lone oak]

https://suno.com/song/df6640f7-e5d8-4a4f-96bd-7233f9d6cd9f

like this?

Only mark the music type in Style of Music
Write other descriptions in the lyrics.

[lone copper]

if i have a 10 second royalty free mp3 song, how can i extend it?

[vale isle]

Is there any way to clean up the lyrics after stitching together several versions of a song?

[lone oak]

10s?

[lone oak]

There is currently no way

[lone oak]

At present, only V3 can keep the original melody only when the first section is expanded.

[cerulean steeple]

Hello. Do v3.5 songs have 20000th frequencies?

[unique stone]

Do you mean 20000Hz?
These should be in the WAV downloads but will be missing from MP3s.

[cerulean steeple]

Yeah, 20khz. Is WAV download available only for paid member right now?

[unique stone]

I think it might be, if you have it it should show on the "Download" menu

[cerulean steeple]

Seems that the highest MP3 frequency from Suno is about 19kHz (at least that's what audio editor told me)

[unique stone]

Due to the way MP3s are encoded, you lose the highest frequencies (lower kbps = more lost frequencies.) you do also lose some of the detail of lower frequencies as well but it is not easily noticed.

[cerulean steeple]

i checked sone other mp3 audio and they have 20khz

mp3 quality loss is not about sound frequency range AFAIK

[unique stone]

It depends on the bitrate/Kbps of the MP3. As the bitrate is higher, the file is bigger and can store more data.

The way MP3 compression works is by removing the least-audible frequencies.

[radiant vector]

Hey, I was wondering, is it possible to specify the chords for some sections of a song in the prompt?

Like when I'm providing the lyric and specifying verse/chorus

[cerulean steeple]

Suno don't even understand a key note if you tell him that. 🤔

[radiant vector]

Yup, that's what I figured

Hope they incorporate it in future 🤞

[lone oak]

The current sound quality of SUNO, even the high-bitrate WAV, seems to be converted from low-quality MP3.

[unique stone]

There is slightly more detail in the WAV, frequencies that aren't in the MP3.

The quality isn't due to the MP3 encoding, it's what the AI model generates. This is why the devs previously did not want to offer WAV because there is very little difference in quality. But people asked for it anyway. So they let us have it.

[warped furnace]

Yeah, I've tried that too, but it didn't work. 😅 Maybe I give it another go.
On Suno's FAQ page you can find a similar suggestion.

https://www.suno.wiki/faq/metatags/

style prompt:
duet
male and female duet
romantic duet

[Mary]
This is my line…

[John]
But now I am singing…

[Mary and John]
There is a chance both of us are singing together…

[warped furnace]

Could be that the genre plays a role, too. Rap tends to use male voices. If you want a duet line by line, the box is full too soon and you don't get you four minutes's worth of lyrics into the box. 🙂

[edgy roost]

I asked a while ago how to do this, and nobody knew. Thank you!

[sharp kraken]

I just got an amazing punk rock duet

I used the Duet style so the singers could swap in and out on the same verse

https://suno.com/song/d4ea2ef9-0675-483f-9fdc-a55b50cb7912

feel free to check out the propting of it, but be careful - my promptingis weird

[atomic wing]

https://suno.com/song/93ddcc92-bc5b-4089-b126-e2d4b11ffe3b

Does anyone else have this issue where the generated song using v.3.5 is shorter than 4min? I even tried extending it but it seems the model didn't forget the end but actually lost the beginning as I can't extend it at all.

[white fossil]

it sounds more like generative artifacts, not encoding artifacts. I always thought it was an inherent limitation of the code to date, not a sampling problem

[lone oak]

At present, the sound quality of generated music is on par with that of encoded music. I think this is a problem with the Suno model.

[white fossil]

that's what i said lol

but the artifacts, while they sound like lossy artifacts at a casual listen, aren't actually that

[lone oak]

https://www.bilibili.com/video/BV1JF4m1P7RZ/?spm_id_from=333.337.search-card.all.click&vd_source=ae0513ef6639f9b82be3e62da5f86a73

I'm sure udio users did it a month ago, suno needs to catch up

[fluid robin]

https://www.alt-codes.net/music_note_alt_codes.php not sure if these would work in suno ai or not

shows all the musical Symbols and how to make them

[unique stone]

Suno songs generally sound to me like the individual sample amplitudes need smoothing somehow. I posted an example a few days ago of how it sounded saved as 8-bit wav, and it's a similar kind of dithering effect.

[lone oak]

Now it is not the gap between AI and coded music, but the platform only hopes that AI can use crude methods to rob the market and get the money from users.🤣

The song above has reached a point where it is hard to tell whether it was generated by AI or software, but the fact is that it was generated by AI.

[sharp kraken]

I just say this: [Genre switch to Drill, ] and viola - we drilling.

[lone oak]

It would be nice if the accompaniment and vocals could be adjusted after generation, it seems some AI tools can already do this

Will V4 really be available at the end of this month?

[sharp kraken]

That one time we made the dnb go 300 bpm by giving it too many alt codes. 🤘 11/10
https://suno.com/song/d7c0f708-04cd-4490-9a73-1ab0db43f7ed

had to use 6/4 ts just to keep up

i use alt codes for everything

just dont tell anyone ;

but ya know... someone could have mentioned that along with the bit about emojis and whatnot

[pure bay]

Problem: compare v3 and v3.5, is their difference only the length of music, not any other aspects (especially the quality of music)?
Since the song that I generate has lyrics, its ending part is always like an abrupt brake, cutting a missing part behind, which means it cannot terminate naturally.
For a natural ending, I have no choice but to extend it with another part that is fully instrumental (no lyrics). However, the extension of v3.5 lasts 2 mins, which is too long.
So if the answer to the problem in bold is NO, then I will just use v3 to generate that last part. Otherwise, I will use 3.5 to expect a higher quality of the melody.

[broken lotus]

they are different.

3.5 has extra training towards lyrical voices. It sings better, but at the expense of losing some of it's ability to prompt sub-genres.

I use v3 AND v3.5 (in the same song). Each has strengths and weaknesses.

[hollow hare]

There are other differences such as 3.5 being much better at reading metatags, though ironically, also worse at ending songs despite the ending tags used (which is why people, in the worst case where the song absolutely cannot end in 3.5, make an extension at the stopping point with a v3.0 extension using [End] as the lyric with the style Short, End to forcefully end the song where they would like to).

3.5 also got better at making some genre but kinda forgot some other genres (like phonk iirc)

[trail badger]

I can't get the feeling I want in version 3.5. He constantly makes songs in multiple musical styles. Slowly and dramatically, the song doesn't emerge. Edm style music is emerging. I get very good feelings in the V3 version. However, the sound quality is much lower than the v3.5 version.

[broken lotus]

start the song in v3 to get the style and sound, Extend in 3.5 for longer clips and better quality.

[trail badger]

Yes. That's exactly what I'm starting to do. I can get that feeling I really want in V3. V3 version needs development. Or that artificial intelligence feature in V3 needs to be transferred to V3.5.

[trail badger]

V3 version can imitate the cultural song style of the countries almost exactly.

[broken lotus]

If you know Stable Diffusion...?

My analogy is that v3 is a 'base model' (like SDXL)
v3.5 is an overtrained 'aesthetic model' (like Juggernaut, et al)

there are people who are happy with the over-trained look of Juggernaut... because it is very consistent in this one area (portrait of attractive people),

but it does not replace the base model because that extra aesthetic training doesn't do the 'regular' stuff very well.

They are technically the same model (at heart), but they don't replace each other.

[trail badger]

I understand. I believe it will become more impressive in the next versions.

[undone bough]

Is there a method to extend a track while making the AI sing in a different style than the original? I love the beat, but the AI keeps replicating the original singing style in the extended version, despite trying various prompts. This beat is gold! Any tips?

[broken lotus]

Might be possible to spam the style box
Female vocal, Female vocal, Female vocal, Female vocal, Female vocal, Female vocal...

or try a metatag
[Female Verse]

or both

[undone bough]

I will try that to "spam" it. Even when I select “INSTRUMENTAL only,” some vocals still sneak into the track. It’s really annoying. Any tips on how to fix this?

[hollow hare]

I found that if you are looking to make a pure instrumental song, use instrumental tags in the lyric box rather than using the instrumental switch.

It avoids hallucination (most of the times, depends on the genre) and pads out the time (as well as giving you control on how/what you want)

Examples:

• RayanAIR's Novo Lands which shows how flexible these tags are

[undone bough]

Very cool, Thanks I will check it out.

[lament cove]

Anyone found a way to 100% inject synth keyboards into hardrock/glamrock (power ballads or not is indifferent) a la Europe style?
I tried a bunch of experiments in the style field, but basically most of the time comes out without synths

[faint lodge]

someone knows if you can get a "proof" of license of a created song? Or is it not possible to let it distribute on soundcloud? I wanted to place it on spotify with soundcloud distribution but i get this message every time :/
"You appear to be distributing generative AI content. Please note that most AI content is currently not eligible for distribution, as is outlined in our Terms of Service. Please only resubmit with a license that fully outlines terms of use. "

[neat field]

theory: new workflow for duets with the new feature coming out soon @alpine venture @wet minnow

1. gen 1x v3.5 female song in a style you like until a voice comes out you like

2. gen 1x v3.5 male in song in the same style with similar lyrics and similar cadence until a voice comes out you like

3. repeat 1 and 2 until the singing is in the same key, OR pitch adjust one of the clips in audacity if you know how

4. apply a simple cut and paste of the songs to create a simple 30 sec- 1min sample of one of the clips voices, followed by the other voice clip

5. prompt the v3.5 model that accepts audio input with the new file, and with prompt containing "duet"

[cold zenith]

i really hope that the V3 model will stay available.

[neat field]

another experiment for the new feature:

• is it more effective to generate new prompts (ie 2x v3.5) text-to-song
  or
• is it more effective to generate 1x v3.5 text-to-song and then 1x feed the output back into the model as an audio clip (of the generation literally just created) and use the same prompt structure
  or
• never text-to-song and always include audio for audio+text-to-audio feature

which method will produce more success for the same prompts over 100 runs? likely depends on how much data is used for audio training vs text-to-song training - so if there is more audio+text->audio training examples in dataset bullet point 2 or 3 may be better (sorry for wall of text, coffee kicked in)

[toxic kettle]

can i somehow make duet song female and male? i tried everything but nothing works

[idle jacinth]

is there a way to make a song end after i extendet it? I don't want ot to go on for a minute, just a few seconds and be done with it?

[end] doesn't help

[untold briar]

Try prompting with V2 (the old one) and use style: "end" and then [end] in lyrics

[idle jacinth]

I'll try thanks

[untold briar]

3.5 in particular doesn't want to end. Which is often a good thing, but not at the actual end of course

[idle jacinth]

it ended, lets hear how it sounds in the full song

[broken lotus]

This is a Soundcloud-for-Artists policy. They don't want to distribute AI generated content.

A certificate of authenticity from Suno isn't going to change their policy.

[toxic kettle]

okay i almost got it but only once verse 1 and 2 its good but later they sing with rong voice

[ornate grove]

i have dabbled in duets but it is extremely iffy especially the longer the song. this is my best duet version https://www.youtube.com/watch?v=aHrQTyH8w5Q

the ending i was not able to get the female anymore so I had to continue from the first part and glue it to end in audacity

[toxic kettle]

https://suno.com/song/f5922475-48f9-461d-9613-4d0d35b844c1 well i got something like this

and i can't rereate this

[ornate grove]

yeah, suno forgot to switch back to female in your bridge

honestly, there is only one way: brute force of gambling

hit the generate button so many times they randomly fall in correct voices

[toxic kettle]

welp i love gambling

[ornate grove]

that is until they bring voice masks and make i possible to have consistent voices, preferably named voices

it seems like the new functionalities will include that to some capacity

[warm parrot]

Suno is like Early Midjourney right now, where it doesn't follow your prompt exactly, adds it's own thing, and has limited control so you just keep pressing generate until you finally get what you want. Makes me worried as what is the incentive to give us more control, if we get what we want in less generations = less credits used = less credits purchased = less money.

[neat field]

open source Soon(tm) Suno is great value so far though

[warm parrot]

Wait Suno will open source? Or some other model is being made, cause I didn't know of this 👀

[neat field]

let Meta cook 👩🏻‍🍳

[warm parrot]

Ahh Meta yeah they're likely to do it 👍

[austere dragon]

anyone have suggestions about how to get Simon & Garfunkel type harmonies?

[ornate grove]

the volume. when it is good enough, it will lure massive amounts of people to use it, instead of the few with either severe gambling problem or severe case of perfectionism. on average less to none re-generations will be needed but total number of generations will skyrocket

and by massive I mean that India surpassed China in number of people and if and when everyone in the world is able to create what ever they want with basically zero effort, potential userbase is in billions

[woven canopy]

are you allowed to like post music you make on suno or not?

[ornate grove]

if you have the sub in suno you can publish it via for example distrokid

other publishers have taken either hostile or difficult attitude to ai music

[woven canopy]

oh so i have to buy the sub

[woven canopy]

who would know xd

it just take a little swoop

then boogsh its unrecognizable

[ornate grove]

yes, free members are not allowed to monetize

[woven canopy]

doesn't matter tbh as long as it sound good dont care if its ai or human made

[woven canopy]

alright thank you

[ornate grove]

well, soundcloud seems to have eradicated most ai music, so they have way

also i think suno uses some kind of audio watermark, so it is traceable anyway

[woven canopy]

do you know what kind of programming language should I learn to actually keep up with these ais and make my own one?

[ornate grove]

no idea. i would ask chatgpt4o

[broken lotus]

'Folk harmony'
...? 😬 🤔
have not tried...

[broken lotus]

Python and Javascript are the ones you need to send API across networks to various bots...
Obviously if you are building a commercial app you will need a bit more to deal with the various platforms you are publishing to....

[woven canopy]

oh alr thank you

what if you create your own publishing platform

site

[broken lotus]

that might be a bit advanced for me...
usually you'd connect into an online service like AWS to run the actual software...

[warm parrot]

It took me so long to find the best way to prompt for Albanian style melodies in v3.5. I thought it was lost cause in v3 just "Albanian" would do it (though not consistently.) But for v3.5 "[Albanisch Balkan melody]" works 9/10 times

[woven canopy]

okay thank you

[nimble sun]

I did it for a ballad using this info: https://www.suno.wiki/faq/style-and-lyrics/duets/

[toxic kettle]

thanks

[warm parrot]

This is cool, I just did this. Thank you.

[keen cloak]

any estimated release date for this new continuing feature?

[scarlet gate]

dead bot?

[graceful condor]

use https://suno.com/

[neat nexus]

https://tenor.com/view/youknow-you-gif-19056787

[raw hinge]

can i update the lyrics of a song and use the same melody?

[soft gyro]

Hello there! Question! I "extended" a song starting with 00:24 but the beginning of the original is gone! Probably from 0 to 24 seconds is gone.
Is the extend function function deleting or altering the original ???? If so, why no warning at all ??? 🤔

[hollow hare]

You can Extend a non-vocal part in order to try to achieve it

[hollow hare]

That is the part 2 you are looking at, you need to combine via get whole song in the context menu of part 2 to get the first 24s of part 1 concatenated

[raw hinge]

thank you paimon

[hollow hare]

👍 For more info, see https://www.suno.wiki/faq/making-music/how-do-i-make-the-song-longer/

[bronze island]

Whenever i refresh the website (The Suno Website) i can't tap create.

[stiff mason]

Hello there! Is it me, or the v3.5 generations create more "robotic" voices? It's OK for many genres like modern pop, where everyone is singing with all sorts of gimmicks and all, but when trying to generate acoustic songs it sounds much worse compared to what v3.0 did

[ornate grove]

i have found some success by adding tags like "high-definition", "HigDef", "Clear sound" but still mostly random

always discard and downvote when mr. robot has snuck to the recording session

real pain comes from having mr robot singing to absolute banger of a melody that you never seem to get again

[tranquil garden]

So if you download the song in WAV format does it actually sound better?

[broken lotus]

if you prefer v3, it is ok to still use v3 – it's very good at 'singer-songwriter' and similar sparse music that is mostly voice....

[stiff mason]

yeah, but 4m duration is so damn cool

[broken lotus]

(not with normal audio equipment, and not with average adult human hearing.... but someone posted an audio spectrum of both and there is some info at very high frequency that is not in the mp3)

[ornate grove]

if i was a betting man, i'd wager that the main thing is compounding of compressions. when you download it less compressed and upload to somewhere like spotify or youtube, they do some compressing. so wav gets less compressed in the process than already compressed mp3

[broken lotus]

I think mp3 got a bad reputation in the Limewire early days when audio was compressed like crazy because bandwidth was dialup, and people were storing thousands of mp3 on hdd that maxed out at 100GB...

Suno's current mp3 are 192kHz 48bit sampling rate... they are technically 'high resolution audio' and better than CD quality....
https://en.wikipedia.org/wiki/High-resolution_audio

That doesn't mean every thing Suno outputs sounds better than CD..., but don't blame the mp3.
😉

[summer shore]

What is the best way to take a song style that you like and "write another one" in the general style (primarily the voice TBH), but to not have to simply "extend" which risks keeping too much of the style and exact riffs of the original song? Is there any way to tell Suno to "reuse the voice and rough style" but effectively do over from-scratch otherwise?

[broken lotus]

not yet...

a new feature has been announced that can build off of uploaded audio...
#🚀┃updates message

it is possible to upload a singing voice (a cappella) or a speaking voice, and prompt a new song by extending on the sample. I think a lot of people will start making songs from a personal voice bank library.... It's not exactly what you are asking for, but the potential is there to create many songs with the same voice.

eta on this new feature is roughly 2weeks
away... it is working in alpha tests, but needs some UI polish

[stiff mason]

Are you sure it's 192 khz, not 192 kbs? The latter is medium quality mp3, it was a somewhat staple in early 00s

[untold briar]

When you extend it, use a very small piece (distance to the start of the clip). You only need a few seconds of a voice to be mostly the same, and Suno won't have the rest of the song in memory.

[summer shore]

That is cool - I wonder if I can find a way to use it with a Suno-generated voice 🙂 I happen to like some of those better than mine. Thanks @broken lotus

I will try it with a shorter clip Jonathan, thanks. I did one with ~15 seconds and it was a bit too close for comfort (bass line etc). I might try ~5 seconds right where the vocals start @untold briar

[unique stone]

Yeah I think it's 192kbps, 48KHz.

The devs chose this because it suits the quality of the generated audio, but there were a lot of requests for WAV. The quality of the model is what affects the audio quality most.

[solar rose]

Why do the 192kBit/s always sound so bad, as if they were at most 64kBit/s, which were also chased through a distortion and a flanger?
Perhaps more emphasis should be placed on quality and less on the speed of creation. What's the point of having a complete 4-minute song ready in 20 seconds if the sound is just bad?
Suno and Udio are worlds apart in terms of quality.

Suno creates such great rock songs with super vocal melodies (better than Udio), but the result is always so strangely distorted that you simply can't listen to it for long.

[summer shore]

I tried extending from 00:03-04 (basically right where the vocals start). I can't seem to shake the general pattern of the song (the riff, hook, bassline, tempo) and preserve the vocals. So now scrambling to think of how to build a super clear prompt of what the vocals should be!

[feral mason]

having trouble ending a song with 3.5, any tips? it just wants to fill up the entire time allotment

[unique stone]

It really does not sound like 64kbps MP3. The classic effect of heavy MP3 compression is that it sounds watery. Which is what I hear with Udio (often sounds like 128kbps or less.)

Suno distorts, which sounds more to me like a WAV that has low bit depth samples.

There are likely many other factors that influence the sound generated. I'm not disputing that there's a difference in overall quality, but MP3 compression is not the issue. It's the AI model.

Both services have their strengths and weaknesses.

[untold briar]

There's a couple of other things you could do. You could try and generate a section of the voice that doesn't preserve the tempo/bassline/etc, like a moment where the music drops out and just the voice remains, not that unusual for a moment in a song. And then use that new sample of the voice. Or you you use a 1s Extend to break the music memory entirely after the voice, generate a bit of a new tempo, and "Get Whole Song" to bring the voice back. Both are pretty tricky though.

Actually first thing I would is just listen to the whole song and see if you find a spot where the tempo or hook isn't prominent, and it's a clear ish voice sample, and then you could try using two clips, one right when that moment starts, extend it by a few seconds, then Extend From the new clip. This is super annoying and hopefully Suno will soon add a feature so you can just snip out any section of audio and use that as a clip, instead of trying to get at it indirectly. (Since you can't set both start and end of Extend Time)

I'm kind of surpised 3 or 4 seconds preserves that much music actually...

Are you sure you didn't get unlucky? Like if you prompt "polka, polka, polka, polka, polka, polka, polka" then that should yank the music around a lot at 3 seconds

Here is an example of the kind of thing you are trying to do I think @summer shore #💬┃general-chat message

[summer shore]

Hey thanks -- I must not quite be using the "extend" type options properly and will study what you wrote + this example a bit closer! I'm familiar with lots of trial and error in Suno by now (some songs take me literally hundreds of credits in experimenting + building!).

But I do agree it would be great to have ways (some type of label/ tag etc.) to preserve only certain aspects of a favorite song and pipe that into the next song. I'm sure they'll get there... 🙂

[untold briar]

I think I'm making it more complicated. If you have a good voice sample in 3 or 4 seconds, that should be enough. I think the only thing you may need to do is chain a few short-ish (3 - 10s clips) together until you kind of break away from the tempo and hooks. Might even just take one more.

[summer shore]

Yeah that is interesting -- like see if I can throw some twists and turns and then use that as my starting point

[untold briar]

In other words, you don't have to start a 'brand new song' in the very first clip of your fresh voice. It does mean on the actual Suno page, the start of the song can be a little awkward because you'll hear the song take some seconds to kind of really start. The final song will contain those small clips used to move away from the original music. So I guess not great for the Suno contest. But if you can edit out that part it's fine.

Unless you can fit the whole song in 2 minutes of course.

[summer shore]

Yep -- all of that leads me to other wishlist items like "trim XX seconds" at start of song, ability to merge 2 disparate/ separate tracks etc. 🙂

My 2 favorite Suno songs are ones where I actually do an intro on one track and then "hit hard" with a separate track. I typically get the WAVs and move to Logic for editing/ merging, mixing etc.

[solar knoll]

I cannot use /chirp anymore? All channels away, why?

[untold briar]

Soon maybe? I often ask about it at least!

[fluid robin]

[Instrumental Interlude] works pretty well instead of [instrumental solo]

[fluid robin]

The bot is not online at the moment u need to create songs on suno.com

[solar knoll]

Ohhhhk thanx

[dusky marsh]

Hey everyone, has anyone else noticed that the first minute of a song tends to sound the best? I'm using the 3.5, which is absolutely fantastic, but it seems like the audio quality diminishes as the song progresses to the end.

[broken lotus]

switch to v3

Extend From Time where you want the ending (the end of a chord progression probably)

in the Lyrics box, type
[end]
nothing else.

[soft gyro]

hello there, I got 3 questionsm, 10x:

1. is there any possibily to select a certain voice ???
2. what kind of attributes exist for voices ?
3. is there a tag which forces the suno "monster" to use a certain instrument ?

[lone copper]

i tried adding reverb, equalizer through Adobe Audible and it helps except vocals still sound low bitrate.

[tawdry beacon]

/bark hello

[loud nacelle]

Is there a way to end a song at a certain spot? 3:53 into the 4 minutes would be perfect spot to cut the song.

[hollow hare]

Chirp bot is retired, use the website https://suno.com

Some helpful starting links: https://suno.wiki
https://youtu.be/lj8utBd_IUM

[hollow hare]

Extend where you want to end with the extension set to v3.0 when using an ending tags (e.g. [end] in lyric box and "end, short" as the style tag) as 3.0 is much more reliable for ending than 3.5

[sharp kraken]

You can extend it into a song with these lyrics "[Immeadiatly Fade to Silence. ] [End. ]" or some variation

i still have used only 3.5 but she does wanna sing more it's true

I dont like switching versions i feel like it changes the character of the sound too much

[loud nacelle]

Thank you both, I tried {End} but it created another minute, I'll try both of your recommendations!

[astral trout]

Guys, is there any plan to build a tab or something sort like it along with the song? That'd be dope!

[sharp kraken]

I founf 3.5 cares alot more about punction so i would definitly END. it rather than END it because, ya know, they might not get that its over until you finish your sentence

[lament cove]

[End]```
Or simply `[End]`, move the seconds up or down if you still get 1-2 minutes of extension

[broken lotus]

scroll up
#📚┃suno-school message

[sharp kraken]

I punctuate evcery phrase i make for suno, and she always just does exactly what i say

but miss even one comma or period and it can make the whole script fail.

[loud nacelle]

go figure, i searched and tried recommendations on {end} but failed to see the one just barely above. 🤦‍♂️ .

Anyways, I got it to work. Thank you all.

[sharp kraken]

1/128𝅘𝅥𝅲 1/64𝅘𝅥𝅱 1/32𝅘𝅥𝅰 1/16𝅘𝅥𝅯 1/8𝅘𝅥𝅮 1/4𝅘𝅥 1/2𝅗𝅥 <- ADV user Reference

you can write your music using traditional notation as well

"𝆑⇡𝆑⇡𝆑⇡𝆑⇡𝆑⇡𝆑⇡𝆑⇡𝆑⇡𝆑⇡𝆑⇡𝆑⇡𝆑⇡𝆑⇡𝆑⇡𝆑⇡𝆑⇡" <- "Why end it when you could always shoot it out of a cannon instead.

[grim elk]

Anyone having an idea how to create a good stuttering voice in suno? Just using "A-A-A-) f.e. does not do the thing :<

[rancid skiff]

Does anyone is currently experiencing struggle in downloading songs? (using chrome)

[naive quail]

Good? No. Bad? Yes! https://suno.com/song/cfbde41c-309f-45bc-ba72-f1508d4946a4

[grim elk]

Well.. not what I was EXACTLY looking for but ... apparently it does work, lol.

[naive quail]

lol

[grim elk]

Maybe gotta try some more to get what i am looking for

[sharp kraken]

you can gate it

ive never done it but just add gated to your vocal like [Gated Vocal, ] "lyrics"

[grim elk]

oooh thank you imma try that!

[sharp kraken]

gates all have a time sequence

you can use these

128𝅘𝅥𝅲64𝅘𝅥𝅱32𝅘𝅥𝅰16𝅘𝅥𝅯8𝅘𝅥𝅮4𝅘𝅥2𝅗𝅥

[lament cove]

clear voice or basically anything with voice in the style prompt in v3.5 render a hoarse growling voice, literally the opposite 😆

don't use it if you want a clear voice

[grim elk]

Where do I put this? In the [ ]? First time trying something more special with the AI so grateful for you tips!

Oh well putting (Glitched Female Voice) intro the style of Music did the thing xD

https://suno.com/song/1b4de584-2d78-4fbc-9b0c-964457b8d706 Just in case anyone would be interested

[final hazel]

Fakers.. Only suno.com and Suno.ai are the genuine ones. The rest are scamers

[coral plover]

yee thought so.

i was looking for the wiki. someone has the link?

[final hazel]

community wiki is https://suno.wiki

[coral plover]

ah haha. thanks a lot 🙂

[turbid kettle]

How would you attempt to prompt for something like this? https://www.youtube.com/watch?v=X7sk5m9cBh0 It seems to not really fit into common genres except maybe a very soft variant of rap, but the voice is almost whispering and the music is slow and it's only using drums and piano for the most part. So far suno keeps going for a american and more aggressive voice when I try for rap but when I focus on few instruments and slower music it won't generate rap and voice sounds more like slow singing too. I think I'm not able to describe it well enough >.<
Any recommendations for tags I could try?

[glacial lion]

Have you tried ethereal, or otherworldly? That tends to give more of an airiness to the vocals.

[neat field]

i download the WAV and manually create a 222kps variable bitrate mp3 version (vs 192kbs) - takes about 2 seconds 13600k probably not even worth the 2 seconds, but here i am

[naive quail]

and v3 seems better at adding noise https://suno.com/song/d61c5e44-a1d5-4a9d-adff-c967fb5599d5

[wheat rose]

I'm getting a ton of songs where its repeating verses for no particular reason... woooo

asking for verse, verse, pre-chorus, chorus... it will actually do verse, verse, pre-chorus, repeat pre-chorus, chorus, repeat chorus. hmmm

[glacial lion]

I found v3.5 repeating the 2nd verse a lot for some reason

[wheat rose]

i just got a bizarre one, good generation if it didn't lsoe its place multiple times... i'm going to try to rescue it with audacity

[wheat rose]

audacity ftw

[royal bramble]

Is there a way to uploaud the music that I created in suno, for my stories and reels in instagram or facebook? Is it possible?

[wheat rose]

not directly through the platforms but if you use your own video editor and put the music in before uploading it you're golden

[royal bramble]

I though that before, but I was wondering if there is an easier way to do that, I would like that. 😅 Well, thanks. 🙂

[unique stone]

Also if you've not created the songs with a paid subscription, be aware that you need to credit Suno (e.g. add text to your video to say the music was made with Suno.)

[royal bramble]

Got it

thanks

[unique stone]

It's good that you're making things with Suno but please don't spam this in multiple channels. Sharing of Suno songs should be in #🐣┃suno-showcase (where I see you also posted this). It doesn't also need to be in general, off-topic and technical-discussion.

[royal bramble]

I think you must be confusing me with someone else. But I understand, I wont spam here.

[unique stone]

The message I was replying to got deleted 🙂 it was not to you

[royal bramble]

Haha, Ok,

[balmy crown]

weird question - is there a way to say "not x" -- in this case, "NO DRUMS"

I assume not, but I wanted to make sure

a lot of times with AI, I realize that bringing something up (even if in attempt to negate it) makes it more likely to be featured

I'm trying to get a lullaby and it really wants to build up and add drums and things by the end

and I'm like "No NOT THE VIBE"

i'm just like "lullaby, soft piano, sleepy"

drums

[rough pelican]

question: can i create a song using copyrighted lyrics for private use / just listening to it?

I know there was a one-time notification when adding custom lyrics, but honestly i forgot what it said.

Does Suno check my private library for potentially copyrighted stuff and "strike" my account?

I'm already careful on what i'm doing. I don't need advice not to copy 1 to 1 etc. Just wondering on the technical side

[neat field]

a capella, trumpet replace trumpet with other instruments not the drums. Results may vary, untested best guess

[mental forge]

hey, I wanted to ask, what is the best practice for producing male vocals in one part and female vocals in another part, or male and female vocals singing together?

[balmy crown]

lol yeah I definitely wouldn't go for trumpet to make my lullaby more mellow

[balmy crown]

there's specific tutorials for this but you can always go with

in the lyrics themselves

[male vocalist]

and whatever lyrics you want it to sing as a guy

[female vocalist]
for the girl parts and whatever

it's not reliable though haha

https://www.suno.wiki/faq/style-and-lyrics/duets/

[mental forge]

Ah thank you, so that might be possible but it happens randomly and may soak up my credits haha

[broken lotus]

just keep it unlisted. It won't show up in searches.
Suno doesn't peek at your private music (that I have ever heard of)

[rough pelican]

thank you! 😄

[sonic rune]

what's this I see where multiple tracks are being fused together in the intro segments and throughout?

[graceful condor]

#🪲┃bugs-web message

Please help

[naive perch]

Is there a way to stretch certain words? Or make Suno sing a certain note longer?

[south kindle]

nooooooooooo!

just kidding. just try something like that.

[glacial lion]

Yes, I don't think negative prompts work. I did get a lullaby-ish song with this prompt in V3 https://suno.com/song/9f24631f-6ead-4a3c-a478-814bd7e445a5

[glacial lion]

You use the extend function (in custom mode) to generate additional sections of the song. You can extend from any point in the song. It is useful because you can cut out parts you don't like. e.g. where it keeps singing the last verse over and over for 2 mins instead of ending 😅

[sonic rune]

I wonder…
If I have a song that I made with a free account, but then extend it and/or finalize it with a Pro/Premier/Custom account, which license applies to the song?

@charred merlin would you happen to know the answer to this? ^

[south kindle]

similarly, let's say someone extends your track. Then where do things lie?

[sonic rune]

wait, users can extend other users' tracks?

[south kindle]

yeah, public ones. weird, eh?

[glacial lion]

I never realized I could extend others' tracks

[sonic rune]

😳

[south kindle]

that should be a setting users can edit on the tracks they create.

3rd party extendable or whatever.

[sonic rune]

yeah, I don't want others editing my tracks without permission

[glacial lion]

Yea, seems weird it is allowed by default

[sonic rune]

especially when I crafted the lyrics myself

[sonic rune]

can it be toggled?

[south kindle]

but worry not, they can't download your tracks, so its fiiiiine.

[glacial lion]

extend - [end] - tada!

[sonic rune]

*yet
I'd give it 1 more month before there's a working 3rd-party downloader tool

[glacial lion]

not that I know of

[south kindle]

just grab the share link, post it here, then download the video.