📚┃suno-school

[spiral bane]

also the vocals so far all sound like its someone siging a shanty at a pub after a few whiskeys

when i say all i mean all but one ive generated so far

perhaps everyons generating and my priority generation is not priority any more lol

ok its broken, its not listening to the style of music section right now 🙂

litterally turned a death metal song into a pop song duet

i'll wait a couple days

[raw moss]

yea its annoying af. otherwise V4 is really good

its every song

[craggy delta]

why is this happen??? 😦

[lunar anvil]

To me it sounds like Reverb Shimmer or Reverb Self-Oscillation. Specifically it is a reverb feedback loop when the wet signal (the reverberated sound) feeds back into the input of the reverb processor, creating a loop of repeated and amplified echoes. Can also be caused when the reverb gain or feedback is so high that it creates an uncontrollable wash of sound that builds on itself, which is known as Runaway Reverb.

[raw moss]

its 100% the call sound from Metal Gear

[rustic gale]

Wow. I only had it in 2 so far out of like 100 tries. U got extreme end. Maybe its tied to some genre or smthing?

on the other hand my style was my nickname XD Sooo i guess not XD

[raw moss]

ive done like 16 songs and 2 dont have it. most of those are remasters

[charred merlin]

Looks like there was an issue with your payment details -- you may need to click the black Edit Billing Info button to resolve any issues -- [e-mail] is really the only support channel that can properly help with that though 🙏

[odd portal]

So far, v4 doesn't work at all for my personas. I hope that gets fixed.

[molten dawn]

There's definitely this persistent, high-pitched ringing sound in the background v4 generations. It's quite distracting. But it seems like a bunch of folks are having the issue so hopefully it's resolved soon.

[raw moss]

me too. but i wanna know who was making V4 and got the Solid Snake phone call

[charred merlin]

Suno is actually developed by the La-Li-Lu-Le-Lo

[fluid sierra]

what does V4 do better than v3.4?

3.5

[charred merlin]

potential instrument / vocal clarity up the wazoo

[fluid sierra]

Hi again @charred merlin Are you official helper here?

[charred merlin]

"official" is such a relative term, isn't it?

[fluid sierra]

If you remember from earlier @charred merlin i managed to get parts of my song sung in the male character, but not throiughout the song

[leaden pawn]

Sadly noticing insane amounts of reverb in v4

[molten dawn]

Aren't we all. v4 is currently stuck in reverb hell.

[empty lichen]

So when I download the stems and put them together in post at least for this song the shimmer is gone

Testing out some more

[astral bay]

That night be worth looking into maybe the stem cutter cuts off the noise

[empty lichen]

Im actually testing more songs with it will report

[molten dawn]

I'll try it.

[empty lichen]

Maybe take something that has more shimmer I took one that had just a little bit

[lunar anvil]

Extends are broken on v4. I am not going to waste anymore credits.

[charred merlin]

I remember you -- always glad to see new-to-me faces hanging out in tech-chat 🙂

Admittedly, most of my music is instrumental on the Suno-side, so I'm definitely looking forward to hearing all of the tips and tricks the community can uncover about v4 -- we can use what we know about v3x as a starting point, but ultimately it's an entirely different model that will have it's own strengths and weaknesses -- #🪲┃bugs-web message (dev quote on the matter)

[astral bay]

Guys i think I found what it sounds like apart from MGS. It's the unreal tournament link gun check this yt out it's the same lol

https://youtu.be/Ifmqg4PWDag?si=PI6PqyyNc3y0WpJ1

[empty lichen]

It's gone in this song too

[astral bay]

With stems?

[leaden pawn]

Do they perform UAT before release or is the community doing UAT? 😅

[empty lichen]

If not 100% gone very reduced

[empty lichen]

Yea

[astral bay]

It's in beta

[empty lichen]

Download the stems of a song that has it and put it together in post it's still there but vastly reduced

[astral bay]

Gonna check out how metal performs if it works I'll respect u even more

[empty lichen]

at least so far in the tests I do. I think it renders the stems again

[wind venture]

On this note, are the devs ever planning on releasing official documentation? It’s been requested by so many community members as long as I can remember.

I know there’s a community sourced wiki (which only one person can edit if I recall). But all information is more or less guesses based off experimentation, rather than concrete info that can be provided by the devs.

[empty lichen]

I'll try it on something heavier too

[molten dawn]

I'm working with something that's pretty heavy right now.

[empty lichen]

Might not work as well on instrumentals

still going to try and use a stem from and instrumental and see what it gives me haha

[molten dawn]

Not hearing as much on the instrumental stem. I'm going to try to remaster the instrumental.

[empty lichen]

It seems to compress it again or something

[astral bay]

Eh I got the stems for a heavier v4 that has it really bad and it's the same. But on another which didn't have it as noticeable it went down more so it kinda works but I guess heavily distorted ones won't do any good

[empty lichen]

There is probably a threshold it can work with

[molten dawn]

That was a complete and total disaster.

Changed everything about the song and none of the changes were good.

[astral bay]

I don't think they really know what they are doing

[lost dagger]

v3.5 appears to be much better for metal. Cymbal crashes sound way more synthetic on v4 when remixing a v3.5 song

[astral bay]

I mean not to sound disrespectful but I think it evolves more along how they develop it so it's not an exact science where I do this and that's the result

That's why I don't think official documentation will happen because it's just so much rng that official might not be relevant

[astral bay]

Yea drums and the overall sound is more stable in 3.5 but 4 has that crispness that you get with real strings. It's quite good at mimicking Distortion and acoustics it's even wider stereo but i think that phasing artifact is exactly the reason of these improvements

[zinc spade]

generate lyrics suno no more?

[wind venture]

I mean even just the most basic info about which styles/genres and metatags actually affect the result would be a step in the right direction. The only reason I can think of them not putting official documentation for that stuff would be because they want everyone to sink their credits into experimenting through uncertainty.

[molten dawn]

yeah ultimately this is a cash grab lol

[zinc spade]

why can not generate lyrics no more ?????

[molten dawn]

One of the best results I ever got included absolutely nonsensical styles: chaos choir gospel chant chorus chomp dog shiba inu roux chirp indecisive indie badgercore caninecore

[astral bay]

Imo it's an LLM and it's so rng based they don't really have the capacity to really understand it

[zinc spade]

why can not generate lyrics no more ?????

why can not generate lyrics no more ?????

[lost dagger]

Question, have you tried the stem extract yet? If so, is it the same as V3 where it just separates the music from the lyrics?

[jade mist]

Remember v3 was trash when available to paid. I think same with v4 beta

[wind venture]

I agree RNG is a major factor but there’s definitely some stuff they do understand which they can provide to their users. They have a better understanding at which metatags will affect the output and which ones won’t.

Midjourney is also RNG but they have massive amounts of documentation and user tips on how to efficiently prompt and use credits. Just hoping to see the same with Suno one day

[sage basin]

in v4 im getting a artifacting in the background of clicks or almost a ring of a old phone

[molten dawn]

everybody is. it's busted.

[sage basin]

ok thanks! just joined the discord to see if its just me or not lol

[unique stone]

V3 alpha? A lot of people preferred that over final V3 tbh

[charred merlin]

You might be talking about the community wiki from wetcircuit -- https://www.suno.wiki/ -- written by one of the old moderators, and an OG Suno/chirp user, so while it might be slightly outdated, they definitely know that they're talking about.

I think part of the issue of writing documentation is that there's so little that works concretely in Suno -- a great example from @polar spade that I frequently refer to is how [boys choir] is more likely to work if you're prompting a gospel song with innocent hymnal lyrics or something, where that same metatag might not be as effective if you're prompting for a black-metal song.

It becomes a lot harder to say for certain "this prompt will guarantee this outcome" when you have NO idea what other prompts will be stacked beside it to impact the weights 🤷‍♂️

[molten dawn]

haha me too

[charred merlin]

argh -- the #known-bugs channel is gone, so I can't link to the picture I had, but are you being told to wait before attempting to generate? Using ~1250+ characters and trying to use the "write about" feature?

[wind venture]

I get that. And I think that’s why I yearn for more user control. I assume Suno has some level of stylization bias to help something sound conventionally “good.” I think it’d be nice to have a model that makes an attempt to do obscure things (even if it it’s less likely to sound “good”) rather than letting its own stylization bias take over.

A simple sliding scale to allow the model to let loose and get weird would go a long way. Not everything needs to sound like a radio hit.

[molten dawn]

Specifying the genre first and then detailing the characteristics within that genre can help guide the generator more effectively.

[charred merlin]

I agree that having more customization options would make for a really exciting experience, and would allow for a lot of crazy experimental stuff as well -- unfortunately tech-chat is more just users helping each other out and talking shop, however the devs DO browse #💡┃suggestions-web if you wanted to get dev's eyes on the suggestion 🙏

[indigo basalt]

I just wanted to ask a generalized question, but will v4 of Suno be released later for the people who just use it for free, or will it only be available for Pro and Premier members? Any and all information is greatly appreciated! 🙂

[wind venture]

Yeah sorry I guess i veered off topic lol. But I’ve left the suggestions in the past so who knows, maybe one day

[charred merlin]

The earlier models were locked behind a paywall for a short time as well -- if history indicates anything, it'll be a week or two 🙏

[indigo basalt]

Thanks for the information! It is greatly appreciated! 🙂

[charred merlin]

Nothing to apologize for! It's not always clear which servers are used in which way, and if people have awesome ideas that I want to see implemented, it's also in my best interest to make sure that those awesome opinions are directed to where the devs are more likely to see them 💪 You rock, dude 🙂

[timid dagger]

I am a content creator and I wrote scripts to automate some of the things I would be doing on the website that are repetitive.
However today this stopped working. I am pro subscriber and I am even paying extra credits each month so I don't understand why I would need to be affected by the recent captcha changes. I am not botting or anything.
Does anyone know if there are any plans to give developers an official API of some sort? It is currently slowing down my workflow significantly and I am considering to find an alternative as a result.

[worn flicker]

So, I haven't heard this yet, but I remastered a song of mine, and it is either the clarity of the vocals or something, but I actually heard the roll of the tongue for an r sound in another language. That was cool.

https://suno.com/song/d1455abc-43c4-449f-81c6-5282ae79945e - it is right at after the 1:22 mark. I have spell casting in elvish and the R rolled like I have never heard Suno do before. Maybe it is because I don't write in other languages, it could have been doing that but I never heard it...lol

[wind venture]

I actually do have a technical question (and this is an example of something I wish Suno would put in documentation).

Is there definitively a difference between the way the models interpret commas vs. no commas?

For example, is Japanese jazz fusion going to generate different results than Japanese, jazz, fusion? I’ve experimented with this myself and don’t really have a conclusion. But I assume the devs have some idea, which is why I wish they’d provide this info to the users who want to try and blend genres

[unique stone]

Suno do not offer an API. Automating song creating using Suno would be against their Terms of Service.

[hollow hare]

If you want to get banned, sure. Suno has no official API. Services claiming so are actually breaking the law in USA by essentially hijacking Suno's Service (see the Computer Fraud and Abuse Act Section B.1, and 18 U.S.C. § 1030(g))

[lunar anvil]

This last generation is causing some major clipping every time the beat hits.

[charred merlin]

oooh, lemme pull something up

https://platform.openai.com/tokenizer This is specific to chatgpt, but gives an example of how AI breaks our instructions into "tokens" it can process -- personally, I've noticed a difference between pop-punk and pop punk, like "blink 182" vs "katy perry with some distortion"

[molten dawn]

If we're just airing grievances, I want to know why tf Suno NEVER listens to me when I tell it to slow tf down. The default BPM is like 190.

[zinc spade]

no when writing descriptions used to come with generate lyrics does not show any more

[wind venture]

Oh awesome, this is a helpful tool. Thanks! It’s a nice step in understanding how the model interprets the prompt at least.

[zinc spade]

what happen to generate lyrics in the custom mode dose not show any more?

[crude jetty]

well, mine are mostly in the 125-127 range

[crude jetty]

it's still there

[molten dawn]

I'm exaggerating. It's mostly the fact that I explicitly include some variation of 'slow' in the prompt and it seems to be wholly disregarded.

[charred merlin]

omg -- the devs fixed the bug where using more than 1250 characters and generating lyrics would give a system error 🙏

[crude jetty]

well, you can get it to do so if you prompt correctly

[molten dawn]

Oh, right. My mistake. I was specifically prompting incorrectly. Silly me.

[crude jetty]

rhythmic is one that works for me. melotic is another

[zinc spade]

example when I put a description of a jazz club the lyrics give me nothing about what i said it give For me nothing related to my description

[molten dawn]

lol I literally just got a 190bpm generation from this prompt: sludgy, heavy riffs, downtuned guitars, thunderous drumming, dissonant harmonies, thick basslines, lethargic tempo, dark foreboding and punishing atmosphere -noise -hiss -ringing -reverb -fast -upbeat

Where'd I go wrong?

[verbal mortar]

hum, whaddya know. I can't make personas at the moment. just keeps erroring out.

[meager frost]

Why doing covers can’t i use the 3.5 model instead of the 4?

[polar spade]

https://www.suno.wiki/faq/style-and-lyrics/should-my-style-prompt-use-commas/ I'm not a dev but know this site exists. It's good for getting things started.

[charred merlin]

I'm calling it now -- Bubbles are the new neon.

[worn flicker]

Oooooo...I'm figuring shit out now baby

[balmy creek]

no cowbell

[worn flicker]

Duets are insane on this v4

[polar spade]

Like giving up cigarettes for whiskey?

[polar spade]

Wait... writing that down in my hook book.

[worn flicker]

I put Duet and it went nuts, it wants me to cue up the vocals...that is awesome.

[molten dawn]

Added "very slow, 60 BPM" to that prompt. Generations produced were 142 and 95.

[wind venture]

I’ve read through this before but I’m a bit skeptical because I feel it’s a bit misleading with how they explained it. A comma’d list doesn’t seem to necessarily cause each word to contribute evenly. The models tend to heavily prioritize whichever words it has more training data on. So if you have a list of 4 genres and some are more obscure, those more obscure ones don’t carry as much weight as the common ones.

[verbal mortar]

LOL omg recurring joke on me, no matter how much i extend to end this song...suno is like "no...we gon keep going"

[worn flicker]

Did you try slow-tempo?

[eager topaz]

Just the same with Stems, only 2 tracks, vocals on one all the rest on the other, with vocal bleed through on the instrumentals

[molten dawn]

is 'lethargic tempo' not specific enough?

[worn flicker]

I would say a bit too specific...lol

How bout "fresh out of 5 year coma"

[worn flicker]

For real though, I think that Lyrics tag weights have been increased. This is what I am seeing. I need to see if tags like [Faster] or [Slower] or whatever versions will work now....

[molten dawn]

"Ultra-slow tempo" gave me 95bpm. Makes no difference.

[worn flicker]

I still say that is too detailed

[charred merlin]

"tempo" by itself could be interpreted as "hey, some fast songs use the word "tempo", let's do a fast song!"

Italian musical terminology isn't guaranteed to be respected, but there MIGHT be something more there -- at least it's trying a different approach, which ultimately is the key

[worn flicker]

Once it is tokenized. Yea, like how the negative can get caught up in there sometimes

V4 is a whole new animal for me. I got experience on 3.5 from what I can tell that was the wrong one to learn on...lol

[molten dawn]

Interesting, I added "[slower] [slow] [slowest]" to the lyrics section and it seemed to make a big difference.

[vapid atlas]

The latest update ruined the site for me. I normally put an idea and song name in the custom lyrics area and then have Suno turn it into a song, and it always works. But the new "Write with Suno" just ruined that. Can someone please get it back to the way it was before? There isn't enough room in it for me to go along like normal.

[worn flicker]

So can I get a bit of credit here...lol

[polar spade]

Purely my opinion from my readings and observation but, are you using phrases or even sentences which might not be optimal. So far, I perceive that the fewer significant words used, the better. Insignificant words such as "the" , adverbs like silently, quickly, less... prepositions, of, in, at, on and so on create tokens that have no worthwhile instruction and so the surrounding valid words of café, film noir, dark, mysterious become voided as well. or It's like not enough valid packets going through the line the the wifi drops. Suno can't mold what you want so it just through in a default. A total guess. But worth trying the minimalist approach a couple times.

[molten dawn]

Yes thank you for your suggestions!

[worn flicker]

Thanks

Lol

[vapid atlas]

Hellllo! Can someone please respond to me?

[worn flicker]

Honestly I just figured the lyric tag weight thing out trying to control a duet that is way over trying

[molten dawn]

Although it only seemed to make half a difference. I got an 85 and 110 bpm out of it.

[charred merlin]

Sorry -- referring to the post from ~2 minutes ago, or had you posted earlier?

[worn flicker]

A difference none the less...hehe

[vapid atlas]

YES! I had said:
The latest update ruined the site for me. I normally put an idea and song name in the custom lyrics area and then have Suno turn it into a song, and it always works. But the new "Write with Suno" just ruined that. Can someone please get it back to the way it was before? There isn't enough room in it for me to go along like normal.

[worn flicker]

Use another chat bot to write your lyrics

[hollow hare]

Use Simple Mode for what you want (make sure Custom switch is turned off).

If you meant you want the old ChatGPT generated lyrics instead, change back to Classic option for the prompt (Click Write About first for the following appearance)

[worn flicker]

Oh cool

[molten dawn]

I'll take 85 though. That's progress. Just got a 65 as well.

[manic vapor]

not wanting to be that guy, but are recent generations failing for others?

[charred merlin]

Please understand that this is a community-run thing, not an official support channel. Sometimes it might take longer than 120 seconds to get a reply 🙏

[primal juniper]

Can you replace section anymore on v3.5? Because when i do that, my song converts to V4 and to terrible sound quality.

[worn flicker]

Hey look...not nanoGPT @charred merlin Tip of the cap to you sir...

[vapid atlas]

I would still go over 200 words if I did that, PLUS, I would have to increase it to get the genra I would want.

[hollow hare]

Usually that is due to some secondary filter being tripped like using an artist name from what I recall

[manic vapor]

even remasters are "stuck" generating and then refund credits

perhaps too busy or something, i can wait

[charred merlin]

Hey man -- to your credit, bark absolutely uses it, and if understanding the quirks of nanoGPT helps you get better results with Suno, that's the key 💪

[hollow hare]

Did you ensure to switch to Classic in the Write About? Seemed to work for me.
Note you need to click Write About first now then enter what you want. Hopefully the following picture helps show you what I mean

[worn flicker]

I will now be looking at ReMi unless they developed it. It would make sense to create one that is custom to your own stuff eventually.

[manic vapor]

i made a custom trained gpt with openai with a lyric set, works pre-nicely

[worn flicker]

I made one for structure and cleaning lyrics up from instructions. I use it for lyrics when I'm testing instead of writing and I need a specific set of syllables.

[manic vapor]

oh nice, gemini seems to understand how syllabic phrases work over gpt tho, so i swap to that if i am trying to get a certain rhythm

although o1-preview works okayish for that too

[random dust]

Thats odd i cant even get in my account

[polar spade]

For fun I use the talking Gemini and try to write lyrics. Eventually, she says "I feel you are getting frustrated".

[worn flicker]

I've gotten chatgpt(custom gpt) to do some nifty things. It took some work to get it there, but I have it rewriting lyrics completely phonetically now. Really helps when doing speed raps, which v4 seems to not be a huge fan of. Or the prompt for it has changed

[polar spade]

You're under "create my account" not " sign into my account"?

[manic vapor]

i found if you put two lines as a single long line, or jam as many words into one line as possible, it does speed up vocalization

[worn flicker]

Correct.

[worn flicker]

I just happened to be writing one today and v4 released just before I made it public, so I reused prompt and it did not do well. I had to swap a good deal up.

[random dust]

I did sign in through that way though

[manic vapor]

yea, for sure, style tags dont seem to work exactly the same, i can't tell if they more accurate, or more diluted, lol

[random dust]

Oh son of the biscuit

[hollow hare]

Do you have any add-ons that may be blocking a captcha?

[random dust]

Nvm one guy is right lol

I was on sign up lol

[polar spade]

Oh. You clicked the "sign in" at the bottom first? To get thr sign in page?

[hollow hare]

Ah 👍 Great detective work Dasvon!

[worn flicker]

I think they moved back to how they were supposed to be. Someone in here told me when I had gathered every style from explore page that they don't match up. I believe it was meant that 3.5 doesn't match up as well and that is where my experience is.

[polar spade]

I burned... I learned.

[worn flicker]

The statement was accurate...lol. I did one that was an anime type style and got some weird rock jazz hybrid in 3.5

[hollow oxide]

so when you keep making songs and songs disappear from your library, if you didnt delte them to trash, are they still gone forever?

[hollow hare]

You sure you didn't see it on the next page? There is a pagination but only if you go to the Library section
(Create section limits the display and hides the downvoted clips)

Unless it was a failed generation in which case it is gone

[worn flicker]

Hey so, this is new. I have always cropped the excess of an extend prior to getting the full song. In v4 once I crop it, I no longer get the option to "get whole song" not a big deal I can crop it after, but interesting that it did that when 3.5 didn't. @hollow hare

[hollow oxide]

you just blew my mind

[hollow hare]

Ikr, easy to miss. You can also click the number and edit the # to jump ahead. (I also remember missing the trashcan in the upper right the first time too)

[hollow hare]

Thanks for the heads up, can you also submit it to #🪲┃bugs-web ?
This may be related: during the alpha, as a mobile user, I noticed that v4 mobile view didn't actually have an Extend or Get Whole Song option! But the other versions did.

Based on that assumption that the menu may have different options, I tried changing to landscape mobile mode and regular desktop mode, and those two had the options, while portrait mobile view did not

[worn flicker]

I'll check. I am always on mobile in Android on my tablet in landscape, but you are correct. Things change all the time due to that, and I can also check if desktop mode changes anything

[lost scaffold]

Tried to remaster a song that I previously removed/replaced the lyrics with my social tags. The remasters came out completely distorted. Is it necessary to have the original lyrics included when remastering to v4?

[hollow hare]

Yes, you need the original lyrics for that part as Remaster closely adheres to it (I wonder if that means we can easily fix our typod lyrics?)

[still wren]

How do we get in contact with support ?

[hollow hare]

Email [e-mail]

[still wren]

Legend thanks xx

[lost scaffold]

Does this also apply for instrumentals where we have used commands in the lyrics box? I tried to remaster a song with the original "lyrics" I used and got similar distortion. I can DM if you need example.

[hollow hare]

I would assume so. I haven't played with Remaster with the original lyrics altered so it is as new to you as it is to me

[lost scaffold]

Dang, ok. Thanks anyway.

[fresh oxide]

v4 gave cleaner vocals but it's destroying everything else. Major artifacts in these!

[fresh oxide]

When you remaster a v3 song in v4 does it keep everything the same but clean up the vocals?

I've heard comparisons and v4 needs to eliminate all the artifacts and fix the timing.

I'm hoping we get to customize pitch correction to ensure it sounds natural and avoids robotic or autotuned effects.

Like Antares

[lost scaffold]

So far it has distorted all the songs I've tried. With or without the original lyrics. I suspect it has to do with the structure of the song, but I've literally only just started messing w/ v4 so still TBD.

[fresh oxide]

This is the best one I've found.

[fresh oxide]

OK. Pls DM me if you get it working.

[lost scaffold]

Got you fam. Sending a friend req so I don't lose you.

[worn flicker]

It doesn't keep the exact same Vocals it follows them closely many times and sometimes it changes them, I just finished one and extended the end to fix the drift. I'll put them here without the embeds.

[fresh oxide]

These are the main 4 I use before I upload to suno so I don't need all the auto-correction and suno often goes nuts with it.

[polar spade]

Same "subjects" seem to be discussed in the Chinese channel as well, so it's not about a cultural difference in how we perceive things. Just checking.

Since I deal more with acoustic, earthy sounds and well kpop and pop in general, there doesn't seem to be as much issue with any overtone noises that others say they are hearing. Now if it goes wild and starts to build at the end like it prefers, then ,yes it's just maddening. For pop , acoustic and some genres, the instruments and vocals have alot of room in the mix, so maybe that helps.

[worn flicker]

That is a VSTi right or some other desktop PC thing addon or whatever, VSTi is specific? That doesn't look mobile.

[polar spade]

Well, the lyrics writer is interesting. I said write a song with no end-rhymes and it DID. good job. But then 8th line down. " Not in the glittery glam
Not in the neon blare" 🤣

[opaque canopy]

ok i'm really trill to play with the new V4 but before the create tab get messy how can i create a folder to put the song im working on in it ....

[worn flicker]

https://suno.com/song/9a0b1c62-343d-480a-ac93-c8350aa75cbb - v3.5
https://suno.com/song/f7a39360-9fef-49d0-a9a7-51dea0b44359 - v4 - notice the R roll in Verse 2 and the insane belting at the end

[fresh oxide]

Yeah, it's a plugin for Universal Audio. It will kee your vocals sounding 100% natural. https://youtu.be/wMoGXs2Sc6g?t=88

[worn flicker]

Ok so it's desktop

[fresh oxide]

Every single album we hear uses processing to hell and back. They even use it when they sing live.

Here's the audio engineering chain:
Reverb and delay effects
Pitch correction
Compression
EQ adjustments
Harmonizing and doubling
De-essing (reducing excessive sibilance)

[worn flicker]

Cool. I have tons of desktop tools. I am not on desktop that often these days, but FLStudio mobile has an interesting amount of mastering tools. Considering how crappy it began

[fresh oxide]

Yeah, it's for the Apollo Twin or any UA DAW.

[fresh oxide]

I can burn these into the original mix or add it later. It's pretty cool.

[patent yoke]

im finding gpt4o is really good at rewriting lyrics but keeping the same rhyme style

but if gemini is good too ill check it out

[worn flicker]

Cool

[worn flicker]

You can get inventive with 4o. I have challenged it to do as well as me and it actually didn't do to bad and I had the scheme tight as hell.

Partially phonetic tight

It still got pretty close.

[rancid fiber]

is the watermark issue with lyric song has been fix yet?

[worn flicker]

Do you mean the actual watermark that actually says stuff about using Suno?

[rancid fiber]

yes that

[broken knot]

It's super depressing that v4 received mostly negative reactions from us users. Jeez, I hope Suno is paying attention to all the complaints and plans to fix the long list of issues... bring on v4.5 ... v4 is a dud. Not ready for prime time. I'm gutted.

[fresh oxide]

Once Suno can do this, it's game over! I won't ever need an audio engineer again.

We need it to be cutomizable

[worn flicker]

Just making sure. Sorry I don't have the answer, I was just making sure that the not watermark being called watermark earlier wasn't spreading.

[worn flicker]

The resources...my goodness...lol.

[fresh oxide]

It's full of artifacts

[worn flicker]

Could you imagine if they bumped up the quality and generated stems separately? The difference in resource consumption and then syncing back up for the rest if the generation...with so many users spitting out 2 at a time...I think that would contribute greatly to climate change in a single day...haha

[fresh oxide]

We'll have access to what the big record producers have kept behind a wall for years. We're already getting there.

[broken knot]

One issue of many. V4 has too many issues to keep track of. The fact that covers and remaster button are still both worthless and useless isn't good. v4 has been a PR nightmare. From the paid shills hyping it for two weeks to the actual release that went over like a lead balloon.

[worn flicker]

Yes but not generating things for earth.

[patent yoke]

Hmmm I am not really experiencing these issues, but im not a musical producer lol

[fresh oxide]

Nothing should be released from beta until it's ready.

[broken knot]

Hold your horses. We need lossless wav. 24-bit/96khz download options. Until we have HD audio downloads, this is just a toy.

Lol

[worn flicker]

192khz let's keep it up there

[fresh oxide]

True, the audio quality is muddy and can easily tell it's AI creaeted.

[worn flicker]

Gotta have that fidelity

[broken knot]

Sure, give us pro users options for either 192khz or 96khz as a 24-bit download. The current wav download is 16-bit and 44khz. CD quality... what decade is this?

[worn flicker]

That is why we have the drift we have now. That is what I mean by the insane resources it would require to pump out songs like it does and do the rest of that stuff at high fidelity

[fresh oxide]

Unless the source itself was recorded or synthesized in HD, it won't deliver true HD when downloaded.

Can't even fix it in post with stems

[broken knot]

Well, if the model isn't able to produce 24-bit/192khz then they need to stop everything else and fix that for v5. If v5 is still stuck at CD quality, then they can go suck eggs. I'll bring up HD audio every other day/week or until Suno responds.

[fresh oxide]

I'm waiting for them to add HD and then upsell it beyond their current subscription offerings lol. These companies always fk us.

[worn flicker]

That's right it would need to at least be a phase of the generation. Like resfix for SD

[fresh oxide]

Facts!

No idea why this isn't a priority. Maybe the audio engineers are fighting back lol.

[worn flicker]

Resources

I guarantee it

[fresh oxide]

I doubt they lack in that department.

[broken knot]

If v5 can massively improve upon fidelity while also bumping output options up to 24-bit then we will have quality outputs. 24-bit option is going to come to Suno or Udio first and whoever adds that will be ahead of the competition

[worn flicker]

You can't pump out AI Gen that fast without insane resources. I mean hundreds of chained video cards

[clear raven]

superresolution techniques with diffusion might solve the 192khz problem, not hard, the main challenge is data of that quality lol

[ashen carbon]

anyone know if Suno watermarks songs so that they can be detected as AI later on?

[worn flicker]

Then multiply it times a million people generating 2 songs at a time

[worn flicker]

Not if it is generated with Suno. Not any more at least. That is from a conversation with a Mod earlier this evening,

[fresh oxide]

Recording in HD lossless is just a software adjustment plus the extra server cost. They could do it tomorrow if they wanted to.

[livid sand]

getting weird clicking/beeping in the background of my v4 songs, at least with personas

[clear raven]

suno's training data mainly comes from spotify or youtube etc, 192khz isn't possible

[broken knot]

At some point, 16-bit/44khz is just going to be destroying the sound quality. V4 we're still in a state where CD quality might get the job done, but 16-bit /44khz limitations would be a joke a year from now or in July whenever they release v5. I expect it to actually sound like radio in v5.

If they can't figure out 24-bit and 96khz at least, we're all screwed and AI music will never be anything beyond a toy.

[worn flicker]

It is the amount of time that it would take to get that quality and pump it out 2 at a time in separate stems and then sync it back up for the full song that takes an entire different encoding process.

[ashen carbon]

wow such lack of patience lol

[worn flicker]

Well technology is exponentially increasing

[ashen carbon]

v4 is a marked improvement as far as I'm concerned and I'm an actual professional music producer

[broken knot]

192khz is probably unrealistic. But 24-bit and 44khz or 24-bit /96khz needs to happen.

[worn flicker]

I'm could be in a couple years, but at the moment they wouldn't be able to make that quality happen

[ashen carbon]

i've spend a couple hours and hundred or so credits and already have enough music to put out an album hahaha

[worn flicker]

Not at the scale they are at now

[fresh oxide]

192kHz audio is easily possible as a software and resource adjustment. They just need to do it.

[patent yoke]

@ashen carbon I'm curious, how are you finding Suno to be helping you with any of your unproduced beats, or beats in the archive, if at all?

[fresh oxide]

Until then it's easy to see where all the music is coming from.

AI identifiable in seconds.

[worn flicker]

It is the sample data that is hard to find, but I don't think it needs that high of quality to produce it. Two different things.

[clear raven]

it's easy to add watermarks technically, and it's also possible to train a classifier to distinguish suno songs from ground truth (even from udio)

[ashen carbon]

I've been generating vocals for my own lyrics for awhile now using Udio, but coming over to Suno now that v4 is out

[fresh oxide]

When I upload my vocals they are crisp and HD. Suno muddies the fk outa them.

[ashen carbon]

inspiration and vocals (so as to not have to use human vocalists)

[broken knot]

24-bit option has to be possible. 24-bit/44khz is a significant upgrade over 16-bit. I'll take anything that is an upgrade. Remaining at CD quality isn't a good idea. Not if this is meant to be a quality product. Suno or Udio might need a "HD music remaster" button but they better be looking into a future beyond CD quality. CD quality is trash. It wasn't even as good as vinyl when it was brand new. Lol

[ashen carbon]

I strip out the vocals and use them on top of my own originals

[fresh oxide]

Nothing beats a human singer.

[ashen carbon]

with v4 they still have some artifacts but way cleaner than ever

[worn flicker]

That could work

[ashen carbon]

Nothing beats a human singer sure, also nothing is more annoying than working remotely with a human singer

expensive, time consuming and annoying

[worn flicker]

That makes sense, like SD you can choose to go further if you like the sound

[ashen carbon]

AI singer is much more convenient, and in some cases, good enough

[worn flicker]

Then use the higher resources after cherry picking

[fresh oxide]

It's part of the process. It's all perspective. But remotely is a pita.

[ashen carbon]

It's possible I've just had really bad luck

but you spend money, then iterate a few times and then have to be happy with the result or shitcan the project

[waxen cosmos]

is there any idea i can generate a preset?

[broken knot]

At some point an AI model will provide vocals indistinguishable from people. Wen? Suno v8? Udio v3? Udio is on v1.6 still lol. Waiting for udio 2.0 and hoping it's a massive update

[worn flicker]

I haven't heard a Steven Tyler or Eddie Vedder. People still take the gold there.

[worn flicker]

That is also true, but it requires original samples so people still win

[ashen carbon]

I think what people are sleeping on so far for Suno v4 is the musicality of the vocals it produces

leaps and bounds above what it was before

[north sleet]

LLM models like Suno do not save audio generations in an audio format. Each generation would be saved in audio arrays consisting of symbols and audio tokens. When the user requests an MP3 or WAV file it is generated from the audio array on demand. I don’t see Suno offering anything higher than 16-bit/48 kHz for WAV in the foreseeable future as anything higher is too resource intensive on compute resources and the egress of several thousand 100 MB files per day would be too costly.

[worn flicker]

In this conversation I have an example of that, in 3.5 and 4. Going to the end of the song the belting is nuts, and she even rolls her R in verse 2

[zealous lintel]

I see a captcha almost every time I try to generate a song, I'm on my regular home internet, it's the same even if I change my IP.

[fresh oxide]

Possibly, but like voice-over, you'll never get the true human element. Anyway, Suno could easily enable HD audio output with adjustments, but it doesn't seem like a priority. It's insanely frustrating.

[livid sand]

guys i can't stop hearing this really faint clicking in every v4 song now. like a watermark maybe? It's so distracting for me now

[ashen carbon]

https://suno.com/song/b224c97e-bf92-4830-b009-925fff97b1f5

the vocal quality on this is exceptional

[fresh oxide]

Yeah, it's full of artifacts and not ready for consumers.

[worn flicker]

Not to be confused with a watermark

[livid sand]

much less AI noise than in v3.5 no doubt but I'll just stick with generating a few dozen versions and finding the best I think

[broken knot]

The best is to find obscure rare tracks from singers you love and then making new songs with the upload features. New Kurt Cobain, new Phil Ochs songs. I've made some new Annette Hanshaw songs. That was difficult. Taking the hd flac from the vinyl rip of a singer from the 1920s. Isolated and cleaned the vocals track, now I can make new Annette Hanshaw songs. And I'm about to do a v4 Annette Hanshaw song after I finish my current Disney movie style v4 song.

[worn flicker]

I am not where I can listen right now. Is it from the very start or does it happen as the song continues?

[knotty drift]

I think ReMi instead of lyrics gave me the kind of promting they use for generate the new lyrics. It was a weird output

[livid sand]

and i think I'll stick with v3.5 for a now. dunno though.

[worn flicker]

That is a lot of lawsuits...lol

[livid sand]

seems to be throughout the song, i can only hear when i listen loud

[broken knot]

Nah, I used a song so old and a vocal track so old it's part of the public domain. As far as Annette Hanshaw goes, that's 100% legal and something I can monetize. The Phil Ochs and Nirvana stuff is mostly just for my own tests/amusement at this point.

[worn flicker]

I'd have to hear it. But 3.5 has its issues but is still awesome

[broken knot]

But to bring back 1920s singers using ai is unheard of. It was really impractical and difficult to obtain an HD audio track of Annette Hanshaw singing. Those 1920s albums are noisy af

[crude jetty]

I think I hear it in 3.5 songs too, but it's faint, so it's not noticable

[worn flicker]

I meant for general use. All the things I am talking about are at whether Suno or other service could pull off the quality needed for production and stem separation at the start and then the complete song.

[broken knot]

But when I effectively made a new Annette Hanshaw song, I shed a tear. That was monumental.

[fresh oxide]

Paying for CD quality is insane!

[broken knot]

It's upsetting.

[livid sand]

i think i just got an ear for it all the sudden. 3.5 has enough "noise" that it's harder to hear. V4 cleaned up the noise but not the clicking

[fresh oxide]

I'm sure Suno reads these comments, so we'll see if they care.

[worn flicker]

3.5 is easier to control because the noise was on the high end

[broken knot]

No clicking on my Disney princess song. Thankfully, I think it depends on your genre

[crude jetty]

as in I think it's supposed to be there but not audible, a watermark, but a nearly invisible one

[broken knot]

Oh, we're half way there
Oh-oh, livin' on a prayer
Take my hand, we'll make it, I swear
Oh-oh, livin' on a prayer

[fresh oxide]

Don't be so cynical ,)

[worn flicker]

From what I can tell, it is still leftovers (drift) from generation. It seems to happen when Reverb would happen or clipping. Anything that is loud. It keeps the would be Echo as a ghost that stacks

[livid sand]

my thought as well but wasn't sure if i was just being paranoid haah

[ashen carbon]

uh-oh

seems I can't generate v4 anymore?

[worn flicker]

Not a watermark

[ashen carbon]

even if I choose v4 it says v3 in the generation. anyone else having this issue?

actually it says v3 for everything now

weird

[worn flicker]

I've read people saying this

Not me, i haven't had the issue

[broken knot]

If I can get an army of guys demanding to add 24-bit download options, maybe they can find a way. The Suno techs... "Grant me one request. Grant me 24-bit/96khz wav downloads! And if you do not listen, then to HELL with you!"

[drowsy finch]

it's a visual bug. the back-end says they're still v4 even though the UI says they're v3

[ashen carbon]

i think it's v4 even though it says otherwise

[ashen carbon]

nice

[fresh oxide]

@ashen carbon

[broken knot]

Suno v4 is like Bon Jovi live in 2024. I need v5 to be like Bon Jovi live in 1986.

We're halfway there

[fresh oxide]

It won't matter if you don't record in HD.

You can't just upscale it with a download.

[livid sand]

model name is "chirp-v4".... artifacts sound like quiet chirps... birds

[fresh oxide]

Video upscaling is easy and cheap but not audio.

Again, they can easily allow us to record in HD but the server cost will rise.

[broken knot]

The output or "remaster" option is going to have to be 24-bit... even if it uses creative upscale. I'm like Conan doing Right Said Fred on SNL at this point. I'm not too sexy for 24-bit/44khz... just anything... I'll take day old donuts.

They have millions of paid subscribers they can afford to go hog wild and take care of premium subscribers.

[fresh oxide]

The remaster can't upscale it that much. It must be recorded in HD.

It can def clean up some mud

But it's leaving crazy artifacts

So it's a wash

I'm sure the v4 remaster is still CD quality.

[broken knot]

Not "recorded" I think the proper term is that it must output in 24-bit. But if it can do a "creative remaster" then it might be able to take a 16-bit version v3 or v4 etc, then create a new output that is legitimate 24-bit

[magic axle]

The v4.0 model renders are reading v3 in green

[broken knot]

Due to expenses, it might be more feasible to have such a creative HD upscale output that generates legitimate HD audio rather than having it be the default

[frosty cypress]

Hey suno, we need merge now, doing stems and remastering stems gives really clean vocal and instrument tracks

[fresh oxide]

Upscaling/remastering 16 to 24 doesn't add any additional information to the original file.

[worn flicker]

It still requires more resources. Remaster may have been a poor choice of words because it technically would be remaking the song...ish

[fresh oxide]

Which means no HD

[drowsy finch]

ya, there appears to be some front-end issue going on. they are definitely v4 songs #📚┃suno-school message

[worn flicker]

HD has become a diluted term

[fresh oxide]

How so?

[broken knot]

16-bit will eventually hit a ceiling and stunt the output quality too much. That's going to be the case with v5. Limited to 16-bit, v5 won't be good enough.

[fresh oxide]

High fidelity better?

[magic axle]

sounded pretty sweet though

[crude jetty]

seriously bot..

[worn flicker]

Well as soon as they decided to use the first part of a resolution to name it than the second for marketing in video...lol

[crude jetty]

hd fhd uhd

[worn flicker]

Yup

Diluted

[knotty skiff]

@worn flicker you still on here bro?

[broken knot]

Yes, it won't literally be a remaster or an upscale, it's a new generation. Yet they would have to label it something that makes sense to noobs. "HD audio remaster" etc

[worn flicker]

Yes....I left for a bit though

[fresh oxide]

Lossless HD formats are considered "high fidelity".

[knotty skiff]

dedicated

[worn flicker]

My wife is pissed...I have a life...lol

[fresh oxide]

Which Suno does not have as it does not record. or generate in HF.

[worn flicker]

V4 is invigorating for good or bad...lol

[unique stone]

Just catching up here and wanted to remind you that uploading audio that you do not own the rights to is against the Terms of Service.

[broken knot]

Technically, lossless 24-bit 192khz might not ever happen due to file sizes. However, I'm going to keep pushing for 24-bit download options. And beg for 96khz. If they are dedicated enough, they'll find a way

[broken knot]

Thanks for the head's up. Again, technically I own the rights to Annette Hanshaw vocal track I uploaded. It's in the public domain as of this year.

[worn flicker]

I was saying that HD was diluted, that was an example. I didn't mean just in audio I meant the term started with 720p then 1080i then p and so on..,it got diluted

[unique stone]

I was referring to: "The Phil Ochs and Nirvana stuff is mostly just for my own tests/amusement at this point."

[crude jetty]

that made sense though. this new naming BS doesn't

[worn flicker]

Yes, once they said 4k I was done...lol

[broken knot]

I uploaded rare unreleased tracks that aren't under any sort of copyright protection. I'm good.

[magic axle]

https://github.com/haoheliu/versatile_audio_super_resolution

[fresh oxide]

You have to preserve the quality when exporting, but the quality can't be preserved if it's recorded in 16-bit/44.1 kHz.

[unique stone]

You do not own the rights to them though, which is a requirement of the Suno ToS.

[fresh oxide]

If this works I'd be blown away!

[worn flicker]

Right you can use AI to up the res though, it is very good at inferring in areas that are that close together

[broken knot]

If it's an unreleased rare demo, it's public doman, there are no rights to the 1975 demos that Phil Ochs made at his friends house. Etc.

[crude jetty]

technically that doesn't matter, the ToS says things YOU own, specifically. just because its in PD doesn't mean YOU own it

[worn flicker]

It is Suno rules not other rules

[broken knot]

Public doman is literally something we all "own" I bet you a million dollars

[worn flicker]

It doesn't mean suno has to accept it

[magic axle]

you'd need to chop the tracks up probably to run on a local vram config.. then reassemble. I think the 4090 handled max 20-30 sec at a time

[torpid thicket]

Why is everything going to V3?????

[worn flicker]

They have to hold off enough lawsuits

[worn flicker]

Visual bug

[torpid thicket]

[magic axle]

or write a script to batch it and reassemble

[torpid thicket]

I'll get the bug spray

[fresh oxide]

Audio doesn't have resolution; only video. Audio has sample rate and bit depth.

[broken knot]

Suno isn't coming after guys uploading public doman material and making private songs that aren't meant for sharing or public release. I'll be very careful.

[hollow hare]

Previously, when some old v3s got remastered, they got relabeled as v4. I can only surmise they were trying to fix that issue when this happened

[worn flicker]

I know, I was trying to shorten the sentence not play nerd war...lol

[unique stone]

You are announcing that you are doing something against the ToS in the official discord of the service though...

[thin flint]

How are people remastering V2 songs? MattVidPro did it with a southern Christmas song.

Remaster option isn't appearing on V2 tracks when I hit the three dots.

[worn flicker]

To be specific, if there is one thing that current generative does well is infer content on short runs, and the distance between notes is pretty short

[broken knot]

It's literally public doman, so it's not. Public doman means we can monetize or do whatever we want with it. Thanks for your concern. Public doman means I own it am allowed.

[crude jetty]

I own it too, right? then you're not EXCLUSIVE

[worn flicker]

That is a mod

[unique stone]

"I own or exclusively control all rights"

[hollow hare]

Guide* DAViD here is a mod

[broken knot]

It says own "or" not own "and"

[worn flicker]

I bow out now

[hollow hare]

https://cdn.discordapp.com/attachments/1069381916492562585/1256404293267882094/Screenshot_2024-03-05_233857.png?ex=673e7d2e&is=673d2bae&hm=1f95130bebda07b1e08c4504509269f05e8df73d9df2b1aab1060c78c746fc5a&
https://cdn.discordapp.com/attachments/1069381916492562585/1267223169979256913/sunoaudiouploadtos.png?ex=673e4c0b&is=673cfa8b&hm=5cdfef54295bdd481597ff740174f8dc8619de1742643aa8c7cf2b3b47bca2bf&

Note the hold to rights part

Regardless, we have done our duty and warned you

[fresh oxide]

Generative models are great at creating content with limited complexity or shorter liek you said.

[broken knot]

Nowhere does it say you can't use public doman material. I technically "own" it. I don't need exclusive rights. It says so right in the terms. I have an lQ above 120. I think I can understand the Tos

[lean pebble]

Is it just me, or I cannot longer change model?

[broken knot]

I'll probably switch to mostly using Udio for uploads soon anyway, since Suno v4 is less lifelike than Udio v1.6

[fresh oxide]

Yeah, because you could upload a real singer's audio and then use his voice. Not cool.

Leaves Suno open to a major lawsuit.

[crude jetty]

when it says "you have exclusive rights" that defines PD very clearly out of the question

[fresh oxide]

Does Udio record in hi-fi?

[hollow hare]

A person who states they have an IQ above a threshold often does not.

Although you are fine using lyrics from public domain works, though ensure it is the original and not from a derivative of the work, the audio recordings may itself still be copyrighted (depending on the year that the audio was made amd potentially what power the estate may have too).

[ashen carbon]

you mean like the one they're currently in court with haha

lol

[broken knot]

I don't think so. There's no true HD AI music yet

[broken knot]

I'm Ashkenazi Jewish and took an lQ test. Around 120. Lol, Ashkenazi iq gap between everyone else is massive. I'll get a Jewish lawyer in on this.

[crude jetty]

which won't help you because it's based on US law

[broken knot]

Study the bell curve. Lol, Jewish lawyers work in the US and are considered the best lawyers money can buy. They're not exclusive to Israel, if that's what you think.

[fresh oxide]

When you upload your song and it's already been run thru your process chain Suno shouldn't mess with it just to add music.

[ashen carbon]

getting a lot of this all of a sudden

[broken knot]

I'll be more careful in the future and if it's not public doman 100% and clearly, I won't upload it. No worries.

[fresh oxide]

This happens when you get too specific with an artist.

[ashen carbon]

oh i had no idea

so you can't put artist names in the prompt?

[fresh oxide]

Yup

[ashen carbon]

(i'm a Udio refugee at the moment)

[fresh oxide]

No, that's theft.

[ashen carbon]

that's theft?

[worn flicker]

Just leave it at terms of service

[fresh oxide]

Yes, you're taking their music and remixing it.

[ashen carbon]

so generating music based on their music is not theft, but mentioning their name is... got it

[daring lava]

Thank you

[ashen carbon]

heh

[magic axle]

Woot they fixed the v4 icon already. Nicely done folks...

[worn flicker]

This is going bad fast

[fresh oxide]

For legal purposes it is.

[ashen carbon]

jk i know what you mean, it's just funny because it's ALL theft

[worn flicker]

Not theft

[ashen carbon]

like this is the most THEFTIEST THEFT IN THEFTING HISTORY

[worn flicker]

Lol

[fresh oxide]

All AI is theft lol.

[ashen carbon]

ok ok

sorry

[broken knot]

I can't imagine being udio only or suno only. If you don't extensivly use both then I question your commitment to AI music... Haha lol.

[worn flicker]

There are obviously general licensing.

Stock photos and the birthday song they finally gave up on

[fresh oxide]

Until it surpasses us which I'm betting it has. Whatever we have access to is nothing compared to what governments have.

[unique stone]

Huh. That seems wrong. It'll usually refuse to accept artist names and suggest describing what you want instead. Using actual material that you don't hold the rights to as an input is a problem - attempting to put in an artist name is protected against and even if it goes through it probably won't give you what you want.

[crude jetty]

I only use suno. I only use suno because that's the platform I was introduced to and care the most about using

[worn flicker]

Now that is a guarantee

[ashen carbon]

yeah I don't actually think the model knows the artist names

[soft owl]

real question though: if i learn guitar and write my own song, and someone later generates a similar song in suno, am i in jail?

[fresh oxide]

Do they offer the same features?

[worn flicker]

It doesn't just use the model there is another piece that does that

[fresh oxide]

No

[worn flicker]

Look at Bark on github. Look at the components, Suno began there and moved to closed source

[fresh oxide]

And music is civil issue rather than a criminal one lol.

[worn flicker]

There are at least 4 pieces happening when you hit the generate button

[soft owl]

but it sounds like a suno song. and their TOS says that my song cant sound like one thats their artificially intellectual property

[fresh oxide]

They can't make any money of criminal charges.

[broken knot]

But you can use Udio to get a decent remaster of a Suno song and Udio can do 100s of genres and subgenres that Suno absolutely can't do. So it's a no-brainer for me. The best Suno remasters use Udio "remix" feature not Suno "remaster" feature. Unfortunately, because switching tracks between the two and using an audio editor to splice shit together is a massive pain in the ass.

[worn flicker]

Now that is something I have an example of. I thought the same thing

[fresh oxide]

That's retarded and unenforceable.

[worn flicker]

Hold up

[ashen carbon]

whoa the R word!!

😄

[worn flicker]

I actually have an example

[soft owl]

even if i made it first? did suno violate my own TOS then?

[fresh oxide]

Is that a bad word or something? I come from X where free speech is back!

[worn flicker]

Of the instrumental being flagged

[ashen carbon]

ok magat

[fresh oxide]

These are ridiculous questions.

[worn flicker]

I played on a keyboard myself into FL Studio a long time ago. It was based on Super Metroid

[soft owl]

i just wanna make sure i dont get in trouble is all. suno sounds litigious

[worn flicker]

I uploaded it to Suno. And extended it

[fresh oxide]

Unenforceable

[worn flicker]

I hit public and it went for a little while a few minutes

Then went back not public

[broken knot]

Why can't Suno do so-cal punk rock, Seatle grunge, or Wall of Sound? Why can't it do songs with both vocals and quality orchestration. These sort of things depress the fk out of me. So I need Udio. What happens when you want to make something that isn't a modern style, you're gonna need Udio for that. Asking for Suno to do 90s Seatle grunge? Might as well ask for pigs to fly.

[worn flicker]

If anyone pays attention I am answering that question...lol

[soft owl]

but like, if I use chords C-G-A-F in a chorus but then someone else makes a suno song that does the same, who owns those chords? i wanna make sure im not illegal

[molten dawn]

nobody has a patent or copyright on a chord or progression. the issue stems from similarity or likeness.

[broken knot]

In many ways, Suno is the better one. But in sound quality, fidelity, and genre accuracy, Udio chops Suno up into pieces

[worn flicker]

You won't get in trouble, you won't be able to make it public, it will stop you from the start if it breaks the rules most of the times

[soft owl]

how similar is too similar? im worried

[molten dawn]

that's for others to decide

[soft owl]

should i use different chords just to be safe?

[worn flicker]

Hey, I am telling you what happens

[fresh oxide]

These aren't serious questions. Why do you keep asking them?

[ashen carbon]

seriously

boring

and stupid

[worn flicker]

Now read above if you want the answer

[soft owl]

TOS sounds pretty serious

[worn flicker]

Now your trolling

[fresh oxide]

It's just unproductive noise. See you guys later.

[ashen carbon]

yep bye

[crude jetty]

I'm tempted to try using a v4 vocal track and the instuments from v3.5 and see how that turns out

[worn flicker]

Later I'm out too

[broken knot]

People who know the logic of Suno techs... why don't they let us create genre accurate songs? If you want so-cal Punk, or Seatle grunge or something, it just can't do it. Suno hates subgenres.

[soft owl]

im not sure where theres a disconnect here. just want to make sure my song that i wrote is safe to play, cause if someone makes a similar one on suno i could get jail right? or do they jail?

[crude jetty]

if it's not used commercially, it's not any kind of an issue

[broken knot]

But I put "punk" in udio and the results sound like The Germs.... that's what I'm talking about. Udio definitely trained on the Germs, Suno did not. 😥

[soft owl]

i dont know how to get in a commercial

[crude jetty]

in basics, it means you can make money off it

[broken knot]

They promised us better outputs. They delivered worse outputs. "I wanna cry"

[soft owl]

but if i write the song BEFORE someone makes it suno, cant i make money off it?

[crude jetty]

the vocals are top-notch, it's the intrumentals that have issues

[crude jetty]

It's not the same law is very clear on that

I like how my v4 song can get remastered. v4 song to v4 remaster? huh?

[broken knot]

Unless you're comparing it to Udio v1.6. Udio has far better vocal and instrumental fidelity and quality than Suno, but Suno makes catchy songs far more frequently. So my new process for the foreseeable future involves remastering my Suno songs in Udio. I shouldn't have to do that, but here we are.

[soft owl]

i just wanted to consult the lawyers here to make sure. so if i write an original song, can someone then make a song similar to mine in suno and claim they own it?

cause i worked hard on it

[loud copper]

can anyone point me in the right direction to talk to a staff member.....all of my generations are stuck at 3:14, i cannot get a new song to generate past that

they are abruptly ending mid word

[crude jetty]

that would be a legal question and I have no legal experience to fall onto

[broken knot]

File it under the bug reports page, bro. I'm sorry to hear about your issue.

[loud copper]

i have, its been like this for a week now

[molten dawn]

[e-mail]

[soft owl]

yeah if someone heard me play my song at a bar and tried to mimic it in suno, do suno TOS prevent me from ever playing it again?

[broken knot]

Email support too. They don't fix things very fast around here.

[molten dawn]

you know how annoying you are? no one here is a legal representative of suno

[daring lava]

Are you using a persona?

[soft owl]

ive never been told that

[loud copper]

yes

[daring lava]

That’s why

You’ll need to extend from there

[loud copper]

is that intentional?

[soft owl]

i saw a lot of copyright stuff in here, just wanted to avoid the jails. does TOS say anything about guitar lessons? just worried is all

[crude jetty]

no, it's not, but that seems to be what's happening

[daring lava]

It has a limit currently

[loud copper]

i guess thats why it threw me off, its a random number

thanks for the info tho

[daring lava]

Happy to help

[soft owl]

Does Suno pay royalties?

like if i make a really good one

ahh i understand, so "Music made entirely with AI may not qualify for copyright protection because a human did not write the lyrics or the music. Writing the prompt does not constitute the creation of the song". so everyone saying these songs are theirs are jail?

why AI doesnt have rights?

[hollow hare]

Different jurisdiction have different laws and, well, many are going to change within the next decade.

[crude jetty]

because AI can't recate only remix

[soft owl]

which jurisdict does AI belong to

cause im worried i live nearby

[hollow hare]

Well, your own is the most important one

And like the other is where the service resides, which can affect matters, so USA (New York and Federal)

[crude jetty]

so mixing v4 vocals with v3.5 instrumentals works well

no more ticking!

stem v4 stem 3.5 use external tool to combine them

[hollow hare]

There is some precedence in actual cases about copyright and the key rule against registration/protection of AI works with US Copyright Office.

More specifically, the clause about 'providing sufficient human effort' (paraphrased).

1. The person with the ape selfie could not copyright the photo because it was not a deliberate effort on their part and, well, it was not taken by a human. So, yeah, there is a precedence there due to non-human involvement

2. Zarya of the Dawn is a comic book that involved AI image generation work. There was a case against the author who lost her copyright protection against the artwork itself because it was made with AI and 'did not have sufficient human effort in its creation'.

The end result is that, while she cannot protect the likeness of her characters and specific scenes of her comic (IIRC), she had still been granted copyright protection for her works title, character names, dialogue/plot/etc., and the arrangement of the work since that required manual effort.

The funny thing is that of the person did used that art for tracing, she likely would have copyright protection...

I touched more about that and provided links at #📚┃suno-school message (at least I hope that is the right link, I had a few involving the same topic)

[soft owl]

but these IQ musicians above said they copyrighted their musics from suno. is that create, or remix?

[hollow hare]

Sorry for the wall of text

[soft owl]

ah that makes more sense. so as long as i write the song 100% myself, without using suno, im not jail?

[hollow hare]

It's not about jail you need to worry about, it is about being sued.

When people submit copyright, more than one part could be protected (or denied). So writing your lyrics can be registered for protection right now but you cannot copyright protect the Output generated by Suno at this time in US Copyright Office

[soft owl]

i just worry eventually someone will make a suno song with similar chords and now i cant play my song?

[solemn lintel]

I've been playing with v4... I actually don't like the songs it gives me. They are consistently worse-sounding than 3.5.

[crude jetty]

as long as you're the sole creative for some part of it. I.E. you wrote the lyrics yourself

[soft owl]

ah i see. so copyright my lyrics so someone cant use them in sunos?

[hollow hare]

To give you an idea of why people want copyright protextion: each infringement can be sued up to $150,000 USD in US Court

[soft owl]

oh wow! so if suno uses too much of my song, do i pay them those or do they pay me those usds

[crude jetty]

vocally it's MUCH better, the intrumentals need work

[broken knot]

It's all we've been discussing all day. V4 was a misstep. Crossing our fingers 4.5 is decent in like March or whatever. Crossing our fingers for some faster updates, but not holding my breath.

[solemn lintel]

I guess audio quality is better, but it just doesn't stick to my personas anymore, and the tunes are simply not to my liking.

[solemn lintel]

Yeah, sorry, I didn't sift through all the messages... guess I should've.

[crude jetty]

I've not had the persona issue myself, and I use personas

[hollow hare]

Er, not quite. According to the TOS, you did grant Suno all and irrevocable rights to your Input. (They are not likely to use it but it is written there).
I can link you to Top Music Attorney's video about Suno TOS about that portion if you want.

Anyways, I believe there is a clause somewhere that basically state you allow others to use your Input if you make the song public or share the link.
And Suno does not guarantee that the Output will be unique to you

[broken knot]

You're fine. It's more like "welcome to the club."

[solemn lintel]

Woo hoo.

OK, as long as I'm not alone. I'm going to continue with 3.5 for this current project, I think.

[soft owl]

so if i put my song on spotify, someone else can use suno can use it to mimic it and then say its not mine anymore? i dont have that many dollars to pay them for the fine

[broken knot]

Then use udio to do your suno remasters? That's what I'm gonna do.

[solemn lintel]

Yeah, I was thinking of doing that. I just have a LOT of songs.

[jovial otter]

Help! Is this a bug? When I try to EXTEND, it does not extend by the 2min default. Also, the song is completely changed and seems like version2 low quality although it says v4. The EXTEND only create another bad quality 2 min song instead of EXTENDING it to 4 mins. Anyone else getting this issue?

[solemn lintel]

I wonder if it cleans up little glitches, too.

[solemn lintel]

I get all kinds of weirdness with Extend all the time. I just keep re-generating it until it sounds OK

[broken knot]

The v3.5 stuff still needs remastering and as long as Suno dropped the ball on it, I'm glad I might be able to do some Suno "remastering" using the "remix" feature on Udio. As wild as it sounds, apparently the method works.

[soft owl]

should i play my original song at smaller bars so suno cant hear it?

i cant afford the fine

[hollow hare]

You will have to dispute it with the distributor. It will not take money in the first instance, just lots of time and headache.

For example, Feline Music had their top-hit Cat, back when Suno homepage had a single trending/top list, stolen at least 22 times before their distributor distributed it #🐣┃suno-showcase message

• and at least 3 songs stolen from Feline Music too #💬┃general-chat message

Spent months trying to get the others down (and get the royalties earned by them)

[hollow hare]

I don't expect Suno be busting your doors down about that lol

[jovial otter]

V4 just came out so saying it happens all the time does not make sense. Also, If I just hit extend mutliple times, this costs me credits and I am not rich.

[soft owl]

did they generate the song originally in suno, or play it on real instruments using their own premise? cause i dont have Suno yet, so i dont know that i have the TOS downloaded on my personal computer yet

[broken knot]

Overwhelming negative reaction to v4 release. I wonder if Suno knew this would happen but just said "screw it"

[solemn lintel]

Sorry, I meant in general, not specifically for v4.

[knotty spear]

Same thing happened to pika a while back.

[broken knot]

Did Pika redeem themselves? Or is it abandoned?

[hollow hare]

But if you do cover songs, make sure you cover your ass and have the appropriate licenses to play those cover songs (see https://dittomusic.com/en/blog/what-license-do-i-need-for-my-music/)

Normally, the bar you are playing at should have that license covered IIRC because they need a license so they don't get sued for allowing copyright song played

[knotty spear]

I can't get any generations, even when I was paying. I asked them about autoscaling and they laughed. Quickest refund I've done.

[soft owl]

but what if i took lessons and sat down and wrote a song myself? how close AI can mimic before its not fair to me? i just cant jail cause of school

[hollow hare]

The songs generated by Suno should theoretically be Suno's own since the work is transformative, but it does not mean your work may be like that.
By chance, you could get a copyright tune and... yeah. Just play the chords in like Shazaam and/or upload privately onto YouTube and they will check Content ID to see if it already exists

[broken knot]

Suno never does v4.1 etc. So we're probably screwed for the next six months while they scramble to fix everything v4 broke.

[hollow hare]

I am telling you, you will not be jailed. Just sued in the worst case scenario if you are playing someone sample or copyright work without the appropriate license

[broken knot]

A lot is riding on v4.5 now and a ton is riding on Udio v2.0 which is probably coming soon.

[soft owl]

oh ok phew. so make sure the song i came up with in my head doesnt sound like previous song, and then i copyright the portions i can copyright? does suno sue me if the computer likes my song enough to mimic? or do i sue suno for mimicking my song?

[jovial otter]

How can they enforce copywrite here in Japan. We have laws here to allow us to use other songs.

[knotty spear]

I can't use it outside of remasters. So I'm not even sure what's going on. Hopefully they squash bugs. My feedback hasn't been touched. :/

[hollow hare]

If you get to the point of suing Suno for something you derived from their Output, good luck with that lol

[broken knot]

Enforced by asking nicely.

[soft owl]

im not much of a suer, but id like to be able to play the song i wrote about my grandpa

[soft owl]

but i didnt derive from their output, i wrote it without computer

[hollow hare]

Anyways, I am not a lawyer despite all the digging I did. If you are concerned with the legal aspect, get in touch with an actual legal representative.

[soft owl]

but does using a mic to put it in computer and then spotify make it the computer's property to then use in suno to mimic?

[soft owl]

i dont use suno, but im worried suno will take my song from spotify and change a few words and claim its the computer's song. and what about my grandpa?

[carmine flare]

almost all the v4 songs i do has this weird choppy hissing sound.

[broken knot]

V4 was a mistake. But it means that the odds are good that v4.5 will be good.

[hollow hare]

You will need to have an actual copyright protection filed btw to be able to sue on the bases of suing someone for using your work without permission and prove they have the intent of doing so.

RIAA is already suing Suno and Udio for what you are talking about and the latter companies argument is fair use and it is transformative in nature, not derivative, which does not violate copyright. Anyways, you can find out more by looking up those particular court cases

[broken knot]

There might not be a fix for v4. It might be too far gone. I've moved on to waiting for 4.5.

[soft owl]

"transformativ" instead of "derivatve"? thats confusing.

how much transforming makes it new? if suno makes the song about my grandma instead of my grandpa, do i owe the USD? can i bring guitar to jails?

[hollow hare]

Not really. Transformative means that the work taken is used intermediately and does not resemble the end product while derivative is... well, derived, and the end product has some relation to the original source.

A 20 year old TikTok rapper got successfully sued for $800,000 USD by Sony for excessively sampling a copyright 1987 instrumental song without a license for example, as well, his sample beats was obviously derived from the 1987 song (Back in March or May iirc)

[soft owl]

yeah sounds like he stole song, he should jail. so resemble = tranform, but copy = derive?

[hollow hare]

Not that simple and something you will need to consult a legal professional as it is heavily dependent on the local state and federal law

For example, you can run across ELVIS Act in Tennesse but not in Ohio (though I would not be surprised if ELVIS Act becomes federal)

[soft owl]

ah understood. thank you for your lawyers, i will call you once im ready to copyright. how much?

[hollow hare]

Like I said, I am not a lawyer. You can look that information up yourself. Was a few hundred for registration iirc?

[soft owl]

you work for the IIRC? are they music lawyer

[hollow hare]

Lol. Iirc = if I recall correctly

[outer cosmos]

Including me mostly everyone is unhappy about v4

[soft owl]

ah understood. i will call them tomorrow to protect grandpa's song. thanks for all your help!

[vapid nymph]

What are the issues ? I haven't had time to mess with it yet.

[broken knot]

What aren't the issues? 😆

For starters, v3.5 makes more compelling tracks and v4 can't remaster them without gutting them. Furthermore, there's a lot of weird noise on most of the v4 tracks that isn't supposed to be there. For starters.

The general consensus is that v3.5 wipes the floor with v4

Suno is already known for only doing like 5 genres. With the update it's down to 3

Some of the biggest Suno contributers are already sworn off v4 and everyone is waiting for v4.5 and hoping for massive improvements and fixes.

[vapid nymph]

Oof. Rip

[long spruce]

Hi guys, I started to use v4 for the first time and noticed that word pronunciation is so much distorted. Have anybody found a solution for this?

[hollow hare]

If the pronunciation is off, do phonetic spelling
Toe-may-toe vs tom-ehh-toe

[molten karma]

why can't i switch back to v3.5?

[hollow hare]

Certain functions likely had the backbone rewritten to be compatible with 4.0 so yeah...

[long spruce]

This is mostly happen with sinhala songs. I didnt try english

[outer cosmos]

same here. 3.5 much better compare to 4

[hollow hare]

It was an example. Suno is a predominantly American company and cannot look at every language.
So yeah, you will have to.do phonetically spelling for sure

[long spruce]

Did you contact anyone from suno?

[hollow oxide]

yeah even in english im finding to many mispronunciations, ruined alot of takes so far im thinking of makingthe song in 3.5 then remastering into 4.0 to help it.

[warm parrot]

I find adding meta tags like : [CD Quality, Crisp Clean Audio, Large Room Environment, Reverb] at the top of the displayed lyrics BEFORE remastering can help it come out better

[fathom siren]

Just spell phonetically, I don't think the remasters are nearly as good as the original generations so far.

[warm parrot]

This is crazy, my ethnic language Albanian got vastly improved, it's pronounces everything correctly now.

Crazy that English supposedly got worse.

[outer cosmos]

Suno should declare alpha users name who give them green signal and good quality feedback.

[limpid spade]

Alpha user feedback mirrors most of the community reactions I've read so far. Alpha users have no say on release timing or authorisation.

[languid junco]

You're certainly on the right track here.

[warm parrot]

👀

[cerulean merlin]

i don't know how they made suno v4 but suno v4 is messing with me

its like its playing tricks on me on purpose

[azure mountain]

I haven't had good luck with remastering on V4, but I have had luck with creating new songs from scratch with it. It does require re-learning metatag prompting to get it right

V4 is actually REALLY good when it does work

I tried that and it had no effect, so might just be a placebo?

Metatags don't seem to have any effect on remasters, but do definitely have an effect when making completely new songs

I suspect the remaster doesn't work in V4 for the same reason I've heard it didn't work for other versions -- the newer model has a fundamentally different "personality" to it, so older songs just don't sound right often with the newer model.

[cerulean merlin]

who here is good with voice meta tags i need help please

[azure mountain]

Yeah admittedly remaster seems like a bit of a failure (as it only seems to work in a few cases)

[azure mountain]

What are you trying to do?

[cerulean merlin]

a duet

[azure mountain]

[Chorus: Male and Female Vocalists] seems to work for V4

[cerulean merlin]

nah 😦 im not doing that its diffrent for me what im trying to do

[azure mountain]

Here's an example with several different vocal changes: https://suno.com/song/38409153-15a1-4fdb-a8fa-6ba09ccba939, any of that what you're trying to do?

I've found that V4 doesn't seem to align the position of the metatags correctly btw, so sometimes you have to move the tags up or down in the song once you've figured out the offset.

[cerulean merlin]

see thats is wrong

[warm parrot]

.... because it uses the lyrics to help with the remaster...

try removing your displayed lyrics entirely before remastering, and watch it completely loose the plot.

example here: https://suno.com/song/23423033-ae13-41ed-b14e-de5d9377eaf4

Removed all the lyrics, it goes crazy.

[azure mountain]

Remaster (in progress now) with completely removed lyrics (although this was an instrumental track to begin with): https://suno.com/song/33434da4-549b-4e4b-8694-dd8b5c68d1d4

Doesn't seem to have lost the plot

[warm parrot]

"although this was an instrumental track to begin with"

The point of the experiment was to show what's in the lyrics effects the remaster

so ADDING meta-tags should help too

[azure mountain]

Here's one with lyrics: https://suno.com/song/7d716ded-338b-4e6c-9fa9-2939cc0abf82

[cerulean merlin]

im making a song about man loves winter and woman loves summer, the first verse is male first but no matter what i did it always generated the female voice NO MATTER WHAT I DID!!! then i got soo sick of it, i just change the vocies so the woman loves winter and man loves summer, and that should make the ai make the female to start but NOOOOOOO it wanted to start with the male i dont know how suno v4 got made but its like its playing tricks on me

[azure mountain]

Maybe try putting female voice first in the prompt instead of male? It often mixes up genders

[warm parrot]

#📚┃suno-school message

idk what to say man, on my side removing the lyrics makes it loose the plot

[lusty oasis]

It flat out refuses to follow vocal meta tags for some reason. ignores them worse than v3.5 for sure

[cerulean merlin]

read what i said

[azure mountain]

Trying a few things...

[cerulean merlin]

i gave up

[warm parrot]

waaaaait a minute, you can't actually fully remove the lyrics. That's why mine has a period still.

Are you purposely being misleading here?

[azure mountain]

No, this song has no lyrics: https://suno.com/song/c2122982-46b4-4f46-970d-16cea94ebdb2

Not even a period

[warm parrot]

🙃 I don't think you understand

[azure mountain]

Do you mean edit displayed lyrics or something else?

[warm parrot]

yes

before remastering

[azure mountain]

Yep and that's what I did with that song

[warm parrot]

so if my song looks like this before remastering

you physically CANNOT save if it has NOTHING in the box

so if you really did remove the lyrics, you'd need at least a period there...

[warm parrot]

okay so that's the difference here, it NEVER had lyrics...

[worn flicker]

3.5 did the same thing. You can read in this chat if you felt like it but the whole writing in sections and or writing non gendered lyrics has come up many times because for whatever reason exactly what you are talking about was happening. In 3.5, I haven't tried in v4 but I assume it remains...lol

[cerulean merlin]

bruhh suno 3.5 worked first try

[warm parrot]

different model, different way to prompt

[worn flicker]

Didn't say always, just that it did

[azure mountain]

Is this what you wanted? https://suno.com/song/6df626f3-f150-4075-9424-8fb5bd212010

[worn flicker]

I know because i was there...lol

[karmic pagoda]

i can't generate songs on suno

its just spins

[worn flicker]

I love duets, and that is what drove me up the wall, so I would swap it in tags and then style then both. It did exactly what you are talking about. Simply put, I feel your pain just not in v4, haven't tried yet.

[warm parrot]

ah I see you had a space, that works.

interesting it doesn't mess up your song

it will consistently mess mine up

[azure mountain]

weirdly if I click "edit" it doesn't show a space, it shows no lyrics at all

[karmic pagoda]

i actually need help tho

[worn flicker]

We are just users

[azure mountain]

I suspect a lot of us are starting to develop confirmation bias (probably including myself) as we learn the new system

[worn flicker]

This is broad so no one answered

[warm parrot]

look I'll do 6 remasters of an english song with the lyrics removed...

I guarantee everyone will become gibberish

[worn flicker]

v4 has issues. I can not defend it. 3.5 had drift but over time. V4 can start with drift (timing issues), so I can't use the same controlling things I used in v3.5. There you could break up the song or change the Tempo to get the timing check to kick in. v4 I am stumped because it starts that way. Nothing to start with to chop out.

[warm parrot]

Here we go, every single one is chock full pronunciation errors.... which does at least prove editing the lyrics does SOMETHING.
https://suno.com/song/47b8ede8-4511-4fac-a2ac-8c37214684cb
https://suno.com/song/cc426d2f-4bdd-4d9a-a755-4cd59e7a18cb
https://suno.com/song/04026dbd-c433-47fb-a05e-43e5f2ea1169
https://suno.com/song/8a3bb170-3fac-415f-a241-c18d6f4d0a2a
https://suno.com/song/dc3ba174-ab49-4267-b29c-2cec78f1a874

original v3.5 https://suno.com/song/47ebb84a-77a3-4ed7-957f-5fbc41c8ae99

[open sleet]

Well this is quite odd

[worn flicker]

I know this wasn't my conversation, but lyrics are heavy as they have been from what I can see. They are insanely heavy compared to 3.5

[azure mountain]

Yep I have definitely noticed that and haven't found a good fix for it yet

[open sleet]

Apparantly my account page can't be found, can search it but it brings me to that when I try to access it

[worn flicker]

Well the issue seems to be something to do with timing. And being overly concentrating on vocal quality over timing. That would make tags much heavier and cause even more drift if that is what is going on. I doubt they trained the model on trained instruments, I doubt they made it to v4 without knowing the basic copy of a copy is bad.

[warm parrot]

remaster with the lyrics intact: https://suno.com/song/b23096b2-df98-4ab8-b91b-af9bc8de78d8

no gibberish

[azure mountain]

Vocal quality was the main issue in 3.5 so I could see them overcorrecting for that in 4.0

[worn flicker]

Bad thing is, I don't think there is a fix. They will need to either do a post processing patch or something to adjust weights. Otherwise this wobble effect will not leave us with usable songs.

[worn flicker]

Exactly, a logical conclusion

[azure mountain]

Adjusting weights is likely very expensive for them and seems a bit impractical immediately after a release, so I'm guessing they'll have to correct for it in pre-processing (converting inputs to embeddings) or something. Post-processing doesn't seem like it would be very helpful when the model outright omits instrumentals where it previously would have included them (in many cases).

[worn flicker]

Yes, I was going from fastest implementation to slower. Right post processing is more than likely what is exaggerating the wobble in the first place.

[long spruce]

Seems the problem is with many languages including english. If new version is coming soon it would be great.....

[worn flicker]

It only goes so far before getting to a full model retrain or a hell of a lot of LoRA style corrections and that doesn't seem like a smart idea since it needs so many more resources

Pre processing, could be it since now the "wobble" drift can start from the first note instead of building over time

[broken knot]

Do most of you guys consider yourself songwriters, or are you just plunking in random lyrics and messing around? Do all of your songs contain the words "echo" and "neon" because ai writes your lyrics? Not me, my lyrics are the whole reason I'm into this. Ive always been a lyricist. How big of a setback is v4 Beta gonna be for us? When do you think they release the final v4 version? Maybe with the public free user release. Do you guys guess that they will take longer to release the final v4 version now, since this is "beta v4" and before today, they had the alpha version. So. Maybe v4 can be redeemed?

[stiff plover]

Ok so I'm gonna write a quick tip here. Let's say you got a REALLY GOOD BANGER. but you want to change lyrics just a bit. So you either replace a section but it sounds different... You then make a persona and change the lyrics a tad, but it overfits and sings garbage. What I did is crop the song a tad, so that some or most lyrics are missing. Then create a persona, then amend the original lytics. Then use the persona you created from the crop, and it will retain the style very very well, but it wont' garble the amended lyrics.

[stiff plover]

I started writing my own lyrics because all AI make supershitty lyrics

[broken knot]

The next version is final v4. It goes Alpha, beta (released today) then final v4. They better delay the final v4 release and fix this shit.

[unique plaza]

why other application like suno ai are down ? brev ai, remusic etc ?

[outer cosmos]

3.5 is ok. 4 is bad

[outer cosmos]

Alpha user giving positive feedback so they released beta. Salute to alpha users

[broken knot]

I write my own lyrics, but sometimes, these days, I often utilize llm to help, but dear God you can't let it actually write anything. Even the newest ai lyrics, released today by suno, it's very goofy.

[worn flicker]

No idea on the time frame stuff. I write my own lyrics unless I am testing something but I will use AI for structure. I am more of a wordsmith than a lyricist. I wrote a song about it...lol. https://suno.com/song/52f4d3a4-6ed2-477e-b517-57e21492441c

[stiff plover]

I use claude and it takes hours to iterate on anything that doesn't sound like a first grader wrote it that is unsingable.

[warm parrot]

Made a V4 feedback thread about this whole thing
#1308679699723194408 message

[broken knot]

No no, in spite of what they want us to believe, only the paid promoters get access to the alpha. They don't determine anything. They simply hype it like mindless drones in exchange for piles of cash and 50k free suno credits per month unlimited etc

This isn't fact, i don't have evidence or proof

It's how I assume this works

[unique plaza]

but they paid to suno for api and or purchasing the palns to use api right ? then why

[worn flicker]

Many people you saw in this channel were alpha testers, so they will correct you I am sure if you are mistaken...lol

[cerulean merlin]

i got it https://suno.com/song/c17939d9-1628-435b-8797-769521ec1963

[broken knot]

But guys like Matt Vidd are NOT providing valuable feedback. They get alpha access so they can make YouTube videos saying how cool it is. I love Matt. Cool guy, but he gets paid. Look at his suno credits, always has over 30k credits. Wtf.

[worn flicker]

I know nothing about any of that, but I do love Timbaland...haha

[broken knot]

Bow before the master of Suno credits... Matt vidd. Now, listen to his song that ChatGPT wrote about a frog and stfu!!

[eager topaz]

Here is one:

https://sunodownloader.io/suno-v4

[worn flicker]

And this makes you download the song?

[eager topaz]

Huh?

[broken knot]

MattVidPro. If you're not watching his hype videos about Suno, you're doing it wrong. Lol..

[worn flicker]

It is called sunodownloader

[eager topaz]

They are stealing, thieves

[stiff plover]

How do I apply an extended song to the old one?

[worn flicker]

No, I just don't want to engage the stuff about people in this channel...lol

[broken knot]

Most people steal. Too bad cro-magnon man went extinct. They never started a war or stole anything. Their brains were massive though.

[worn flicker]

I was asking if it made people download the songs after generating. That is all

Great way to spread malware

[broken knot]

What stuff? I really am a fan of MattVid Pro AI but I wish I always had 30k suno credits. I'd use that up every month.

[eager topaz]

I have no idea. I saw a video on YouTube with the link. I'm not interested in stealing!

[worn flicker]

Right here is where I diverged

[broken knot]

The first song I ever made on Suno was a v2 and it still holds up. It's a masterpiece of pop/dance music.

[worn flicker]

Ok, it was just a question as to whether they were thieves or thieves delivering malware

[eager topaz]

I've some great tracks done wih V2

[eager topaz]

Well, I wouldn't risk putting my CC details on there!

[broken knot]

Well, it's a bit tongue in cheek. You know, I love Suno a lot. And I understand it's a business but I like to see how the sausage is made.

[worn flicker]

No one said you did. I just asked a question...lol. i get it, you don't use it.

[worn flicker]

I would love to see all the guts...I feel you.

[broken knot]

How many guys wish they could get 35k free Suno credits a month as well as a paycheck once in a while to hype Suno. MattVidPro AI reached Suno God Mode.

[worn flicker]

I've been researching how to fix something I have no access too...lol

[hollow hare]

They did not pay. Suno has no official API offered for use/sale (except maybe MS Copilot due to their partnership).

Services claiming they have access to the API so are actually breaking the law in USA by essentially hijacking Suno's Service and Resources (see the Computer Fraud and Abuse Act Section B.1, and specifically 18 U.S.C. § 1030(g)). Pretty sure legal is going after them soon, if not already - or those based in USA at least

[warm parrot]

That could be it too

[broken knot]

https://youtu.be/29upYN5q3i0?si=vtzbjlygQCOnwvDz

That video... classic.

Best Suno promoter by far. Accept no substitutes.

Dude said "best music generator" and when Udio releases v2 he will say it for them. 😆

[worn flicker]

I've heard of Matt, I was saying that I wasn't saying anything about alpha testers and whether they got compensated for their time. I am getting majorly misunderstood this eve,,,it must be bed time...lol

[broken knot]

You can tell he's a good guy though. Matt is good people.

[worn flicker]

You know I just don't feel like correcting the typo...lol

[broken knot]

Lol

[worn flicker]

Aight... I'll fix it...makes me look dumb...lol

[hollow hare]

I can tell you that I did not get compensated for alpha testing personally. Not sure about the others
(Matt is likely enrolled in some sort of special Media Relations Program which gives lots of credits for promotion - a lot of companies do this)

[broken knot]

My first song v2 on Suno was such a masterpiece that it set a very high standard for me. But I never release songs on this website. I'm saving them for Spotify and YouTube eventually, I guess. Need to remaster some this Ole suno stuff. Gonna literally have to buy a premium Udio. I go premium Suno and Udio, I just have 20 accounts. Lol

[worn flicker]

That is what I was saying is many people that frequent this channel were alpha testers and they would surely clarify the whole compensation part for you.

[broken knot]

Doesn't need clarification. Who the fk am I to demand clarification? I demand 24-bit download option then I immediately back off and stop demanding shit.

Making AI stand-up comedy is still in its infancy. I'm guessing v4 is way better at Stand-up comedy ai but I haven't tested it yet.

[worn flicker]

Ok. Did that go wrong as well...I am just gonna have to stop talking.

[broken knot]

Nothing went wrong.

You guys are cool.

Suno guys and gals. Ftw

Have a nice night either way.

[worn flicker]

Well let's hope some v4 gets patched up because it is capable of some crazy good sounds. Have a good one

[broken knot]

This is only the beta. I've verified that already. It can only get better from here.

Postponed final version of v4 is way better option than a rushed out final v4.

[worn flicker]

Yep, I was here oddly. I have been here longer today than I have been...ever...haha..don't know why...

[broken knot]

Because it's v4 day. That's the whole reason I joined this discord today.

[leaden lark]

remember guys, they are paying attention to liked and dislikes

[worn flicker]

Yea, it was nuts

[broken knot]

I'm too wild for discord. Lol.

Anyway, I have faith that A: They will postpone the final v4 and we will chill with this beta for a a couple months. B: Around Christmas we will get the real v4 and it will be happy holidays.

[worn flicker]

Hmm...here's hoping.

I'm gonna sleep now..,later

[broken knot]

Gn guys .

[high cipher]

My bros and sises, after wrestling for about 6 hours, I'm now convinced v4 is a solid and remarkable update including the remaster feature, however there are weird issues which are possibly linked to launch day's traffic.
In my pro account, v4 didn't run properly and remasters were horrible, but in my 2 other free accounts, the free remasters were so beautiful I was brought to tears.
So I assume, waiting for a while and not blasting those credits (like I did 😅) is the best strat.

[grand plank]

Hi, why im unable to use v.4? when I remastered or create with v.4 model. then the result is always. v3.5

[olive zenith]

https://discord.com/channels/1069381916492562582/1308503194938118164

[desert birch]

Latest v4 model have Huge problem with foreign languages. For example Hebrew. In v3.5 I fixed this with "Nikud". In v4 it doesn't help.

[meager frost]

I'm still forced to use v4 while doing covers, i can't choose the 3.5 model anymore

[cerulean merlin]

i have an issue

[olive zenith]

And some can't choose the v4 anymore

[meager frost]

😂😂😂 perfect

[azure crane]

I created a remaster of a piano piece. The result: Almost every note had distortion. Was hardly what I had expected from V4. I suspect aggressive limiting…

[frail grove]

V4 doesnt adopt the characteristics of the personas, i cant seem to be able to select v3.5 in the dropdown menu without the output still being v4

[frail grove]

Lucky you

[small pulsar]

does anyone have any experience with element synapse?

[tacit sequoia]

My V4 remasters sound super tin-y and flat, it's quite odd. The vocals are mostly an improvement but in my testing thus far it has absolutely butchered everything else

[small pulsar]

somehow so far i have hit a goildmine of remastering old songs

[tacit sequoia]

Has it maintained the intensity level of the stuff you made originally? A lot of what I've done is on the heavier side, and the remasters cut the oomph by like 50% lol

[small pulsar]

the only thing i dislike just a little is one particular song that was good in itself in terms of the ai singing on the beat perfectly somhow became off-beat and mostly false notes

[small pulsar]

for me most of them sound better even as in better voice etc

[tacit sequoia]

Huh, well I assume the kinks will be worked out sooner rather than later

[small pulsar]

instruments are a little iffy

i have heard it mostly with rock (or big thumpy bass) where the guitar becomes a bit flat and with lots of bass it tries to put more bass in

[patent yoke]

Any prompt suggestions for covers to retain the same beat for uploaded audio?

[tacit sequoia]

Yeah for me the guitars sound like they've been muffled or put through a filter (in a bad way)

[copper shard]

Same, the styling and pacing of 3.5 seems better than v4, and remastering that style/pace into v4 seems to give better vocals, so it's becoming a "best of both worlds" kind of thing

[patent yoke]

Ive noticed remasters of old songs made with pure suno is better than covers or extensions of uploaded audio

To he fair some of my uploaded audio is Muffled so that effects it

[azure crane]

Strange, when I do a remaster it creates one in stereo and one in mono.

[small pulsar]

yeah well i am still sad that the one song i wanted to work didn't. kinda salty over that lol

[halcyon harness]

brooo.....

[small pulsar]

yes bro whats up

[halcyon harness]

they broke my workaround

with the new update

[small pulsar]

spill it 🍵

[halcyon harness]

why would they do that

[small pulsar]

depends on what the workaround is

[halcyon harness]

You used to be able to extend a song, crop the extension and then still use it to combine into a whole song.... but now once you crop the part 2 it blocks you from using it to create a whole song

[small pulsar]

hmm

[halcyon harness]

I used to use this to generate a part 2 with a section of silence at the end.... crop out the silence and any padding time I needed then combine into a whole song, and then use replace section to generate an end to my song of the exact length I wanted

[small pulsar]

oh yeah i always hate it when you think a song has ended. and then the ai goes haywire and starts speaking alien

[hybrid vine]

I have some songs in French and with v4 the result has a strong american accent, it doesn't support foreign languages?

[small pulsar]

i think it is mainly focussed on english as i once tried making a dutch song but i had to choose my words VERY carfully

bc most of them woul just be mispronounced or it would just straight up skip whole verses (3.5 that was tho)

[hybrid vine]

also, is there any change to the "Get stems" feature with v4?

[gaunt skiff]

Hi Suno team. I try to find this topic but i dind't find it. I would like to know how can we force suno to sing portuguese from Portugal. This is something that any AI app stroggles. They always sing or talk with portuguese from Brazil.

[azure crane]

Why mono when remastered?

[small pulsar]

ffs

[small pulsar]

as i wanted to say... that is the difference i hear? do you anso know which one is wich?

[hybrid vine]

the one that sounds flat => mono

[dusky moat]

Guys, a little help please. I had an attempt of the song in v4 model, that I had listened to start to finish while it was rendering. It is exactly what I need, but it is displayed as stuck rendering, even though it's been 8 hours already. I can't play it anymore and tgere is no way to interact with it.

Is there some way to recover it?

[jade mist]

Do you think most of the bugs are related to the fact V4 is in beta? I've had a hard time to get a key change for example. Also the constant annoying artifact on some songs

[small pulsar]

don't you love it when you ask suno to make drift phonk and it makes a full on country song? xD

[small pulsar]

if it is in beta i would be almost certain of it. how long is it out now? i jyst saw it this morning

[stuck obsidian]

I don't understand how I'm supposed to answer this.

[hollow hare]

Be more direct, e.g. see my comment at #📚┃suno-school message

[stuck obsidian]

...I described a topic, hit 'surprise me,' and that message appeared.

Ohh okay. I thought I was supposed to answer it somehow.

[hollow hare]

Using new Write With Suno prompt

• Click 'Write with Suno' first then type something in the prompt.
• Change the lyric model if needed.
• Write what you want in the new field. For example: "A rap song about the great German Sheppard named Nard Dog who stole some bacon in the style of a AABA rhyme "
• Finally, click 'Write Lyrics' (though I highly recommend copying your prompt before submitting in case you want a different result)

To see a picture example, check out #📚┃suno-school message

[small pulsar]

so that is how that works.... was trying to figure it out lol

[hollow hare]

Yeah, it was confusing at first. I found out by accident lol

[small pulsar]

your tip works wonders

one thing i found out is that if i create a song at 3.5 then remaster it sounds better than just at 4

[dark fossil]

hi, does download the wav only can 1 by 1?

[mint heron]

No. Same

[hybrid vine]

😔

[lime vapor]

Am I the only one who has a problem with the v4 generations? I have Pro, but even when I remaster or create a new song, it uses v3.5 instead of v4 (it's 100% selected)

[mint heron]

Yeah, I have same problem. Wrytten to the support. Have you tried remastering? did you alos getr 2 second files?

[lime vapor]

I have tried. Same thing. It takes v3.5, but the length is the same as orig

[mint heron]

For me remasters are broken.

[pale mica]

ders difrent style name for rhyme eh

[hollow hare]

Yep. Common is AABB and ABAB. But you can also define something like AABBAACCAA

Or tell the AI something like:
Make verses AABB rhythm
Make Chorus ABAB rhythm
Make bridge a ABCD rhythm

You can also say something like "write the lyrics as if it was from an award-winning lyricist", or storyteller, etc.

[hollow hare]

I think the issue at the moment is that there is a bug where the wrong visual label is applied but the model is fine. (If the remaster sounds clear, very likely it is v4 despite what the current label says)

[lime vapor]

Can't say it sounds clear

[mint heron]

I whill recheck, But I can here difrence in v4 by this metal ringing bug. I never hered it. But maybe I was just not tried inuf.

[knotty skiff]

I've been meaning to try a AABBCCDD or ABABCDCD for a while
Most of my newer stuff has been AABBCC

i started out with AABB and ABAB

[pale mica]

simply amazing 🤩

[knotty skiff]

does anyone know what the tag or genre that is least impacted by artifacts currently?

anything rock or rock adjacent seems to be plagued with clicking - or rather with drums it seems lol

[hollow hare]

Pretty much anything excluding loud genres ahould be fine. So no Rock, many Dance/EDM, and Metal (esp.)

I am surprised I basically went through 2.5 posts worth of various genres when demoing v4a and missed the shimmering issue (despite having 3 different types of metalcore. I heard the shimmer but did not realize how bug the issue was): #🗣️┃chill-chat message

[knotty skiff]

any prompts that people have tried make any difference? like the [reverb] etc.?

[hollow hare]

I think some people were talking about placing [CD quality][High Quality] at the top of the lyrics before remastering? I didn't try it though

[knotty skiff]

hmm... it does seem strange to do, but i guess anything is possible with AI

[broken knot]

Been using suno v4 all night, working on rock music tonight and everything is rap-influenced now. In genres where it shouldn't be. I'm all for rap and any other genre, but not when I'm doing a 200 word prompt for some symphonic goth rock. And the really good song I made on v4 just ends up going into a terrible piano solo for no reason in the middle of the rock music. V4 has bugs. Lol. All my personas are kind of ruined too, as V4 sounds nothing like the v3.5 vocals.

It helps to exclude rap but it's weird that I need to in such an unrelated genre

[kindred shard]

It can be easily replicated to any song, genre or whatever with these tags followed by each other. Credits to sushileaf, from where I copied these.
[Foley Percussion]
[Textured Crunchy FX]

[broken knot]

You gotta get lucky af to get something nice out of v4. It's way harder to get good tracks out of v4 than 3.5

[knotty skiff]

lol dude... I've been trying to get some gothic metal going and it's been a little bit of a disaster... vocals like a soft rock/pop artist and slow tempo are the main culprits... and then there's.... the shimmer

[knotty skiff]

they wanted consistancy... and they got it lol

[graceful condor]

My thoughts are to directly type lyric or copy/paste from an app like Google translate

[broken knot]

No shimmer on my "Modern symphonic goth alt-rock vampire pop, ballad, pop, dark wave" lol but I don't know, I'm listening through a jbl Bluetooth speaker wtf do I really know until I put it under a lens. Lol. But I had and my personas all set up so that with 3.5 I could instantly get a masterpiece. But now the personas don't work anymore on v4 so I dunno, currently it takes me hundreds of credits in v4 to get something decent. It's a credit eating machine. V4 feels like gambling. Lol

[knotty skiff]

please link the song i'd like to see

[broken knot]

I have probably the most extensive set of audio repair tools that pretty much anybody has and I can get rid of unwanted noise if I need to. Worst case scenario.

[broken knot]

Well, the lyrics are a masterpiece. So it's "too hot for suno" and I'm plan to do a YouTube channel once I have a hundred or so songs remastered and ready with perfect enough audio

[knotty skiff]

I've struggled to master my WAV files to eliminate peaking and background artifacts, but i'm abysmal at it

[broken knot]

You need very very specialized tools that cost 100s to 1000s of dollars each lol

[knotty skiff]

totally offtopic... but why 100?

[broken knot]

Because it's a good amount of songs to start out with a YouTube channel if you do a wide range of genres and music from different eras or it sounds like 60s music, the stuff on Udio. Best of my Udio songs. Best of my Suno songs and I have all these ideas that's all

[knotty skiff]

i've been using ultimate vocal remover and davinci resolve... you're probably correct I'd need to invest in a paid software to get real results

[cold zenith]

When i cropped, we can't get the whole song a

[hollow hare]

There are some free mastering services out there, like Bandlab (online service, even has free plan includes basics vocal splitting).

But, yeah, for the good stuff- usually you gotta pay and pray

[broken knot]

Ultimate vocal remover isn't the ultimate at all

MVSEP is killing it

Free and online

[knotty skiff]

no it's just free lol

[broken knot]

Great cutting edge stuff. I've tried 30 different stem extraction methods

[broken knot]

Mvsep has exclusive technology literally has audience removal. So useful. They always come out with the newest models and ultimate vocal remover has old stuff.

[knotty skiff]

any of us in here experienced producers? what are the main software you're currently working with?

[knotty skiff]

might give it a go at some point

I'm sortta going through a writing lyrics phase atm and haven't been mastering much

[idle heart]

Anyone else getting a barely noticable sound in their V4 remasters that sounds like an old fashioned film projector? I can't unhear it now that I'm aware of it, and it's on EVERY generation. I'm thinking it's Suno's 'watermark'

[knotty skiff]

yep, come sit in our support group for shimmer survivors, here's a badge

[broken knot]

I know of 3 or so free mastering tools online and I have thousands of dollars worth of stuff in my Adobe setup. But I just want to streamline this shit and make it sound good. Being able to always remove any unwanted noise is a good thing but hope these suno songs will get cleaner and cleaner. You know they never are generated as YouTube ready, you can take a v3 or v4 song then add some gain to it. Hope that's all most of these needs is a little bit of eq and gain. But you couldn't get v2 to remaster well and even 4 isn't necessary easy. It's 16-bit 44khz and that's not what you want when mastering. You want 192khz 24-bit or at least 24-bit 96khz. So I'll be asking for 24-bit option download for as long as it takes or until the good folks at Suno say it can't be done... if they want to do it they can do it. Some say, since they trained on CD quality were stuck at CD quality. If so make a deal with someone to train off of HD data or find a way to do a creative "HD music" button that gives us legit 24-bit because you can only get so far with CD quality

CD quality is amaturish and it's nearly impossible to make it sound any other way

[cold zenith]

hello devs. there's a problem with the crop, now they're completely isolated from the base piece. You can't put them together in one piece like before. This is detrimental when you've made several crops, with several generations. This no longer works on V4. I don't know about other versions.

[knotty skiff]

I don't really want to touch on the whole AI training debate but... labels and artists are concerned about AI especially at the moment with the lawsuits underway, it's going to be the hardest sell to get that training deal for upping the quality

I've been considering just collabing with some of my band friends and splitting anything made, produce it together it'll have the quality and it's a starting point as a style reference

[broken knot]

Then they need to do something. AI faux 24-bit 44khz with more of an open sound. Something. But it does seem like they need to train off of 24-bit why would that be impossible for a company taking in millions of dollars. They need to immediately get on access to enough 24-bit training materials and harness ai technologies to make the most of the training data. Do it the same underhanded way you did it last time. Just get it done

The amount of money they make, they can hire a few bands make a few recordings they own

[knotty skiff]

yea I think that's part of the reason they brought in timbaland

qualiy samples equals quality output at the end of the day

[broken knot]

Buy albums. Get 24-bit any way you can. I don't care if you have to do a smash-and-grab

[mint heron]

not havving 24 bit is the leesed of suno sound problems. They need to first fix what avagrage user could here. Vocoder thing is less and less, but stil is there. Sometimes songs just turn out with artifacts and noize, etc.

[broken knot]

As in buy albums from bands before the albums are made. Pay local artists across the country if they're good. Get the 24-bit 96khz wav training data set and then release to us a model that can churn out 24-bit/96khz. That would be absolute madness

[oblique falcon]

Service is getting worse and worse for me.. I can't even generate stuff

tried v4.. takes forever then says it got stuck or something and refunds my credit.. tried v3 and it says a system error occured.. tried v3.5 clip gets stuck and it refunds my credits

[broken knot]

I'm hoping they work towards 24-bit models alongside the 16-bit 44khz model and of course it's one of many things Suno needs but it's a massive issue. 16-bit celling will be hit by Suno v5.5 -v7 or so and then 16-bit will be even more ridiculous.

[knotty skiff]

Sorry I went away for a sec and thought about that... it's a bit reductive to say that the users don't also have an integral part on the quality of the output product

[knotty skiff]

more options are nice to have for those who are willing to take the extra steps in the process

but I agree with @oblique falcon there needs to be a basic consistancy for the casual users that isn't plagued with artifacts and tonal shifts and that appropriately heeds to prompts

[oblique falcon]

for me v3 was the last usuable version.. I hated v3.5 and well I can't get v4 to work right now even

[broken knot]

I want the option to listen to my suno tracks in 24-bit and 96khz. Generating something like that would be a massive achievement and milestone. V4 improvements show me that by the time v6-v7 come around, 24-bit will be really beneficial. If they can make a v6 or v7 model that had a 24-bit Generation method, even if only 44khz, it's better than nothing. 24-bit music is significantly better. V4 isn't good enough yet for it to matter. 16-bit is good for v4. I'm considering v10 lol

[knotty skiff]

3.5 was my jam sorry friend

[oblique falcon]

I tried to make similar music that I made in v3 and it wouldn't do it

[broken knot]

They'll turn it all around. Suno isn't going anywhere. It's new technology. Kinks still being worked out

3.5 with personas was bad ass.

[oblique falcon]

Will my credits disappear at the end of the month? I have 2005 I can't even use

[mint heron]

we all hate new models from the start, but I think bekoz old thethods duz not worked, where they worked before. I had a hate phaze, then acceptance, then I forgot about v3 at all. And I huess suno is actualy silently updating there newest models, But maybe not.

[oblique falcon]

I've been trying on and off over a wek now

[broken knot]

Unfortunately they don't carry over and I'm not a fan of that at all. I try not to ever think of the horror of losing credits

[knotty skiff]

I found I had to change up the tags and style choices or really be specific for it to get close to V3 output, V2 had it's limits and V3 had it's issues similar to V4 infact... V3.5 wasn't perfect but a good foundation to build from that was a good balance of broken and working

[oblique falcon]

That's bullshit if they disappear.. their service won't even generate a song for me.. I'm going to write them a second time

[broken knot]

Take screenshots if you can. Crossing my fingers your issues get resolved

[oblique falcon]

in v3 I was typing 80s fast ska and it was generating me some good stuff.. v3.5 it was generating other genres of music not even ska sometimes

[knotty skiff]

don't remind me... my computer was at the repair shop last month, i don't want to think of those lost like 3-4k credits

[broken knot]

I got disillusioned with the lack of next generation models coming out and took like 2 months off my annual contract. No pausing it is bs

[oblique falcon]

someone mentioned personas.. I haven't tried them.. is that where I can make music that all sounds like the same band or something? I'd love that

[broken knot]

People want to wait for the new models and I get disillusioned easily. It's almost where it needs to be but it's not. It's close but oh so far away

[knotty skiff]

meh... i just had a bit of writers block and other stuff going on, i'm by no means "in the industry"

[broken knot]

I never have writers block when I try to write songs I write bad ass songs. Usually many in a row now that Suno and Udio made it so easy.

[knotty skiff]

I think that's why everyone was so hyped for the remaster feature

[oblique falcon]

what the hell is that cover song maker? Is it useful for anyone? I put in a song I want to make a cover of and it says I inputted copyrighted music.. of course I did.. how can I make a cover if I didn't?

[knotty skiff]

yea that's pretty much what they do... they have their issues though, very limited occassions where they work well

[kindred shard]

It's not that like of a cover. It is like "give a little modified version of my input song"

[oblique falcon]

I've been making my friend upbeat ska songs.. here's Public Enemy the rap group: https://www.youtube.com/watch?v=kJBofkPsBuA -- they were coming out good at first

[knotty skiff]

somebody's not shy

[oblique falcon]

that's using v3 model

[broken knot]

We were really hyped. If we get lucky, udio will update with a v2 and make it way easier to upload a 4 minute or shorter track into Udio to master it there. Right now people use the Remix feature and can get a more straightforward remaster from the Udio "remix" then the Suno "remaster" but it's two minutes sections and splicing the song together in an audio editing tool. It's a lot of effort, but I'll probably wait for Udio v2 to go premium with Udio. They are on v1.6 so I assume v2 is around the corner with 4 minute upload/remaster ability

[knotty skiff]

I think it really does depend on the style you're using as to how well it generates, as a rocker AI really struggles to get it solid

[oblique falcon]

I used to type "80's fast ska" and boom was like magic generated what I wanted.. I haven't generated anything useful in awhile

[knotty skiff]

I tried Udio way back when Suno V2 was here and didn't like the shortness of output, i just didn't like it... might return to try one day, I'm a Suno fanboy all day though