✨│ai-help

[hexed dew]

no on colab

it doesnt work

[latent kettle]

Try to use your model without index

[ocean oracle]

how can i make my model into a url

[snow hazel]

I keep getting this error

Idk what I'm doing wrong

[snow hazel]

idk if it's the pth that I downloaded

[rare goblet]

Ayo? @snow hazel level 5 !!!

[verbal oxide]

Try download latest applio version

[snow hazel]

How do I do that?

nvm

[burnt zealot]

what is this error?

[tidal scroll]

is the ilaria rvc meant to take as long as it does? bc i've been timed out like three times now

[dawn peak]

was just abt to say this

[hollow vigil]

so is it just me or i cant train lol

[tidal scroll]

ive tried others and they work well

only ilaria refuses to give me my output file lol

[hollow vigil]

i just had my gpu take 99% and it train nothing with a batch fil of 3 on a 4060 ti which it shouldnt be

[tidal scroll]

DAMN

[hollow vigil]

idk y

[burnt zealot]

this is on a 4060 ti and it goes to 99 bc i start training not bc of window manager etc

[desert cradle]

i need help with the google drive link

[burnt zealot]

and it doesnt train anything

[tidal scroll]

yeah to me it goes way over the estimated finish line and just times me out after a while

[rare goblet]

Ayo? @tidal scroll level 1 !!!

[hollow vigil]

wdym estimated finish time

[rare goblet]

Ayo? @hollow vigil level 1 !!!

[hollow vigil]

theres none

[tidal scroll]

im using the online v

[hollow vigil]

ohh

okay

is trainning down rn?

[hollow vigil]

is that what u think

[tidal scroll]

i dont understand much ab these things, only the v bare basics riP

this is the first time ive had issues w anything rvc tho

[hollow vigil]

mine is trainning fine now

[tidal scroll]

im gonna try applio

[hollow vigil]

but i train locally

[tidal scroll]

i usually do as well, but rn im on my tiny laptop tryna work sumn out T-T

[molten pecan]

You upload it to hugging face. We have a guide of it

-huggingface

[azure marsh]

huggingface.co upload Guides

English guide

Link: https://rentry.org/fdg_guide_new
Credits: FDG

Spanish Guide

Link: https://rentry.org/AI_Hub_guide_esp
Credits: Julia (ailen2091)

Hindi guide

Link: https://rentry.org/AI_hub_guide_hi
Credits: Enes

Turkish guide

Link: https://rentry.org/AI_hub_guide_tr
Credits: Enes

[molten pecan]

Hehe I'm in the credits :3

[snow hazel]

I think it updated. What do I do now?

[molten pecan]

Yes. It depends on the size of the audio

And if you duplicated the space

If you haven't done it yet, do it

[tidal scroll]

idk what this means

[molten pecan]

Oh okay

I'll explai

You see the three dots above in the page?

[tidal scroll]

ill open it rn

[molten pecan]

Click them and hit Duplicate

Make sure you set it to Private. Then hit duplicate

And wait a few minutes for it to load

[verbal oxide]

It should be fixed in newest version

[tidal scroll]

i need an acc on huggingface right?

[molten pecan]

Yup

I thought that part was on the guide of how to use Ilaria RVC

[molten pecan]

.

Read the above messages. Hope that fixes it!!

Ping me if anything

[tidal scroll]

i used to do this locally so thats why i wasnt used to this T-T

tysm for the help!

[molten pecan]

What's the matter?

[molten pecan]

yw!

[verbal oxide]

Download here https://github.com/IAHispano/Applio-RVC-Fork/releases/tag/v2.0.0

Windows.bat

[desert cradle]

i'm on this website n i'm trying to put selena quintanilla ai model to my fave song with this huggingface link so i need help with that

[rare goblet]

Ayo? @desert cradle level 1 !!!

[snow hazel]

I did that already

[molten pecan]

Which HF space exactly?

Ilaria RVC?

[desert cradle]

HVC

[molten pecan]

Are you getting errors on the Applio installation too?

[molten pecan]

Can you please send a screenshot?

I never heard of it

[verbal oxide]

Hmm idk much, but the updated version is fixing the inference path model

[desert cradle]

@molten pecan

[molten pecan]

Can you send the link of the model here?

In the chat

[desert cradle]

i can

[molten pecan]

Do it please

[desert cradle]

wait in here or

[snow hazel]

Don't think so

[molten pecan]

Yeah in here

[desert cradle]

https://drive.google.com/file/d/1RIdDN-6nE_emNDEp22iCzaBpzPSOfmyK/view?pli=1

[verbal oxide]

Maybe try delete all applio folder and download again in github

[molten pecan]

Have you duplicated the space?

Do you know how?

[snow hazel]

How do I do that?

[desert cradle]

no i haven't

[verbal oxide]

Just delete applio folder on your Drive then download again

[molten pecan]

Click on the three dots that are above the space

[verbal oxide]

https://github.com/IAHispano/Applio-RVC-Fork/releases/tag/v2.0.0

[molten pecan]

Then "Duplicate this Space"

Make sure you select it as Private

And click Duplicate

Wait for a few minutes to load and try again

That will prob fix it

[desert cradle]

google drive?

[molten pecan]

Btw you gotta be logged in HF to duplicate a space

[molten pecan]

Google drive what?

[tidal scroll]

small question: lossy input files work on rvc right?

[rare goblet]

Ayo? @tidal scroll level 2 !!!

[molten pecan]

lossy quality you mean?

Yeah they do

[desert cradle]

i don't have hf

[molten pecan]

Make an acc then

It's pretty easy

[snow hazel]

Do I do 1,2 or 3?

[desert cradle]

i did, now what.

[molten pecan]

Go back to the space and do what I said above

[molten pecan]

.

[verbal oxide]

1 Nvidia support

[desert cradle]

this is so confusing, i'm so new at this

[molten pecan]

It's pretty easy to do so don't worry

I'll send you my guide

[desert cradle]

okay

[rare goblet]

Ayo? @desert cradle level 2 !!!

[molten pecan]

https://rentry.org/HFguide_en

[desert cradle]

am i supposed to selena quintanilla as ai model?

i duplicated how do i name it or just keep it the same?

[molten pecan]

You don't have to name it

You can keep it with the same name

[molten pecan]

You can use any model you want

[desert cradle]

okay

how long am i supposed to wait?

[molten pecan]

It depends on how long the audio is

And the settings that you applied

[desert cradle]

well, it says it's building

[molten pecan]

Oh that will take a few minutes

[desert cradle]

yeah

okay, it's done building

[tidal scroll]

((@julia ur advice fixed it tysm for existing :)))

[molten pecan]

yw!!!!

So happy to hear that ^-^

People don't even say thank you so I appreciate it

[snow hazel]

It just closes

[molten pecan]

Ping me if anything!

[desert cradle]

what am i supposed to do now?

[molten pecan]

It's in the guide

Keep reading it

[desert cradle]

okay thanks

i will keep you updated

[molten pecan]

Aite

[sacred bone]

I know in general they were saying RVC is better for high end, Okada for mid tier, but what sets them apart really?

[rare goblet]

Ayo? @sacred bone level 2 !!!

[sacred bone]

(Not over 90000, only 2 X_X)

[molten pecan]

Try asking on #🔍│help-w-okada

[sacred bone]

I wasn't sure which one I should drop it in since it's asking about both as a pro/con sorta summary without going crazy XD

[desert cradle]

what is index?

[molten pecan]

The model

[desert cradle]

is it a zip?

[molten pecan]

The index?

[desert cradle]

OH WAIT

[desert cradle]

yes that

[molten pecan]

Well uhh

A model has two files to work

A .pth file and a .index file

Usually models are compressed in a .zip file

I hope that answers your question

[desert cradle]

okay

[molten pecan]

I don't know what you were trying to ask

[snow hazel]

when I tried to install it again it did some stuff than closed

[desert cradle]

-Then go back to the main tab and click Refresh and select your model in "Inferencing Voice" and "Auto-detect index path".

-Below it, select your audio file. If you want to convert multiple audios at once, you can do it in the Batch tab.

-Set up the pitch and the settings in Advanced Settings however you want, and you are good to go.

-Click Convert and wait. Keep in mind, if you leave the page for a long time it might lose the connection and it will give you an error. I recommend to come back to the page a few times while it renders.

-To download it the output audio, you click on the three dots on the right of the audio and then Download.

Note: The conversion won't be instantaneous. A 1 minute-long audio will take around 5 minutes. If you want to make a faster space, go to the original space, click on the three dots, Duplicate this Space, and in "Space hardware" you select another hardware. Keep in mind that it's paid and you'll be charged per hour.

ignore the note, what does it mean go back to main

[desert cradle]

oh, sorry about that

[rare goblet]

Ayo? @desert cradle level 3 !!!

[molten pecan]

To the first tab

[desert cradle]

okay thanks

[sacred bone]

I think they're asking what the index is, I kinda take the .pth to be the core of the model, the intended audio signature, and the index is maybe references to the model within, so it can better find the process you're hoping to export, which 'might' reduce processing time, or help slighlty with quality as a side effect? But I've also heard indexes aren't really required either? So maybe I'm just totally wrong X_X.

[desert cradle]

yeah this

i think i did something wrong...

[sacred bone]

RIP

[molten pecan]

Can you still click on Convert?

[desert cradle]

yes

[molten pecan]

Then try again

It might still work

[desert cradle]

okay

[molten pecan]

That's interesting. I'm not familiar with the whole process of making models

So I appreciate the info

Thansk

Thanks

[desert cradle]

it still say error

[violet heron]

HuggingFace Spaces?

[desert cradle]

yes

[violet heron]

Did you duplicate the space

[molten pecan]

Do you have a stable intention connection?

[molten pecan]

Mhm

[desert cradle]

i did all that already

i set it private

[violet heron]

Did you select your song and vocals

And do the vocals file name have no spaces

[sacred bone]

Oh I'm a complete noob to it all, that could be completely incorrect as it's just an assumption, but an index generally helps with sorting/recalling information. So I just hope that applies here, I could be wrong for sure.

[desert cradle]

let me see

[molten pecan]

Gotcha

[desert cradle]

no

[molten pecan]

Thanks either way

[violet heron]

Index usually acts as the accent

[molten pecan]

Then remove any spaces

That the file name has

[desert cradle]

okay

i remove all the spaces n it still tells me it's error. i think i did something wrong.

i used vocalremover.org to remove vocal then after that remove instrumental

[violet heron]

I sometimes use that website to and haven’t had the problem

Are you sure you selected the model

[desert cradle]

yeah i'm sure

[molten pecan]

Maybe the problem is the model

Or the audio

So try using another model and audio

[violet heron]

Model shouldn’t be a problem if it imported properly

Can you share a screenshot

[molten pecan]

It happened a few times that people got errors with a specific model only

It's weird

Or maybe the space couldn't support it or something

[violet heron]

Try a different model

One that’s hosted on HuggingFace

[desert cradle]

okay

it still says error but thank you guys for helping me tho, i appreciate it.

you 2 don't have to waste your time helping me, it's fine right now but i tried... thank you!

[noble dawn]

What does connect to runtime mean ?

[peak bone]

help wehen i select my mice always gives me this error

´`Looks like there's a bit of a problem.

unknwon message

If you clear the information being managed by this app, it may be recoverable.

Error
unhandledrejection
no error stack
NotSupportedError: AudioContext.createMediaStreamSource: Connecting AudioNodes from AudioContexts with different sample-rate is currently not supported.
504/t/et/this.setup@https://9336-34-125-73-244.ngrok-free.app/index.js:2:1268166´´

[zenith hazel]

getting "You cannot currently connect to a GPU due to usage limits in Colab." error in AICoverGen google colab. Anyone know how to fix it?

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

🇺🇸 English (main)

https://docs.aihub.wtf/

​

Translation by country

🇧🇷 Brasil (PT-BR)

https://docs.aihub.wtf/v/brazil

🇪🇸 Spain

https://docs.aihub.wtf/v/spain/

🇫🇷 France

https://docs.aihub.wtf/v/france

🇩🇪 Germany

https://docs.aihub.wtf/v/germany

🇯🇵 Japan

https://docs.aihub.wtf/v/japan

🇰🇷 Korea

https://docs.aihub.wtf/v/korea

🇵🇱 Poland

https://docs.aihub.wtf/v/poland

🇷🇺 Russia

https://docs.aihub.wtf/v/russia

🇮🇹 Italy

https://docs.aihub.wtf/v/italy

[remote verge]

Perhaps a silly question. But you know how you can pick up and keep training the same model by filling in the same details in the gui.

Can I add more files into the dataset and train the same model? For example lets say I trained with 20 min dataset upto 500 epochs, can I add 10 mins more of dataset and train the model further or do I have to retrain a new model from beginning with the new 30 min dataset?

[finite galleon]

Retrain a new one

[remote verge]

Thanks for clearing it up

[radiant flare]

i need (would like) command line usage of the software... gimme that rvc -I -m bob_dylan -f /path/audio.wav -o /path/output.wav

[finite galleon]

Applio has a cli

[remote verge]

Overtrained yet?

[finite galleon]

Looks like its overtrained at 40k

But again hearing your models checkpoints one by one is the best way to know which models are the best

[plush compass]

https://colab.research.google.com/github/ardha27/AICoverGen-NoUI-Colab/blob/main/CoverGen_No_UI.ipynb Are there any nowebUI Colabs other than this Colab?

[dawn kayak]

why does my model randomly make like sounds [most of them are weird moaning sounds]

[proper shale]

Your place might be particularly noisy? Any noise coming from outside can be detected as "voice" if you don't change the S Thresh. I suggest increasing it if this is the case.

[proper shale]

RVC Disconnected too, but that's for training models

[proper shale]

Hmmm, I'd say no right now, but there is 2 worrying signs here - kl stopped descending and is trending to stabilize and loss/g/total is ascending and never coming back. So I think you might need to be careful going forward, because it's likely that overtraining will be happening soon

[pseudo mountain]

Hello, is there anyone who can help?

[proper shale]

The spot highlighted in green would be the safest spot, at around 60k. The ones in yellow are the ones that could be used but I wouldn't be sure if they'd be better than the 60k version of the model.

[proper shale]

What's up?

[finite galleon]

Isnt it over fitting

[proper shale]

Kl and Mel seem to be going down still

[finite galleon]

G/total just goes up after 60k

[proper shale]

So it is still not the optimal point for the model

[proper shale]

That is a sign that it could be starting to overtrain, but not exactly right now

• D/total also seems to be going down

[modern oxide]

160k steps

is it like 5 hours dataset

[finite galleon]

G/total and d/total not going down at the same time is bad right

[pseudo mountain]

Hello, I want to work in this field with curiosity and contribute to people. How can I use the models here?

I am sharing a sample code

import torch
from your_tts_model import YourModel # YourModel must be the class that defines the structure of your model.

Load the model

model_path = 'path/to/walterru.pth'
model = YourModel()
model.load_state_dict(torch.load(model_path))
model.eval()

Process text and convert it into sound waves

text = "Hello, this is a test message."
audio = model.synthesize(text)

Save the audio file

import soundfile as sf
sf.write('output.wav', audio, 22050) # 22050 is the sample rate of the sample audio file.

and what is RVC?

I want to do model training myself, but I couldn't find a proper reference source.

I aim to make a text-to-speech study in which sounds can be produced naturally and to train a model and use it. Can you show me an example of a work that I can use as a reference to help me in this regard?

[rare goblet]

Ayo? @pseudo mountain level 1 !!!

[proper shale]

Yeah. And if d/total starts to plateau, that's a sign

But again, keep this in mind: wait everything out.

[proper shale]

RVC is an AI platform that allows you to convert audio files of a speaker to another speaker. Uses Python and even torch too

[modern oxide]

how many epoch's is that?

[finite galleon]

So you want to do tts to rvc?

[pseudo mountain]

Actually, yes, I am trying to train a model myself, using GAN and autoencoder.

So how can I use the ready-trained models here?

thanks
@finite galleon
@proper shale

[modern oxide]

for tts? no idea tbh

[proper shale]

^ yeah

it doesn't really support tts natively

[finite galleon]

You would need a third party tts software

[pseudo mountain]

@modern oxide yes for tts

[proper shale]

BUT you can use it with audio files

[finite galleon]

And then synthesize the tts result with rvc

[modern oxide]

usually people just synthesize text with external tts, like eleven labs or strawberrytts and then just inference the result through rvc

or styletts

whatever it's called

styletts is an open source local thingy, eleven labs is technically free site but with limitations

[finite galleon]

https://youtu.be/MckT7z7W_qM?si=Eb3TAyZgqA__cK6m

[modern oxide]

i'm sure there is a way to implement rvc models directly into tts, but i've never seen anybody do that aside from eleven labs

but they don't share their tech as far as i know

[remote verge]

1500

21 min dataset

[modern oxide]

what batch size did you put?

[remote verge]

4

[modern oxide]

zamn

hopefully your dataset is pretty clean

[remote verge]

yeh its me

[compact oyster]

RVC link?

[modern oxide]

#1159949278270193734

[remote verge]

ty, i tried 60k and it sounded good

[rare goblet]

Ayo? @remote verge level 3 !!!

[wheat peak]

I'm trying to run rvc for the first time and when I open the file it gives me a command screen telling me to press any key to continue, but when I do that I think it just crashes or it just closes itself. What do I do?

[radiant flare]

i thought applio was a bot, but ill check it out 😄 thanks 👍

[proper shale]

I'd say try the others too, even though they might be starting to overfit

[waxen agate]

uhmmm sorry to disturb but it keeps saying termination behavior can be configured in advanced profile settings and this and it wont start. does anybody know what happend here?

[gleaming roost]

when i try to launch applio it says i dont have tensorboard even though i do have it

[harsh peak]

helloooo

[proper shale]

Hello, state your issue here if you have any

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

🇺🇸 English (main)

https://docs.aihub.wtf/

​

Translation by country

🇧🇷 Brasil (PT-BR)

https://docs.aihub.wtf/v/brazil

🇪🇸 Spain

https://docs.aihub.wtf/v/spain/

🇫🇷 France

https://docs.aihub.wtf/v/france

🇩🇪 Germany

https://docs.aihub.wtf/v/germany

🇯🇵 Japan

https://docs.aihub.wtf/v/japan

🇰🇷 Korea

https://docs.aihub.wtf/v/korea

🇵🇱 Poland

https://docs.aihub.wtf/v/poland

🇷🇺 Russia

https://docs.aihub.wtf/v/russia

🇮🇹 Italy

https://docs.aihub.wtf/v/italy

[old hawk]

whenever i click refresh voicelist it just refreshes indefinitely

using local web

[copper pebble]

which one?

[old hawk]

idk how can i check

its my first time

using local

[copper pebble]

you doing inference right? can u link me that pls

[old hawk]

idk what funny words are u saying

[proud elbow]

is there some error messages in the console?

[old hawk]

nope

[copper pebble]

u doing covers?

[old hawk]

yes

i mean i want to

but cant

[copper pebble]

can you send me the link of that?

[old hawk]

bcs error

[hot lagoon]

what is
RVC-Studio ?

[copper pebble]

seemz outadet dunn9

[old hawk]

ok

[copper pebble]

did u try ilaria rvc?

[old hawk]

what is that

[copper pebble]

hf space for covers

https://discord.com/channels/1159260121998827560/1167844377046040769

heres a guide

[proud elbow]

are you using mangio 23.7.0? mine is at that version and it works fine

[copper pebble]

duplicate the space

[old hawk]

https://github.com/Mangio621/Mangio-RVC-Fork/releases

i downloaded

it from here

[hot lagoon]

How dose the Demucs work?

[old hawk]

yes

using that

version

[copper pebble]

@old hawk if ya just need to do some covers use this

[old hawk]

i tried alot of tutorials

they work but

its just

i cant continue the tutorial after adding the model in logs and weights

bcs when i refresh

it just

refresh indefinetly

its still refreshing

[proud elbow]

have you tried updating nvidia driver and rerun mangio?

[copper pebble]

it isnt the one u ursing rn its another web

[old hawk]

my drivers updated

[old hawk]

oh

is the download link in the guide

[copper pebble]

its just a website u dont need to download anything :3

[old hawk]

oh

damn

[jolly ridge]

Is there a way to fix that ?
PS : I use CPU & I use french language and I'm on laptop
image link cuz no perms :c
https://cdn.discordapp.com/attachments/1032236285554855976/1179404303560085594/image.png?ex=6579a8da&is=656733da&hm=e2b15196589736e526f1b64c0f54e0c4c6ad9e318e399fc4d9979eb747a19a5e&

[old hawk]

ill try it

i hope it works

h

[copper pebble]

duplicate the space for no queue

[old hawk]

idk what that means too

e

is there a queue

on the website

ill probably understand when i use the website

[copper pebble]

https://huggingface.co/spaces/TheStinger/Ilaria_RVC?duplicate=true just use this link thats gonna make a copy of that website

but its just your private one

[old hawk]

oh

[rare goblet]

Ayo? @old hawk level 3 !!!

[proud elbow]

my driver is still 537.58 and it's known stable on me, the latest is probably not always stable tho

[old hawk]

o

[copper pebble]

if u just need to do covers use that pretty good

[old hawk]

can i also

make models

on that

[copper pebble]

no

:((

but there are some online alternative

[proud elbow]

I'm using original RVC for training

[copper pebble]

like the easygui colab or the disconnected one

@old hawk if ya want to train u can find some online alternatives here #1159949278270193734 #1159513888199540817

[cerulean hedge]

how to use other people model in mangio rvc fork?

[old hawk]

ok

thx

is there a way to

dark mode

my eyes are burning

lol

[copper pebble]

ahhaha

yh just go to setting and you should find it

[old hawk]

nvm

found it

yes

thx

[jolly ridge]

gl

[old hawk]

ty

[cerulean hedge]

Anyone?

[copper pebble]

can u link me what u using

[cerulean hedge]

https://github.com/Mangio621/Mangio-RVC-Fork/releases

[copper pebble]

just need to do covers right?

[cerulean hedge]

yeah

[rare goblet]

Ayo? @cerulean hedge level 1 !!!

[jolly ridge]

Am the only who i can't convert with RVC WebUI rvc model (pth + index) into a onnx file ?

[rare goblet]

Ayo? @jolly ridge level 1 !!!

[jolly ridge]

yay

[copper pebble]

. @cerulean hedge just use this

[copper pebble]

https://discord.com/channels/1159260121998827560/1167844377046040769

[cerulean hedge]

ty

[hot lagoon]

what is
RVC-Studio ?

[jolly ridge]

if someone got a idea ?

[cerulean hedge]

Does Ilaria RVC work with both RVC V1 And RVC V2?

Ilaria

[copper pebble]

nvm i dont think so

[woven vessel]

Hello. Apology for going into other conversations. I'd like to ask shortly if the RVC crash during the conversion is a known problem. Admitting that my PC with 4 GB of RAM and HDD might not be ideal for the program.

[copper pebble]

im sorry its not enough..

im almost completely sure that its not enough

[woven vessel]

Alright then. At least it's clear what's the issue then. Thanks. : )

[cerulean hedge]

for some reason my pitch sound very weird

every time I generate voice it is not even accurate at all

[finite galleon]

How wierd?

[cerulean hedge]

The pitch is -12 btw

[finite galleon]

Is the model female and the original audio male?

[cerulean hedge]

nope

[rare goblet]

Ayo? @cerulean hedge level 2 !!!

[finite galleon]

Model male original female?

[cerulean hedge]

The original audio male and the model itself is also male

[finite galleon]

Dont do -12

Female model and male audio = 12
Male model and female audio = -12
Male/female model and male/female audio = 0

[cerulean hedge]

What pitch is recommended? The model voice is very deep

[finite galleon]

Because your trying to infer a talking audio just try each semitones

-6, -7, -8, -9, etc

If your trying to make a song cover which needs to be in the same key as the instrumental -12 is the only option so that its the same key

[cerulean hedge]

May I ask you something?

[cerulean hedge]

Do you know mangio rvc?

[finite galleon]

Whats the problem

[cerulean hedge]

how to use other people model in mangio rvc fork?

I'm really confused.

[finite galleon]

Isnt there a url model extractor on one of the tabs in the ui?

Oh wait thats applio

[proper shale]

put the pth in the weights folder
make a folder in logs with your model's name and put the index there

[cerulean hedge]

Okay thanks.

[proper shale]

np! ;)

[distant turtle]

https://github.com/IAHispano/Applio-RVC-Fork Does anyone know how to start the RVC applio program?, what should I press, it's been updated and everything seems to have changed, Thanks

[noble dawn]

Anyone got a working collab

[old hawk]

the funny

anyways its fine now the colab is good

[candid lichen]

Yes

[noble dawn]

What’s the recent one

[noble dawn]

Cuz I can’t tell if this is working I’m using this one

[candid lichen]

Google colab right?

[noble dawn]

Yes

Is that a working one ?

[candid lichen]

Wait are you training a voice?

[noble dawn]

Nah model

Wait wym voice ?

[rare goblet]

Ayo? @noble dawn level 2 !!!

[candid lichen]

Cause that one I think it's ment for voice training

[noble dawn]

Ohh

Can u help me Rq

[candid lichen]

I can try

[cerulean hedge]

Ugh my model doesn't work.

[noble dawn]

Thank u 🙏

[cerulean hedge]

It shows error.

[noble dawn]

Ima dm u If that’s cool

[candid lichen]

Ok

[noble dawn]

Just dm

[proper shale]

probably didn't select the audio correctly?

[distant turtle]

https://github.com/IAHispano/Applio-RVC-Fork Does anyone know how to start the RVC applio program?, what should I press, it's been updated and everything seems to have changed, Thanks

[cerulean hedge]

I didn't even press convert and it already said this...

Never mind, I restarted rvc and it worked.

Wtf

[rare goblet]

Ayo? @cerulean hedge level 3 !!!

[cerulean hedge]

Now I tried again and the voice don't even start.

I can't hear anything.

[old hawk]

-colab

[azure marsh]

All working colabs / spaces ☁️

• AICoverGen-NoWebUI, by Ardha Google Colab
• RVC Disconnected, Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab
• RVC-HF, by r3gf Huggingface Spaces
• AICoverGen, by r3gf Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces
• Advanced RVC Inference, by neuclya Huggingface Spaces

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, -local, -overtrain, -realtime, -rvc, -help

[old hawk]

does anyone know how can i make the AICoverGen-NoWebUI less airy and more vocalized, the models choke alot on it

like any configuration

[shy zealot]

**Method 1 🔎: ** The search accuracy is 100% 💕

KAF (KAMITSUBAKI STUDIO)

Upload the <t:1684949968>
Server uploaded: AI HUB (RIP)
Channel uploaded: AI HUB (RIP) Post
Profile Owner: taesankyu

Author

Nero

Discord User

taesankyu

Internal ID

81

Language

EN 🇺🇸

Type

RVC

Size

0 MB

Extractor

N/A

Epoch

620

Steps

0

Powered by ModelAI.net, Made with Love ❤️

[proper shale]

Maybe it's the song that's problematic?

[old hawk]

i tried 3 song, first had that walkie talkie type effect on lyrics so i also thought its the song thats the problem but when i switched to a clear song it still chokes alot of the time

3rd is converting right now

[proper shale]

Have you tried different models?

[shy zealot]

**Method 3 🔎: ** The search accuracy is 40% 🥺

Arijit Singh

Upload the <t:1684549110>
Server uploaded: AI HUB (RIP)
Channel uploaded: AI HUB (RIP) Post
Profile Owner: clubbedsam

Author

clubbedsam

Discord User

clubbedsam

Internal ID

61

Language

EN 🇺🇸

Type

RVC V2

Size

0 MB

Extractor

RVMPE

Epoch

450

Steps

0

Powered by ModelAI.net, Made with Love ❤️

**Method 1 🔎: ** The search accuracy is 100% 💕

Madara Uchiha

Upload the <t:1698458165>
Server uploaded: AI Hispano
Channel uploaded: AI Hispano Post
Profile Owner: ligth0723

Author

Ligth

Discord User

ligth0723

Internal ID

19533

Language

Unknown 🇺🇳

Type

RVC V2

Size

0 MB

Extractor

Rmvpe

Epoch

500

Steps

0

Powered by ModelAI.net, Made with Love ❤️

[cerulean hedge]

"in audio_to_temp_file
temp_dir = Path(dir) / self.hash_bytes(data.tobytes())
AttributeError: 'NoneType' object has no attribute 'tobytes"

Any fix?

Context: I'm using mangio rvc fork

Wtf

I tried it and it taking forever

[proper shale]

I think you selected the wrong path

[shy zealot]

I'm sorry, I've tried everything 😭
It seems that I can't find any model with your search, please try another query.

[hot lagoon]

How to fix?

E:\Mangio-RVC-v23.7.0_INFER\Mangio-RVC-v23.7.0>runtime\python.exe infer-web.py --pycmd runtime\python.exe --port 7897
Traceback (most recent call last):
File "E:\Mangio-RVC-v23.7.0_INFER\Mangio-RVC-v23.7.0\infer-web.py", line 79, in <module>
conn = sqlite3.connect('TEMP/db:cachedb?mode=memory&cache=shared', check_same_thread=False)
sqlite3.OperationalError: unable to open database file

[shy zealot]

**Method 3 🔎: ** The search accuracy is 40% 🥺

Minigame Voice [FNaF 2/SAVETHEM]

Upload the <t:1687897980>
Server uploaded: AI Hispano
Channel uploaded: AI Hispano Post
Profile Owner: thekelpek

Author

Kelpek

Discord User

thekelpek

Internal ID

580

Language

ES 🇪🇸

Type

RVC V2

Size

0 MB

Extractor

Harvest

Epoch

300

Steps

0

Powered by ModelAI.net, Made with Love ❤️

**Method 1 🔎: ** The search accuracy is 100% 💕

Atif Aslam (Pakistani Singer)

Upload the <t:1690733588>
Server uploaded: AI HUB (RIP)
Channel uploaded: AI HUB (RIP) Post
Profile Owner: clubbedsam

Author

clubbedsam

Discord User

clubbedsam

Internal ID

2537

Language

EN 🇺🇸

Type

RVC V2

Size

84 MB

Extractor

null

Epoch

280

Steps

0

Powered by ModelAI.net, Made with Love ❤️

[hot lagoon]

Need help

[brittle wing]

-colab

[azure marsh]

All working colabs / spaces ☁️

• AICoverGen-NoWebUI, by Ardha Google Colab
• RVC Disconnected, Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab
• RVC-HF, by r3gf Huggingface Spaces
• AICoverGen, by r3gf Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces
• Advanced RVC Inference, by neuclya Huggingface Spaces

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, -local, -overtrain, -realtime, -rvc, -help

[shrewd axle]

guys, how to found models?

[copper pebble]

Information for new members!

Voice Changer help #🔍│help-w-okada
Cover & training help #✨│ai-help
Searching for models #1175430844685484042/#🔍│find-models
web site:http://weights.gg/

[frank kraken]

https://cdn.discordapp.com/attachments/1179329672891355199/1179451305421455532/Screenshot_20231129-212759_Chrome.jpg?ex=6579d4a0&is=65675fa0&hm=f2f5953d70fc6043d1121a2fb52d76cd5dfae676715179403134cddbfb2dc525&

[rare goblet]

Ayo? @frank kraken level 1 !!!

[half cove]

the amount of botspam ive just seen

[frank kraken]

Can anyone help me pls

[copper pebble]

yo

are doing covers??

cuz i cant recognize whatcu using

[frank kraken]

Collab

[copper pebble]

yh i meant which one

[copper pebble]

??

[frank kraken]

Yes

[copper pebble]

https://discord.com/channels/1159260121998827560/1167844377046040769 ilaria rvc (huggin face space)

https://discord.com/channels/1159260121998827560/1175615544037212200 AiCoverGenNOwebUI (colab)

[frank kraken]

Which acc to u is better

Hugging face or collab

@copper pebble

[copper pebble]

well everyone say that the colab its better so

i never used it tbh always used the 1st one

[hot lagoon]

Can anyone help

[latent kettle]

Can anyone explain on which principals RVC works?

[remote verge]

Should I do 32 Hop length or not worth it?

[delicate wharf]

uhmmmmm, I'm using rejekts' easygui on colab and I'm done training but there's no weights folder but instead two D_ and G_ .pth file, how to make it into a zip for storing model?

[rare goblet]

Ayo? @delicate wharf level 1 !!!

[tranquil scaffold]

Hey can someone help me with setting up the voice changer dm me

[full turret]

-colab

[azure marsh]

All working colabs / spaces ☁️

• AICoverGen-NoWebUI, by Ardha Google Colab
• RVC Disconnected, Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab
• RVC-HF, by r3gf Huggingface Spaces
• AICoverGen, by r3gf Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces
• Advanced RVC Inference, by neuclya Huggingface Spaces

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, -local, -overtrain, -realtime, -rvc, -help

[hot lagoon]

?

[wheat adder]

can anyone help me with applio rvc issues?

[proper shale]

what's up?

[molten pecan]

Beat me to it

[hot lagoon]

How to fix?

E:\Mangio-RVC-v23.7.0_INFER\Mangio-RVC-v23.7.0>runtime\python.exe infer-web.py --pycmd runtime\python.exe --port 7897
Traceback (most recent call last):
File "E:\Mangio-RVC-v23.7.0_INFER\Mangio-RVC-v23.7.0\infer-web.py", line 79, in <module>
conn = sqlite3.connect('TEMP/db:cachedb?mode=memory&cache=shared', check_same_thread=False)
sqlite3.OperationalError: unable to open database file

[wheat adder]

Traceback (most recent call last):
File "D:\Philipp Diesler\Documents\Applio-RVC-Fork\infer-web.py", line 16, in <module>
import lib.tools.model_fetcher as model_fetcher
File "D:\Philipp Diesler\Documents\Applio-RVC-Fork\lib\tools\model_fetcher.py", line 3, in <module>
from tqdm import tqdm
ModuleNotFoundError: No module named 'tqdm'

Applio

I always get module not found errors

[proper shale]

did you install all the modules?

it's in requirements.txt

[wheat adder]

I ran the applio installation

So yes

[proper shale]

...wait

does your Windows username have a space on it

[wheat adder]

yes

[rare goblet]

Ayo? @wheat adder level 1 !!!

[proper shale]

that

that is probably the problem

[sand iris]

https://cdn.discordapp.com/emojis/790214199544512573.webp?size=48&name=kek&quality=lossless

[proper shale]

😭

[hot lagoon]

💀

[brittle wing]

forgot how to fix

[sand iris]

Sorry to do it to you, but probably don't use Mangio nowadays lol

[brittle wing]

related to .index file right?

😭

[proper shale]

something wasn't selected properly...

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

🇺🇸 English (main)

https://docs.aihub.wtf/

​

Translation by country

🇧🇷 Brasil (PT-BR)

https://docs.aihub.wtf/v/brazil

🇪🇸 Spain

https://docs.aihub.wtf/v/spain/

🇫🇷 France

https://docs.aihub.wtf/v/france

🇩🇪 Germany

https://docs.aihub.wtf/v/germany

🇯🇵 Japan

https://docs.aihub.wtf/v/japan

🇰🇷 Korea

https://docs.aihub.wtf/v/korea

🇵🇱 Poland

https://docs.aihub.wtf/v/poland

🇷🇺 Russia

https://docs.aihub.wtf/v/russia

🇮🇹 Italy

https://docs.aihub.wtf/v/italy

[brittle wing]

Same here

[trim brook]

UVR5 users, should I tick GPU Conversion if I have a strong GPU but I'm doing other things in the background?

@molten pecan maybe you can enlighten me, my friend? 👀

[molten pecan]

I'm not your friend but sure

Turn on GPU conversion

that will surely make it faster

[trim brook]

Thanks!

[boreal grotto]

Having a bunch of issues in cutting off

[brittle wing]

hey how do i make a rvc model using like 5000 of 1 second files

[vague sapphire]

does rvc v2 disconnected auto remove background noise or do i have to do it myuuself?

also for the datasets thats used do i have to resize them?

[small wolf]

Using Applio with an RVCv2 model:
https://www.weights.gg/models/cloam7n1p00v2wsj78gpy3b2h

and the output keeps on sounding more like my voice than Cave Johnson and nothing like the samples, is there anything I can do to improve this?

So far I've tried:

• changing pitch settings
• using tts
• adjusting advanced parameters

[rare goblet]

Ayo? @small wolf level 1 !!!

[small wolf]

does Applio just not work with v2 models or smth?

[inland osprey]

when im trying to launch applio it just says
Starting Applio-RVC-Form with Conda default compatible with all platforms...
Python Press any key to continue . . .
Anyone knows why?

Pressing any key results in a "Couldnt find the directory" error

althrough i have conda

[silver egret]

Mine says No module named 'torch'

[elfin bay]

Keeps loading, has been happening with multiple ways of ai covering

Don't know what's causing it

[sleek island]

does anyone got saul goodman's voice model?

[rare goblet]

Ayo? @sleek island level 5 !!!

[elfin bay]

Just keeps going

[green shell]

bro you are making an ai song off of the damn web

[rare goblet]

Ayo? @green shell level 1 !!!

[green shell]

do you have a brain bro

that 0.9 is just a visual glitch

that has no representation of how long it will take you

it could take up to 30mins

cause your using a WEBSITE

[tame mica]

could be worse, just try waiting

[elfin bay]

Bro no need to be a bitch, it used to be way faster the way I'm doing it right now, that's the reason I'm asking

[rare goblet]

Ayo? @elfin bay level 2 !!!

[green shell]

de u know if rvc gui by tiger14n is legit and safe to use?

https://github.com/Tiger14n/RVC-GUI/releases/

[rare goblet]

Ayo? @pallid jacinth level 1 !!!

[trim brook]

Any idea what I'm doing wrong? https://prnt.sc/3RpwUjy8Udb2 I think it worked the first time I did it

[rare goblet]

Ayo? @trim brook level 2 !!!

[pallid jacinth]

its not loading man

[elfin bay]

It timed out💀

[tame mica]

LOL

[pallid jacinth]

the voice ai shit isnt loading

[trim brook]

The model?

[pallid jacinth]

no the whole menu

[trim brook]

Yeah idk I'll probably restart the whole thing and see if it works

Cuz it worked the first time around

[pallid jacinth]

yo

the voice changer aint working

no sound coming out

[hazy magnet]

how to stop training the voice model on collab rvc disconnected

bcs i think i made too much epochs and it might overtrain

i've set 300 epochs for a 6 minute audio clip and i think thats too much

[proper shale]

just click on the stop button and export the model

oh and train index

[proper shale]

probably not

it really varies

[dawn bobcat]

does anyone know why my trained models in Applio aren't showing up? I have them in the logs/weights folder.

[molten pecan]

It may be in the Results folder

[dawn bobcat]

lemme check

[full turret]

-colab

[azure marsh]

All working colabs / spaces ☁️

• AICoverGen-NoWebUI, by Ardha Google Colab
• RVC Disconnected, Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab
• RVC-HF, by r3gf Huggingface Spaces
• AICoverGen, by r3gf Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces
• Advanced RVC Inference, by neuclya Huggingface Spaces

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, -local, -overtrain, -realtime, -rvc, -help

[rare goblet]

Ayo? @boreal grotto level 1 !!!

[dawn bobcat]

I can't seem to find a Results folder, is it in the Applio-RVC-Fork directory?

[rare goblet]

Ayo? @dawn bobcat level 1 !!!

[hazy magnet]

do i need to download ALL .pth files from the google drive folder or only the one with my model's name to export my model?

[rare goblet]

Ayo? @hazy magnet level 1 !!!

[green shell]

do u know if rvc gui by tiger14n is legit and safe to use?

https://github.com/Tiger14n/RVC-GUI/releases/

can anyone check please?

[copper pebble]

just the one with ur model name

@hazy magnet u trained index right?

[hazy magnet]

i think so

[copper pebble]

bro thats pretty outdated

[hazy magnet]

im just a begginer so idk

but i've been following the tutorial

[copper pebble]

well it was the last cell

[copper pebble]

perfect

[hazy magnet]

so yeah, i did it

[copper pebble]

just the added.index its needed

and ofc the pth with ur model name

[hazy magnet]

okay i did it

i just need to upload it to huggingface

[copper pebble]

-upload

[azure marsh]

huggingface.co upload Guides

English guide

Link: https://rentry.org/fdg_guide_new
Credits: FDG

Spanish Guide

Link: https://rentry.org/AI_Hub_guide_esp
Credits: Julia (ailen2091)

Hindi guide

Link: https://rentry.org/AI_hub_guide_hi
Credits: Enes

Turkish guide

Link: https://rentry.org/AI_hub_guide_tr
Credits: Enes

[copper pebble]

dunno if u need it but theres a simple guide :3

[hazy magnet]

thanks!

[copper pebble]

@green shell https://discord.com/channels/1159260121998827560/1167844377046040769 ilaria rvc (huggin face space)

https://discord.com/channels/1159260121998827560/1175615544037212200 AiCoverGenNOwebUI (colab)

if u need to do inference (covers just use on of these two) the RVC GUI works but its outdated

[green shell]

what.. im not trying to make a cover im wondering if it's safe

ohh i see

[copper pebble]

well its safe

[green shell]

but those dont use gui i dont know how to use it without gui?

[copper pebble]

but its made for covers eheh

[green shell]

are the other rvc easy to use or does it require computer knowledge like command prompt

[copper pebble]

well a bit but there are many guides.. but 1st watchu trying to do?

[green shell]

just make an ai cover of some song

thats it

i was hoping to download something quick, safe, secure and high quality

and easy to understand

[copper pebble]

well you can do ai covers with these

u dont need to download anything

and there really easy guides on them

[green shell]

they are extremely slow using the web public servers

like were talking 20 mins

[copper pebble]

mhh no

maybe ilaria

but not aicovergen

it takes like a max of 10 minutes

[green shell]

which one is ai covergen?

[copper pebble]

ilaria takes like 300s per 2 min audios

[green shell]

oh i see

[copper pebble]

https://discord.com/channels/1159260121998827560/1175615544037212200 AiCoverGenNOwebUI (colab)

[copper pebble]

even tho it varies

it can also take just 150s or even 350s sometimes

also depends on your internet connection

@green shell tbh they are both good

[hazy magnet]

erm what does extraction method mean?

[copper pebble]

Rmvpe/ Mangio-Crepe ecc.

[native bobcat]

can someone take a look at the logs of my loss for the model im training and let me know if continuing training would make sense

[rare goblet]

Ayo? @native bobcat level 2 !!!

[green shell]

alr ima try it

ty

[proper shale]

they're pitch extraction methods

[proper shale]

send the graphs

[copper pebble]

was writing it but thank u mj :3

[native bobcat]

i dont have a graph

[hazy magnet]

okayyy....

[native bobcat]

im using the default rvc github locally

[proper shale]

oh u gotta launch tensorboard then i suppose

check the pins, there's the tensorboard install .bats

and launcher

[green shell]

is clone and install a 1 time thing for aicovergen

[rare goblet]

Ayo? @green shell level 2 !!!

[native bobcat]

will i be able to use it even though its already been trained or will that just be from now onward

[proper shale]

yea i think u can

[proper shale]

locally, yes

[hazy magnet]

im sorry but how do you know which one was used in the training

[green shell]

like on the collab?

[copper pebble]

what did u use for training?

[hazy magnet]

rvc v2 disconnected

[rare goblet]

Ayo? @hazy magnet level 2 !!!

[proper shale]

nope, it's per session

only locally it's once

[copper pebble]

well if you sure u didnt touch anything its RMVPE

[native bobcat]

the ROD download link in the pin is a dead github page

https://github.com/grvyscale/AutomatedOvertrainingDetection/releases

[hazy magnet]

okay, thanks<3

[proper shale]

na i have pins for 2 bat files

#✨│ai-help message

[native bobcat]

i se

[copper pebble]

btw its late here so im gonna go to bed bye yall

[green shell]

what part of the google collab end part do i say whos voice i want

there is no part that asks

[native bobcat]

is this le graph you are looking for

[proper shale]

yeah, but in the scalars tab, and with smoothing set to 0.95 or something

[native bobcat]

heres 92

[proper shale]

hmmm this doesn't tell the complete story but i think u should keep going

[native bobcat]

its just hit 350 epochs on the 12 minute dataset which im gonna just test the different models it made every 50 epochs and then if its bad just add more data

[proper shale]

fair enough

[green shell]

what do i do it kind of just stopped here

[proper shale]

well it did it

now just go on the colab file explorer and find it

[green shell]

where do i go to see it

o ok

yo wtf this shit is sick

you programming mfs sick as fuck

nah thats crazy lit

[proper shale]

lol im not one of em

u gotta thank the engineers for that

[green shell]

i dont think ill ever understand this bro

ima just download the rvc gui cause it looks easy to use

why not try it out

[violet heron]

Outdated

It doesn’t have RVMPE so all the covers you make will sound bad or really raspy

Or both

[green shell]

are we talking like really bad?

cause everything else is sooo hard to understand!!

[violet heron]

I would use Ilaria RVC or applio then

[proper shale]

the one by tiger is very outdated

[green shell]

what

[green shell]

(for myself)

[violet heron]

Applio is a fork of Mangio RVC made to easy to use

good local method

[green shell]

ok where do i find?

https://github.com/IAHispano/Applio-RVC-Fork?

?

[violet heron]

Yep

[green shell]

kk

[violet heron]

https://github.com/IAHispano/Applio-RVC-Fork/releases/tag/v2.0.0

[native bobcat]

damn brah this is the best model ive made

and yeah its not overtrained

[proper shale]

yooo nice

[violet heron]

Did you duplicate the space

[violet heron]

Works but it’s outdated

[violet heron]

Most people don’t have access to computers, and another thing is HuggingFace runs off servers and not the users device

[violet heron]

#🔍│help-w-okada

[elfin bay]

I used computer too same problem, also I don't think so how do u do that

[violet heron]

3 dots then click duplicate this space

You need a HuggingFace account

[elfin bay]

Ah alr thanks

[hazy magnet]

Thank you for your help ❤️ i finished the model!

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

🇺🇸 English (main)

https://docs.aihub.wtf/

​

Translation by country

🇧🇷 Brasil (PT-BR)

https://docs.aihub.wtf/v/brazil

🇪🇸 Spain

https://docs.aihub.wtf/v/spain/

🇫🇷 France

https://docs.aihub.wtf/v/france

🇩🇪 Germany

https://docs.aihub.wtf/v/germany

🇯🇵 Japan

https://docs.aihub.wtf/v/japan

🇰🇷 Korea

https://docs.aihub.wtf/v/korea

🇵🇱 Poland

https://docs.aihub.wtf/v/poland

🇷🇺 Russia

https://docs.aihub.wtf/v/russia

🇮🇹 Italy

https://docs.aihub.wtf/v/italy

[willow raft]

i installed the applio rvc fork but i don't know how to open the client, do i have to open python and type a command?

[zenith hazel]

getting "index -1 is out of bounds for axis 0 with size 0" when its 50% through generating an output

[remote verge]

Also I do want to confirm does a lower batch size = higher quality? At first I thought this was the case but searching through messages theres conflicting info. Some say its based on how much vram I have. If it is based on Vram could someone recommend me a batch size for 24gb vram?

Because I've just been using batch size 4 lol

[hidden laurel]

how do i make ai covers bruh

there is nothing good

[green shell]

why does it say this

[hidden laurel]

bat files always get fake marks

[green shell]

i want to download these things but ngl its kinda sketchy cause its source code

i wish reddit had more "trusted" posts i guess

like some sort of verification

[violet heron]

It’s open source don’t worry

[green shell]

that feels like something to be more worried about

[rare goblet]

Ayo? @green shell level 3 !!!

[green shell]

has someone actually went through the code before?

[violet heron]

Yeah…

That’s why a bunch of forks and online ports exist

[green shell]

oh ok

[formal void]

what do i do after this?

[willow raft]

i keep running into errors both trying to run run rvc via collab and locally,

when i try to run the trainer locally it gets stuck in the step2a processing data forever and doesnt progress, the bat window doesn't say any errors or anything

when i try to run it via collab it keeps instantly deleting the zip file, and the code says
"Dataset folder already found. Wiping clean for import operation...
Error in verifying dataset: There is no spliced.zip in /content/drive/MyDrive/dataset"

[strong belfry]

the common factor being your dataset?

can you ensure its in the proper format locally?

I cant comment on colab

[willow raft]

honestly i have no idea

[strong belfry]

But locally, processing a 30m dataset takes a bit

how long did you wait

and was there any system usage during that time?

[willow raft]

it's around 15 minutes of data, last wait it went to10 minutes

[willow raft]

i was using the computer for other stuff, yeah

[strong belfry]

as in was there any system usage from the local trainer to indicate it was still working on processing?

drive usage/cpu usage

[willow raft]

oh, i didn't check that

[rare goblet]

Ayo? @willow raft level 1 !!!

[strong belfry]

I would check, but if you have a decent cpu and drive it should only take like 5 minutes to preprocess and maybe 15 to extract features

[willow raft]

i have a ryzen 7 5800x and a 3070ti, that's why it going for over 10 minutes without any progress seemed weird

[strong belfry]

Yeah seems odd

did you by any chance click on the cmd window and accidently pause the program?

[willow raft]

i don't think so

i'm gonna try doing a local training again

in this step, do i keep the quotation marks from the Copy as path copy? and by any chance, is my computer's name having a space in it a problem for the path file?

[strong belfry]

I dont think you need it, But seems like you are using a old fork that had known problems with spaces in dataset location

[willow raft]

uh oh

[strong belfry]

well its not old

just i dont remember mangio ever fixing that problem

just put the dataset somewhere else

and hope

[willow raft]

i'll try, thank you 🙏

it's still not doing anything

550 seconds have passed and nothing moves/changes

[hidden laurel]

maybe u got a shit butt ass gpu

[willow raft]

no, when i use rvc gui to use other people's models and when i use ultimate vocal remover they all take under 30 seconds to do the thing

i'm having a user error somewhere but i have no idea what i'm doing wrong

[oblique zephyr]

silly question but where and how do i make AI song covers from AI Hub?

[eager vine]

What is better? pm, harvest, dio, crepe, mangio-crepe, rmvpe or rmvpe_gpu??

[violet heron]

RVMPE

RVMPE_GPU utilizes your GPU more

[steady mantle]

yo can my pc run this?

my specs are :
amd ryzen 5 5500
rtx 3060 8gb
16gb ram

[willow raft]

it should be able to, yeah

[rare goblet]

Ayo? @willow raft level 2 !!!

[steady mantle]

bet

[stray lantern]

may i still use google colab?

[strong belfry]

rvc disconnected i think works

[molten pecan]

You need to log in

Wait are you using hugging face?

[molten pecan]

Yeah of course

[stray lantern]

how do i use that?

[rare goblet]

Ayo? @stray lantern level 1 !!!

[molten pecan]

You don't need the server's permission

[stray lantern]

i heard its bannable or something?

idk i wanna stay safe

[molten pecan]

There's a guide on #1159513888199540817

[molten pecan]

Use a burner acc

[willow raft]

no idea where I'd log in to, but I fixed it by moving the folder locations and also renaming the folders

[molten pecan]

Ah I thought you were using HF

[brittle wing]

help

[strong belfry]

with?

[hollow vigil]

should my gpu be at 99% on a 4060 ti making models with 3 batch size

[native bobcat]

yes

training uses your gpu

[hollow vigil]

but a 4060 ti? at 99% with only a batch size of 3

like my 3060 was at a batch size at 8 and it wasnt even at 99%

and a 4060 ti is 50% faster

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

🇺🇸 English (main)

https://docs.aihub.wtf/

​

Translation by country

🇧🇷 Brasil (PT-BR)

https://docs.aihub.wtf/v/brazil

🇪🇸 Spain

https://docs.aihub.wtf/v/spain/

🇫🇷 France

https://docs.aihub.wtf/v/france

🇩🇪 Germany

https://docs.aihub.wtf/v/germany

🇯🇵 Japan

https://docs.aihub.wtf/v/japan

🇰🇷 Korea

https://docs.aihub.wtf/v/korea

🇵🇱 Poland

https://docs.aihub.wtf/v/poland

🇷🇺 Russia

https://docs.aihub.wtf/v/russia

🇮🇹 Italy

https://docs.aihub.wtf/v/italy

[median socket]

Yes, you could also even make models although it would be quite time consuming due to 8gb vram.

Inference doesn't require too much VRAM although it will certainly cause performance issues in graphically demanding games

[proud elbow]

which 4060 ti variant? 8 gb can go up to batch size 6, and 16 gb can go 16

[hollow vigil]

8gb

[rare goblet]

Ayo? @hollow vigil level 2 !!!

[hollow vigil]

but if i bring it to batch size 6 wouldn’t it go over 8gb

bc it’s like 7.5 at 4

[proud elbow]

it can also depend which RVC you are using, the original one uses less VRAM than Applio

[stray lantern]

could someone please assist me on a model? im trying to download it but its a weights model

[glad zealot]

Wdym

From weights.gg?

[stray lantern]

yes

i am trying to put it in google drive but im not sure how to

[glad zealot]

You already downloaded it right?

[stray lantern]

yes

i downloaded the zip file

[glad zealot]

What are you using it for? Inference/cover or voice changer?

[stray lantern]

cover

[glad zealot]

Which covergen or mangio/applio/mainline

[stray lantern]

hm? i dont sem to understand your question

[rare goblet]

Ayo? @stray lantern level 2 !!!

[glad zealot]

Which program? The ai covergen? Or the other 3?

[stray lantern]

im using huggingface

i already have everything set up the vocals and stuff

i downloaded the zip file

[copper pebble]

u just need the download link of the model

[stray lantern]

how do i get that?

[copper pebble]

where did u find it?

the model ofc

[stray lantern]

https://www.weights.gg/models/clohwa21l004aws7gk8gjeemo

[copper pebble]

just a momnt

[stray lantern]

ok, thanks!

[twin quiver]

bro all these are viruses

[copper pebble]

https://huggingface.co/C0ttontheBunny/Springbonnie/resolve/main/Springbonniefnafmovie.zip

[copper pebble]

oh my god not again 😭😭

[twin quiver]

what?

[stray lantern]

is that the link?

[copper pebble]

you dont how many peoples say shits like that

[copper pebble]

yh its the download link of ur model

[twin quiver]

bc its true prob

[stray lantern]

its sats its potensially dangerous but its safe to download yes?

[copper pebble]

ive been using ai from this server for almost 3 month
(still alive) @twin quiver

[copper pebble]

yh thats safe dw

[stray lantern]

thanks!

alright, i downloaded it, now what do i do?

[twin quiver]

well most are

bc i checked like 6 girl voice changers all were viruses

😭

[copper pebble]

u dont need to download it bro

[stray lantern]

do i just put the link on "Choose your model"

[twin quiver]

https://huggingface.co/Gigrig/GigrigRVC/resolve/main/MarineHosho_RVC_v2_e670_s87100.zip?download=true
is this even safe

[copper pebble]

no in the "download models"

[copper pebble]

mhh yh its just a model..

[stray lantern]

OHHH ok thanks

[twin quiver]

so?

[stray lantern]

do i check the index settings?

[twin quiver]

@copper pebble so?

[stray lantern]

all good?

[copper pebble]

did u duplicate the space?

[stray lantern]

no....how do i do that?

[copper pebble]

just click on the 3 dots above on the right and duplicate it

it will take some minutes

[copper pebble]

its made of a pth and a index and its safe like the others

[stray lantern]

it says "space hardware" do i change anything?

[twin quiver]

no idk how to use it

[copper pebble]

#1159513888199540817

[twin quiver]

its saying open with "reccomeneded applications"

then i

[copper pebble]

#1159949278270193734

[twin quiver]

clicked discord

idk what to do

[copper pebble]

ok ive got like 5 minutes the i have to go

[twin quiver]

ok theres no guide

[rare goblet]

Ayo? @twin quiver level 2 !!!

[twin quiver]

ok wheres okada

how do i download

girl voice changer

[small wolf]

cross-chat hell 💀

[copper pebble]

@twin quiver https://docs.aihub.wtf/voice-changer-realtime/w-okada-voice-changer-realtime

heres the guide for okada

sorry i went out of signal

[twin quiver]

@brittle wing @small wolf

[small wolf]

ok so...

[rare goblet]

Ayo? @small wolf level 3 !!!

[brittle wing]

Start the application

[small wolf]

according to the guide...

[twin quiver]

oh

vb cable right

[small wolf]

ye

it's available for mac too, just follow the links in the guide

[twin quiver]

downloading it

done

[small wolf]

that was fast

[brittle wing]

After you installed the program, just restart your pc

[brittle wing]

I like your "about me" design btw

[small wolf]

thanks :p

[brittle wing]

Yep just wait for it

[twin quiver]

then?

[brittle wing]

Np

[brittle wing]

Then restart your pc when it's done

[twin quiver]

[brittle wing]

I'm an AI cover maker so I don't know Okada but I can kinda understand what you need to do till now

[small wolf]

you might be able to help me tho :p

I've basially just been finding that whenever I try inference, on most models it just sounds like the original voice doing a really, really bad impression :/
also weights.gg for some reason sounds like, significantly better with the same model

Using applio, will upload comparison in a second

ok nvm so in the case of these vocals weights.gg AND applio both sound like trash :/

but when trying with my voice they are distinctly different (weights.gg sounds better)

(and ofc with some other people's voices it does seem better)

I mean, this is to be expected to an extent, I just wasn't really expecting it to be so bad...

[stray lantern]

help, when i try to duplicate space, it says "invalid password" how do i fix that?

[brittle wing]

I use RVC-GUI for my AI covers and I experience the same problem with human voice models, so that drove me insane and I decided to use a non-human voice model like Aatrox so it never even sounds distorted

[small wolf]

yeah but... what if I want to sound like another human...

isn't there some kind of training step where I can provide MY voice saying the same stuff in the dataset...

(yes I know how dumb of a question this is I just don't see why this ISN'T a thing)

[brittle wing]

That's a problem I also try to find a solution for

[small wolf]

:/

at this rate I might just try to get a transformer model to do voice changing idek anymore lol

[brittle wing]

I use chill vocals for the human voice models so it doesn't go crazy in energetic or high-pitched vocals, otherwise I just use Aatrox and make Aatrox covers

[small wolf]

eh, fair

[brittle wing]

The voice model quality also matters

[twin quiver]

ok my power went off bro

[brittle wing]

But Idk which voice models are high quality like 5 stars

[twin quiver]

pls tell me girl voice changers

[brittle wing]

I just try to make favorite characters sing lol

[small wolf]

fr

[twin quiver]

what do i open

[brittle wing]

Voice models? Like I said look around in #1175430844685484042 if you're done installing the voice changer program

[small wolf]

the runhttp or whatever

just run it, okada comes included with some egirl stuff iirc

[small wolf]

ngl I still don't get how RVC even works lol
like it shouldn't be able to train on ONLY the character's voice

and yet...

[small wolf]

last one ig

[plush compass]

It doesn’t matter whether you extract the index first or last, right?

[twin quiver]

Booting PHASE :main
PYTHON:3.10.9 (main, Jan 15 2023, 23:00:56) [Clang 14.0.0 (clang-1400.0.29.202)]
Activating the Voice Changer.
0%|▏ | 768k/181M [00:01<07:52, 400kB/s Booting PHASE :main | 1.17M/361M [00:01<08:21, 754kB/s]
93%|█████████████████████████████████████▉ | 167M/181M [04:33<00:18, 754kB/s]
45%|██████████████████▎ | 162M/361M [04:33<05:52, 594kB/s]
36%|██████████████▋ | 129M/361M [04:33<06:21, 637kB/s]

39%|████████████████ | 142M/361M [04:33<04:20, 881kB/s]
80%|████████████████████████████████▉ | 139M/173M [04:33<00:39, 915kB/s]
40%|████████████████▌ | 139M/345M [04:33<09:43, 370kB/s]

it says this

[pulsar oak]

sound devices are chosen correctly

[brittle wing]

The RVC tutorial videos I watched are so simple that I got hyped, even their own results that they share at the end of the video are perfect but they definitely fix their vocals with some kind of program at this point

[small wolf]

fr

[small wolf]

apparently some people are refering to the feeding the AI pairs of audio

I too would like to use this architecture random medium article

feeding the AI with only the target voice was never a sound solution imho

base just wasn't generalized enough for input voices

[brittle wing]

But they don't tell how they fix it, they act like it's totally RVC's job

[small wolf]

exactly

[small wolf]

POV: The linkedin article only talks about "advanced training methods" like BRO running train.py with a gajillion arguments is NOT advanced

[stray lantern]

can someone help me?

[rare goblet]

Ayo? @stray lantern level 3 !!!

[stray lantern]

please?

[small wolf]

what's the issue?

[twin quiver]

@brittle wing @small wolf

BRO

[brittle wing]

Keep waiting till they're all 100%

The helpers should help but I guess they aren't active rn

[twin quiver]

bro

[rare goblet]

Ayo? @twin quiver level 3 !!!

[twin quiver]

its js stuck

like that

[sand iris]

what do the logs say

[twin quiver]

NVM

NVM

[sand iris]

Oh good

[twin quiver]

its works

its works

[sand iris]

Pls tell me you didn't close it when it was downloading earlier

[twin quiver]

no

btw how do i make it girl like

@sand iris what index do i do