📚┃suno-school

[woeful yoke]

I think you can play around a trial period with it

[worn flicker]

You are looking for the less full Instrumental prepped for vocals

[woeful yoke]

I suggest you watch this video first:
https://www.youtube.com/watch?v=BrozmGMxP20

[worn flicker]

I can't say Suno is up to that precision yet. I can see a market for it. Have you tried Udio?

[woeful yoke]

that video got it sold to me... and I haven't regregretted

[glacial scarab]

Yea i don't doubt it's useful.
I just think right now it's not "worth it" if I (mostly) publish on suno

[worn flicker]

That is where I end up. Since I don't put my stuff on YouTube or anything else. Suno reignited my music love, but in a way that let's me make stuff without a ton of setup. More of getting that tune in your head down on paper.

[flint haven]

✅ Warning logged for lamprey_99436. I couldn't DM them. || Rule 1 - please be respectful. Suno has a very helpful community and mods are not appointed to be go-to FAQ people, they are tasked with moderating the server. This is primarily a community-ran server

[rotund scarab]

Hey I have a question, is it possible to transfer credit from one account to another? I have two accounts that have credits on them and I would like to see if it's possible to transfer them both over to a brand new fresh account. Starting from scratch with an unused account would that be possible?

[woeful yoke]

Suno was the reason why I rebuild my "studio" again too... even made a lot of upgrades, including getting myself Adam Audio Studio Monitors so I can finally produce/mix/master the same as I used to do when working in the field back in 2000's

[unique stone]

I think my questions may have been missed in the above discussions. I wanted to determine if there's a possibility that your account could be affected by a false-positive for something that is used to prevent fake Suno apps/websites from functioning properly. At first glance it doesn't look like that's the case, but I would recommend e-mailing [e-mail] to ask them to check if your account has been flagged in some way. @rotund scarab

[worn flicker]

Hell ya boss. Gotta get it!

[woeful yoke]

I kind of gave up trying to balance IT work and music production... all tracks took too long. now things move faster as SUNO helps me get unstuck

[worn flicker]

Same

[woeful yoke]

I don't get it when some people are saying that AI makes musicians obsolete

[worn flicker]

Music has obviously always been the hobby

Oh not even close

[woeful yoke]

I see it as a co-producer, not as someone who takes my "job"

[worn flicker]

I haven't heard a Steven Tyler or James Hetfield on Suno anywhere. No Lzzy Hale, no where close. Musicians are original.

[flint haven]

There is no transfer of credits system available @rotund scarab - You can ask Support if they can do it but I doubt they will, especially if one is on the free plan.

[woeful yoke]

Yep. I got hooked back in 1996 when I saw a cool dos program named Fast Tracker 2 (ft2.exe, baby)

[dapper crescent]

Is there any good way to make a appropriate duet

[woeful yoke]

I had a 8-bit mono sound card from creative labs

😄

[chrome stump]

I believe it’s a way to give you ideas, and in fact, since I’ve been using it (issues with the tracks aside), it has unlocked some inspiration.

[worn flicker]

Ooo...remember the Assembly Clubs, like Future Crew?

[woeful yoke]

well... multiple passes... or you mean like different singers?

[worn flicker]

Extend extend extend is the best way

[rotund scarab]

No both are paid accounts, but that would be great if possible. I can't figure out exactly how to contact Support, I've looked everywhere that's why I'm here on discord. Apologies for my frustration I think it's stems from having so much trouble this morning on the program, and then I felt like I was getting convoluted answers that didn't really get to what I was asking. I'll try to be nicer next time... 🙏

[warped furnace]

I think so, but it's mostly hit or miss. There are some suggestions you can look up using the search. For example working with extensions.

[dapper crescent]

Yes with male and female for each verse

[unique stone]

Nice. I had an 8-bit mono cheapo SB compatible 😅
Tracker music blew my mind the first time I heard it (Second Reality demo), having only heard OPL3 synth MIDI files before.

[woeful yoke]

Yep... And before getting to PC, I had been playing with and hacking ZX Spectrum 48k with all those audio cassettes 😄

I even had a little program that could record 2.8 seconds of 1 bit audio

[flint haven]

E-mail is [e-mail] @rotund scarab

[woeful yoke]

😄

[woeful yoke]

well, at least it sounded like a crazy over-saturated audio preamp transistor that is about to blow your speakers... more like extreme bit crushing... but hey, it was able to play me back something from that 48k mem

[worn flicker]

I remember being blown away by the Wolfenstein 3D on the pc speaker then when sound blaster happened. I liked to fell out my chair

[chrome stump]

Thank you so much for the help, sismai. Now I have to run. Have a good continuation!

[worn flicker]

Np

[warped furnace]

I had some luck with these prompts, if you'd like to look at one of my duets.
https://suno.com/song/441645ef-2633-4073-84eb-73fc2132e8e8

[woeful yoke]

same here... our school got a couple of first computers and then after school we were in th computer room and Wolf 3D gun sounds scared the hell out of a cleaning lady who arrived in the evening...

we had the speaker blasting gunshots over the whole corridor 😄

[worn flicker]

Aiiieeeee

[woeful yoke]

all these memories remind me that we're getting old now 😄

well, maybe only physically...

[worn flicker]

I know and tech moves so fast you have to pick a path instead of just going with computers...lol

[woeful yoke]

mental state is still somewhere around 23yo right now

[worn flicker]

Since I'm back in college that is literal for me ATM...lol

[woeful yoke]

Mu mom was 44 when she finally went to college... so it's never too early (or late, depending on the perspective

actually we both went... I was 24 then... so we had some lectures even together 😄

[worn flicker]

Well then I'm ahead. Don't hit 44 til Jan and I only have two classes left.

My mom went back in her 30s

[dapper crescent]

It was beautiful but not perfect duet at all

[worn flicker]

Oh I'm sorry. #1314766265897717780 message

[warped furnace]

Just an example. If you want it to be precise it's best to work with extensions. 🙂

[worn flicker]

I love duets

I first came in here looking to see if I could get to Broadway level of 6 simultaneously running lyrics. Sadly, not able, but duets are best done with extending.

[gentle ridge]

Are there any ML architects or enthusiasts here that could help me to understand positional encodings in transformer models? Like I almost got, but not quiet, Im confused

[worn flicker]

@woeful yoke someone is speaking your lingo

[worn flicker]

I can explain small amounts, but Harly knows some stuff. I can point you to videos

[woeful yoke]

did chat GPT refuse to explain this to you? 🙂

[gentle ridge]

Didnt refuse, but definetly also didnt do a great job haha.. But watching videos also didnt really explain what I was looking for

[worn flicker]

I do love how chatGPT never gets frustrated no matter how tiny of a question you have.

[gentle ridge]

dude called my Jumbo since the beginning of the conversation for some reason

Fair point, Jimbo! Let’s tackle this in the simplest possible way, focusing on why both SIN and COS are used for a single word’s embedding rather than just one function.

wtf

[worn flicker]

I normally tell it to take a role, as a tutor, or I tell it to explain it to me in star wars terms

[gentle ridge]

That also explains my question btw @woeful yoke 😄

[woeful yoke]

okay, what's your question?

I would help you but... my ML journey started in 2016 with tensorflow and shit... and then later I had a pause for a few years.... now picked things back up... been able to replicate most of the smaller projects (including Karpathy's nanoGPT) but... I still fully don't grasp all this 100% to be able to give lectures.... 🙂

but I can try

[worn flicker]

See...lol

It predicted that Jimbo would be the best way to refer to you. That is some flawed weighting...lol

[gentle ridge]

well to be more specific, I do not understand why the positional encodings make use of both SIN and COS functions alternatively for even and uneven dimensional index per token embedding

[woeful yoke]

I have building a small GPT-like audio AI right in front of me... doing research and testing of the data that I want to tackle itself

[gentle ridge]

Sounds really nice, which kind of audio is it generating?

[woeful yoke]

from what I understand, use of both sine and cosine functions in positional encodings in transformer-based language models is a way to capture both the magnitude and direction of the position

Meaning that you need to be able to determine both the direction angle and also how strong the signal is. There's some really interesting relationships and meanings being encoded into both of them

[gentle ridge]

Im currently making a custom midi dataset to train a transformer on it for a semester project in university

[worn flicker]

Ahhhh

Midi baby

[woeful yoke]

well... one example would be it helping to understand the order of the words in a sentence. as the order of words can change the meaning of a sentence.

[woeful yoke]

nice.... and I have future plans with midi too 😛

back in the day I was doing a lot of electronics stuff... making a microcontroller generate/respond to MIDI codes was "fun" 🙂

[worn flicker]

I love not knowing stuff but hate being behind...lol. I know the words, but I am not at the point y’all are at and I am a know it all, so that sucks...lol. my brain can't handle it all.

[gentle ridge]

Yeah I get that it makes sense, since the model processes the tokens at the same time so it needs some encoded information in the embeddings itself. But I dont understand what the PROs are for using both sin and cos instead of only sin for example...

[woeful yoke]

as long as you keep on evolving and learning, you're still going to be ahead of most of the peoople

[worn flicker]

I'm going to backseat and read this conversation

[woeful yoke]

as far as I remember math lectures.... um...

[worn flicker]

0-255, OK now back seat

[gentle ridge]

I mean, at that point it probably doesnt even really matter, why it is like that but it just bothers me this whole dam evening lol

[woeful yoke]

feck this, I asked the friend that lives inside my GPU here:

[untold briar]

It's over my head, but I'm not sure encoding position like that is actually good, assuming that's the standard absolute position that GPT-2 class models used. Later papers were like, "Oh if you just do nothing, the model figures it out anyway?" https://arxiv.org/abs/2305.19466 (But even later encoding schemes to actually seem to be better)

[woeful yoke]

I remember something about angle between since and cosine....

so, the thing I'm working on right now... with that I don't have to worry about that...

I'm using the embeddings that is being internally used by DAC already 🙂
all you need to do is query that specific layer during inference

it took me a while to understand that for audio tokens alone, I can and don't even need to use a text embeddings model like you need with text tokenizers

[gentle ridge]

Hmmm I see, looks like I have to dive a bit deeper into the maths here, but thats definetly the best answer i got yet

[worn flicker]

Dude knows some shit...lol

[woeful yoke]

also, yesterday meta dropped a paper and code on us which shifts the whole paradigm by enabling us to use binary text in LLM without tokenizers at all.

[woeful yoke]

to be honest... I don't concern myself with half of the math that I don't need nor understand.. I know how to implement most of the stuff from scientific papers and associated repos... that's enough for me... I rarely need to dive into theory of how it all comes together in the lower level. I'm not a PhD 😛

[untold briar]

position encodings https://arxiv.org/pdf/2104.09864 section 3.2, note that this is Rope, I think the standard cos/sin absolute is from like 2018, but this is building on that

[worn flicker]

I'm still at this level https://bbycroft.net/llm

[gentle ridge]

First of all I love how they called it NoPE lol.
So I might as well just leave the PE away when training in the future? But why did people even implement it then lol

[woeful yoke]

I know a lot about a lot of stuff ... but not too deep. I can use google, if I need to understand more... at one point my brain was already smoking from trying to understand how the high level implementation of Transformers works on torch/python level. I had been using Tensorflow for years, so it was tough at first.. as initially needed to learn to use/understand Torch itself 😄

[worn flicker]

I would ask my professor. I guarantee you she would know the answer, but I don't know enough to ask the question...definitely not enough to get into the sin and cos parts

[woeful yoke]

There were visual deep explainer series on how transformers work 3blue1brown channel, I think

[worn flicker]

https://www.3blue1brown.com/

Yup

She showed me that

[woeful yoke]

multiple long and detailed episodes... like I said, I have an understanding of principles but if I'd need ot get deep answers I'd refer to those who have made it easy to digest 🙂

[woeful yoke]

okay, I see 🙂

[gentle ridge]

Yeah thats what I always say to myself as well but then I end up again digging my head into those low level formulars that I dont understand anyways lol... To be fair I want to make my PhD in that area tho, so i probably have no choice >.<

[worn flicker]

Hey at least you shouldn't have an issue with a Thesis. Unlike English PhD candidates

[woeful yoke]

well.. if you are really keen to find this out... I can take some time and go over and study it a bit tomorrow... so I can explain... at work I'm quite often mentoring junior enigneers on shit I had just learned myself 😛

at least I used to 🙂

[gentle ridge]

Definetly enough to research on for sure haha

[gentle ridge]

I´d love that dude! You seem to be really smart at finding the right knowledge if you look out for it haha

[woeful yoke]

I usually know that I know about something... or that I've seen it... and then if I really need, I go and find it again... I don't have infinite capacity for garbage in my head... after all.... I work as a Network DevOps these days... so all ML stuff is still on hobby level. Quite far from Networks eithe way

[woeful yoke]

okay... perhaps we could even do a screen sharing session somehow... so we go over and explore this together.. I usually learn things faster when I have to also relay the concepts to someone else too..

[untold briar]

this is all attention is all you need says: https://proceedings.neurips.cc/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf

[worn flicker]

Ahh back to my world...lol

[gentle ridge]

Yeah I read that and didnt understand it lmao

[worn flicker]

I meant to reply to the network part

[untold briar]

I don't either, it feels a lot like "this feels right to us, and it seems to work?"

[gentle ridge]

For sure, tomorrow is gonna be tough for me I think but we can do it next week 😄

[worn flicker]

I hate that I know each word separately...haha

[untold briar]

like a lot of early tokenizers. just throwing stuff out htere

[gentle ridge]

Yeah 100%. I feel like I know exactly how they felt writing this.
"Yeah the shit im about to write is gonna sound so good, but most ppl wont understand it anyway"

[worn flicker]

Yall flip my geek switch and it pisses me off that I have other stuff I have to do....

[gentle ridge]

we all do man, just gotta take the time for it haha

[woeful yoke]

that damn paper!
I don't know if you saw this the other day but... my ML experiments in 2017 culminated with me literally asking around in ML communities about whether it would be possible to improve sequence modelling, to get LSTM or something like that to start pay attention to stuff... So I was literally facing an issue wheer I felt that no matter what I do or how much I train, my problem was still not converging, and I had even identified that I need attention-like mechanism.... and I couldn't figure it out how to do this, or whether it's even possible

I gave up.... and then 2 months later Google dropped that faithful paper!
But back then I didn't know anyone who could translate the paper into a concept I could try to implement in python, so I gave up once again! 😄

but the freaking timing!

[untold briar]

it wasn't that long ago that simple changes to encoding methods really blew up context lengths (not in a way that is 100% as good as having actual longer context, but still in a useful way)

[rotund scarab]

Thank you 🙏

[worn flicker]

When things were switching from 32-bit to 64 and crap didn't have drivers, I found myself for the first time asking who writes the drivers, why can't I? I learned quickly why not. But, that was when I learned how little I actually knew

[untold briar]

I think it is possible to fine tune bark to use relative positioning, and thus allow for those longer context 'hacks', back when bark first came out was hoping that might happen. i can't remember if anyone actually demonstrated fine tuning a model like that, probably just not worth the effort, but when I tried to research whether it would work, it seemed possible from my limited understanding of asking other people smarter than me

[woeful yoke]

yep...

I usually spend one morning each week going over last weeks ArXiV 😄

that's why I also know that CosConvNN paper dropped... and it revolutionises CNN's, making it 73% faster and less demanding in terms of compute... in addition to getting better performance out of audio feature self-supervised learning/extraction.

[worn flicker]

Damn it. I have to go now. Don't make any breakthroughs without me

[woeful yoke]

I already can already forsee an increased iference quality/speed in the near future... at least for audio/speech stuff, if not more....

[gentle ridge]

To quickly just repeat on what you said before... So you said that sin and cos isnt even really implemented anymore in modern transformers?

[woeful yoke]

Definitely makes sense... But realising that your're a little bit affected by the dunning-kruger effect is actrually a good thing. True wisdom is in realising that you still have a lot to learn 🙂

[untold briar]

i think they are still the most common kind of embeddings, just build on the concept to make better versions that work better for longer context and stuff. but there are totally different embeddings that don't use sin/cos, I think "alibi" is example

[woeful yoke]

soon we can open a AI/ML-themed pub for maniacs 😛

actually, I, once again, managed to sidetrack from my coding stuff in here. It's hard enough to look at my new balance and not do anything at the moment, as I'm trying to move along with a project and not lose focus . All those new credits screaming at me 😄

my current objective is to literally try to understand if the pretrained embeddings inside the DAC could be directly shared by the GPT side of things. Based on Vampnet's implementation I'd say I can hook into them directly and should be able to model the sequeces. if it works non-autoregressively in vampnet, it should definitely work autoregressively for preciese sequences.

[charred merlin]

You might not have seen my previous message;

A ‘2x sound quality increase’ sounds like a significant improvement.

Would you mind elaborating a bit on the methods or metrics you’re using to measure this improvement? For instance, is this a subjective evaluation based on listening tests, or can you share how you’ve approached measuring the noise reduction? It would be great to understand how you’re quantifying the difference.

[drowsy orchid]

Hello. I have a problem. A week ago I bought premium through the android app. I was logged in with a phone number and the payment was through the play store. I still do not have premium status. No one from suno responds to emails.

[charred merlin]

The support team has been flooded with messages recently and has been delayed getting back to people — at the moment, the only way to get official support is through email, so you’ll need to wait for a response.

[drowsy orchid]

It's sad that I have to wait so long for any answer.

I think it will be faster if I refund the payment and buy again.

[buoyant tinsel]

Oh No, got the fake song thing again. I filled out the fake song google form and the problem got fixed and after creating another 10 versions got the same message again. Please help.

I've been trying to get the sound effects to work in the intro but nothing works, does anyone know anything about this? I tried to get the intro sounds many times and while trying the system gave me a captcha and the fake song problem.

[Intro: Ocean sounds]

[Verse: Fingerstyle Guitar, Gentle Voice]
The tide rushes in, my breath runs fast
A moment feels endless, it’s too much to grasp

Been learning better prompting through this guide:
https://howtopromptsuno.com/making-music/sound-effects

[unique stone]

I'd suggest contacting [e-mail] and explaining that you've already filled out the form and had your account fixed but that it's happened again. Could be something related to your internet connection, e.g. if you're using a VPN.

[buoyant tinsel]

Thanks @unique stone, just sent the email.

[polar spade]

Did you upgrade to premium from a free status or pro status? If you have a receipt in the play store, does it NOT show you have a present live subscription?

[polar spade]

I didn't know there is a refund option. Oh, I take that back. I think it will refund on the day it charges you if you cancel that same day.

[nimble hemlock]

Does Suno provide any sort of API? I'm interested in pulling data about my library and presenting it on a custom webpage for just myself.

[twin bough]

OK sighs How do I get this suno thing... switch between male and female vocalist.

I have tried

[Verse 1]
[Female Vocalist]

And

[Male]

It's just NOT doing it.

[unique stone]

Not at this time. Generally any access outside of using their website and official apps is forbidden.

[nimble hemlock]

Ah ok

I just wanted to be able to look at data about my library. See most popular, most played. Also, was thinking of adding a "shuffle" option to give me text for things I want to post in the showcase. Right now, I'm keeping a manual list and it's not sustainable.

Is an API planned at some point?

[unique stone]

I've heard of people asking Suno directly about it (probably via [e-mail] ?) but I haven't heard of there being any movement.

Part of the problem is likely that I expect a large number of API users will simply make a Suno-like app and sell it (and then Suno have to provide support for that API as well as compete with a service that does pretty much the same thing as itself.)

[polar spade]

Include in your understanding, that Suno uses the styles prompts, metatags AND lyrics to work out what's going on in your song.

Simly telling it to switch in the tag prompts often isn't enough. Think of it as percentages, weights and links. So if you suggest 1) a literal need for a mix of vocals in style prompts , then it's helpful to sneak a style that often has mixed vocals. You could use "ensemble rock" for instance as in rock operas, musicals, etc. often using multiple vocalists. Suno gives it a small weight and will nuance the song with harmonies, choirs, or even, without asking change up the vocals.

Then, putting [female vocal] metatag, and [Male vocal] among other variations then specifies to suno ".... and this would be great if you'd specifically add a gender vocal here...." For me, suno always does the opposite, but the intent is there. It tries.

Then also, vocal changes work best if the lyrics or feel of the piece lends itself to duet, choirs, vocal changes, etc. Lyrics with I and us and we.... lyrics asking questions and responding... lyrics discussing or exposing a though or idea.

Just an anthemic song or an epic Viking passage would be harder to force a change of vocals.

If you could generate 65 million outputs, a few would be exactly as you wish, so it's just a chess game to find that balance between all three to get results in just a few less generations.

I spend a bit of time moving a prompt up and down the list of prompts to see where it triggers what I want ( such as getting a duet feel) without changing the whole feel of the song. So it can take many generations as well, depending on how many and in what position are your style prompts.

And occassionally, you get lucky and suno hits a home run within just a couple of goes.

[twin bough]

OK.. this is a LOT to read So give me a second. Wasn't expecting all this.

[nimble hemlock]

The problem is that some genres don't inherently "lend themselves" to a particular sex. I like EDM and EDM can work well with male and female vocals. But sometimes the lyrics can make a difference on who you need singing them.

Personally, I have also found that any attempt to specify sex will always result in you getting the opposite of what you specified. Even to the point where I specified "male" when I wanted female, and I got female vocals. So... /shrug

Also, "if you generate 65 million and a few would be exactly as you wish," 😖 is not Chess... it's chance.

[twin bough]

In using.. for example... Esemble Rock... when I actually want...Synthwave, or Retrowave (which are similar) it will then give ms Ensembel Rock instead. I listen to a TON of synthwave and that genre has a TON of both male and female vocals. Gunship and The Midnight for example.

"Then also, vocal changes work best if the lyrics or feel of the piece lends itself to duet, choirs, vocal changes, etc. "

The thing is, most (not all) of the lyrics are generated using ChatGPT and then I fiddle with them. So I'm to the point.. honestly... what I'm going to jsut say eff it and just suck up the fact that it's ALWAYS giving me a female vocalist and jsut wait another few years for them to figure out a better way to switch between genders. 😦

[twin bough]

THIS!! YES!! 100% true. EDM I would LOVE to have both genders... and this bloody ai can't figure it out. It's frustrating...

[polar spade]

Glad you confirm. Suno does the opposite of what we want and switches the gender.

The lyrics and feel / style can hinder getting vocals to change.

And yes... the idea was to infer chance... in 65 million generations. Such hyperbole comes with poetic license, as Suno is the one who is specifically using every combination it can think of and trying not to repeat.... but you're right. From our perspective... sheer luck.

[twin bough]

On that note guys, I'm sorry but I really have to head to bed. I'm exhausted and running on fumes. Feel free to ping me here and I will check the message. when I wake up. 🙂

[nimble hemlock]

Sorry, I didn't mean to nitpick. It was an attempt at lighthearted commiseration 🙂

[polar spade]

Off to lala here too. Until then. Later gaters.

[leaden pawn]

I must say, I am sick of seeing this, 4th time today...

[potent fern]

Hi hope its not too off topic but I generated a song with 17th century words from a poem and made it into a song with suno. I'd like to publish it on distrokid. I'm pretty sure the original music to the poem is now lost.. In distrokid I put it in as a cover and wrote the poets name. But this is a very unknown guy only known by a few experts on local poets from this part of italy.. Did I do the distrokid part right?

[rocky sinew]

Getting sick and tired of not beeing able to download WAV Audio (I´m paying). Anyone else that has this problem? Support are not responding...💩

[polar spade]

@twin bough I'm cherry picking 4 Duets. Using words like ensemble, broadway, opera, gospel and CCM, to nudge duets.

In v3.0 using ensemble or broadway, it really recognized the prompt and add alot of choral group singing, so that was out , no good, for a synthwave song.

I got a duet 5 times out of 22. But were first runs interestingly both, after I "cleaned and washed" the create page, by just running a dot ( . ) of nothing, so it played a piano piece. What this does, if anything, who knows. But it would gen a couple of duets once then did all male vocals for the next few outputs. So I cleared the create window, put in new text and prompts, and got a duet again. Just a placebo illusion, likely, until I can repeat consistently, but I'm doing that more often, thinking it helps.

Anyway, here is a group. Using ensemble and others after a list of 80s terms didn't affect the genre, ( using "opera gospel" did) but hopefully nudged a duet in the vocals.

An observation, one all male did have lower octave male on first verse and appears another male vocal on second... another had mid-range male vocals on 1st verse and falsetto (Charlie Puth style) on 2nd verse.

This one it sings the first verse twice male.... before going female on verse 2 and duet on the chorus... https://suno.com/song/567e72b6-9ce1-4455-b9bd-6aa428589af5

This one pulled a more hip hop feel from the "gospel" in the prompts, likely, so not good synthwave but still went duet..
https://suno.com/song/893cf909-aad5-4714-9aac-95e0816d8075

A more retro sound ( duet) https://suno.com/song/f5b60c92-051c-4933-83f7-b4b75ce02f3e

And Similar (duet) https://suno.com/song/2889aa00-52ae-44d8-8973-6391b948b9b7

Out of 22 outputs (11 create gens clicks) i got 5 duets, 22% , with small prompt changes each time and trying v4.0 and v3.5.

Had I made no small changes and just clicked Create 10 times, I'd guess 1 (of the 20) would be a duet.

[charred merlin]

If you're on a chromium-based browser (opera, chrome, edge), try on Firefox, otherwise email [e-mail] with the song links that don't provide .wav files

[polar spade]

To compare: with Android Samsung Fold 5, using chrome, I'll say I have NOT had any problem with downloading WAV/MP3 files.

It's a great help too, with any questions about issues, to first note what we're using and level of subscription(free, pro, prem -- you did say you're paying, so thanks) : ios, android, chrome, browser, tablet, abacus, etc.

[polar spade]

And in a shorted prompted one, I got a duet again, first run , with copy and pasted lyrics. https://suno.com/song/7cfa05d4-4e66-4966-8c99-34c0569edff5

Interestingly, the 2nd verse male vocal sang "Harry" instead of "Mary " on a first run (prompts typed in a cleared create window), but pasted in old lyrics. Where it triggers Suno to mix text like that, I've not figured out.

[drowsy orchid]

From free to pro

[brisk delta]

Would paying for a subscription and not getting credits be a tech issue?

[charred merlin]

Discord isn't an official support channel, you'd need to contact [e-mail] -- but no, not a tech issue.

[naive quail]

As harber mentioned, you should contact [e-mail] for issues like that.

This channel's name can be a bit confusing and is often the first stop for issues because of that. The channel is really for discussing the technical aspects of Suno prompting as well as the underlying technology that make Suno tick.

[brisk delta]

Thanks. They sent me an email and said to try discord. I guess I'll wait longer then and contact my credit card company to dispute it. Thanks again

[naive quail]

I keep seeing references to them redirecting people back to the discord. I'm curious if that was that an automated response, or was it actually a person telling you to come here?

[charred merlin]

I know they recently started sending out an automated reply to support requests, does it tell you to ask Discord for help? I haven't seen the message they're sending out

[polar spade]

Edit: i see the "Active account" notice. No date though. Email receipt is the way to go.

I was mentioning that your email would have the official receipt and every monthly renewal receipt after that. That would be the document best used when addressing billing with questions about your account.

Just search "suno" in emails.

[brisk delta]

It was definitely an automated email

[charred merlin]

It still shows 50/50 credits on their account page instead of the expected 2,500/2,500, presuming that screenshot was taken just now.

[naive quail]

Ahhh. Ok. So it's probably a case of "We're currently slammed with support inquiries and our response times are delayed... in the meantime, try the discord."

we should start billing them

[charred merlin]

Looks like furrybeans and lumine have their work set out fo them 😉

[naive quail]

...or just point at furrybeans and lumine and then hide

[woeful yoke]

Ok ladies and gentlemen... We have now arrived at the destination which I predicted almost 2 weeks ago. I was hoping we would never get here, but ... here we are. The sound of all models is completely screwed up now. Only exception being songs extended from uploads or from V2.

That being said, I took a couple of songs made with both 3.5 and 4.0, extended the tracks to V2, collected, and from there you can do covers with v 3+ or. Do remastering on that V2 rendition of your song, and you will get decent results which aren't dynamically compressed to death with a generous sprinkle of shimmers and noises on top.

I predicted that this would happen. Because I actually understand the root cause. But for some reason people were still hung up on me daring to say that v2 generates the best quality sound. 😎

[naive quail]

It makes sense now

[charred merlin]

Thank you for sharing the context -- appreciated 🙂 Hoping they're able to get back to you asap!

[brisk delta]

Thanks for all the feedback

[polar spade]

Peer-to-peer: isn't that, in my offering suggestions, just the blind leading the blind? Lol

[woeful yoke]

In a way that's really funny considering that we are dealing with AI experts here, yet they can't be bothered setting up a support agent bot, or something that would make 99% of common issues automatically resolvable with self-help from the end user.

[brisk delta]

Lol

[naive quail]

@brisk delta If you don't get an email reply after a few more business days (not counting the weekend), keep an eye out for Mike B here on the discord. He's the official community manager and might be able to poke someone in billing to look at the issue. He probably gets a ton of DMs but he very definitely responds to some of them some of the time. 😄

[hollow hare]

That would be nice to have in this server. Be like: try the bot first then us

[polar spade]

Yea. No date. No knowledge if the 2500 were used up yet before renewal. Hard to tell.

[charred merlin]

Would be nice to have some form of direction or contact with the support team, but I suppose anything in this channel is based on our own personal curiosities.

It's a puzzle that we want to understand, whether or not we have any means of following up.

[woeful yoke]

@hollow hare at the end of the day, you can have both a bot here, and then AI agent also replying to emails and/or there could be live chat with AI support agent. These kind of things are quite common already and not really rocket science. I imagine that 99% of issues are recurring and exactly the same from a
The Support Desk's standpoint. Whichever buttons the human needs to press to resolve a customer's issue, an AI (RAG) agent (with a couple of allowed function calling commands) can do the same.

[drowsy orchid]

They haven't answered my emails for a week.

[polar spade]

Just don't use Gemini to respond. It's always:

"I can see you are frustrated. I apologize. I'll try to be more responsive in my answer next time" Always a neutral response. Lol. One time Gemini said "Rest assured.....etc". Grrrrrrr..... The code word for "we have no guarantees but this script phrase seems to help end the conversation."

[charred merlin]

Remember when some Chevrolet dealership rushed to add an AI assistant to their dealership website, and it agreed to sell a 2024 Tahoe for $1?

[terse tendon]

Guys, is there anyone knows how to fix the "this is a fake app" issue?

well thanks for replying, but I'm using the suno.com website. This issue also happens on my ios app

[charred merlin]

Something in the automated system flagged you the same way it does people who do something suspicious -- it doesn't mean you did anything wrong, but reaching out to the team can help them figure out WHY the system is flagging you and try to prevent it for yourself and others.

[terse tendon]

thanks, I'll fill out the form 👌

[charred merlin]

Discord isn't official support, so we're just offering informed guesses on the matter, but we've been trying to figure out if it might have something to do with VPN usage, or shared network stuff

[charred merlin]

awesome; I hope they get back to you asap -- I know they've been swamped lately though 😅 From what I gather they're amenable to compensating for the downtime that you were impacted by the watermarks, but it still sucks in the meantime when you're in the mood to create

[terse tendon]

Well I guess it's not the VPN usage caused this problem. Because I signned up for a new account, and this watermater issue didn't happen 🙂

yes, I can't access suno without VPN cuz I'm currently at mainland China 😦

[charred merlin]

VPN usage might not cause an immediate issue, as it is like the internet wearing a disguise -- but perhaps prolonged or frequent use of the VPN might appear "suspicious" to the network if the disguise is different each time?

I think we're theorizing the same, @cedar thistle

[paper obsidian]

Is it just me having this issue or when you use a persona from a song and use close to the same lyrics they won't notice word changes in lyrics. Like I would change a word to something else but it still sings the previously used word

[terse tendon]

Thanks guys, I think this VPN stuff could be the problem... Guess I'll have to wait for the reply from the team 😦

[charred merlin]

I'm not educated enough on different VPNs, but I used to use one from Canada to watch Netflix content from America or different countries -- eventually the services would recognize that I was using a VPN and would give some kind of error message, but there were still companies that seemed to provide a stable service -- it might also be something to contact the VPN provider regarding?

[terse tendon]

Hey guys I think the watermark issue is somehow solved 🤔 I changed the VPN IP and suno somehow worked

🎉 Thank you guys!!!

[sand bane]

what country vpn do you use
sorry if my english bad

[terse tendon]

Well the watermark issue will happen if I enter the lyrics.... ：（

[terse tendon]

Hong Kong, Singapore and Taiwan

well I just need some guitar motif, the watermark issue actually won't be a problem for me anyway. LOL

[charred merlin]

afaik, personas are intended to keep the same vocal and instrumental style without outright repeating the song -- meaning the instances of it giving the exact same output would be something like "context regurgitation".

If you're leaning into using persona to change a specific word, from what I gather, you're giving it the original song to use as context which might make it hard to deviate too much from "what it's hearing as context".

It's not a perfect solution, but the easiest way I know to change 1 word is to use "replace section" and try to recreate the surrounding line/verse

[paper obsidian]

Basically I am just trying to recreate the exact same song but I fixed the lyrics. But it still keeps some of the words. Like its not seeing that I replaced the words

[naive quail]

On the duet front, I thought I'd take on Pur Ti Miro... I got my arse handed to me. lol. Getting the two to alternate every line is gonna take a LOT of extensions.

So I took some liberties with the arrangement and got a decent one-shot: https://suno.com/song/f44166cd-bf77-4d0d-8462-b2bb3338b27d

[charred merlin]

And to achieve this, are you using the "replace section" feature to change a small section of your finished song, or are you trying to use Persona to output the exact same song, but with the lyrics replaced in a small section?

[paper obsidian]

I think the latter. Since I am not going for exact replica. But I want the same style and voice which is why I made a Persona from the song to try and achieve that

[charred merlin]

I'm not quite sure that's how persona is intended to be used -- so while it's not impossible that it would work, it's like using the clawhook-end of a hammer to try to remove a screw

[paper obsidian]

I think I am just going to try and create multiple times and see if I mange to brute force it to work

[charred merlin]

That's certainly one way to go about it if you have the credits to blow

[paper obsidian]

I do

I think it might also be just giving me issues since the Lyrics are Icelandic xD

[charred merlin]

Just as long as I shared that it's intended to copy the voice and style, not necessarily act as a means of recreating the same song -- if anything, personas have an issue with the original content bleeding into new generations, so it might be harder to get it to use the context-clip it listens to when creating a persona song, and then ask it to recreate that all except for one specific section -- like getting it "on-rails" then trying to change the direction

[polar spade]

Does the VPN not supply the choice of USA, Canada, UK, France, etc. cities as a VPN node? May be a Chinese based VPN company, I assume?

[paper obsidian]

Honestly forgot to read this thing. Thanks for pointing it out
There is one great thing that I did manage to do with a Persona. I managed to make a persona that always has close to same first 19 secs every time I use the persona. These are the 19 secs I manage to get in 95% of cases

[charred merlin]

I'm wrestling with a song that gave me hallucinated gibberish lyrics, and I'm trying to turn them into the "correct lyrics", but the route-one methods of covering/personas don't always like to budge -- sometimes it can yield a beneficial change, but it definitely feels like a slot machine experience trying to get it right without a more involved approach

[charred merlin]

That's a shout to wake you up, if I ever heard one 🤘

[paper obsidian]

I may or may not be taking Animated Children Movies and writing Horror style lyrics and making Metal type songs xD. So yeah constantly remaking the same song till I get the right background music can be a type of hell xD

[umbral scroll]

I have a problem because I invited my friend, he registered from my link, he made 10 songs, but I didn't get the free credits that I should get after he made these actions. Is it possible to do something about this.

[brisk delta]

@cedar thistle sorry Alexander's, I was busy at work. It is tied to the same account. It won't even let me resign up because it says I already have subscribed

[polar spade]

A few have brought this up, but as no one has returned with news of how it was resolved for them, the only suggestion is to contact [e-mail] or perhaps [e-mail] to discuss it.

[weak sage]

Hi, personas is not available. Is it coming back any time soon?

[hazy cairn]

So sick of the vocal quality in suno right now, feels pointless to try make something

V4 is not in beta, its in alpha. This month feels so wasted to pay for, if I had known it was this I was paying 30usd for I would have waited with subbing

And v3.5 is so much worse now

And funny is that a team member said that v.3.5 quality wouldnt be affected, but it is, it really is

Which I noticed early

[dapper sluice]

Hello all, did anyone find a solution to songs ending mid word and such? Like a prompt to fix that?

no rush

[woeful yoke]

Guys... I'm going to contact suno devs somehow. I was going over songs from March 2024 and I now think that there's a 50/50 chance, that V3 wasn't screwed up after all but rather they might have a bug in the code that renders the songs during inference... But in order to potentially confirm this and propose a test fix, I really need to ask some engineering questions that the regular customer support really can't answer...

My main point is that I personally hope that there's that bug as the potential fix can improve the sound of all the model versions without having to retrain

The thing is still bugging me that there are two different ways I could replicate the sound issues suno has been having for a while now, not just this past few months. And I'm going to give up. Try to figure out as some things can only be answered or determined by talking to actual Suno engineers.

[hazy cairn]

Im going try ask for some kind of compensation from support, I got 6000 credits left with my sub ending 17th, and I feel really disappointed. It would be one thing if V3.5 was as it was before

each generation sounds the same

same vocal problem

[woeful yoke]

I actually got really excited right now as I suddenly started seeing the light at the end of this tunnel

[dapper sluice]

sounds hopeful @woeful yoke

nevermind on my above question. I uploaded a longer clip than usual and the exttends with the lyrics make it run over time limit

lol I really should take a gander at he workflow channel. I don't know what I am doing. I am composing by he seat of my pants lol

[woeful yoke]

The main point is that all the degradation in audio quality is a side effect of the Band-Aid but that was put on V3 some time after March. If I'm right about the root cause now, the Band-Aid could be removed, and then they would be able to regenerate everything that was generated with version 3 as those tracks now become higher quality so they can fine-tune version 3.5 or v4.0 straight. And this should then also return the sound quality where it used to be at the start. Actually even better because for me, the Band-Aid was kind of useless as I wasn't satisfied with the sound quality V3 to begin with 🙂
Anyways, I hope I can establish some kind of dialogue with the engineering at suno. Today I also identified and tested a couple of bug fixes for the mobile web UI on Android/Chrome.

Sometimes it feels really unusable

[left acorn]

can we use more than 1 persona to do duets and such?

[unique stone]

I don't know how much this will help but during one of the v4 feedback threads your efforts were mentioned to Mike B (customer support.) I added this reply. I'm not sure how open they'd be to discussing things outside of the company, but let's see if anything comes of this.

#1315841002577199144 message

[dapper sluice]

no and [duet] is hit or miss mostly miss

[left acorn]

ah hopefully in future updates they make it work better for that then

[hazy cairn]

I hope we can get some kind of compensation for how terrible this month has been on Suno. Im writing to support about it now. It's like if you had a subscription to Netflix and suddenly one month you get audio artifacts or the video lags, its just not acceptable

If V.3.5 would have remained the same as before I wouldnt complain at all, I just wouldnt use V4, but now theres no options

[charred merlin]

The devs say v3x hasn't changed, from what I'd seen.

[hazy cairn]

Yea thats a lie

[charred merlin]

[unique stone]

I specifically asked one of the devs a few days ago about this and was told v3.5 has not changed.

[hazy cairn]

Its just not from my experience, look around on the discord, many says the same thing

[hazy cairn]

Yea I asked it also, same answer

Its not me just saying this, many users has said the same thing

[charred merlin]

With such compelling evidence to back it up as well.

[hazy cairn]

Some has speculated that Suno removed copyright training data which was the reason 3.5 was so bad now

[unique stone]

I mean I asked one of the people that work on the models themselves, someone who would be involved in actually making such changes.

[charred merlin]

"speculated" -- not sure what you want to do with baseless theories in a technical forum

🤷‍♂️

IF they're removed copyright data, wtf do the meme prompts still work? snoop dog style, snoop style, doggy style, or adding DJ Khalend in the lyrics

[hazy cairn]

Yea, Im just bringing it up. That other users has been speculating WHY it is so bad now

Im not saying theres any truth to it

But the point im making is that its just not me saying this

I noticed the decline early on, and now I see many say the exact same thing

[unique stone]

The point I see often mentioned is that shimmer is now present in v3.5 when it wasn't before. Except it was always there, but the improved clarity of v4 makes you notice it in older songs (because once you hear it, you can't unhear it...)

[hazy cairn]

I just feel like I cant get close to the vocal quality I was able to do the month before, no matter how long I try

It worked great until V4 was added

[charred merlin]

all of the baseless conspiracy theories without any attempt to back them up is why I left social media -- starting to feel the same way 😂

"just saying -- awful convenient, huh?"

[hazy cairn]

But this is not baseless when many says the same as I am, that V3.5 is worse after V4

[charred merlin]

A lot of people say that the world is flat, but even a dedicated forum doesn't make it so -- show me any piece of evidence, not just saying "others are saying it" -- give me ANYTHING to discuss instead of our heartfelt opinions. Give me anything to test.

"all copyright training data was removed and now its trained on v4 songs" -- that sounds like a great reddit post, but unless there's a shred of evidence to point to corroborate it, it's just fantasy talk.

[hazy cairn]

....

[waxen maple]

yeah, i also noticed that aswell. and compared to v3.5, the v4 sounds much better. maybe that is the reason. you think v3.5 is bad now, because we have a something new to compare to. 🤷‍♂️

btw. i noticed that the shimmer is gone now in the last songs i made today. or maybe i just stop hearing it xD

[unique stone]

Hmm... Ok question: Do you get better/worse results if you aren't using a persona?

[hazy cairn]

I have still used the same persona, still bad results. Ive used this persona plenty of times before and had many good results, now I cant get anywhere close

I dont get better results without a persona either

[woeful yoke]

Well it all comes down to balancing any potential risks with potential benefits, doesn't it. I'm finding it really difficult to come up with a potential risk to them that might arise from discussing how exactly they implemented a certain piece of open source code (Which, by the way, was developed by the company I work for). There's no secret sauce, nor is there something proprietary in that part of the system, it's presume.

After all, suno open-sourced Bark and the parts kf the architecture that I'm talking about, should be, in principle, the same 🙂

[charred merlin]

That's not where I see the onus falling in this matter. I'm sorry you feel like things are so fundamentally different, but if you feel like you're being outright lied to by the staff/developers who are attempting to talk directly to the #1308499132356362383 forum, I don't know what you're hoping to really discuss here.

[hazy cairn]

No one from the team respones there

[charred merlin]

Only the head of community support, Mike B

[hazy cairn]

You just see people sad that quality is so bad in that forum part right now

It would be one thing if it was just me that complained that v3.5 is worse, but its many more users

But it dosent matter if I use a persona or not anymore, the results are just way worse

[woeful yoke]

Guys, I personally think the biggest problem is that there is no status page. Also no showing us specific versions or builds of models that have been deployed at any given time. It would be really helpful because otherwise people waste a lot of credits trying to get something to work without even realizing that the model has changed underneath.

[hazy cairn]

100%

[woeful yoke]

My main point is that people wouldn't make a big deal out of seeing a model build number change suddenly... It's just if you don't know that and we must go by the field then we end up with arguments about whether it's actually changed or not...

[hazy cairn]

I have many tracks to back up with what I was able to do

[woeful yoke]

I mean take any piece of software be that real software or softer as a service, You're always shown some kind of information about what version we are dealing with...

[gritty badger]

@woeful yoke exactly, I don't want to gamble by spending my credits and hoping there has been a wee upgrade behind the the curtains and hoping that it's not just a rumour

[hazy cairn]

It would also be good if we would know when updates would roll out, so we could plan subscribing or wait

[gritty badger]

Need to be a bit more transparent about these updates

[hazy cairn]

I started my subscription like 2 days before V4 rolled out all sudden

[unique stone]

I think it's more of an issue of how much they as a business want to disclose really about how things work.

I mean, there's the obvious way to be able to get involved which would be to look for a job opening - but I get the impression you're not looking for them to employ you but are a user with knowledge/experience that wants to just help make the service better.

In any case... Hopefully one way or another this will lead to a positive outcome.

[gritty badger]

The midjourney discord is a good example of this transparency. They keep people informed of every little change and improvements. Here is bit more of a guessing game

[unique stone]

I think they use the same for cover/persona. I may be wrong though. It feels like the functionality underpinning both of these is of a similar nature ("make something that sounds like X but different")

Knowing about every change would be nice, but the danger there is that the team might say "we made this update" and then users will blame every tiny defect on that update.

[charred merlin]

From his work developing Hipster, I assume Alex speaks from experience?

[unique stone]

I only ever looked into this for covers tbh.
Also cover and persona both have the reduced output length which leads me to suspect they operate similarly.
Sure-fire way to find out: prompt for theremin 😄

[hazy cairn]

But major updates like v4 should be informed when they release for sure

[unique stone]

"Today we rotated the domain protocols for even wear"
~ oh so that's why I'm not getting female vocals

[hazy cairn]

Did they beforehand say the release date of V4? I must have missed it in that case

[gritty badger]

I mean, just because a big majority of the users in this community doesn't understand the engineering part, doesn't mean they should be kept in the dark. Not hearing updates regularly causes more frustration to be honest. The sense of "at least they are doing something to fix it" is better than "It has been weeks and we still have nothing" mentality

[hazy cairn]

Yea but thats a big difference

If I had known the month I subbed would be the month of v4 I would probably have waited

We need dates

Not ninja updates

Build up can be 2-3 months before

We need dates so that we can plan if we should sub or not, if its going to harm the service, like in my experience it did

[peak frigate]

Good afternoon. I'm using the official app from play market, and the suno.com website. But at any generation I get a notification that I am using a fake app, although it is not true.

[hazy cairn]

But it is different now

Im not alone in saying this

[charred merlin]

Apparently that's a lie because people in #1308499132356362383 have banded together, sorry.

[hazy cairn]

Yea dont trust what the user base is saying

I brought it up early, as I noticed a difference early

Then time went and more people said the same thing

[charred merlin]

"the customer is always right" means "we should try to keep the customer happy", not "the customers are 100% factually accurate, even when they're admitting it's speculative"

[hazy cairn]

Just try make something with the quality of the tracks I posted now

[unique stone]

Yeah I get that. Just having a more regular reminder of things like "we're still working on shimmer, progress is being made, we're aware of the issues you're having"... I guess the announcement about "in 2 weeks the model will be better" followed by a period of silence doesn't help this.

Sometimes I do try to give my own reminder that things being worked on if I am aware of that being the case.

[charred merlin]

Are you arguing that Suno is now incapable of producing songs with comparable audio quality to the track you shared?

[hazy cairn]

Yes

[unique stone]

That sounds like your account has been incorrectly flagged as using a fake app for some reason (if you use a VPN, or there's a billing issue for example.)

Please fill out this form so the team can address this: https://sunomusic.typeform.com/to/zGsJSzsf

[hazy cairn]

I have many more with a similar quality

[charred merlin]

And if a track was shared that exhibited comparable audio quality, you'd concede your earlier points? Or is this an arbitrary request made to bolster your claims?

I asked for proof that anything had changed, and yet you're asking me to prove a negative.

[bronze sail]

Is anyone having issues with regenerating videos? Every single time I try with any song now, I only get "Video Generation Timed Out, Please Try Again".

[hazy cairn]

Im bringing proof what I was able to make before, but no matter how hard I try now it all sounds consistantly as bad

And this is after V4

[charred merlin]

I'm sorry you feel that way -- if it's such a night and day change, surely this will be the only thing people can talk about and they won't last much longer.

It was fun while it lasted.

[hazy cairn]

Well check the V4 feedback forum and see how many are dissapointed

Im not alone in saying V3.5 is worse now

[charred merlin]

Are the discussions there more or less nuanced than these exchanges?

[hazy cairn]

Well you cant prove me wrong, but I can show what Ive been able to make now and I can show other people sharing that feeling

[gritty badger]

If you are using the web browser, I had the same issue before. Apparently after requesting the regeneration, if I exit the browser and do something else on my phone and come back again, I get that message as well. But if I request the regeneration and stay on the page for 15-20 seconds and then download the video, the process seems to have done its job. I think exiting the browser before the generation ends in the background causes it to time out. Just try waiting a bit and then check if the regeneration has been done

[bronze sail]

That's what I've been doing. Usually if I remain on the page without interacting with anything, it usually does it's thing. However, it seems to be stuck on this constant loop of not working properly.

[charred merlin]

Check the amount of users happily making music with good audio quality, who recognize that not everything is as neatly compartmentalized as you're currently making it out to be.

I'm sorry that you're having such a bad time with Suno, and that you feel like differences in opinion amount of calculated efforts to lie to you. 🫂

[gritty badger]

Hmm maybe clearing the cookies etc might reset things. Try doing that

[hazy cairn]

I been looking at what people has been making, the same issues are there that I hear, the vocals sucks

[lilac garden]

Is it just me or the extends on V4 lead to different voices? Anyone has anytips on this?

[hazy cairn]

its not just you, this has been brought up by many in the feedback channel

[lilac garden]

Argh! 😦

[obsidian frost]

hi, is there a way i can insert a instumental bit after my lyrics?

[lilac garden]

[Instrumental Interlude], or if you want specific instruments [Guitar Solo] for example

[lilac garden]

And is not just the extents, is also the Replace.

[restive yoke]

Hey, guys, I suddenly can't listen to songs on discord. What may I have done? What could I check

[charred merlin]

Can you elaborate? What happens when you click the play button on a song over in #🐣┃suno-showcase ?

[restive yoke]

Nothing, and the blue isn't moving
Edit: It works again

[gray summit]

Hi all. To me it looks like Folders will take some time until they gome. It doesn't seem like there is such a think on the top of the backlog. My question would be: How do you organize your stuff? Playlists? Any other idea? Playlists have the issue you always have to add new generated stuff to a playlist in order to keep it organized. So I wonder how you guys doing it?? Any idea is highly appreciated

[lost scaffold]

Trying to distribute some tracks through SoundCloud but in order to be approved, they are requesting documentation showing I have the right to distribute. Has anyone been successful in obtaining this documentation? I have already reached out to Suno support but would like to hear any testimonials.

[charred merlin]

You don't get song-specific documentation from Suno -- you show the timestamps from when you'd made the songs (you might need to show individual extension segments if requested, or they might accept the "full song"), as well as the receipt from your Suno subscription showing that you had a paid subscription when you made the track

[charred merlin]

I use playlists to organize stuff after the fact, but my method is using a randomly generated word for each "session" -- pick a new random word each time you start exploring a new thread, or each day, or whatever works for you -- that way you have a unique keyword to search for when trying to find all of the related segments

It's been neat going back to look at the "original sessions" after I've finished up a song, because I can see what other songs were made during that same creative sprint

[leaden lark]

I know that perfectly well, I thought less organized. I have pages of just gens that I never used but have the same lyrics from stuff that I eventually made public. It’s cool seeing the ideas I was going through before I stayed on a particular gen

[naive quail]

I do much the same. If I don't like a gen, it goes right into the trash.

I prepend "Intro" to any tracks that I think might make a good intro and have a couple of other tags like that that I'll use - like when there's an interesting hallucination or SFX.

My library right now is about 25 pages and needs some pruning. My trash has been up to 1700 pages and is probably several hundred pages right now since it's been a few days since the last wipe.

w00! only 153!

I'm slowing down 😄

...can probably thank v4 for that. lol

gen > shimmer. gen > shimmer. gen > shimmer.... Walk away in irritation and pick up a guitar.

[gray summit]

this is genius. I will use specific words for each song. thanks

[naive quail]

They need to release workspaces already, tho I'm already dreading the bugs. 😄

(the Workspaces menu item has appeared a couple times on the Live server)

I see you taking a sudden interest in your fingernails...

haha

[wild spade]

hi guys, i have the shimmer problem when i make a remaster from 3.5 to 4. Expecially the last part of the remastered song, those high frequencies makes me angry

[naive quail]

we don't have a secret handshake, but welcome to the Shimmer Club

[prime anvil]

Ringing*

😄

[wild spade]

sorry ahahah yes

[prime anvil]

Shimmer is old

[wild spade]

Theres no solution right?

We'll just wait the developers

[naive quail]

Nope. All we can do is wait

[wild spade]

Ok................. sad

[prime anvil]

not atm, there's a lot of stuff going on under the hood, but since they're not saying when they release an update we can only listen closely and hear if we "see" any difference

[naive quail]

and try to stay under the shimmer's radar

[prime anvil]

the big problem of ringing is that it's able to make your ears bleed for so much HF is using

[wild spade]

Indeed

[wild spade]

OMG

[prime anvil]

this image was done by @woeful yoke analyzing a song

[wild spade]

My dog is crying for those HF

And i don't have a dof

dog

[prime anvil]

https://www.youtube.com/watch?v=Yi4UXqAIM0I

this is what i mean for ringing xD

go at 1:20

[wild spade]

this is ASMR in comparison

LOL

[hard sable]

Heya got a quick question if anyone knows. If you make a persona with shimmer in the song, are you forever cursed with shimmer when you use that persona?

[modern pecan]

Is there any way to force the gender of the singer? Putting it explicitly in the style description doesn’t seem to work

[warped furnace]

No, not really. You can put in style and then in the lyrics box [male voice] but it's no guarantee that'll work. 😁

[modern pecan]

Thanks, hopefully this will be a future feature

Curious, could personas work for this?

[warped furnace]

I think it might work. But I've never used persona. Sometimes the style of music leans towards a gender. Or the lyrics sound female or male (at least for the AI).

[untold briar]

Style: "male, male, male" Exclude Style: "female, female, female". Persona is a good suggestion as well, and you can use both.

[warm pendant]

so I came here to see if anyone was having the high frequency issue with later parts of songs.... seems I'm not alone... It appears for me mostly after extending. I'm not getting the "shimmer" so much anymore but this high frequency thing is ruining otherwise great takes on my latest songs...

[dapper otter]

Is there any trick to consistantly create a duet and alternate male and female voice in the order wanted consistantly

[worn flicker]

Suno loves to flip genders. If you want to write gendered lyrics (he/her) then I would do it in sections. You can get it to flip mid lyric but it is hard to control which vocal will go.

If you do it in sections you can just use tags and extend each time it goes wrong.

I did a little tutorial here: #1314766265897717780 message

[glad badger]

when remastering a song, does Suno take into consideration any alterations added to the original song by editing the lyrics? For instance if I add something like [Reduced Instrument Volume and Noise]?

wasn't sure if there was anything to play with for that or not... I know it's not as transformative as a cover, but it definitely does some sort of nudging when it remasters, one way or another

[worn flicker]

It takes Minor changes to lyrics into consideration. I just remastered a song that I had to go back in and put the phonetic spelling in the original lyrics before remastering because it would mispronounce the Italian word "pace" since obviously it would say the English version. It will hold the same sound and you can get Minor pronunciation changes consistently, but as far as tags go, I don't think remaster will go that far from the original.

[glad badger]

that's what I've noticed as well... back to extending then! 🙂

[worn flicker]

Yup

[eager topaz]

I would say "yes". That's how it is for me.
I think it would be the same as uploading an audio file with background noise. Suno would see it as part of the composition. My thoughts, anyway.

[pale bear]

I took a music-making class with Suno in my school today, and I signed up with my school ID. But the whole thing doesn't go in. Shouldn't Suno allow us to do our school domain?

[untold briar]

If you had a class with Suno, that must be a special event. So your situation may be unusual. Are you saying that you logged in at school and created a personal Suno account. Then you created songs. Then after, you went home and tried to login to Suno, but the same ID you used at school no longer works?

[odd delta]

How can i get my invoice or bill for Suno AI subscription??

[polar spade]

Normally, it should be emailed to you each renewal date.

Oh, I would be talking about your receipt of payment. I do not know of a literal invoice document, other than your subscription renewal date noted on the app/site under your profile.

[tiny cobalt]

what's the problem with making a separate payment plan for the API? It's profitable suno...

[mortal kindle]

Is there a way to download a regenerated image for the cover at 1024x1024 size instead of the initially generated image when creating a song?

[pearl sorrel]

Help, I haven't received my daily free credit for two days.

I'm pro subscription

where can I ask about it?

[regal dagger]

Hello, all my songs start with the words about fake app, although I use your official site. Can you tell me the reason why I can't fully use the subscription for which I paid money?

[bronze lodge]

I don't know who to turn to. I can't write songs. I paid for the month and spend my money listening to the commercial at the beginning of each song. It's disgusting. Do something.

[trail bone]

why the Hell does Suno keep refuding my credits if I try to extend a song...? Im a free user! I know Suno is not perfect but please look into this! I LOVE SUNO ❤️ it works sometimes and not

[autumn vault]

Is there a news when we got the fix for the sound quality of V4?

[slender crest]

I don't know if anyone else has this problem, but Suno is generating songs without vocals, only instrumental... and I don't have the instrumental only option checked. It also often happens that Suno abruptly ends songs by cutting off the lyrics, these things consume my credits and I am paying a subscription. 🥺 😭

[hollow hare]

Subscription members receives their free daily credits (at refresh time) once all Subscription credits are exhausted

[flint haven]

Free Credits Refresh

Refreshes at 0000 UTC+0 (<t:1713222000:t> standard local time)

[hollow hare]

Can you link to one of the issue songs so we can see the prompts? You can DM it to me if you want

[hollow hare]

It is still undergoing various adjustments and shadow updates. I believe the team will post in Announcements once everything is done to a sufficient degree

[hollow hare]

That sounds like your account has been incorrectly flagged as using a fake app for some reason (if you use a VPN, or there's a billing issue for example.)

Please fill out this form so the team can address this: https://sunomusic.typeform.com/to/zGsJSzsf

[hollow hare]

See my reply above

[hollow hare]

We moderators and guides are not part of the Suno team so we cannot say on their behalf. If you wish to know, please ask [e-mail]

[molten ledge]

hi, question. i've generated a song and I don't like the singing but the music is dope, can I somehow ask suno to get rid of the voice and just give me the music ?

[hallow elk]

now, i dont have anything against gay, at all... but like, a part of producing you HAVE TO deal with stereotypes in anything, and with that, if i have anything remotely love lyrics, it's incredibly biased towards gay or female in the jazz genre.. even if i write male singer 600 times, it still doesnt listen, we should be able to write what ever we want in which ever expression.

"straight male" as a prompt is alternatively too condescending... id rather it just give me ya know all male but different styles, but like cmon, do i have to write that?

[hallow elk]

yeah theres a option for that....

[molten ledge]

I do not see what you mean but maybe you cant ask for "things that make somenone sound straight" whatever it is. like maybe deep voice or something

[hallow elk]

its called stems

[molten ledge]

personally I hate "auto-tune" turned up to 11

[molten ledge]

i'm gonna loook into it

ty dude

[hallow elk]

np

..... im writing straight to see what happens anyway

... same thing

[molten ledge]

just ou of curiosity would you share a song that sound "gay" ?

also fyi straight jazz is called ... blues 😂

[hallow elk]

lololol

this might be the straightest one i have lol...https://suno.com/song/0dc20b79-9b2c-474b-9c39-fe433c8b55c6

[molten ledge]

beside the fact that the voice seems genderless half of the time I do not hear the gayness in that song.

[hallow elk]

yeah...

[molten ledge]

better ? : https://suno.com/song/53640f3e-f5c2-4067-9993-ec79b0387a74

[hallow elk]

a little better, i appreciate it, but the common stereotype is the, ya know "show", the enunciations, phonetics are still like that a little bit... maybe ill type one with exluding styles "show"

[molten ledge]

yeah sry, I know nothing about that.

[hallow elk]

yeah, we're only just assuming at this point

[molten ledge]

I guess that's frustrating cuze you're like " I love jazz" and suno keep tellling you "bro that's gay". IA ar bias as F, garbage in, garbage out

[hallow elk]

yeah.. im afraid of that...

there's also probably plenty of gay people who like chet baker and buble... to just a common male voice

[molten ledge]

had to look it up

[woeful yoke]

Guys, has anyone noticed that music generated in certain styles has more shimmering and is much more noisier than music generated in other styles?

It's not just random. To me, there are clear signs that specific styles are affected more than the others. Or am I imagining this?

[hallow elk]

i want to say, that when you're expirementing it shows up more often, but like songs with sustain particularly, and~ complicated amount of instruments, and or~ electronic to complicated beats, the renderer will throw those artifacts

like, the more randomosity you put in comparable to the amount of music thats out there for it, the "shimmering" is a node of invention...

[woeful yoke]

Okay, but let's forget about v4 and the newer shimmering for a moment.

Let's only look at artifacts and noise present in higher frequency bands in almost all songs generated with v3 and v3.5 models.

[hallow elk]

speaking of artifacts, excluded styles seem to leak as an included style..

[hallow elk]

i... of... like it haha

[crisp sparrow]

Anybody ever had trouble getting an orchestral kind of sound out of Suno? I've been trying to make a song featuring just lyrics, an acoustic guitar and a violin, but after a countless amount of attempts, it's just not happening. Style prompts & meta tags I've tried include Violin, Orchestra, Orchestral, Strings, Symphony, Symphonic, Classical... literally not one has worked, and all I ever get is the guitar. It's like Suno doesn't even know what a violin is... 😢

[naive quail]

Suno definitely knows what a violin is: https://suno.com/song/6440bf2b-1b84-45e9-a569-fd68ad76b5e7 You could also try fiddle.

Orchestra and Symphony are likely to pull in a ton of other instruments, so you'll probably want to stick to calling out just the instruments that you want in this case. If you're having trouble getting the violin to show up, try putting it at the front of the style list... and maybe even have it in there twice as well as in Tags within the lyrics box.

This one has a singer, guitar, and violin. It goes in a weird direction, but that's just Suno doing it's thing when left unattended and fed some prompt noise. https://suno.com/song/c0351580-c603-449f-abe1-2ee5f1b854a6

[tiny falcon]

Yes. for me especially instruments with synth, wave, ambient, rain.

Good morning @naive quail

[naive quail]

I got up for work... did my normal morning routine and was almost ready to head out to the car when I remembered that we're on holiday break until the 6th of January

[tiny falcon]

Violin is hard to get for me 🙂

[tiny falcon]

wth. You're a workaholic man. lol

[naive quail]

I haven't prompted for it much, but it may be one of those instruments that's easily overwhelmed by more inclusive styles. 'might need to overprompt for it, or break some of those bully styles down to their core elements.

[mild yew]

Suno incorrectly flagged one upload as containing voice (and it does not) so I can't make the songs I created from it public. Is there anything I can do to change that classification?

[naive quail]

Nope. All you can do is try to tweak the audio clip. 😦 I've had that happen a few times with guitar uploads.

Suno probably thought they were screaming children 😄

[mild yew]

Mine was a recording from a modular synth...

Pads and arpeggios

[crisp sparrow]

Thanks, SixStringToker. Violin was definitely the first thing in the style list, I even wrote it in all capitals as I'd read somewhere that it could help. Hadn't tried putting it in the style list more than once though; I figured including too much stuff would probably be counter-productive.

(P.S: "Lightly salted" as a style amuses me! 😆 )

[naive quail]

re: "Lightly Salted" - I'm always trying to force hallucinations / weirdness and have a bad habit of leaving that stuff in when I'm testing other things. Often bites me in the arse. lol

[acoustic vigil]

Hi there, I would really like to be able to make music with Suno AI for my voice, but I'm a baritone and I don't know how to tell the AI to make me maximum tracks up to E4 or F4, because G5 is already high and sounds pretty high, I don't know how to make it make me minor tracks, although I write there under style of music ,,minor" it still makes them for me and still too high for my voice. Can you help me, please? Thanks!

[raven turtle]

Yep, the layered Synths and vocals in synthpop and electroclash really get that shimmering.

[naive quail]

I've heard tracks from others with baritone voices but I don't recall seeing if they actually prompted for that, or just got lucky. It's likely very strongly influenced by the rest of the style prompts.

If you do ever get a deep voice, persona that sucker so that you can reuse it! I have a "Dark Lord Kevin" persona that I made for an evil voice test that's actually given me some gems when I use it for music tracks: https://suno.com/song/f3c6397e-505c-42c3-9a65-52bfb24c5d43

[raven turtle]

I've had the same issue. Best thing I've done is either upload audio of myself and get it to extend from there so I get generations in my range. Or go with "male baritone vocal" and swap out the vocal using Kits.AI with a voice clone of my own voice.

[acoustic vigil]

can u teach me please?

Do u have instagram?

I'm beginner

I'll be thankful to you

I generated a beautiful song, but the range vocal it's too high, but the song is masterpiece. I'm singer

with ,,male baritone vocal" unfortunately it's same problem. I think the high note it's A4 and it's exagerated for me like baritone 😦

[raven turtle]

You could try:
Export the wav.
Take it into Audimee or Kits and it can analyse the wav and you can choose a lower range AI vocal to replace the main vox. You drop the range down an octave or two. You have to fiddle with it but it's doable. Or export an instrumental version of the track and you record yourself over the top of it.

[acoustic vigil]

so download entire song and after import in kits?

[raven turtle]

Yep. That's one way of doing it. Or Audimee. If you've Logic Pro or another kind of DAW you could download the wav, split it in Fadr, and import the instrumental version and sing over it in your DAW.

[acoustic vigil]

I have Cubase

not a problem, i have studio

the problem it's juist range vocal too high, i want to key change but without affecting the quality

[raven turtle]

Try Fadr. It can get you an instrumental version pretty clean. Drop it into cubase and away you go

[acoustic vigil]

fadr to transpose the key of instrumental?

becaus suno it offers me stems for the song

like instrumental

[raven turtle]

Do you need to transpose it? I just thought it was too high an octave?

[acoustic vigil]

no no

i just want -2 semitones

-2 semitones it's ok but affects the quality of the song...

and i don't know how to do it in Suno

[raven turtle]

Yeah, you won't get Suno to do that.

Might get away with it if you split the stems and transpose the "bass" and "other" stems down, keep drums as is, and throw out the vocal stem. But... Hmmm. Not easy to keep integrity of it.

[acoustic vigil]

yes, but i don't know how to do it :))

existing applications for that?

[raven turtle]

Once you split the stems in Fadr, Cubase should allow you to transpose the audio stems you want

[mild panther]

I recently deleted my Suno account with the username donfares without fully realizing the consequences of this action. I would like to kindly request your assistance in recovering my account, if possible.

[gritty badger]

I usually work on Rock and Metal genres and 90% of them suffer from the shimmers. I think especially the cymbals get affected by it.

[hollow hare]

You will have to get in touch with [e-mail]. No one on this server will be able to help you with that

[charred merlin]

I'm relatively certain that if the account was deleted using the "Delete Account" feature on the app, it might not won't be recoverable -- just to set expectations 🙏

[hollow sapphire]

That! That's boring!
In the term delete, there is? Delete!
Yes! So not having a backup like on the Windows OS,
It seems complicated to be able to recover it.
The only possible reality is to make a new account ^^
So ... 🧙‍♂️

[viscid badger]

Out of interest, does anyone know if there are plans to add advanced features like download mp3/wav, extend etc. to the Android app as well?

[hollow sapphire]

Everything is explained here ^^
*This is the first version of Suno for Android. We wanted to get it into your hands quickly so you can start creating today. While this release includes our core features, we’re just getting started, and there’s much more to come! *

#📢┃announcements message

[viscid badger]

Perfect, thank you!😍❤️

[stoic mauve]

Kann bei suno in der Basis Mitgliedschaft meine Songs nicht teilen warum

[dreamy magnet]

leute hört jemand meine musik

https://suno.com/song/6266f129-a74b-4512-932b-7f1ced8d8e77

[crisp sparrow]

Gave up. Going to stick with the guitar only! 😆

[naive quail]

I support this choice!!

[woeful yoke]

That's what you get when your high quality training data is lacking and you also make wrong audio pre-processing choices which accidentally end up making the already kind of poor quality training dataset even worse...

And later the model will be predominantly generating Songs with poor audio quality, as that, what most the training data contained....

However, we know for certain that those models are still fully capable of generating good quality audio as well....

This is an example of relatively clean and artifact/ noise free audio v3.5 just gave me.
https://suno.com/song/8eb4f062-10f7-4140-9800-a853d60f5e4a

If your condition is a generation with high quality audio (in other words, you extend from high quality source, audio or song) then it's almost guaranteed that you will get the rest of the song as clean as that.

This particular statement holds true for both version 3.5 and 4.0 right now. Namely, there's very little difference between how this particular song sounds in both models. But at least with this experiment I'm able to show that 3.5 /4.0 aren't inherently noisy.
In fact, I have a strong suspicion that 4 is basically actually 3.5 with additional layer on top that's doing the 4.0 mastering (it also seems to be activating when using 4.0 to create a cover and also it seems to be activating during when you use the the replace selection tool).

If you extend a song from 3.5 to 4 or back the opposite direction. The results are essentially the same. But things get even more interesting when you extend a 4.0 mastered track into 3.5. the result will be exactly the same as was this song on 4.0 right before you remastered it. That's why I'm thinking that 3.5 and are essentially one and the same model which also explains why at one point both started degrading in quality at the same time.

And here is the same song in 4.0
https://suno.com/song/c6bd0801-0523-4854-bb96-0682b5e41f74

When I open both up in audio editor then both waveforms look essentially the same.

And here is the one that was used as the starting point:
https://suno.com/song/ce8c460e-e02f-4e5c-87bd-83bdb97a1efd

I burnt through about 1000 credits right now to do this analysis to figure out how those models behave 😂

Because I have no songs to make anymore

[gritty badger]

You might actually have a good point here. I tried to experiment by making a song with either V3.5 and extend with V4 or vice versa and tried to see any differences and didn't get much except the shimmer was very little with the V3.5 and the sound was dull. But when I exported my song and fiddled with its EQ's and then imported back to Suno and tried to extend a bit more, the extension followed the quality of the uploaded version. So yes, there is something to it.

[woeful yoke]

By the way, even if the training data contains a lot of really noisy music, then if they were willing to change on how they are changing the music before training and especially at the very moment when the model is generating a song, you could already start getting better resulting songs out of current models, even without retraining anything. Althugh, retraining would just make those improvements more permanent.

I will try to get someone from sunk to talk to me because there is still one mistake. They are making human with version 4 and especially with remastering. So even if you get sharper and cleaner audio, there's a cumulative effect that another catastrophe will happen if outputs from version 4 would be used for training 4.5 and above.

I wish I could prevent them from repeating the same mistake over and over again...

[primal sierra]

@woeful yoke Hi. I read that you found a way to fix the broken suno.ai . I saw that you used a neural network. Can you tell me where you got a lot of music to train this neural network? I also want to train a neural network to fix broken songs, but I can't find a large enough number of songs for the neural network to learn on.

[woeful yoke]

You have been misinformed but the thing is you don't need much. I'm just facing a dilemma that instead of creating something to clean up what AI generators generate? I'd rather help them to clean up the generated music at the source,

There are some things you can't really easily fix without training a whole music generation.

And also the reason why I don't need much training data is that I'm not looking to cover the whole audio domain, but rather tackle very specific audio issues created by very specific source, meaning that you don't need to train a huge model. The downside would be that it would be kind of useless for processing or improving. Anything else generated elsewhere

[primal sierra]

I agree that it's better to teach suno to generate good music right away. But while it's broken, I'd like to somehow improve the songs I already have. So, I'm looking for ways.

[primal sierra]

I took my 3 songs that were generated before the high frequencies appeared in the second half of the song, remastered them, got exactly the same songs, but with terrible high noise. And I tried to train a neural network on this data. But it didn't work. The output was terrible quality. So I wonder if I took bad data, or not enough ))

[woeful yoke]

Also, I didn't finish it because I realized that my time would be better spent trying to find a way to get them to fix another bigger problem. So whatever I was able to clean up was kind of part of my experiment but that specific shimmering artifact is not common anymore. And the rest of the noise and other quality issues can be fixed only at the source. V4 remastering alone is not going to fix those. There is a change needed in how all models generate the basic tracks. One specific change can be achieved without even touching the model weight, but rather implementing a non-ai-based simple audio Post-Processing right after rendering. Using that, it's possible to undo certain bad things that has been trained into those models. And then that particular post processed audio can be fed into version 4 remastering or used as part of training data for future models and gradually things will get getting better and better

The reason why I'm not being more specific right now is that I haven't been able to have a chat with them yet. If nothing good comes out of that I will implement my suggestions as an independent too instead,.so it would be usable by everyone, everywhere.

[primal sierra]

I tried to remove the high-frequency noise using izotope de-reverb. It doesn't work.
I noticed that this high-frequency noise appears mainly on the difference between the right and left channels, I selected the central channel, the left channel and the right channel. And I tried to remove this noise from there, it didn't work very well either.
In general, I'm confused, I don't know how else to fix the songs created by suno. But in general, I really hope that they fix their model (I hope this is still possible)

[primal sierra]

It would be great if you posted your solution somewhere. I would be very grateful. I think many would even agree to pay for using your developments.

[austere dragon]

Anyone else getting [prompts] within the lyrics box being sung?

[vivid stag]

When i use my voice in suno it doesn't keep it and instead it uses some other voice or it only uses my voice for the hook

Anyone know how i can fix it

[charred merlin]

Happens more frequently when you overload the metatag -- 1~3 words seem to be the most effective, anything beyond that start increases the likelihood that it gets ignored or sung as lyrics

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

[silver wind]

Hebrew dosent work properly on v4 are you working on a fix??

[queen knot]

I changed the background image on a song, and when I click regenerate video it's not using my new background. How do I video of a previous song with a new image I uploaded to it?

[naive quail]

When you say that it's not using your new background, do you mean the video downloaded to your phone or computer, or the video as displayed here on discord? If the later, it's because Discord has cached the original video.

[charred merlin]

You can download the video to see if it's been properly updated, but Toker brings up a good point about discord caching -- you can add a / to the end of the song link to make discord cache a separate version (or at least that used to work, should still work)

#1308617723328204880 is also a megathread regarding that exact issue 🙏

[naive quail]

the slash trick has never worked for me. That only works for me when I initially post a track link before the video has rendered and I want discord to reprocess the url once the video is done.

[queen knot]

The video downloaded to my computer does not show the new background image.

[charred merlin]

@naive quail lemme test this out

https://suno.com/song/654a81b0-1c8a-4016-8120-962d127cbaed

[naive quail]

The video generation is probably timing out in this case. I often have to retry it several times to get it to generate the video

[queen knot]

Ok thanks I'll try again later

[charred merlin]

https://suno.com/song/654a81b0-1c8a-4016-8120-962d127cbaed/

[queen knot]

I use Brave Browser I'll try it in a different browser

[charred merlin]

Oh yeah -- it only updates the video display image, but the actual background stays the same

I reported that to Z a while back (note how I said "should still work")

[naive quail]

Yea

[queen knot]

Ok yeah it worked on Chrome

Not on Brave

Thx

[naive quail]

must be a plugin or something since they're both chromium

or just luck. It does often take me several tries

[queen knot]

I've been having a lot of weird issues with Brave lately like even my Amex Gift Cards I got from work sometimes I can't log in on Brave but I can on Chrome

Noted

[stiff torrent]

Any option to download my library via FTP?

[naive quail]

Nope.

[stiff torrent]

Would be helpfull

[naive quail]

It still exists! I heard someone else mention it a year or two ago! 😄

[coral epoch]

I'm sure someone could do a userscript/extension for a browser to manually step through a Suno library and download each song - but yeah I very much doubt any sort of file interface is coming

It's very unlikely the music files are even saved within directories as you know them

[charred merlin]

Someone may have created a song with the steps to that exact thing in the lyrics section

[naive quail]

coughs

[charred merlin]

Harsh pull?

[quartz jewel]

Yo, I’m trying to make a song like the meme: https://youtu.be/rdi7G1hY4-w?si=85EEx4I_5ee9_Bqd

How would you guys go about writing the Style?

Wanting it to have the accent and everything, I’ve been trying for half an hour or so and it’s always been an American accent and not the same instrumental style

I want it to be raag Bilawal style but it won’t work with that

[woeful yoke]

Lol, I think I just made the V4 remastering processor crash with my 10-minutes long track experiment 😂

I wanted to see how far can I stretch the thing before I lose either the original bassline or its ability to sing or perhaps the drift towards the noise will end the song first.

Ooh, it actually finished successfully. Looks like it was the UI that was still spinning

So, basically the result is shit... But hey, manure is the fertilizer for the future.

https://suno.com/song/2d14a416-ac91-4329-9f48-1c55d39e7ff9 😆

Interestingly, throughout the quality drifts between acceptable and noisy, but only around every two minutes. It's probably your extend shorter segments. The quality would stay more consistent.

[ornate spire]

How do you get replace section to actually blend the replacement into the existing segment??

[woeful yoke]

I have no idea why I'm wasting credits on this kind of crap 😆
But somebody's going to do it to test its limits

[coral epoch]

It was doing a good job of that a few weeks ago and now it isn't. That's how it should just be working

[ornate spire]

so frustrating. I think a song is perfect then notice a choppy skip somewhere in the middle

[woeful yoke]

Obviously you can't do it right now... They mistakenly remaster the newly generated segment which will stand out from the rest of the track

[ornate spire]

Just gonna have to extend from before that point I guess

[woeful yoke]

At least with 4.0 you can't maybe try selecting 3.5 when recreating a section

[coral epoch]

It's really funny when you're trying to replace a 10 second segment with the keep it the same length box selected and you end up getting 2 minutes anyway

[tranquil estuary]

Sucks just as much as when I get the perfect vocals except for a mispronounced word. There seems to be no way to just edit that word without the instrumental completely changing for that part

[woeful yoke]

That's weird by the way. + I don't really understand how this could even be because let's say I'm a developer + if I know that there's a slight chance that the model isn't going to generate the end of sequence token to tell the system to stop generating, you could still measure the number of tokens and the length that has already been generated + stop the process when your newly generated segment has already gotten longer than the range it was supposed to generate.

Completely relying on the model to tell the generator to stop doesn't seem to be a very good strategy until the model is still in alpha/ beta

[ornate spire]

neural networks kind of tend to do their own thing. Adding too many constraints can limit their creativity quite a bit

I would love for them to give us the ability to set and modify parameters to fine tune segments. Give us a scalpel instead of a sledgehammer

[coral epoch]

There are certain circumstances where creativity isn't desired

Remasters and segment replacement would be two examples

[woeful yoke]

Yeah you're talking about reinforcement learning or something like that which may affect the ability of larger language models to reason and so on...

But when it comes to audio generation then I wouldn't call training the model to spew out EOS token at the appropriate moment, too much of an ask 😂

I've noticed that quite often it keeps on going when there's an issue with your selected range and lyrics. If there's a larger gap between, let's say two choruses and you're replacing something around there, then it might run past the first one and try to reach the other one..

But from what I understand, the recreate section part seems to be pretty bad right now. Most likely related to its inability to properly transcribe existing vocals to match them to the lyrics.

For example, when I did purely audio segment replacements earlier today, I never had this issue of model going further than needed.
It already started but I want it and then it where I wanted it. The only problem was that it changed the sound of the original which is a completely different matter

So yeah, replacing lyrics in already generated song is a little bit more challenging than just replacing parts of the song. That is mostly because it tries to "listen" to the existing lyrics in the song and then when it matches what transcript says then it starts regenerating the replacement instead of that. Let's now say that it started replacing, but for some reason could n't correctly transcribe the last word or last sentence after which it was supposed to stop so it keeps on generating and listening because it's still hoping to finally encounter the phrase at which it should stop.

At least that's my understanding of how it works because If you're trying to recreate sections of vocals in a check where your lyrics transcript doesn't match what's in the actual song, then it's definitely going to fail. Even their guide says that.

[hallow elk]

uhhh... it glitches playlists and personas to where i cant delete absolutely everything, playlists cant be permanently deleted for example

i also cant remove my account

[ornate spire]

I don't think the issue has anything to do with lyrics to be honest.

I'm getting problems and I am only working with instrumentals

[prime anvil]

i lost all control of a song at 6min extension, voice was hardly recognizable xD

[worn flicker]

Anyone know if there is a max character or time for regenerating the video or if the regeneration is just either extremely bogged down or broken?

[worn flicker]

The choppy part happens on a one shot generation?

[worn flicker]

Ah I read higher, so it is on extends and whatnot. I have the same issue when trying to splice in too directly. I instead go back a little further as much as I can stomach to get to a natural break. Are you custom instructing the instrumentals?

This is tech discussion, not support. No Suno employees in here.

Umm no...lol. we don't know them. We are users like you

You are speaking to no one...lol

Lol

[clever monolith]

hahaha sorry, this app is pissing me off

i canceled 5 months ago and they still charge me/ghost me

[unique stone]

It sounds like you have been trying to get a response from support for quite a long time... Send a DM to @steady kelp please, he should be able to look into this for you.

[urban sonnet]

can't you just block them from your bank's side?

[clever monolith]

i appreciate the help, ill delete the review if i can actually get some quality help

[clever monolith]

problem is i want my money back

[urban sonnet]

fair demand. Hope you will get a hold of Mike, 🙂 Best of luck.

[grim shuttle]

Hello! I have a question if anyone can help me, I spent a while without logging into my account and today when I logged in, all the songs I had created disappeared, did I really lose them all or is there something that can be done?

[steady kelp]

not to sound super defensive, but the only emails to us were in the last couple of days. no worries though, all resolved!

[charred merlin]

Not defensive at all -- more informative 🙂

[clever monolith]

yes youre right^ i wasnt trying to be deceptive. i really just dont know what im talking about obviously

i hope yall all have a good day.

[charred merlin]

Yourself as well 🍻

[hollow hare]

Are you logged into the correct account? View your profile and confirm the handle.

[ornate spire]

I'm getting these arbitrary "abrupt choppy skip" moments in the middle of an extend... Starting to get pissed off

At this point I'm begging for a 3 second long "credits refunded" segment to just end this shit

Wow.. no matter where I extend from in this song it just chops into the extension.. doesn't line up at all.

Basically have to throw the track away even though I'm 4 minutes into it

[tame venture]

Orrrr take it into the daw and chop chop

[woeful yoke]

If a song doesn't want to end, just leave it as is. Get your full song and then just use the crop tool to stop the song at the suitable step or bar.

• Once you put your desired length, you can go back and use the recreate selection tool to select the ending of the song without lyrics and just have it generate a nee ending.

The replace selection tool usually works pretty well if you tell it to generate the same length as selection but don't try to change or add lyrics.

But it's kind of 90% guaranteed that if you want to change the 0 existing lyrics, it may start replacing but end ul overshooting the length timestamp, so it keeps on going until it runs out of default length around 2 minutes.

[ornate spire]

The song doesn't have lyrics to begin with.

I was having the same chopping issues with replace section too

[stiff torrent]

Yes, extending does not work always.

[woeful yoke]

What do you mean extending doesn't work? I thought this year is quite the opposite that it works but it doesn't want to end songs?

[ornate spire]

It's just on this particular song, it's like it has a virus

[stiff torrent]

Oh you can extend, but cut and continue not done in intelligent manner

[ornate spire]

When you extend it makes an extension but it doesn't actually flow with the existing song. Like they just copied and pasted 2 completely different segments back to back

[woeful yoke]

One thing you could try is to merge already existing parts.. get the full song till the last part after which you wanted to finish. And then try to extend from the full song already. It definitely wants to when it figures out that the song it's extending is quite long already. It seems to be blind to the fact if you already extending one short part after another

[ornate spire]

Yeah I tried that too, believe me lol.

Ended up having to go back to the original generation and basically start over. Somewhere along the line it developed a disease

[woeful yoke]

Well that's kind of easy to fix. Just don't extend from the very end, but rather find an appropriate place from which to resume. Make it extend from like 5 to 10 seconds before the end the last part.
For example, sometimes I would try extending a 2-minute track from 1:50

[ornate spire]

Normally this works. But it didn't work on this song. It was literally messing up extensions from different points spanning over 2 minutes

[woeful yoke]

That disease is them trying to improve the sound of the model but unfortunately they they are training on music with wrong loudness. It would take time for model to finally learn the new new correct volume results and gaps between songs were to insert the end of sequence token for generation to stop.

It's just a noise floor. It's a little bit higher now. That's why I know the reasons why I need to discuss with them because they don't seem to have any idea about music mastering. Mixing and mastering and editing are all three different things. Right now they are making mistakes and those mistakes have caused many of the issues without them even realizing it, I think. Actually, I am pretty sure that they haven't figured it out yet. Otherwise, they wouldn't keep on doing the same thing over and over again, and hope for a different result.

[north lintel]

@hollow hare

[woeful yoke]

Try killing both the app and your web browser and try again

[hollow hare]

Ah, it is similar to the issue I had with the default app being Opera.

Try changing the default to other browsers like the internal/initial Browser, Firefox, etc. That should solve that problem.

Once you succeed in logging in, you can change your Default App back

On my Android, you can change the default app by:
Settings > App > Change Default Apps

[woeful yoke]

I've never encountered this

[north lintel]

I got a OnePlus

[hollow hare]

I would not be surprised if it is a specific issue with certain browsers. When I was using Opera for the Google Sign-in, after logging in, it would redirect to a page saying 404 (on the day of release).
Changed it to Firefox and it worked (after trying a few more things)

[north lintel]

Thank you it worked 🙂 🙂 🙂

[north lintel]

God bless you ❤️ 🙂

[hollow hare]

Great to hear! You're welcome and have a great day:)

[north lintel]

You too 🙂

[runic rampart]

I purchased the Pro plan and paid for it on Google Play, however accessing Suno with the browser my active plan is the Basic plan, in the Android app the Pro plan appears, however it works as if I don't have it.
I have tried contacting support via email but have not heard back yet.
What can I do now?

[halcyon sparrow]

i like cover my own songs ;)

Absolut-Fr4nk = Mute Box

[violet mesa]

This one is based on my own songs , lots of tries but absolute awesome outcome. 50/50 my work and suno
https://youtu.be/uigwRSOHe48?si=aGNSNePvoAxcKJb_

[halcyon sparrow]

sounds good and for extended version good job. mean the arrange FINE bro

https://soundcloud.com/mute-box-aka-fr4nk/mute-box-crescendo-trpping-extended-version

@violet mesa v3,5 or v4 think 4 or?

erm 3,4

3,5

[violet mesa]

4

[halcyon sparrow]

ok ....thought 3,5 cause vocals backing

interesting is structur and stil in end

commands and all

[violet mesa]

Took me at least 30 intents

But I finished all in my DAW

[halcyon sparrow]

i do same always....

:D

thats maybe prob with suno

[violet mesa]

,, This is the way "

I freaking wish midi sheets from suno for our songs

[halcyon sparrow]

it would be so awesome to have 4 tracks ....melody,bass,drums,vocal

to export from suno

[violet mesa]

Yep .. even 6 to 8 tracks...

[halcyon sparrow]

in clear

more than better

[violet mesa]

I would pay for that

[halcyon sparrow]

if iam programmer i would do a button split track

and export

[violet mesa]

Or, suno DAW integration.. imagine

[halcyon sparrow]

yeah also my idea

++++ fruity loop plugin++++

[violet mesa]

Yep

[halcyon sparrow]

just intrumental cause the split in 2 tracks ok but not pro

studio always miami style

[violet mesa]

Listening 🎶 to your song.. love it.. great 👍

[halcyon sparrow]

i created so much just for commands and spend so much credits cause i think its the future

ahahahhaaha

[violet mesa]

90 of the songs are destroyed because the f... shimmering.. but it seems less the last days

[hollow hare]

Did you verify if the account logged on the Android is the same as your main one on the browser?
You can check by viewing your profile handles (a surprising amount failed to realize they had created a new account)

[violet mesa]

Same.. but in the most awesome outcomes... shimmering destroy the song.

[proud folio]

HI , I have suno pro, made about 30 songs , I log back in and more than halve have disappeared ! , anyone elese have this issue ??

[unique stone]

The songs I like most on Suno have derived from insanity. I have one that was meant to be a sound effects test of a haunted mansion, but it turns into some insane singing followed by a really beautiful haunting piece of music.

A more recent one started as what appears to be someone leaving a voicemail while being tortured and then v4 was like "you know what, how about we follow that with a complete banger?"

[hollow hare]

Did you check the next page? There is a pagination system (create only shows your latest creations, you can access the rest in Library)

(See the yellow-orange outline)

[hollow warren]

you can't upload audio and make it a Persona

[coral chasm]

can't download or play for 4 days already. What to do?

[hollow hare]

What happens when you press play? Nothing?

Also, have you tried using incognito or another browser in case an extension is interfering with the play function?

[coral chasm]

not a single song starts, and errors when downloading any song

It used to be fine, up to 1000 songs were created and there were no problems

[violet mesa]

I should test that next time

[pliant idol]

The pagination search only works for 12 result pages- after that it gives no results. Try it and see.

[coral chasm]

it didn't help, the site was broken over the weekend and still is

Even on the phone nothing works

[charred merlin]

Yes you can, just not if it contains vocals 💪

[hollow warren]

I tried and it is not working, it doesn't open the window

[charred merlin]

I can't help but disagree, seeing as how I've done it multiple times, but you're allowed to believe what you'd like 🙏

[hollow warren]

okay, it worked when I cropped it... weird?

[charred merlin]

Sometimes clips without vocals can be flagged with a false positive (the same as clips that SHOULD be flagged can be let through with a false negative) -- maybe you clipped out the "offending section" that was being mistaken for vocals, or cropping acted like a "second pass" that corrected the false positive

I know that cropping the entirety of the song works to bypass some of the restrictions, like uploads being limited to v4 extensions -- might be related

[hollow warren]

it was 15 seconds of silence xD now 10

[charred merlin]

That was the loud silence you got rid of 😛 Just being a smartass -- glad you got it working 😄

[hollow warren]

that's silence lmao

[charred merlin]

...how is 15 seconds of uncompressed pure silence 1.26mb?

...I suppose there's really no harm in compressing the everloving hell out of it at that point

[unique stone]

It's a WAV file? 😄

15 seconds x 44.1KHz x 16 bit = 1,323,000 bytes

[charred merlin]

Or it still has the same number of sampling points, but they're just zeroed out -- still information SORT OF

[unique stone]

Yeah it's just zero, 1.3 million times

[charred merlin]

That clicked for me

[unique stone]

silence is loud
silence is also big

[charred merlin]

More like a straight chain

[charred merlin]

😂 audacity defaults to .wav, and I couldn't be assed to swap the filetype 😛

[hollow warren]

I'm not sure but I notice a slightly difference. It still doesn't sound 100% clear but I feel the noise is reduced. Did you try it for v3.5 or v4?

[split slate]

Can anyone help me please?

I created songs, then saved links to them and deleted the account to create more for free, but when I wanted to listen to the tracks, their links did not open and I was very upset.... please help, is there any way to restore them? I have their links

I think there won't be any more songs like this, there are some really great tracks with great lyrics...I'm totally in despair

Please, guys 🥺

[unique stone]

As far as I am aware, if you delete your account, all of the songs you created on that account are also deleted and cannot be recovered. You could try e-mailing the support address above, but they may take a few days to respond.

[charred merlin]

Hopefully you downloaded the .mp3 from the songs 😬

[split slate]

😭

[split slate]

I forgot about that 😞

[unique stone]

I would also not recommend trying to exploit the free credit system in this way.

[thick bramble]

Hi, does someone know why my card is rejected?, I called my bank and they told me everything is ok with it, but suno rejects the card, it is mastercard

[brazen sparrow]

I recently subscribed to the Pro plan for the first time. I was wondering if I can upload songs made during the Pro plan directly to social media platforms like TikTok or YouTube without editing or processing them. I am not skilled in editing, so I want to know if posting the original downloaded versions is acceptable, even if they may look less polished compared to professional songs. I recently posted some on my new YouTube channel, @thgieeight4103. Is that okay, or do I need to add any attributes or do anything else? Or is uploading like this not allowed?

[brazen sparrow]

thank you

[karmic cedar]

help

[slender crest]

Hello, I’m having quite a few problems with Suno correctly interpreting my song, as it ends abruptly and cuts off the lyrics. Suno is supposed to have a 4-minute limit, but strangely, my song always gets cut off around 3:17 to 3:20. Also, it doesn’t seem to pay much attention to the [music style]. I don’t know what to do, and I’m limited right now because I don’t have any more credits and I’m wasting them trying to get the song right.

I’m also using male voice avatars, and even though I set the [music style] to male voice, sometimes it still interprets the song with a female voice... I wouldn’t mind if the song didn’t end abruptly though... 😭

By the way, is there anyone who can help with this problem in Spanish by any chance? It's easier for me than English, I even prefer Chinese or Japanese...😅

[silk sorrel]

Also having issues with songs stopping at 3:23~

[obsidian needle]

The suno music allow use for nft music?

[untold briar]

If you're using a Persona or a Cover, the maximum length of a song is less than 4:00. It's not always 3:23, it depends on the specific Persona or Cover. But if you are using those features, it's probably working normally.

[slender crest]

that's how it should be, but with the new song I've written, commands included, they seem to be ignored, it's the first song that's giving me a big headache. I even kept modifying and changing the commands, but Suno keeps cutting the song. I even added time commands.😫

[torpid thicket]

What is going on with Vocals in V3.5? They sound crappy with every generation I do

https://suno.com/song/deadb485-7db9-4f0f-9b7a-6741b76859a6

https://suno.com/song/73b7bf5e-6fed-43fb-a205-a3ff021fbe2a

What's with the robotic undertone, dude

[slender crest]

That's true if you don't use a voice avatar and if you don't add time commands, at first it was like that for me, but as I added commands I realized it wasn't as random as it seems. 🤷‍♀️

[slender crest]

At first I used that version and it was very difficult to get a song tuned so that it didn't sound robotic, then I simply switched to version 4. 😒

[neat bear]

Hi there, you little ponies! What's up?

Best kind of arrangement to be used input for Cover? I used only bass, piano and harmonica, got mixed results concerning the melody and chords accuracy. Maybe about 50 % of Covers did not have the same melody.

In this example, I did not have vocals on my demo. Which is super cool that Suno can at all take the melody on put my lyrics successfully on right notes! Would it be better to have vocals on your demo too?

[eager topaz]

Covers is Beta, therefore very hit and miss. Personas is also Beta Exclude Styles is also Beta. V4 is also Beta. All very hit and miss

[woeful flame]

I have 45 pages in my library/history. Any way to Zip up all songs for download?

[woeful yoke]

But besides vocals, what's the other, even bigger problem with tracks you generated?

[austere panther]

Anyone having issues with songs not generating?

Both IOS and Desktop

[unkempt fractal]

It's been an issue since mid November 😉

[austere panther]

First time it's happened for me.

[empty cloud]

Just gonna leave this here, the "shimmer/reverb" issue is already present in 3, 3.5 aswell i guess we should stop liking songs with it inside to learn suno that we dont want it. Sucks that it's hard-coded.

[unkempt fractal]

Is it possible to generate only one song instead of two?

[lucid anvil]

audio upload no longer working

[ornate spire]

Is there a good style tag for "not fast" but without making it super slow

[cosmic oxide]

I'm trying to change lyrics and it's not working, what should I do?

[ornate spire]

Have you tried prayer?

[cosmic oxide]

What do you mean?

[ornate spire]

Try praying. Because I've been in that situation and could never get it to change the lyrics

[cosmic oxide]

That's a sad joke actually. So nothing can be done?
What about trying to recreate the same song with different lyrics? possible? same sound.

and I understand the "support" team is very active and helpful as well.

[ornate spire]

The solution is to "replace section" with different lyrics.

If it's not working, then there's not much you can do.

You can try expanding the size of the section you're replacing

[runic rampart]

Yes I'm sure is the same account. And finally PRO version isn't work both sides. I think the only way at this point is to open an incident on Google Play Store

[polar spade]

No . Simple answer. Same as Udio. There's a method to the process.

You ever have issues with any old voice bleeds in the instrumental stem showing up in the background, after adding new lyrics and running outputs or a suno remaster clears those up?

[slender crest]

It’s true that the [persona] option has great potential, as it’s incredible when used correctly. Today, I discovered that, for some reason, my songs were limited to 3:20 when using the voice avatar. Something must be interfering to cause this, as it had never happened before. Now, without the voice avatar, the songs can reach up to 4 minutes, but the randomly generated voice, even though it’s the V4, sounds somewhat distorted and metallic.

I suppose I’ll save this challenging song for later, when the voice avatars are more stable. 🙄🤷‍♀️ 😭

[polar spade]

One gen for 10 credits instead of two? The madness. Lol. I want 8 gens for 10 credits.

With over produced songs, I doubt it's ever a noticeable thing. But I haven't tried with an acoustic folk song yet. Just guitar and vocal. Ok. Thanks.

[polar spade]

Reminds me of Walmart being OK with selling non-organic bananas 🍌 one at a time. But ....they want organic bananas to be sold as bunches (still price by weight) to remain in the purple tape, for some reason. They don't like if you take off the tape, take a banana or two and re-wrap the rest. Lol.

So you gotta leave the tape on the gens. Can't split them half for half the price.

[ornate spire]

I think it's so people don't buy organic bananas and pay for them as non-organic

[polar spade]

And yes. I know the reason: so you don't buy the organics at the cheaper non-organic price. Thy don't trust me. 🤣

[polar spade]

GMTA

[polar spade]

Soon they be locking them behind a glass cabinet and only let staff hand them out. 🤣

[woeful yoke]

Oh nice!, I just wasted time on finally writing a post under v4-feedback, only so I could see it automatically getting deleted by some kind of a system...

It happened so quickly that I couldn't even read the reason why it was deleted because the freaking thing deleted the reason after a couple of seconds too.

[gritty badger]

Weird.. What was it even about?

[woeful yoke]

It was about my findings. It was about me offering them a technical fix of the root cause of most of the problems, after me spending weeks and thousands of credits only for analysis purposes in addition to replicating the issues in my own lab.

So I guess it's not really needed

[charred merlin]

You’ve DM’ed Mike B, the head of community support, right?

About wanting to share your technical insights regarding what they’re doing wrong over and over?

[woeful yoke]

No I posted in V4 feedback. It's not common for me to send unsolicited DMs to people who I never have had any communication with.

[charred merlin]

I understand you did that — I’m asking if you’ve also reached out to the head of community support

[gritty badger]

But Mike is there for that. Write him all your insights so that the devs can get their hands on the info

[charred merlin]

You’ve been repeatedly mentioning how you want to get in touch with the developers — the head of community support is a good start, right?

[charred merlin]

@steady kelp — you might need to make the first move if they’re uncomfortable DMing you, but want the devs to take notice

[polar spade]

You could have done something as minimal as used 10 dashes in a row to separate paragraphs and Woodie will block it. So check with a mod to see if they can track it to un-delete.

[gritty badger]

I'm guessing @woeful yoke misses most of the mentions around the server. Maybe he didn't notice the open invitation for a collaboration to fix the issues. Lets hope this caused an awareness 🙏

[polar spade]

And as always.... copy first before posting. Always a pain, often necessary, but always a frustration saver.

[gritty badger]

yeah I agree, I had a couple of posts getting deleted like that. Now I don't take chances 😅

[polar spade]

I think the blocking reason on mine once was something like, "newsfeed" layout or something. I made the post too organized.

[gritty badger]

Woodhouse is a sneaky ninja 🤭

[polar spade]

So be careful making bullet points. Lol

[charred merlin]

My favourite is attempting to use subscript (tiny text) and having him remove the post because "you're yelling".

[gritty badger]

I saw you @flint haven

[polar spade]

No footnotes or bibliography allowed here.

[charred merlin]

I'd post that citation, but it's written in subscript

[gritty badger]

It's a good way to summon him, like saying Beetlejuice 3 times 😅

[remote oak]

on that topic...

what is the background behind woodhouse occasionally being a conscious character (for example, creating its own music on suno)? is it controlled by a moderator? a dev? maybe a group of people have access? quite mysterious, really, or maybe im just uneducated

[unique stone]

It's primarily a bot that does auto-moderation (a lot of the stuff that gets handed by it you won't see because it gets deleted immediately, e.g. bots trying to post scam links.) Sometimes he becomes sentient and will talk or write music. Bit worrying really. Pretty sure this is how we end up with Terminator.

...Nah he can be controlled manually. Sometimes it's to do moderation, other times just to have him talk about prunes (his fruit of choice.)

[remote oak]

...careful, if woodhouse IS conscious, we shouldn't offend it with conspiracy simple misinformation

i'd still be curious as to who has access to woodhouse, as its messages seem pretty frequent (assuming that its controlled by one person or a small group)

[flint haven]

Quietly observing

[charred merlin]

mods and guides

[remote oak]

it all makes sense now... (im never trusting a guide again (jk, they're guides, of course im going to trust them))

[flint haven]

https://tenor.com/view/archer-fetch-a-rug-gif-14497579

[naive quail]

Uhhhg... selected 100+ pages for deletion only to find out that somewhere on one of those pages is a track connected to a persona that can't be deleted, so it won't delete any of the other selected files.

YEA

caps!

yea

I have a ghostscript that selects 20 pages at a time. Saves some clicks... until I hit that one track that can't be deleted, then I have to tediously delete up to that and then past it.

[charred merlin]

The amount of work you put into getting Suno to meow properly literally astounds me.

[mellow ermine]

Have they fixed treble quality degradation yet? And shimmer?

[mellow ermine]

Delete all your personas. They are just genre information anyway. Like "Metal core, Female vocals," etc. Because I did a duet persona and the persona never generated the duet again.

You can also click on create, then persona's, then the "open" icon, and see what songs are linked to it. But I'd just move them to trash.

[charred merlin]

They autopopulate the Style of Music section wen they're re-selected, but the personas are more than just pre-saved collections of genres -- it uses the selected clip as context, which is why you can hear the clip bleed through on some generations (this varies in frequency between personas)

[velvet yoke]

I have a question regarding the commercial use of songs generated through Suno AI. As per my understanding, without a subscription plan, I cannot distribute the generated songs commercially, and any unauthorized usage might be detected.
If I generate a song without a subscription, then purchase a plan after a few days and download the song as a WAV file, will I be granted the rights to use and distribute it commercially without facing any copyright issues?

[mellow ermine]

That didn't happen with me. I wasted like 200 credits trying to get it to do the duet like it did in the original. I was not happy.

[mellow ermine]

Nope. The say the pro plan does not grant comercial licence retroactively.

https://help.suno.com/en/articles/2425729

[naive quail]

They're more than that. They are very useful for reusing voices (or close approximations to the original voice).

Knowing what songs are linked to them doesn't help in this regard - a song used to create a persona cannot be permanently deleted. Personas can't be permanently deleted at all.

[mellow ermine]

Oh so move to trash doesn't actually delete them?

LOL

Broken

[charred merlin]

If you email [e-mail] and let them know what song(s) you're requesting, they've been known to grant the licenses if you subscribed shortly thereafter 🙏

[naive quail]

The trash is just a separate library of files. Hell, even permanently deleted files don't immediately get deleted, they just get moved out of your trash library and put into a queue for mass deletion. I dunno how often that mass deletion script fires on the Suno server, but it doesn't seem to be very often.

[mellow ermine]

I have like 60 songs over a year that I edited for about 14 hours each in Audacity.

Do you think they will let me usae them?

[velvet yoke]

they said this

Hey there,

Thanks for making music with Suno! With all of our recent updates, we're experiencing longer than expected wait times for email replies. You can trust we'll follow up as soon as we can, but we wanted to send a few helpful tips:

If you haven't checked it out yet, we have a Knowledge Base with common questions over at https://help.suno.com/

We also have a Discord server for peer-to-peer help:

If you've reached out already, feel free to follow up in the thread with the same email you sent! Sending another new email rather than a threaded reply may result in replies from multiple agents and a longer wait, so please hang in there! We'll do our best to reply with helpful information quickly!

We look forward to responding, and thanks so much for your patience!

❤️ Suno

[mellow ermine]

So basically, if you want to delete files, don't make personas with them.

[mellow ermine]

Yep a bot replay. I got that about a week ago

[charred merlin]

That's a bit of a different situation to what the other user had mentioned -- but can't hurt to email [e-mail] and ask

[velvet yoke]

ok then i have to wait

[mellow ermine]

Hey I'll pay for all the months that I didn't have the pro plan for. Upgrade all those free moths to pro plan. Even double. But if they don't, I have a years worth of work that I can't use. And my entire year plan is down the drain'

[charred merlin]

Yes -- the email support team are the only people who can grant ownership of songs made on a free plan -- other users can only give advice 🙏

Hoping they will get back to you before long 💪

[mellow ermine]

It destroys my songs for my feature film too. So it pretty much shuts down that prohject as well.

[hollow hare]

Should have done your due diligence

[mellow ermine]

Yep but how many others are asking the same thing?

[hollow hare]

Anyways, nothing we can do since we are a community ran server. Anything official related must be through [e-mail]

[mellow ermine]

And that's the ones who learn about the issue

[charred merlin]

https://discord.com/channels/1069381916492562582/1294423389515284552

[hollow hare]

Not that many considering the number of users. Heck, Suno's Help section on the website (links at #1294423389515284552) even covers that scenario

[charred merlin]

The help.suno.com content is basically a user-friendly way of approaching https://suno.com/terms

I was thinking of tactful ways to explain the matter, but I trust the guides have this covered

[mellow ermine]

Yeah I wasn't thinking about comercial use back then

And I didn't know if Suno was wrth sticking with

[hollow hare]

Since you are interested in the commercial aspect now, you should check out Suno's TOS at https://suno.com/terms. If you want a video explanation, see Top Music Attorney (links and useful timestamps at #📚┃suno-school message)

[mellow ermine]

But then when I relised I needed the comercial use with the pro plan, and that there was no getting the comercial rights, I was gutted. I haven't made a song since.

And I find listening to my songs make me HATE. And so I tend not to listen to my songs anymore either

Yeah pretty devastating. Going from on top of the world ready to release to, wait, I just wasted an entire year. But that's life. So back to playing GTA5 and throwong molotovs at crowds 😄

[rustic lance]

Hi there. I am having some problems with my SUNO account and I don't know if I a in the right place to find some help to solve effectively and soon.

am*

[hollow hare]

Hello, what issue are you experiencing?

[rustic lance]

I have been using your SUNO services for a while now, but just today it didn't recognized my account.

And it limited me to a few songs, only. I have been paying monthly to be able to carry on with my own music project.

In fact, it only shows me today's songs and not the ones I have been working on previously.

[hollow hare]

I should note I, along with the other guides and moderators, are not part of the Suno team. This is primarily a community-ran server

Hmm, this sounds like you are logged in to the wrong account. Have you tried logging out and in to your main account? You can confirm by viewing your profile handle

If yes, have you tried checking that the account you are logging into on the correct one? The subscribed account should have gotten an email from Suno with the receipt of their membership.

If yes and the account is one and the same but the handle is different, you will need to contact [e-mail] with as much detail as possible (about the situation and confirming your identification)

[bold spade]

Hello.
You're even kidding me! They blocked my access to my paid account from 6.12.2024 (plus I burned more than 200 credits in vain) and you're still asking me to pay for the next month?!

[hollow hare]

You will want to contact [e-mail] if you cannot cancel the subscription due to the account being banned (assuming that is the case)

[rustic lance]

I've tried logging out and into my main account but it doesn't work. It seems to be as if it didn't exist anymore.

[hollow hare]

Highly unusual. Email [e-mail] with information as I mentioned before.

Though support is under high load right now. Last I checked, response time was like 7 business days

[rustic lance]

The point is that trying to reach either the billing dpt. or the support one, I just receive a email back in which it is explained I have to contact this way, by Discord.

But then again, you say you cannot help here. It is disheartening, the least.

[charred merlin]

The head of community support seems to be open to DMs over discord if you wanted to "skip the queue"

[hollow hare]

Unfortunately we cannot help with anything directly related to Suno accounts here. As the email stated, the discord is more of a peer-to-peer (community) help

[bold spade]

I just ask you to restore my access to my paid account!

[daring talon]

v4 is impressive

[hollow hare]

As I just stated we cannot help you with that Dave Kvant. Most of us are regular users with no power or support over accounts at Suno. You will need to wait for a reply from [e-mail]

[rustic lance]

Ok. But I look foward to keeping my account without changes or modifications. I've lost my music project which were saved there!

Thank you, anyway.

[bold spade]

Where else can they write here? In what topic?

[charred merlin]

@bold spade when you posted the exact same message in 3+ channels, the head of community support personally reached out to you and asked you to get in touch.

[bold spade]

Okay, I'll wait.

[charred merlin]

You don't need to wait 🙏 , @steady kelp is asking you to send them a message over Discord so they can help you (they work for Suno)

Hey Mike, long time, no chat -- hoping all is well on your end 🍻

[steady kelp]

thanks for helping point people in my direction, and for helping so much in general!

[steady kelp]

sent you a DM so you don't have to try to find me 👍

[rustic lance]

Just another more question. Why one person here has said to me I had to keep on waiting for help (up to a week time!!) and you have just helped the one who arrived next after me?

Does it work like that?

[graceful condor]

Just DM @steady kelp

He should help

If your issue is urgent and you have been waiting for some time for a support answer

Mike is always your go to

[steady kelp]

DM me, we'll resolve it!

[rustic lance]

I need to know if my account has been shut down because when logging in I was relocated to another account I don't know and in which I haven't got any of my musical pieces saved

[graceful condor]

We should do better and make it so it's not a free helpdesk, correct answers get 10 credits

There are ways, but that was a joke, haha

[polar spade]

#📚┃suno-school message. @steady kelp

Have you DM'd Mike as you requested?

I will say from personal experience, it appears a new account was inadvertently created. Made by mistake. I did the same thing. So it's a matter of knowing how the "real/original" account is logged into. By email or phone or other. Keep that in mind that it's an easy work through probably.

[polar spade]

These two screens look similar. As one tried to log in, we accidentally try on the "Create Your Account" Screen by mistake, and it gives an error. We think it's a mistake, so we put in a phone number, or email, differently from our normal log-in routine. And violá, it create a new account and sends One time code. We enter and think we have logged in now. But it's a new account, showing a random ID profile and only 50 credits.

It happens. No worries. I hope your issue is that simple. The fix is to cipher which way we originally had logged in.

Does that seem close to your experience 🤔?

[graceful condor]

According to one of the musicians I talk to, v4 version flagged a song for moderation, while v3.5 completely ignored it. Is it a bug, @steady kelp?

I can send the person to message you directly

"One on an extend and one was a single gen. Extended on v4, single gen 3.5". Looks like it happens when Extending a Song only

[naive quail]

Anything is better than what they've stuck us with!

[rustic lance]

I don't understand this procedure 😦

[steady kelp]

id need more context to give any sort of firm answer. they can DM me with details or write to [e-mail]!

[graceful condor]

Thank you very much

[lyric sonnet]

Can we have a longer "Song description" in non-custom Create? Struggling with 200 characters to suggest a story

but what i can squeeze in 200 works very well 🙂

[fresh bolt]

Hey please someone help me

My problem is here: https://www.reddit.com/r/SunoAI/comments/1hgg9gk/the_real_suno_tells_me_its_fake_help/

I just can't do anything, if i generate something, it just burns my credits and i get this weird thing telling me that i'm using fake suno app, which is not true...

[naive quail]

#🪲┃bugs-web message

[fresh bolt]

I filled out the form and sended it... I mean, i hope i'll get a refund or something

I just burned out 200 credits on these fake ad songs

The first ones was just a ordinary bunch generation, then i discovered a bug and tried every day test if it is gone

[graceful condor]

That's odd. Are you on some weird VPN/proxy server or something?

[naive quail]

There's not really much we can do about that - we're just users like you.

You can try reaching out to Mike B here on discord - he's the official community manager. He probably gets an overwhelming number of DM's, but he definitely responds to some of them some of the time. 😄

[graceful condor]

That's actually kind of genius....some websites do actually have this model, where users can upvote "This solved my issue" type of thing. That said, the number of users seeking help far outnumbers those savvy enough to deploy it.

[graceful condor]

That would boost the numbers of people helping here

[graceful condor]

Your account was most likely flagged for API use or manipulating the website, such as using a VPN or payment.
Please, use this form:
https://sunomusic.typeform.com/to/zGsJSzsf

Scratch what I've said about the form

[graceful condor]

The only thing left is to wait. If you don't get a response soon, you can always message Mike B

[steady kelp]

DM me please!

[graceful condor]

People would definitely abuse it in some way, creating fake questions and then approving answers to get credits

[woeful yoke]

Absolutely unbelievable what is happening.

[fair tide]

Here’s the email I just sent and still haven’t gotten a reply from anybody this is the second email:

Thank you for addressing this issue. I appreciate your attention to it.

I can’t be the only one experiencing this frustrating problem. When I sign out of your app on my iPhone, Apple seems to hide the app or my email address, making it impossible to sign back in. Since your app is considered a third-party app by Apple, it puts the responsibility on your developers to address this.

This issue extends beyond just the iPhone app—I couldn’t even sign in using the web app. Please understand that this is a complete nightmare, and it needs to be fixed as soon as possible.

Thank you for your time and efforts in resolving this.

[graceful condor]

Oh I didn't mean cash - I just meant "attaboy/girl/nonbinary friendo" type of points

[steady kelp]

what's happening?

[graceful condor]

Yeah once you introduce a cash motive you wind up with all kinds of crazy stuff

nahhhh - similar to reddit, you would get zero moneys for helping in this hypothetical scenario 🙂

[fair tide]

Yo, does anybody in here know how to sign back into the app if you use Apple hide my email which I didn’t mean to use it’s a complete nightmare. I can’t stay in Apple. I hate Apple. I hate them so much.

[steady kelp]

so on iphone, you control the settings to have your email address hidden, they use an apple private relay email instead, which you should also be able to identify in settings

[fair tide]

Yeah, Mike I know that now it’s a little too late for that

I found the email that they use the temporary email or whatever it is and I tried to sign in with that and I could not

[steady kelp]

OK, hang tight. I'll DM you in a moment

[fair tide]

I think I misunderstood you. I apologize for that.

Thank you

Suno is awesome, by the way, they give you a lot of free credits for everything they do and so I’m only complaining to the developers that man you gotta get away from Apple and make it easier to sign in and out and easier I can’t sign in on the Web app at all cause I don’t know my address and that’s my fault for not paying attention and just clicking shitmy bad but please make it easier because all of my saved songs on my phone and the phone app iPhone app

My bad

[sterile kindle]

Hey I have a very frustrating problem. I just subscribed to Premium via the Playstore and the money was already taken. But the premium subscription is not working and when I press on v4 it tells me to subscribe to premium. But when I press on my current subscriptions it tells me premium is the active Plan.

[graceful condor]

Is your version up to date? V4 is currently available to everyone

[sterile kindle]

Yes I just downloaded the App two hours ago. I already used my free songs and now I wanted to subscribe

[graceful condor]

Is it possible for you to log in on the Web version and check if you have the Credits there?

If you've subscribed, you should get the Credits

[sterile kindle]

I already did that but there wasn't any active subscription at all

[graceful condor]

How long has it been since you've subscribed?

[sterile kindle]

Maybe one hour

Its also an active subscription in the Playstore and I already payed

[charred merlin]

@sterile kindle Might be a good idea to message [e-mail] -- fellow users can try to help and give advice, but none of us are staff here beyond Mike B, who seems to be taking an increasingly active role.

[sterile kindle]

Okay thank you

[mortal belfry]

one month and the shimmer stuff is still aggressively there... ruining most of anything I create...

[prime prairie]

Still can't sign in to the freaking Android app! I am NOT happy about this!

[spring ginkgo]

I like making my own beats. Here’s a cool vocal stem made with suno @stone vault

[random prawn]

Shimmer seems to be in chats everywhere. https://share.zight.com/JruQXdwN does everyone elses songs look like they have bullet holes in them?

[cunning mulch]

So, does Suno ever generate anything like a .lrc file when generating the lyrics video?

[hollow hare]

1. Ensure your app has the latest update
2. Getting an error 404 or issues with redirecting when signing in (mainly via Google)? Change your default browser.

I had that issue with Opera and switching to Firefox fixed it and I heard the initial Browser works as well. To change the default, go to Settongs > Apps > Change Default Settings (name and location may change base on your device/firmware). After signing in, you can change your default browser back.

[woeful yoke]

That's the SUNO way of fixing issues by putting Band-Aids on them instead of dealing with the root cause 😉

[random prawn]

they havent fixed anything or hidden anything. But we will see if they can figure out why this distortioni s happening

[prime prairie]

I've done all of that!

[hollow hare]

If you did switch to Firefox as the default - then I'm out of ideas then, other than trying to use another browser, sorry.

[woeful yoke]

The good thing is that my opinion of the team has slightly improved. For a while I actually thought that they don't really care. But over the last couple of days, I've come to realize that they still have not even figured out why things got so bad and why they are getting even worse in the future - because the Band-Aid is not going to hold.

[prime prairie]

I do use a VPN. Does that make a difference?

[graceful condor]

Have you tried reinstalling the app?

[graceful condor]

Yes, try disabling it

[prime prairie]

Yes. A few times.

[summer trout]

hey all, is there a way to change the tempo of a song to shorten a bit? I have some great songs that drop right at the 4m mark and probably a 4:15 or so.. or at least make the song fit better without dropping off at the end

[prime prairie]

Ok. I'll try that. Thank you!

[graceful condor]

Fingers crossed that it helps

[woeful yoke]

It started in march already. I figured it out because I recreated the same issues in my lab.

[graceful condor]

You can add slow or fast to styles to change their tempo

[spring ginkgo]

More suno a capellas

[random prawn]

what is the theory as to what these dots (artifacts) (distortions) are?

[wispy totem]

bro

[random prawn]

lol

[wispy totem]

ignore the lyrics

it messed up my song so badly on V4

did u listen to the whole thing

its pretty funny tbh

just ignore the lyrics

[woeful yoke]

Most of them are noise, namely small, audio-token-sized features. And the problem isn't that the holes came to be but rather that the rest around the holes started getting brighter over time due to an implementation flaw in the audio pipeline.
When you train or fine tune your model on your own bad outputs. Then this problem starts amplifying itself and getting exponentially worse with each new model

[radiant egret]

Is there a way to remaster songs on the android version?

[woeful yoke]

No

[random prawn]

this sounds like model collapose.