✨│ai-help

[marble forge]

no

[proper shale]

maybe try an audacity re-export, as .wav

use ElevenLabs or something else that gives decent TTS results

[earnest viper]

using style tts2 or xtts v2 with voice cloning should improve results by a lot right?

[proper shale]

Ooor maybe that, yeah

[crystal gull]

#1 RVC Realtime states I have to have an index file
#2 RVC Realtime gives me this error when starting audio conversion

  File "threading.py", line 980, in _bootstrap_inner
  File "threading.py", line 917, in run
  File "C:\Users\BumiMandias\Desktop\RVC1006Nvidia\gui_v1.py", line 653, in soundinput
    with sd.Stream(
  File "C:\Users\BumiMandias\Desktop\RVC1006Nvidia\runtime\lib\site-packages\sounddevice.py", line 1800, in __init__
    _StreamBase.__init__(self, kind='duplex', wrap_callback='array',
  File "C:\Users\BumiMandias\Desktop\RVC1006Nvidia\runtime\lib\site-packages\sounddevice.py", line 898, in __init__
    _check(_lib.Pa_OpenStream(self._ptr, iparameters, oparameters,
  File "C:\Users\BumiMandias\Desktop\RVC1006Nvidia\runtime\lib\site-packages\sounddevice.py", line 2747, in _check
    raise PortAudioError(errormsg, err)
sounddevice.PortAudioError: Error opening Stream: Illegal combination of I/O devices [PaErrorCode -9993]```
Any clues

[slim geyser]

tts is weird with rvc, since they're basically just inserting a tts voice into rvc to infer that

[slim geyser]

you need an index file in its little area, but you don't need to use it

[crystal gull]

Ah

[slim geyser]

could be some random index file

[proper shale]

that's when you use different audio type stuff

there's MME, WASAPI, shit like that

use the same type on both

[crystal gull]

Ohh

Gotcha

Time to test out the voice ey

Oh wow it sounds pretty damn good

Not bad!

Definitely nothing amazing but surely great for my first try

Er first try whilst I know what I'm somewhat doing

[slim geyser]

that audio file is messed up

can't play it at all

[manic plaza]

i'm using the 40k pretrained model (i use the same one for all my models)

[proper shale]

Normal pretrains get the job done 95% of the time

[crystal gull]

Discord issue

Works fine for me

[crystal gull]

https://vocaroo.com/1aZaDnDFNcTG

[slim geyser]

okay the audio file works when I open the discord app

but not on the discord website

amazing

[slim geyser]

def not bad for a first try at all

[crystal gull]

https://voca.ro/132RtIkMn03M

First one is the dataset Second is the ai voice obviously

Perhaps I could have better settings but the voice itself is pretty dang sweet

[slim geyser]

oh gawd that background noise, not wonder uvr did not handle that very well

[crystal gull]

It was a different dataset

That background noise is because im playing it through like

OBS and I recorded it all weird

That's not directly from the set

[slim geyser]

oooh

[crystal gull]

well ok I just raised the audio on the original video and there is actually background noise .-.

but it seems to have handled it well

[slim geyser]

I'd try raising the extra inference time to max, and lowering the pitch to like, 12 through 10.

Rvc is pretty good at handling some white noise, but only like soft fan kinda white noise, otherwise it's going to infect your model

[brittle wing]

can someone help me with using the mangio rvc thing

[crystal gull]

Yeah the dataset could 100% be better after raising the volume I can definitely hear issues

[slim geyser]

ye, another thing with uvr, it's not 100% consistent at dealing with white noise. If you listen back to the audio file you'll probably notice moments where it doesn't really catch it all

it's particularly bad with that if the white noise is decently loud, or changes in volume

like background music

and it's also not very good at getting rid of noise if its happening while someone is speaking at the same time

it'll like blend together

[crystal gull]

Guess I just gotta go and find better datasets then

[proper shale]

what's up?

[brittle wing]

its just not working

idk what it is

[proper shale]

wait what are you trying to do rn

and what step are u on

[brittle wing]

uh i downloaded it already

im trying to use it but idk

[crystal gull]

🤦‍♂️

Dude is giving no examples or information of use

[proper shale]

https://docs.aihub.wtf/ I suggest reading the Mangio part inside RVC > Local then

[crystal gull]

  File "threading.py", line 980, in _bootstrap_inner
  File "threading.py", line 917, in run
  File "C:\Users\BumiMandias\Desktop\RVC1006Nvidia\gui_v1.py", line 653, in soundinput
    with sd.Stream(
  File "C:\Users\BumiMandias\Desktop\RVC1006Nvidia\runtime\lib\site-packages\sounddevice.py", line 1800, in __init__
    _StreamBase.__init__(self, kind='duplex', wrap_callback='array',
  File "C:\Users\BumiMandias\Desktop\RVC1006Nvidia\runtime\lib\site-packages\sounddevice.py", line 898, in __init__
    _check(_lib.Pa_OpenStream(self._ptr, iparameters, oparameters,
  File "C:\Users\BumiMandias\Desktop\RVC1006Nvidia\runtime\lib\site-packages\sounddevice.py", line 2747, in _check
    raise PortAudioError(errormsg, err)
sounddevice.PortAudioError: Error opening Stream: Invalid number of channels [PaErrorCode -9998]``` uhhh anyone experience this before?

No matter what input & output I use this happens now

[brittle wing]

thank youi

[proper shale]

Huhhh what is this

[crystal gull]

Exactly..

[proper shale]

Yw :)

[crystal gull]

I get it when starting audio conversion

For RVC's Realtime ))

[proper shale]

I suggest a restart and using a different type of audio API or whatever it's called

[slim geyser]

looks like an error with your audio devices

[crystal gull]

Oh I switched to mme only and it works now

[slim geyser]

do they all have the same number of channels?

oh

[crystal gull]

this is so upsetting that the voice sounds so damn good but there is background noise

[junior halo]

Would this be considered done..?

[slim geyser]

many such cases

sadge

[proper shale]

Check mel and kl

[crystal gull]

pure sadness

perhaps ill pay this person to give me a version of it without background... 🤣

[junior halo]

This mel?

[slim geyser]

lmao. def not a sus request

[crystal gull]

No... Not at all..

[proper shale]

Still going, but I'd recommend stopping training and exporting model

So you can return later

[junior halo]

If you say it's not done yet, why stop now?

[slim geyser]

what do you say if they ask, "but why tho?" xD

[brittle wing]

its very difficult to understand

[junior halo]

I don't mind waiting for all 500 epochs to wrap up.

[proper shale]

Cause Google Colab limits sessions

And it could make you lose the model

[junior halo]

It does, but I have enough compute units.

[crystal gull]

UhhHhhhhHhh I have a medical condition that makes it so I can't hear anything but background noise if it exists?

[proper shale]

Ah

[crystal gull]

👌

[proper shale]

Alrighty then

[junior halo]

(I bought em a long time ago)

[slim geyser]

slicccck

[junior halo]

(foolish I know)

[slim geyser]

so slick haha

[crystal gull]

Actually you know what I could do, to possible clean it up. Run it through steelseries's ai thingy

[proper shale]

You can always ask good ol GPT if you're confused

[crystal gull]

wouldnt work in the longrun but maybe ill try it if i can't find any datasets without background noise

[proper shale]

Or you can explain the issue and we can try to help ya

[slim geyser]

wait that's actually big brain

why haven't I thought of that

fuk

[crystal gull]

what running it through the steelseries ai?

lol

[slim geyser]

it's really good at separating noise

[crystal gull]

i use it for my mic so i thought why not rvc too

[slim geyser]

way more than uvr honestly xD

[junior halo]

My plan for now is to wait for it to finish training to 500 and test out different iterations of the model where either g or d graphs dip and see which one sounds best.

[crystal gull]

welp ig im on the hunt to find a new dataset without background noise

[rare goblet]

Ayo? @crystal gull level 8 !!!

[proper shale]

Normally the latest one sounds the best

[slim geyser]

https://tenor.com/view/nuh-uh-beocord-no-lol-gif-24435520

(listen to your ears)

[proper shale]

Yes, this too

[slim geyser]

graphs are cool and all, but your ears are your most powerful tool to decide which checkpoint is best

[proper shale]

But, either way, if mel and kl are still going down, model is improving, so it's most likely the best version of your model

[brittle wing]

chat gpt?

[proper shale]

yeah

[brittle wing]

hm oka

[junior halo]

Don't tell me you've never heard of GPT as we speak in this AI hub server

[crystal gull]

||no offense to noobies but the vast majority of people that touch chatgpt are morons and don't know how to properly use it ||

I swear it's like people ask chatgpt to do one specific thing, like "create me a unity game" or "code me a minecraft mod that does this" instead of asking it questions along the way and helping it explain to you what to do

[junior halo]

Precisely.

[crystal gull]

It's not necessarily wrong to ask it to make something but it just doesn't understand specifics rather than small portions, hate the people who try chatgpt to code and go "CHATGPT SUCKS" because they don't know how to properly use it

Chatgpt / Copilot is amazing with code if you know how ai works 👌

[junior halo]

Is this not common sense?

[crystal gull]

It should be

But it isn't

[junior halo]

??

That blows my mind.

[crystal gull]

I see people all the time asking it to make huge projects or code a whole thing. As if it understands that

It's like people don't understand that ai is based off things already created

[junior halo]

It's like asking your mom to make you a sandwich only to get upset when your mom put ham instead of turkey

[crystal gull]

^^^

[brittle wing]

no ive heard of it its just i dont use it if im confused about something

[rare goblet]

Ayo? @brittle wing level 2 !!!

[brittle wing]

and i still dont know what im doing wrong

[junior halo]

Take advantage of that.

[crystal gull]

Well what step are you at.. What exactly are you trying to do at the moment

[brittle wing]

make an ai voice cover

[crystal gull]

No shit sherlock.. Respectfully

But what STEP

There is no button called "make an ai voice cover"

[junior halo]

Have you considered looking up a tutorial on YouTube? Those were exactly where I got started.

[brittle wing]

ig step 3?

[crystal gull]

Of what doc? The ai hub docs?

[brittle wing]

yes

[crystal gull]

So your at the inferencing step and setting up RVC

Are you trying to install it locally or on a colab?

[brittle wing]

wait i alreaedy downloaded it

[crystal gull]

You also have to consider, do you have a powerful enough gpu or cpu for what your trying to do. If not, you'll have to get someone else to do it. Or use a collab (cloud option in the docs)

[brittle wing]

i have a good cpu and gpu to be able to make it

[crystal gull]

What gpu?

[slim geyser]

little quick test here, loud ass music was in the background of this, but steelserires catches most of it. Like you can still hear some of it, but you could easily cut those parts out. BIG BRAIN

[crystal gull]

I am indeed a genius

[brittle wing]

how do i check that in the system settings

[crystal gull]

..

Open up task manager, press performance, press gpu

[brittle wing]

woopsies

[crystal gull]

Shows in the top right of that page

[brittle wing]

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• Ilaria RVC, by thestingerx Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab
• 🆕 UVR5 NO UI for Google Colab, by Eddy Google Colab

How to use?

🤗 Hugginface Spaces

• Applio, by IA Hispano Huggingface Spaces
• Ilaria RVC, by thestingerx Hugginface Spaces
• RVC-HFv2, by r3gm Huggingface Spaces
• AICoverGen, by r3gm Huggingface Spaces
• Advanced RVC Inference, by r3gm Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces

More commands: -audio, -local, -overtrain, -realtime, -rvc, -help

[brittle wing]

lemme see

[junior halo]

right click task bar, task manager, performance

[crystal gull]

slowww

[junior halo]

damn.

[crystal gull]

😛

[brittle wing]

okay i checked it

[crystal gull]

Gotcha and what do you have?

[brittle wing]

AMD Ryzen 5 3600 6-Core Processor

[crystal gull]

.. Thats a cpu

[brittle wing]

thought itd be helpful

[crystal gull]

Ah

[junior halo]

go to the gpu tab

[brittle wing]

NVIDIA GeForce GTX 1650 SUPER

[crystal gull]

PF

I'm not sure what collabs have

[brittle wing]

is that good

[crystal gull]

But I don't think you're doing too well with a 1650 super and ai

[brittle wing]

hm

well is that the problem?

[crystal gull]

Well that isn't THE problem that's just the first step in deciding whether your using cloud or local, I believe the cloud gpus are faster than that? I don't know 100% as I've only ever used local

[brittle wing]

im using local i think

[crystal gull]

Perhaps @slim geyser knows?

[crystal gull]

Yes if you downloaded it onto your pc then that's called local

But the issue is I don't know if that gpu can create an ai cover in a reasonable amount of time

Versus a collab and if it would be more worth to run on a collab

[brittle wing]

i have that then

[junior halo]

I would recommend trying the colab. Doesn't take too long either.

[slim geyser]

wut

know what

[junior halo]

https://colab.research.google.com/github/hinabl/AICoverGen-Colab/blob/main/Hina_Mod_AICoverGen_colab.ipynb?pli=1&authuser=1#scrollTo=NEglTq6Ya9d0

[crystal gull]

Mares do you know if it'd be more worth for him to run in a colab versus local on a 1650 gtx?

Idk colab specs

[junior halo]

In my experience, it's not bad.

[slim geyser]

You should be able to train with that, just pretty slowly.

Cuz you gotta pay for cloab if you don't wanna get cut off so, not ideal if you already have a card that can train

[crystal gull]

Ah right

[brittle wing]

wait you have to pay for it?

[slim geyser]

yeah?

[junior halo]

No

[crystal gull]

dont even

answer that-

show is reading like 2 words from every sentence

[junior halo]

It's "freemium"

[slim geyser]

they cut you off after a couple of hours

esp with rvc stuff

[brittle wing]

no im not

[crystal gull]

show you can train locally if you don't want to get randomly cut off by colab, just keep following the steps

[slim geyser]

because using collab for rvc and stuff like that is against the spirit of colab in the first place

[junior halo]

You can use the colab for a good while at no cost. Once your free processing units run out, just switch to another Google account or make another.

[brittle wing]

im not trying to train im trying to use the voice conversion

[slim geyser]

they don't wanna give free compute to a bunch of little timmys making skibidi toliet ai covers

[crystal gull]

yeah thats what i mean*

oh well shit oh

If you already are using a pre-made vooice

voice*

You can do it just fine on your gpu

I thought you didn't already have one*

[brittle wing]

i did

[slim geyser]

idk if they fixed it, but google had blacklisted rvc notebooks in the past, where they only worked for long enough if you paid for compute

[brittle wing]

ive been trying to know why i keep getting an error everytime i use the thing

[crystal gull]

Using the mainline?

Would be nice if you showed the error

[brittle wing]

it just says error

[crystal gull]

just.. error

When starting a program? When clicking a button? er?

[junior halo]

In my experience, it works just fine. Both the hina mod voice conversion and the v2 disconnected.

[brittle wing]

im probably looking like an idiot but

[quasi ether]

I want to know how to make my own RVC models (mainly so I can put it through my voice so I can basically do a overcomplicated miku situation), but all I really know is to use other models, I forgot where I got my RVC engine, and for some reason I cannot supply a screenshot, so oof...

[junior halo]

Here's a brilliant tutorial to get your started: https://www.youtube.com/watch?v=tnfqIQ11Qek&t=717s

[rare goblet]

Ayo? @junior halo level 7 !!!

[quasi ether]

Thanks

[junior halo]

But first, you must prepare a dataset properly.

Where are you sourcing your chosen voice from?

[slim geyser]

huh, guess they manged to dodge it somehow

[junior halo]

Sure did. Except you can't just run it overnight. It'll give you the boot for inactivity.

[quasi ether]

For me, I'm gonna start with my voice, so I can just record me saying a lot.

[rare goblet]

Ayo? @quasi ether level 1 !!!

[crystal gull]

Frankly.. I have no idea, how did you start the program.

Looks like the api having issues maybe?

[brittle wing]

i opened the mangio rvc, then go-web.bat

[rare goblet]

Ayo? @brittle wing level 3 !!!

[junior halo]

Record yourself speaking and/or singing in a variety of pitches. Avoid straying away from your "normal" voice. Record at least 5 minutes of audio. Then, truncate silence using these settings:

[quasi ether]

Though I won't need to now, Here's a screenshot of my RVC engine

[junior halo]

(assuming you are using Audacity)

[quasi ether]

Which I do! :D

[junior halo]

Are you doing this locally?

[quasi ether]

I think so? Though if things go wrong, I have the video

[crystal gull]

No idea never seen that happen before, sounds like you started it properly. Perhaps a pc restart?

[junior halo]

What's your GPU? Just to double check.

[crystal gull]

Is there any command prompt with errors in it?

[quasi ether]

(Hold on as I try to find the GPU of my laptop)

[crystal gull]

task manager > performance > gpu

[junior halo]

right click task bar > task manager > performance tab

DAMN IT

[crystal gull]

😉

i type 120wpm and i wasn't even trying

stinky

[brittle wing]

from when i was trying to like use the ai voice conersion

[crystal gull]

i gave you like a 3 second headstart

[crystal gull]

ahhhh

[junior halo]

I have very improper typing skills, or lackthereof.

[crystal gull]

LOL

im a little too good

[quasi ether]

Though just to say, I have a second faster computer, though I have no access to it as of RN

[crystal gull]

can i let you borrow some skills pls

[crystal gull]

gpu aint good enough ey

[junior halo]

yes plz

[brittle wing]

well my cpu work with it?

[crystal gull]

ok they will be delivered tomorrow

[crystal gull]

your cpu is fine but your gpu aint shiii

"RVC SHIT" what a real folder name

[brittle wing]

lol

[crystal gull]

I'd do a cloud install using collab

Since your gpu isn't up to the task it appears

[brittle wing]

i dont wanna pay for it

[crystal gull]

colab is free

but has limited use-time

[brittle wing]

hm

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• Ilaria RVC, by thestingerx Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab
• 🆕 UVR5 NO UI for Google Colab, by Eddy Google Colab

How to use?

🤗 Hugginface Spaces

• Applio, by IA Hispano Huggingface Spaces
• Ilaria RVC, by thestingerx Hugginface Spaces
• RVC-HFv2, by r3gm Huggingface Spaces
• AICoverGen, by r3gm Huggingface Spaces
• Advanced RVC Inference, by r3gm Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces

More commands: -audio, -local, -overtrain, -realtime, -rvc, -help

[brittle wing]

which one should i use?

[junior halo]

From what I have read online, it looks like this gpu isn't too bad performance wise.

Not sure if it can train okay?

[crystal gull]

that.. isn't.. a gpu

[junior halo]

oh

[quasi ether]

Oh, Oof, Hold on, Messed up on my end

[crystal gull]

pf

[junior halo]

gpu tab

[crystal gull]

oh

fuck

nope

colab it is for you!

[junior halo]

colab time

[crystal gull]

unless you have a tab called GPU 1 after it

[quasi ether]

I knew it would come down to Colab.

[crystal gull]

both yalls gotta use colab 😭 lol

[junior halo]

Colab ain't too bad.

[brittle wing]

which one is that ?

[crystal gull]

i feel bad for those who dont have the gpu to local, though i waited so long for my upgrade. I was sitting on a 1060 6gb for FOREVER

[junior halo]

I've been meaning to upgrade for several months now.

Don't have the money.

[crystal gull]

yo dj what should show use for ai covers ey?

[junior halo]

Colab suffices for now.

[crystal gull]

do you know?

[crystal gull]

psh if you ever need something done faster than colab my dms are open ;p

[junior halo]

I don't understand your question.

[quasi ether]

I'll make sure to when I have access to my better computer, send the GPU here just to see.

[crystal gull]

"show" is trying to make ai covers, do you need a specific colab setup for that or could you use something like applio?

I think aihub suggested using llaria?

[junior halo]

Colab works brilliantly.

It doesn't eat up as many compute units as training.

[brittle wing]

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• Ilaria RVC, by thestingerx Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab
• 🆕 UVR5 NO UI for Google Colab, by Eddy Google Colab

How to use?

🤗 Hugginface Spaces

• Applio, by IA Hispano Huggingface Spaces
• Ilaria RVC, by thestingerx Hugginface Spaces
• RVC-HFv2, by r3gm Huggingface Spaces
• AICoverGen, by r3gm Huggingface Spaces
• Advanced RVC Inference, by r3gm Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces

More commands: -audio, -local, -overtrain, -realtime, -rvc, -help

[brittle wing]

so i clikc the first one?

[crystal gull]

yes but which should show use 😭

[junior halo]

I use Hina: https://colab.research.google.com/github/hinabl/AICoverGen-Colab/blob/main/Hina_Mod_AICoverGen_colab.ipynb?pli=1&authuser=1#scrollTo=vC4gLMHI9xb3

[crystal gull]

Try out what Kah uses ig

[brittle wing]

cant sign in

[crystal gull]

What

[junior halo]

can't sign in, can't use it.

[brittle wing]

it keeps redirecting me to sign in

[crystal gull]

So sign into a google account

Or make one

[brittle wing]

okay i got it

[crystal gull]

thats it i think im gonna go find a random girl at my school and pay her to stand infront of my mic and say shit for 15 minutes straight

im throwing hands, i cant find any good data

[livid hazel]

where would i use a tts with custom models

[junior halo]

I have four words for you:

raw

videogame

audio

files

Pick any cute sounding character. Pluck the audio from the game files directly or an online archive.

[crystal gull]

what video game has an e-girl talking for 15 minutes

im gonna need to play that whole game

[livid hazel]

valorant

[rare goblet]

Ayo? @livid hazel level 2 !!!

[junior halo]

https://www.youtube.com/watch?v=acXbuRR4ZoM

[crystal gull]

best e-girl

[junior halo]

warframe copypasta moment

[livid hazel]

do you need a good egirl voice model?

[crystal gull]

yes

[junior halo]

OOH

[crystal gull]

lol

[livid hazel]

bruh

[junior halo]

Overwatch??

[crystal gull]

good point

[junior halo]

Yeah?? :D

[livid hazel]

delta can i dm you/

[crystal gull]

yes

[crystal gull]

🤔 perhaps kiriko

[slim geyser]

https://tenor.com/view/cooming-coom-aaaaaaa-im-gonna-coom-gif-25545869

I would not use audio of exaggerated characters if you wanna use it for real time. You'll just end up sounding like a cartoon character

Unless you're into that sorta thing, I guess???

[junior halo]

Junkrat

best e-gorl

Female Drifter voice lines maybe? (warframe)

[slim geyser]

ok, unironically one of the best egirl data fonts is vtubers doing reaction content on youtube

because they're reacting to a video they got no background noise of their own

just clip the parts where they pause the video

boom

[junior halo]

Genius.

[slim geyser]

yeah there's ton of them

not really surprising considering how brain dead easy it is to react, but yeah

good for finding nice n' normal feminine speaking data as well, since not all of them do the cringe high pitched vtuber thing

[quasi ether]

Just one last question, what's your reconmended epoch for around 4 minutes 40 seconds of training data?

[rare goblet]

Ayo? @quasi ether level 2 !!!

[quasi ether]

:D

[junior halo]

Start with 250 epochs. Keep your eye on the graph.

[quasi ether]

Ok

[crystal gull]

Anyone have a simpler way of realtime audio to microphone besides voicemeeter banana?

It's so bloated and has so many extra features i dont need

[quasi ether]

I followed the tutorial...

[sharp glacier]

it's not Picking up volume no matter what i do.

[rare goblet]

Ayo? @sharp glacier level 2 !!!

[quasi ether]

I'mma restart.

[junior halo]

You need to run the cell that downloads the pretrained model.

[sharp glacier]

i don't understand.

[junior halo]

Which doesn't say in the tutorial, I'm aware.

I was talking to Scratcher, sorry.

[sharp glacier]

oh

[junior halo]

Nor do I have experience with live audio conversion.

[thin stump]

Turn on Sup 2 and change f0 detection to rmvpe. It's probably voicemeeter's fault or your mic isn't enabled for speaking

[quasi ether]

I'm training RN!

[sharp glacier]

whats fo detection i don't see that

[violet heron]

Click start

Instead of harvest use RVMPE

[sharp glacier]

i got passed that part but

[junior halo]

50 mucking epochs to go. yay

Though from the looks of it, the loss/g/total graph doesn't look like it will go any lower again.

The other two graphs are still going down.

Shouldn't be a big deal. I'll see how he sounds.

[sharp glacier]

issue i can't solve

it's not my pc its a new laptop..

[proud elbow]

you dont need voicemeeter, VAC could work without issues, whose download link can be found in pinned guide in #🔍│help-w-okada

[rare goblet]

Ayo? @proud elbow level 40 !!!

[quasi ether]

So I have finished training my model, but how through the method I did could I get a ZIP out of it so my RVC engine could import it?

[charred hound]

Anybody know why I can't uninstall RVC? I'm the administrator, I shouldn't need my own permission to uninstall but it refuses to give me access to this folder specifically.

[sharp glacier]

i don't see it

guy's i'm sorry i'm so stupid..

but i'm not good with pcs

[quasi ether]

NVM

[proud elbow]

you should find it in the rickroll guide: https://rentry.co/VoiceChangerGuide
(ctrl+F VAC if not sure)

[proud elbow]

simply dont use that old application, get newer and better one

-local

[azure marsh]

🖥️ Local stuff

• 🍏 Applio, by IA Hispano GitHub
• Mangio-RVC-Fork, by Mangio621 Huggingface
• RVC Studio, by SayanoAI Huggingface
• AICoverGen, by SociallyIneptWeeb GitHub
• Replay, by Replay Team Website
• Original RVC, by the RVC-Project team GitHub
• GPT-SoVITS, by RVC-Boss GitHub

Need some help? 🤔

You can find more info on the #1159513888199540817 channel. If you can't find your answer, feel free to ask for help in #✨│ai-help. Credits to Faze Masta and Antasma for compiling these links.

More commands: -audio, -colabs, -overtrain, -realtime, -rvc, -help

[sharp glacier]

the issue is the guides for w-konda

i never found the one for the other thing

[sharp glacier]

so i'm confused i only need w-o and the vac right?

[violet heron]

RVC is portable

It’s not a install, you just delete the folder

[thin stump]

you can uninstall voicemeeter now

[charred hound]

Yeah by uninstall I mean delete the folder and it will not allow me to do so.

[thin stump]

never follow youtube tutorials since they're old

[charred hound]

The only folder on my PC that does this btw.

[violet heron]

Can you delete other folders

What shows up?

[violet heron]

Don’t use RVC-GUI

It’s very outdated

[sharp glacier]

okay banana is uninstalled

[violet heron]

Use original RVC instead

[rare goblet]

Ayo? @sharp glacier level 3 !!!

[charred hound]

Says I need permission from my own god damn account to delete it and also refuses to let me modify or change permissions for the RVC folder at all.

[rare goblet]

Ayo? @charred hound level 1 !!!

[quasi ether]

ok

[sharp glacier]

i opened the file and see a bunch of stuff what do i click..

[thin stump]

You don't have to click anything. I don't remember doing that

[proud elbow]

if you cant open the guide unless using VPN, here's the link for VAC
https://software.muzychenko.net/freeware/vac470lite.zip

[charred hound]

So I guess I just have a 10gb folder of shit that I'm never using again just sitting on my desktop now.

[violet heron]

Did you download vac lite or vac trial

[violet heron]

You should download vac lite

[thin stump]

i mean leave the folder alone

[violet heron]

Try restarting then trying to delete

[quasi ether]

Currently getting rid of old GUI, and installing OG

[violet heron]

Or delete its contents one by one

[sharp glacier]

i got what was sent to me by MJ earlier today..

[thin stump]

which is VAC

[sharp glacier]

i guess? it's lite i have no idea

i don't understand pc....

[violet heron]

Run setup64

[sharp glacier]

Fucking bullshit

THERES NOTHING WRONG WITH MY PC!1

[charred hound]

Windows more like windowlicker

[sharp glacier]

does anyone have one that bypasses this shit

my pc's not bad

and it's fine

tired of people saying "your pc is to low grade"

:/

[proud elbow]

run the setup64 (the 64a is for ARM cpu)

[sharp glacier]

okay what now

[proud elbow]

open the voice changer and set output to the VAC (Line 1 iirc)

[sharp glacier]

alright let me get to that

witch one theres 3 line 1s

[junior halo]

How do I manually export a version of the model at a specific step count during my training?

Which files do I even need? The index and pth, right?

[sharp glacier]

but theres 3 line 1s

[junior halo]

I see an option to export it at a specific step/epoch count, but it says to not do so once the training is complete, which it is.

[sharp glacier]

the top 1?

okay what else?

[junior halo]

Should I try it anyway? Will it mess anything up?

[sharp glacier]

input as what?

oh

mk

alright now how do i link it to dc?

or games

what about output?

alright thanks btw

[junior halo]

I dunno which exact pth and index files to save.

[quasi ether]

How would I put in an audio file and get an ai audio file out?

[junior halo]

The last what?

If I have to find a particular pth file, which index file do I save with it?

[thin stump]

added_index

[violet heron]

Wrong file

You open go web

If it’s not there now it is

If you don’t trust downloading from me you can download from the repo

[quasi ether]

Ok, I would like to ask how to put in custom AI models in Web GUI?

[violet heron]

Put the .pth in the weights folder

Which is in assets

[quasi ether]

Ok

[violet heron]

And put the .index in logs

[quasi ether]

Thx

[sharp glacier]

so i try to extract this but it doesn't work

https://huggingface.co/Blocktoast64/The-AI-Bakery/blob/main/Rouge (prime).zip

it's in the server like all the voice files i went to downloads hit extract all browse desktop ai voice folder and extract but it only keeps transfering as a zip.

i think it's broken

[tawny sundial]

whats the command for the doc

[thin stump]

try without the blob https://huggingface.co/Blocktoast64/The-AI-Bakery/tree/main

[open mountain]

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• Ilaria RVC, by thestingerx Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab
• 🆕 UVR5 NO UI for Google Colab, by Eddy Google Colab

How to use?

🤗 Hugginface Spaces

• Applio, by IA Hispano Huggingface Spaces
• Ilaria RVC, by thestingerx Hugginface Spaces
• RVC-HFv2, by r3gm Huggingface Spaces
• AICoverGen, by r3gm Huggingface Spaces
• Advanced RVC Inference, by r3gm Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces

More commands: -audio, -local, -overtrain, -realtime, -rvc, -help

[rare goblet]

Ayo? @open mountain level 1 !!!

[sharp glacier]

now i'm having yet another issue

i can't hear my youtube audio.

and idk how to fix it. i got discord working and games but for whatever reason yt now won't play audio only fix i found was uninstalling things. but does anyone know the corect fix

[alpine basin]

I need help trying to train a new model, I keep trying the tutorials but they wont work for me

[thin stump]

you tried the guides here? https://docs.aihub.wtf/ and they don't work

[alpine basin]

Yes but I think I’ll poke around at it more tomorrow, I’ve been up trying to do it for the last 3 hours - I’m not sure why but when it gets to the GUI part it won’t go through

[rare goblet]

Ayo? @alpine basin level 1 !!!

[alpine basin]

Idk I’m new to anything so I’m probably doing something wrong

[thin stump]

I'll let someone else answer that. I'm not familiar with training on GUI

also you can post in #1192011222023950368 with a pic attached

just in case someone sees it

[alpine basin]

I’m open to learning any way of training but yeah I’ll try again - thanks dude!

[sharp glacier]

yea... idk how to fix yt.

[rare goblet]

Ayo? @sharp glacier level 4 !!!

[proper sapphire]

how do you tell you're overtraining

[compact sinew]

g total loss went up basically

[proud wagon]

How do make an AI cover?

[proper sapphire]

where do you check that

[finite galleon]

But seriously you could check #1159513888199540817 and find out there

[proper sapphire]

I set it to 40000 SR

idk where it's getting 48000 from

[glossy plank]

Hi,
is there a passable to make AI sounds crispy with high pitch voices?

[proud elbow]

you had set it 48k in preprocessing before, so you have to start over on it

[proper sapphire]

is there no way to convert from 40k to 48k without restarting completely

[proud elbow]

you have to either continue with 48k or start over completely with 40k (or vice versa?)

[proper sapphire]

what does tranpose do again

[proud elbow]

pitch shifting in inference

[proper sapphire]

How do you check whether you're overtraining

[rare goblet]

Ayo? @proper sapphire level 3 !!!

[pastel oak]

https://docs.aihub.wtf/rvc/resources/epochs--tensorboard/

[proper sapphire]

thanks

[pastel oak]

Please use the most up to date version of RVC to avoid these small mistakes

https://rentry.co/RVCRealtimeGuide

[toxic thunder]

hey uh, I'm using AICoverGen-WebUI, by SociallyIneptWeeb (modified by Hina), and I noticed there is always a little bit of the original persons voice in the background in the output audio. Is there a way to prevent this?

[wanton sail]

FileNotFoundError: [Errno 2] No such file or directory: '/content/drive/MyDrive/rvcDisconnected/filelist.txt

how can i fix that

[proper sapphire]

On average how long is an epoch supposed to take

[wanton sail]

i used that colab before and it didnt finish now i guess ill use applio

rvc disconnected always gives errors i bored

[rocky fjord]

is it better using Server Audio or Client Audio?

[rare goblet]

Ayo? @rocky fjord level 2 !!!

[wanton sail]

you must use client

[rare goblet]

Ayo? @wanton sail level 1 !!!

[rocky fjord]

mmm...

why?

[wanton sail]

'cause you dont have a server or vds, that happens on local network so client

[proper shale]

It really depends, if server works better for ya

Go for it

[proper shale]

Autosave enabled

[proper shale]

Depends on the length of dataset

And batch size

[wanton sail]

i disabled it but nothing change

[proper sapphire]

what does batch size exactly do

[wanton sail]

am i need set it from 0

[rocky fjord]

actually it is

because when I use Client then forward to Discord, my friends voice is somehow also got converted

but when I use server, only my voice got converted

[proper sapphire]

I have a batch size of 2 and dataset that's 12 minutes long

[proper shale]

I wouldn't really recommend using that if you're making AI covers, since it can fail on the vocal extraction part. Use MVSep to extract vocals and prefer other Colabs instead, imo that's the best option

[proper shale]

Mmmm that'll take a while per epoch

[proper sapphire]

what does batch size do

[proper shale]

Trains x number of samples at once

Batch size 2 trains 2 samples at once.

Those samples are 3 or 4 seconds long...

[proper sapphire]

oh

[proper shale]

yeah do everything again tbh and see if it works this time

[proper sapphire]

is having a batch size of 1 bad for learning rate

[proper shale]

Yeah since it'll change parameters every sample

Use 2 and beyond

[proper sapphire]

Im looking at my tensor board but for some reason loss/g/total isn't updating but everything else is

[proper shale]

Sometimes you do need to click the refresh button again

[proper sapphire]

alr

i set it to auto update

but it just didnt work

I'm at 56 epochs

how much more do you think I'll need before it plateaus

[proper shale]

Mmmm I suppose it'll keep going for a while

Check mel and kl

[proper shale]

Mm

Yeah it's probably still going

Though I'm kinda worried it's getting high kl and mel values

Test out and see how it's going, though. Values don't mean much if it sounds good yk

[proper sapphire]

what do kl and mel represent

[placid holly]

i have a problem with "RVC Realtime"

every time it infer voice, for the first word it glitching/cut
-for example i count 1 to 5, that 1 will cut/glitching, rest are fine
-next i stop speaking for several second, and tried speak again, it happen again
-only happen everytime on first word after silence (at start of it inferring)

any guess what happen or solution?

Windows 11
i5 11400
RTX 3070 Laptop

[rare goblet]

Ayo? @placid holly level 9 !!!

[proper shale]

Kl = how different the model is from the dataset.

Mel = how well it's reproducing clarity from the dataset

Lower is better

[proper sapphire]

i see

i have a pretty clear dataset

hm

[proper shale]

Increase extra inference time perhaps

[rocky fjord]

/what's Thresh do

[wanton sail]

it worked !

omg

[proper shale]

Mic sensitivity slider, basically

[placid holly]

i did, max to 1.50

still there

[proper shale]

yoooo

[rocky fjord]

So, lower are good?

if my surrounding a bit noisy

[proper shale]

If you have bg noise, maybe increase

Higher = less noise being converted

[proper shale]

Can you share your settings here real quick?

Just wanna see what else you set up so maybe we can change something and make it work better

[rocky fjord]

Ah I see, thanks

[proper sapphire]

is there a way to stop training on this version

[proper shale]

Perhaps it's doing that because of index?

I'd say mess around with the loudness factor and response threshold stuff, too

Haven't used RVC realtime yet, so I'm hoping it can be fixed like that

[proper shale]

You can close command prompt, as long as it isn't saving the D/G files you should be fine

I see your save frequency is 10, so, every 10 epochs it'll save the D/G files

If you're on epoch 60 for example, I suggest waiting until epoch 61 to close it

So you can be sure everything is fine

[rocky fjord]

really, I'm confused

Sometimes, the voice from other people in voice channel are got converted by the RVC

although the input is Microphone

[placid holly]

it fixed it, but sometime it still there, but not so common

[toxic thunder]

Okay thank you

[placid holly]

ah sorry, it still there, it back again

[proper shale]

Feels like mic bleed

[proper shale]

Not really sure why this is happening tbh

[proper shale]

You're welcome :)

[proper sapphire]

yeah it's really bad rn

I'll probably leave it on while I sleep

[proper shale]

Oh that's probably your dataset

[proper sapphire]

My data set seemed fine

No background music

no other people talking

[pastel oak]

This happens if Extra is above 1.01~ish i think. It has been brought up before but not fixed yet

So feel free to lower it to that, or around that number, and try around

It only sounds muffled in the beginning if the voice changer has been idle for ~5-10 seconds

[proper shale]

Perhaps it sounded too muffled?

Anyway, keep training and see how it goes

[placid holly]

i'll try sour solution later

[proper shale]

Maybe it's RVC screwing up, too

I don't doubt that

[proper sapphire]

is having a little echo bad

[rare goblet]

Ayo? @proper sapphire level 4 !!!

[proper sapphire]

i can send you what i'm using

[proper shale]

Mmmm yes

Oh yeah don't really post datasets in here

Forgot to tell ya

[proper sapphire]

oh

[proper shale]

Taking a quick listen, there are some noisy samples inside

The first few seconds seem decent but have some slight bg noise...

[proper sapphire]

yeah idk what that is

it's like a film reel

[proper shale]

Did you use any type of isolation for the dataset?

[proper sapphire]

no

[proper shale]

Mmmm

[proper sapphire]

should I?

that's just audio i clipped together from youtube videos

[proper shale]

I suggest doing so, maybe use MVSep Demucs DNR, that seems to remove noise and background noise quite well

https://mvsep.com - also create an account for lower queues

[proper shale]

Yes

[proper sapphire]

ok

[sturdy sluice]

How do you prevent voice cracks and make ai covers that sound more natural?

[frank musk]

Use an audio with quality and a model with quality, then you can edit it using any DAW you like to sound more decent

Any RVC link that's working?

[sturdy sluice]

I don't know what DAW is

[rare goblet]

Ayo? @sturdy sluice level 1 !!!

[sturdy sluice]

I'm kinda new

What do you mean RVC link?

[tame mica]

bruh

digital audio workstation

like fl studio, reason, garageband

[proper sapphire]

uhh

[tame mica]

convert the audio maybe

[proper sapphire]

i tried both wav and mp3

[tame mica]

strange ... wav and mp3 works for me

[proper shale]

Oh wait maybe it's too long

Try separating 5 mins of ur dataset then, it should work

[tame mica]

iirc mvsep will cut audios more than 10 mins

[proper shale]

Try to use the lead vocals only, voice cracks can happen with poorly isolated lead vocals or when you're using all vocals at once

[proper shale]

Didn't do that for me though... 😭

[proper sapphire]

if I use an altered dataset should I restart or continue from what Ihave

[proper shale]

You'd have to train from the beginning

[proper sapphire]

doesn't work

[proper shale]

Huh... that's weird

Did you export inside Audacity? Hmmm

[sturdy sluice]

I followed the guide on the website about isolating voices. Is there a better way to isolate them?

[proper shale]

You should be fine for the most part...

Though I think you should use better models nowadays

If you're on MVSEP - use BSRoformer (most recent model type) to get the best vocals, and use Ultimate Vocal Remover HQ on model type "UVR-BVE" to get the background vocals and lead vocals

[placid holly]

ok it fixed it, thank you

mine better at 2.99, lower better but i lost accuracy

[sturdy sluice]

So BSRoformer is for getting the instrumentals and UVR-BVE is to get lead and non-lead vocals?

[proper shale]

yeah BSRoformer gets all vocals and the instrumentals

[sturdy sluice]

Alright, thanks. I have never used MVSEP, so I'll give it a try

[proper sapphire]

Alright I'm going to leave it training for 300 epochs and see how far that gets me

[tropic trail]

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• Ilaria RVC, by thestingerx Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab
• 🆕 UVR5 NO UI for Google Colab, by Eddy Google Colab

How to use?

🤗 Hugginface Spaces

• Applio, by IA Hispano Huggingface Spaces
• Ilaria RVC, by thestingerx Hugginface Spaces
• RVC-HFv2, by r3gm Huggingface Spaces
• AICoverGen, by r3gm Huggingface Spaces
• Advanced RVC Inference, by r3gm Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces

More commands: -audio, -local, -overtrain, -realtime, -rvc, -help

[wind ice]

Anyone got Evelyn?

[gentle merlin]

-help

[azure marsh]

Command has expired

[gentle merlin]

-local

[azure marsh]

🖥️ Local stuff

• 🍏 Applio, by IA Hispano GitHub
• Mangio-RVC-Fork, by Mangio621 Huggingface
• RVC Studio, by SayanoAI Huggingface
• AICoverGen, by SociallyIneptWeeb GitHub
• Replay, by Replay Team Website
• Original RVC, by the RVC-Project team GitHub
• GPT-SoVITS, by RVC-Boss GitHub

Need some help? 🤔

You can find more info on the #1159513888199540817 channel. If you can't find your answer, feel free to ask for help in #✨│ai-help. Credits to Faze Masta and Antasma for compiling these links.

More commands: -audio, -colabs, -overtrain, -realtime, -rvc, -help

[rare goblet]

Ayo? @gentle merlin level 1 !!!

[gentle merlin]

-gui

[azure marsh]

https://cdn.discordapp.com/attachments/1122285248844144733/1203460490475343953/caption.gif?ex=65d12cec&is=65beb7ec&hm=bd2fb8d010006dd7c6e3c1c67d3ae846fd1478e1a3124c544c31b43086fe54aa&

[errant prairie]

Is there any way MMVC (AI voice changer) can work with AMD GPUs?
Or only cuda cards?

Thanks

[rare goblet]

Ayo? @errant prairie level 1 !!!

[errant prairie]

Shuld just dump it in a folder with ai voicechanger i guess? Right?

[chrome epoch]

How this fix??
TypeError: Cannot read properties of null (reading 'modelSlots')
TypeError: Cannot read properties of null (reading 'modelSlots')
at i (http://127.0.0.1:18888/index.js:2:1305771)
at Object.updateServerSettings (http://127.0.0.1:18888/index.js:2:1306003)

[errant prairie]

kk

[chrome epoch]

Mod?? Admin??

[proper sapphire]

what's a good mel and kl

[noble crater]

You're sure you followed all the steps correctly? Also you can't change the folder name when the model finishes to train(or gives an error)

[crystal gull]

Yeah the issue was fixed I think I was trying to retrain under the same name without deleting the previous files

After a few restarts and name changes it fixed

[wet elm]

Web Server Launch Exception, DLL load failed while importing beatrice_internal_api
how to fix?

[noble crater]

And it trained correctly?

[crystal gull]

Yup

[noble crater]

LikeCat

[worldly oasis]

Hello i have some trained and full rvc i think ( i mean i have the voice that i want) but i cannot make it to song that i want does someone can help me

Does someone can help me

[maiden remnant]

can anyone help i have amd graphic card and it wont recongnise it at all it keeps using cpu

[proper sapphire]

i take it this means im overtraining

[tranquil raven]

what is this?

[wanton sail]

bro colab time is over and i couldnt find .pth files

it was 500 epoch, i guess rvc disconnected time was not enough for this epoch

[rare goblet]

Ayo? @wanton sail level 2 !!!

[tame mica]

-overtrain

[azure marsh]

How do I know if my model is overtaining?

All-In-One Guide on how to make a good model

This guide explains how the D and G files works and much more: https://rentry.org/RVC_making-models

Credits: LUSBERT

Automated Overtraining Detection (AOD)

Will be available soon in #1159513888199540817

Credits: grvyscale

[glad zealot]

what colab are u using?

[proper sapphire]

MVSEP keeps giving me an error saying "This is not an audio file'

I've tried both WAV and MP3

[wanton sail]

rvc disconnected

[carmine hearth]

At the end of the colab runtime, in the case of RVC Disconnected, any models that were not manually saved will disappear.

[muted whale]

Its normal that it uses almost 60% of my CPU when im using it?

[worldly oasis]

Hi i have a model but i don''t know where to set it to make a AI Cover

[proper sapphire]

Does anybody have an issue of MVSEP not letting you upload any audio

[worldly oasis]

Where i can use the model to make the song

[slate bramble]

how can i fix this error in collab? (GPU Check)

Exception                                 Traceback (most recent call last)
<ipython-input-23-74d578b798b3> in <cell line: 21>()
     23 
     24 else:
---> 25   raise Exception("No GPU detected; training cannot continue. Please change your runtime type to a GPU.")
     26 gpus = "-".join(i[0] for i in gpu_infos)

Exception: No GPU detected; training cannot continue. Please change your runtime type to a GPU.

p.s. i have tried changing runtime doesnt fix it

[rare goblet]

Ayo? @slate bramble level 1 !!!

[glad zealot]

run out of gpu, use a different account or wait 12-24 hours

[slate bramble]

OK thank you!

[glad zealot]

depending on what colab wants

[slate bramble]

ive been working on this voice, do u have any recommendations on where i can find a code to help prevent timeout?

if not tysm anyways 🙏

[glad zealot]

google-

[worldly oasis]

Sorry do you know where i can use already a model to make a AI Cover

[azure marsh]

Suggestions for @worldly oasis

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• Ilaria RVC, by thestingerx Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab
• 🆕 UVR5 NO UI for Google Colab, by Eddy Google Colab

How to use?

🤗 Hugginface Spaces

• Applio, by IA Hispano Huggingface Spaces
• Ilaria RVC, by thestingerx Hugginface Spaces
• RVC-HFv2, by r3gm Huggingface Spaces
• AICoverGen, by r3gm Huggingface Spaces
• Advanced RVC Inference, by r3gm Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces

More commands: -audio, -local, -overtrain, -realtime, -rvc, -help

[slate bramble]

do u have the model?? if not and u want to make a AI voice use that^

[worldly oasis]

I have the model

[slate bramble]

what ai cover are u refering to

u can use text to speech or RVC n say the words/sing wtv

all i will say bc idk if i can even help you 🙏

[slate bramble]

ok i used chatgpt it told me a code lol but when i start the code no other cell will start if its active advice?

main issue i keep timing out thats why im having this gpu issues 😭

[worldly oasis]

i done the vocals and i have the full model

Is it free

Before i used Google Collab but they ruined it

[rare goblet]

Ayo? @worldly oasis level 2 !!!

[worldly oasis]

Yea but it's not the same as before i already have the model just i need it to place it to some music but don't know how

[slate bramble]

use a software editor??

😭

like how ppl make yt videos

or FL studio

[worldly oasis]

is there any more easible way before i used google collab but now it's not the same

[slate bramble]

ok if i understand right u are trying to make a song right?

[worldly oasis]

just to put the model in some music to make it with that voice

[worldly oasis]

yes like Justin Bieber played by Selena Gomez

[slate bramble]

do u have the audios?

[worldly oasis]

Yea

i make the model

[slate bramble]

u have the voice lines already ??

[worldly oasis]

Yes

[slate bramble]

so u just want to add music and voice audio together?

[worldly oasis]

I need to sound like her sing song that's not her

[slate bramble]

ooo wait ok we are talking ab model sorry i cant understand very well

[worldly oasis]

i mean that song already is played by someone and i need it to change that sound with mine

[slate bramble]

you need help training it

[worldly oasis]

no no it's already done

just to place it

[slate bramble]

u dont know how to use RVC?

guys am i slow here?

[proper sapphire]

What mvsep option is recommended for removing artifacts for datasets

[slate bramble]

😭

erm

[worldly oasis]

Oh sorry guys i don't know how to describe it

let's say i have Sel's voice and want's to be on JB song but don't know how

i already have done training vocal model

[slate bramble]

is it possible forsomeone toanswer this if not itsok T_T

[proper sapphire]

What mvsep option is recommended for removing artifacts for datasets

[worldly oasis]

yesss but howww

[humble verge]

@junior halo

[violet narwhal]

Hi guys, I need help with my gpu, i can only chose CPU and nothing else

Intel® Core™ i7-4600U CPU @ 2.10GHz

[rare goblet]

Ayo? @violet narwhal level 1 !!!

[violet narwhal]

:0

isn't it?

[proper shale]

oh no

[violet narwhal]

TToTT

Sooo

i should download online version?

[proper shale]

You'd have to use Colab yeah

[violet narwhal]

I stayed woke up until 3 am yesterday trying to fix it

thanks a lot

[muted whale]

If I choose the gpu option, it is normal if it still uses cpu instead of the gpu? (it uses like 30-60% of my cpu when im using it)

[proper shale]

You're welcome

[proper shale]

If you're on AMD, that can occur

[muted whale]

yea I use AMD

[proper shale]

Have you exported your downloaded models to ONNX though?

[muted whale]

nope

Im new on this so I don't really know how it works

[rare goblet]

Ayo? @muted whale level 1 !!!

[muted whale]

[proper shale]

Definitely do that, since it'll help performance wise. It'll use CPU though, that's just how the AMD version is, sadly

https://rentry.co/VoiceChangerGuide - check the "uploading models to AMD" section

[muted whale]

And what is this supposed to do?

[proper shale]

Basically, converting to onnx makes it actually use your GPU

[muted whale]

It will make it work better? I mean, when I import a new voice, this one just gets really laggy and it doesn't sound natural

[proper shale]

Yes, it'll make it not laggy, at least

[stiff goblet]

Which one is better for medium quality English dataset? Titan or Original pretrain?

[muted whale]

and the voice will sound more normal?

[proper shale]

Just go for original

[proper shale]

Ehhh, not so sure about that, but you can always configure the Extra, Index and Tune parts to make it better

[muted whale]

what does the index do?

[proper shale]

Applies the model's accent to your voice

(or tries to, yk)

[stiff goblet]

@proper shale should i go for 32k or 40k training by looking this spectogram ?

[proper shale]

32k

[stiff goblet]

thank you

[proper shale]

since it doesn't go a lot beyond 16kHz

If it really went beyond that, up to 24kHz with actual, usable data, yeah you'd go with 40/48K

[alpine basin]

can somebody help me, I keep trying to make a model but it never lets me get past this part and I've been following the guide to a tea

[stiff goblet]

it barely goes to 22kHz

[proper shale]

and it feels like it's not really data, more like filler

dithered stuff

[proper shale]

Have ya linked it with Google Drive yet? Hmm

[alpine basin]

yeah

[proper shale]

That is... weird tbh

I suggest switching up Google accounts, putting that zip in the rvcDisconnected folder, and trying again

[brittle wing]

What does RVC stand for?

nvm its in the channel desc

[proper sapphire]

Is it supposed to be like this

[patent horizon]

Hi, i wanted to ask for some help with the installation of python requirements which doesn't seem to work for me, it crashes with many logs and this error

ModuleNotFoundError: No module named 'distutils'

[random stump]

When training using Applio, how can I manually analyze graphs to know if I'm overtraining?

[lusty plaza]

With the RVC can you put the google drive link into 'Inferencing voice' or is there another step?

[violet narwhal]

Hi

i would like to know if this has something to do with some errors going on in collab version

[viscid moss]

https://docs.aihub.wtf/rvc/resources/epochs--tensorboard/#usage-guide

[viscid moss]

What colab are u using?

[violet narwhal]

ngrok

i runned it again and it didn't work

https://github.com/w-okada/voice-changer/blob/master/Realtime_Voice_Changer_on_Colab.ipynb

this is the link i went through

[viscid moss]

That's outdated

[violet narwhal]

where can I find another collab?

[viscid moss]

#🔍│help-w-okada message

[odd shale]

Did you download the correct W-Okada version that matches your GPU?

Which GPU you have?

[random stump]

Oh lol, I've been using Applio Colab and damn I'm lucky it say's the runtime disconnected but also that it's waiting to finish the current execution, and that actually just continues to train the model

[violet narwhal]

Hi, im trying ilaria RVC but it doesn't uploads any model

would you mind helping please :c

mi model dont appears

[manic plaza]

This might be a stupid question, but my voice dataset is in 48khz. Do I use the 48k sample rate
Thanks for helping!

[brittle wing]

yes

[manic plaza]

thanks

[proper shale]

Not necessarily, check the spectrogram

Normally I just use 32k, 48k can bring noise and doesn't bring any big benefits

[proper shale]

Use the most recent one instead: https://colab.research.google.com/drive/1mHKTGH5e3SAyDSBss1KtiYRbDdQzwSMs

[west fox]

could this be a possible point of over train?

[proper shale]

Check mel and kl. seems to still be going

[west fox]

yeah it's still going

[compact sinew]

no

[odd shale]

Nope.

[rotund knot]

Hello.

When I try to download my model from the EasyGui colab (gradio), I get an error message :
FileNotFoundError: [Errno 2] No such file or directory: 'assets/weights/My-Voice'

What shall I do ?

[vapid pewter]

i have everything downloaded but it doesnt pick up my mic, what should i do?

[maiden remnant]

can anyone help i have amd graphic card and it wont recongnise it at all it keeps using cpu

[manic plaza]

alright, thanks

[meager pollen]

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• Ilaria RVC, by thestingerx Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab
• 🆕 UVR5 NO UI for Google Colab, by Eddy Google Colab

How to use?

🤗 Hugginface Spaces

• Applio, by IA Hispano Huggingface Spaces
• Ilaria RVC, by thestingerx Hugginface Spaces
• RVC-HFv2, by r3gm Huggingface Spaces
• AICoverGen, by r3gm Huggingface Spaces
• Advanced RVC Inference, by r3gm Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces

More commands: -audio, -local, -overtrain, -realtime, -rvc, -help

[bronze moth]

mine just says "cpu", i doubt that the rx 580 is supported but if it is, tell me what im doing wrong.

[violet heron]

Refresh

[violet heron]

First part explains it

Depends on cuda for GPUs so you can only really infer

[stiff goblet]

Which one is better for training? Applio or RVC Disconnected ?

[maiden remnant]

im using the ai voice changer but it only runs on cpu cuz i have amd and not nvidia

cuda is on nvidia

[weak flame]

Hugginface Spaces works in amd?

Because, I have rx 580

[maiden remnant]

same

i have rx580

[weak flame]

LoL

[maiden remnant]

u got the same issue?

[rare goblet]

Ayo? @maiden remnant level 1 !!!

[weak flame]

At least with applio yes why it told me only nvidia graphics xd

[maiden remnant]

MMVCServerSIO_win_onnxdirectML-cuda_v.1.5.3.15

someone said it might work on this version

though i cant try rn

[weak flame]

luck bro

[rare goblet]

Ayo? @weak flame level 1 !!!

[violet heron]

Did you download the directml verison

[violet heron]

HF Spaces runs in the cloud so your GPU doesn’t matter

[maiden remnant]

yes

[maiden remnant]

now i downloaded older version and il see if it works

[white jasper]

how do i do this

[maiden remnant]

VoiceChangerV2 Initialized (GPU_NUM(cuda):0, mps_enabled:False, onnx_device:CPU-DML)

see

[tranquil raven]

How do I monitor my voice with rvc?

[rare goblet]

Ayo? @tranquil raven level 2 !!!

[tranquil raven]

Are there any video tutorials to rvc's gui

[rare goblet]

Ayo? @violet narwhal level 3 !!!

[violet narwhal]

:C

[violet narwhal]

Ilaria is not working also :c

[low shard]

You finished your daily google colab gpu

either wait till tmr or use another google acc

thats the reason why its not working

[violet narwhal]

how much should I wait

[low shard]

either around a day, or use an alt google acc

[maiden remnant]

hey

[rare goblet]

Ayo? @maiden remnant level 2 !!!

[maiden remnant]

how can i use amd gpu with ai cuz the cuda shi wont allow me

[uncut cairn]

its not really of a question related to rvc but i want to try to make music using it, i can do that but i cant seem to figure out how to put the background music on the voice over, everybody else's seem so perfect

[restive tusk]

https://github.com/Vali-98/XTTS-RVC-UI

how do i get this working

i got it running

i keep getting error though after moving some models into the rvc folder

[dull orchid]

-local

[azure marsh]

🖥️ Local stuff

• 🍏 Applio, by IA Hispano GitHub
• Mangio-RVC-Fork, by Mangio621 Huggingface
• RVC Studio, by SayanoAI Huggingface
• AICoverGen, by SociallyIneptWeeb GitHub
• Replay, by Replay Team Website
• Original RVC, by the RVC-Project team GitHub
• GPT-SoVITS, by RVC-Boss GitHub

Need some help? 🤔

You can find more info on the #1159513888199540817 channel. If you can't find your answer, feel free to ask for help in #✨│ai-help. Credits to Faze Masta and Antasma for compiling these links.

More commands: -audio, -colabs, -overtrain, -realtime, -rvc, -help

[proper sapphire]

What is a good Mel value to have

and kl

[proper shale]

Kl should be around 1 to 0, mel should be around the mid 10s

[proper sapphire]

i see

[proper sapphire]

my total/g is weirdly bumpy but my mel keeps going down

should i train it more

[proper shale]

mel keeps going, it's still improving

[proper sapphire]

alr

[merry cloud]

So, how do I convert wav to pth?

[proper shale]

you train a model, using the .wav as dataset

https://docs.aihub.wtf/

has everything you need to know about training and RVC in general

[stiff goblet]

Which one is better for training? Applio or RVC Disconnected ? I'm using Google Colab

[merry cloud]

thx

[proud elbow]

change the extension
training the model as pth from wav files as dataset is not called "conversion" tho

[glass stratus]

H guys, just wanted to ask what is currently the best quality audio cloning open source repo/colab going around now adays, since a lot of variants are coming up

[proper shale]

yw :)

[brittle wing]

Test

What's the best inteference time and sample length setting for rvc real time voice changer

For a RX 570

[proper shale]

either applio or mainline, can't go wrong with em

https://docs.aihub.wtf/ for guides n stuff

[proper shale]

test and see what works best, tbh

[brittle wing]

So I have to do trial and error

[brave garnet]

Voicechanger Settings (Okada)

⠀

Settings for AMD GPUs

Don't forget that your models needs to be converted in ONNX!

F0 Det.: rmvpe_onnx (suggested for all series)

7xxx XT cards: 112-128 chunk | +16384 extra
6xxx XT cards: 128-192 chunk | +16384 extra
5xxx XT cards: 192-256 chunk | +8192 extra

RX 580: 192-256 chunk | +8192 extra
RX 570: 192-256 chunk | +8192 extra
RX 560: 256-384 chunk | +8192 extra

Advanced Settings

Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low

⠀

Made with love by Ilaria 💖

[brittle wing]

I already know this list exists

[proper shale]

this is for Okada

[proper shale]

yeah, tbh

[brittle wing]

"Sample length: The realtime voice changer works by sending small chunks of audio for quick conversion, then stitching them together.
Longer sample lengths feed in longer chunks, making the stitches less obvious and reducing GPU requirements but increasing output latency.
On a low end GPU, setting this too low will make the GPU unable to keep up and produces stutters.
On a high end GPU, setting this too low will cause warbling as an artifact of stitching many overly-short chunks together.
Equivalent to "CHUNK" in w-okada."

"Extra inference time: How much old audio to load into each chunk.
The extra context usually improves voice quality for the generated chunk but is more demanding for the GPU.
Equivalent to "EXTRA" in w-okada."

[rare goblet]

Ayo? @brittle wing level 1 !!!

[brittle wing]

Found this on some sort of youtube comment

[proper shale]

yea this is right

[glass stratus]

K thanks✌️

[proper shale]

You're welcome :)

[brittle wing]

If I set the extra interference setting to the max, is that fine for my RX 570 8GB variant?

It seems to be using 35-45% of my GPU usage

[proper shale]

Setting that to the max kinda screws things up

[brittle wing]

In task manager

[proper shale]

Doesn't help much

And also breaks the start of words

[brittle wing]

Do you think 3.25 should be a sweet spot

It's giving this spiky graph on the 3D encoding

[proper shale]

See if that works out

[brittle wing]

Do you think -50 response threshold is too sentivive

I know the fact that setting it to -40 literally makes the client pick any sound from your mic

[proper shale]

Depends on your mic

[brittle wing]

Does this setting even does anything

Like does it even use your CPU threads

[proper shale]

Only when using harvest as your pitch extractor

[brittle wing]

What if I'm using rvmpe

Is it basically useless

[proper shale]

Yup

[brittle wing]

Should I just set that setting to 0 or something lol

[proper shale]

yeah, ig

[ionic oak]

Hello, I have the voice of a person that I would like to use on RVC, but it is only a voice message, how do I do the index and the other thing I don't understand?

[proper shale]

so you want to clone that voice, right?

https://docs.aihub.wtf/ has everything you need to know about training n stuff 🙏

[ionic oak]

It's a friend who authorizes me to take this voice to troll someone however it's a 19 second voice message, how do I put it in the index and the other thing? so yes clone I think

[brittle wing]

Is a RX 6700 XT somewhat better than the RTX 3060 12GB variant

[proper shale]

Yeah you have to download that voice message (or ask them to send a new one) and do the steps to train your model

[brittle wing]

If both are used in RVC

[proper shale]

For AI? nah

[brittle wing]

What about a RX 6900 XT

Does it stand a chance

[proper shale]

Since NVIDIA stuff is just plain better for AI

[brittle wing]

Against the RTX 3060 for AI

[swift thicket]

i cant find it

there is nvidia or gpu install links but i cant find the downloand link of the app