✨│ai-help

[half cove]

😭

[spice wharf]

idk the momo one

[half cove]

yea i think i found the right one

[spice wharf]

lol

[half cove]

for some reason its still not downloading 😭

wtf

[spice wharf]

damn

Any other rvc thing i could use that would possibly work with the model?

[half cove]

dw

it was me

@spice wharf https://huggingface.co/poiqazwsx/test123/resolve/main/momo-by-leelo.zip?download=true

try that

[spice wharf]

Aight

[rare goblet]

Ayo? @spice wharf level 2 !!!

[spice wharf]

dam

still doesnt work

wait nvm

its worked now

thanks alot

had to just press download again lol

[dull yacht]

when i look at models what does epochs mean?

[half cove]

how many times it went through the dataset just so you know more epochs doesnt mean its a better model

[dull yacht]

oh ok was just wondering

thanks

[spice wharf]

dam now i got an error while converting lol

does the red mean its too big?

[dull yacht]

what does it mean by batch size?

[spice wharf]

wait nvm my bad

[native surge]

hello their i tried to download a voice and uploading it and test it out but for some reason im still a robot...

[patent pasture]

usually equvalent to amount gb in GPU, if I am right

[dull yacht]

what does it mean by rvc v2 and v1

[rare goblet]

Ayo? @dull yacht level 1 !!!

[patent pasture]

use v2

[rare goblet]

Ayo? @patent pasture level 8 !!!

[patent pasture]

it's better

[dull yacht]

how do i choose it

[patent pasture]

Are you going to train a model? Do you use colab or on local PC?

[half cove]

use v2

[half cove]

use 8

[spice wharf]

could i get other colab links

[half cove]

-colab

[azure marsh]

☁️ Google Colabs

• 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
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab

🤗 Hugginface 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

[dull yacht]

im on local pc and im not creating one im trying to use one

[spice wharf]

thanks

[half cove]

then dont worry bout it

[dull yacht]

oh ok

[half cove]

stay on the model inference page

[dull yacht]

okie doie

[patent pasture]

@dull yacht, what is you GPU?

[dull yacht]

3070ti

[patent pasture]

okay

[river cairn]

I found a model in search-models that I would like to upload to w-okada's voice changer, but the Download button takes me to Kits AI. Is there a way to download models from Kits AI?

[rare goblet]

Ayo? @river cairn level 1 !!!

[half cove]

you cant download models from kits ai

[river cairn]

Got it, thanks!

[cyan spear]

As soon as I click start_http.bat it opens for a split second and just closes

I’ve never had this issue before

[sand iris]

#🔍│help-w-okada message

[brittle wing]

yo guys I am done making a voice model
but how can I use it in rvc v2?
like how to get the link

[cyan spear]

@sand iris

[sand iris]

you sure you did that in the same folder as on the video ?

[cyan spear]

let me go try again

does this mean it worked

[sand iris]

nah

but we got an error to work with

[cyan spear]

alr

[sand iris]

try renaming MMVCServerSIO_win_onnx...17b to something like voice-changer

[cyan spear]

in the cmd menu?

[sand iris]

the folder path is too long when that error happens, usually

[sand iris]

From file explorer

[cyan spear]

so just rename this

[sand iris]

yes

[cyan spear]

now what do i do

[sand iris]

try doing the same cmd thing

and if it works this time, run start_http by double clicking as usual

[sand iris]

Well fuck

[barren acorn]

well i cant FIND GIRL VOICE GOOD

[rare goblet]

Ayo? @barren acorn level 1 !!!

[barren acorn]

AAAAAAAAAAAA

[cyan spear]

i keep having issues with this crap

[sand iris]

As a very last effort, maybe you can try re-extracting the voice-changer zip somewhere on the C:/ drive ?

It gets a loooot of issues on anything other than C:/ usually

[cyan spear]

it worked earlier

but ill move it to c drive

not working either

do i have to delete it and reinstall it

this is like the third time

[sand iris]

Originally I thought the win32 error thing was due to long filepaths

but apparently not

I don't understand why it's happening

[cyan spear]

neither do i

it worked fine last night

but now all of a sudden its fucking up

im just gonna reinstall it

for the third time

[ebon maple]

i got mine working but none of the voices sound even close to what theyre supposed to 😭

ive tweaked the pitch everywhere but cant get any male voices to sound good

[barren acorn]

@sand iris man the voice i use sound weird

[fathom mason]

#🏙│ai-images

[sand iris]

idk

some models sound ass

[proven lily]

hello is there a way to put these filesinto streamer bot to make the custom voice work on there

[light pagoda]

idk

[spiral osprey]

I tried to make a voice model with more than 3 or 4 audios, I put in 1157 epouchs, I spent almost 7 hours doing it and he couldn't finish it. He said that there was a limit of use since it had already been 12 hours before I did it. Can someone help me?

[proper shale]

You have to wait for 24 hours

[pallid hare]

Hello there! I'm new on using RVC, but I've been trying to make a Linkin Park-inspired song, but it's very hard to make Chester scream, especially in a way that sounds good. It almost never does, and when it does, the pitch is all over the place.
Now, keep in mind that it's my first time using RVC, and I'm not good at singing at all. In fact, I believe I might've screwed up my voice after many tries of singing the same part over and over again, so I don't know what should I do. I also tried pitching my sample so the tone would sound more like Chester's but it still didn't give me any meaningful results. What should I do?

By the way, as for the models, I've tried using three of them: 1k Epoch 40k Steps, 79k steps and a HT Era one which I got from weights.gg.

So yeah, I don't know if there's a way I can trick RVC without actually having to scream or not.

[molten pecan]

JESUS 1k epochs?

Uhm how long was your dataset?

[molten pecan]

@viscid moss 1k epochs no es normal, no?

[viscid moss]

No es común pero talvez tiene mucha dataset para llegar a ese punto

[molten pecan]

Tener un dataset así de largo es necesario?

[viscid moss]

No, con 15-25 min es suficiente

[violet heron]

I know why

You can’t open the the ui since your using a free colab one

You have to use no ui otherwise google terminated your runtime

[spiral osprey]

but they told me it was 12 hours.

[spiral osprey]

scream

[pallid hare]

I tried checking the places where I've downloaded these but I couldn't find much info about it unfortunately, but I'd assume at least one of them (The second one, with over 467 MB) might've been at least an hour long

[rare goblet]

Ayo? @pallid hare level 1 !!!

[pallid hare]

I've downloaded the first two on Hugging Face, the HT one was on Weights.gg

[molten pecan]

Try using mangio-crepe algorithm

Lower its hop-length by 20 with each attempt until you find the right spot

If that doesn't work, then I don't know

[half cove]

Mangio crepe with a 1hr dataset this shit bout to take a while

[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

[proven lily]

can someone help i have ryzen 5800x when i use voice changer my cpu goes to 100% and bricks my hole pc

[molten pecan]

He was inferencing 😭

I think

[proper shale]

Well it wasn't

[half cove]

💀

[topaz kelp]

Is there a link?

[sweet igloo]

when i try to run start_http there's nothing

[violet heron]

Nothing shows up?

[sweet igloo]

the black screen pop off for like 1 sec

[violet heron]

#🔍│help-w-okada message

[glass bloom]

hey

can someone help me

[glass bloom]

hey would you be willing to ho[p in a vc

im quite confused

[glad zealot]

its mostly meant for quick ai cover making, not the best, the ai probably thinking that the voice is either background music or secondary singer, youd have to manually do it if you want a better result on some songs

[sacred jungle]

how do i get the voice changer

[violet heron]

-realtime

[azure marsh]

This interaction has expired, use the command /guides realtime if you wish to see it again.

[violet heron]

First one

[outer grove]

anyone know why my rvc just has me hearing myself and no output coming through the mic

[zealous shuttle]

since i can't send image anywhere i just gonna send here, is this true 48khz? i download it from spotify downloader

[forest sluice]

You should right-click the audio file, go to properties

[neon radish]

help 😭 ive been trying to add a new model to use for inference by adding a shareable link, but it says there's been an error

[zealous shuttle]

and then?

[forest sluice]

you upload to hf?

[zealous shuttle]

my properties was like this

which tab?

[molten pecan]

use ilaria audio analyzer

google "ilaria audio analyzer hugging face"

upload the audio and it will tell you all the info about it

[neon radish]

im not sure what hf is but i dont see it in the version im using :'DDD so i think not
i would send a ss but discord wont let me

[rare goblet]

Ayo? @neon radish level 1 !!!

[proven hill]

E-GIRL VOICE MODEL SALES!!

DM ME FOR DISCOUNT PRICES!

E-Girl, Mommy, V-Tuber, Anime girl and much more!

Wondering what models?
You find them here: https://ko-fi.com/ilariaowo/shop

INCLUDING FREE ONES!! :swaygirl:

I also do commissions if you want aspecific voice!

I accept the following payment methods:

• Ko-Fi
• PayPal
• Steam Gift Cards
• Amazon Gift Cards
• Throne

Grab yours today!!!

[forest sluice]

huggingface

-upload

[azure marsh]

huggingface.co upload Guides

English guide

Link: https://rentry.org/fdg_guide_newer
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

[forest sluice]

Oh sorry, I misremembered stuffs. The Properties thingy only show bitrate, not sample rate

ffprobe -v error -select_streams a -of default=noprint_wrappers=1:nokey=1 -show_entries stream=sample_rate input.wav If you have ffmpeg installed

Or Ila's audio analyzer as julia said

[spiral osprey]

Before, when I made one it was already more than 24 hours ago, and then I made a model and it couldn't finish after 12 hours. I made another one and it stopped again, so I didn't understand. Now I know that you have to wait a little more than 12, like 24 hours. But I had already waited before when I did the first one.

[karmic heron]

i got a dataset with 34 seconds, how many epochs i put?

(not a singer)

[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

[karmic heron]

im saying how many epochs not how to make a ai cover

[wintry palm]

-colab

[azure marsh]

☁️ Google Colabs

• 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
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab

🤗 Hugginface 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

[muted sonnet]

why isn't my gpu showing i downloaded the cuda

[arctic shoal]

how do i change the epochs and the steps

[muted sonnet]

gpu(dml) doesn't have the drop down

[vocal cipher]

what are the python requirements

[violet heron]

You don’t

Unless your training a model

Then you do “total epochs”

[brittle wing]

If you use the official rvc, it comes with everything needed, if you use mangio it also comes with everything needed, if you use applio no idea never touched it, never will, but it should tell you on its page

[hybrid spire]

Quick qs, if I've already separated vocals & cleaned up audio, but it's got like bird noises in the background, what can I do from there?

[vocal cipher]

i can get rvc models to speak but not sing. pls help

[rare goblet]

Ayo? @vocal cipher level 1 !!!

[forest sluice]

Try Audacity

https://www.youtube.com/watch?v=Eyqgu4N9rho

This demo has bird sounds as an example

[autumn hill]

https://cdn.discordapp.com/emojis/1097657388863082586.webp?size=48&name=onk_ai_wtf&quality=lossless

what is it blabbing about

[autumn hill]

https://cdn.discordapp.com/emojis/1029669479615045642.webp?size=48&name=4ytra&quality=lossless

[paper nymph]

is it possible to use rvc as local on mac

[autumn hill]

oh forgot it doesn't show my emotes lol

[sand iris]

What's your GPU?

[sand iris]

Yes

[paper nymph]

How

[rare goblet]

Ayo? @paper nymph level 1 !!!

[sand iris]

So it's a little more complicated

But essentially

https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/main/docs/en/README.en.md

There's a run.sh script

[autumn hill]

[paper nymph]

Only this?

Using M1 btw

[sand iris]

Yeah. Should work most of the time, but the script can be buggy

[sand iris]

What do the cmd logs say

like fully

[autumn hill]

oh i deleted it to see if somehow it was an installation issue but it was basically this

[sand iris]

Might need a https://developer.nvidia.com/cuda-downloads

[autumn hill]

man.. ok

[woven furnace]

well i'm pretty sure i messed up because the other voices to use are voices i never put into it

😭

[paper nymph]

oh it looks simpler than ı expected

but ıdk anything about terminal

Can you explain it as if you were explaining it to a 5 year old child?

[old oracle]

-colab

[azure marsh]

☁️ Google Colabs

• 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
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab

🤗 Hugginface 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

[sand iris]

sh is the program that interprets shell commands

Here you are asking it to interpret their install script

the install script installs python, sets up a folder that stores all the python external software it needs, installs them, and then runs RVC

[patent pasture]

OK, thank you

[paper nymph]

ok ı get it thanks

where should I do extract this file

in which directory

[sand iris]

rvc beta.7z contains the windows executable

The instructions are kinda flawed

you can "download code" on the GitHub page

[storm ermine]

So I'm using the RVC WebUI, how do I import my model in

[rare goblet]

Ayo? @storm ermine level 4 !!!

[sand iris]

What does it display right above ?

[sand iris]

did you unzip everything?

oh wait

you're supposed to execute run.sh from the RVC folder

you're executing it from one folder above

[paper nymph]

yeah

[rare goblet]

Ayo? @paper nymph level 2 !!!

[paper nymph]

so I'm executing file from the original folder

on desktop

but then the RVC folder will run under desktop?

isn't desktop a folder

[sand iris]

Yeah, but executing it from desktop leads to this

You see the Voice-Conversion-WebUI-main folders? you gotta run run.sh from there

[paper nymph]

bruh

[dull yacht]

which is better out of the section where it says crepe crepe_full etc which is best out of all of them

[paper nymph]

ı think ı m too uninformed to do this

[sand iris]

wait a sec

you can run cd Desktop/Retrieval-based[...] (press tab to complete the folder name), and then ./run.sh

[forest sluice]

cd to your rvc folder

Man, should open another PR to fix this

[sand iris]

This script is balls

just breaks if python 3.11 or 3.12 is installed

[paper nymph]

It's working now I think

[sand iris]

Great

[paper nymph]

downloading something

[sand iris]

Downloading the base models to run RVC

[paper nymph]

wow it installed tons of site-packages

[forest sluice]

You know what's even more balls? https://github.com/IAHispano/Applio-RVC-Fork/blob/main/go-applio-manager.sh#L40 blud's confused between cls from cmd/powershell to clear

[paper nymph]

still waiting for it

[sand iris]

As expected

[forest sluice]

Do you have homebrew installed btw?

https://brew.sh

[paper nymph]

yeah

[sand iris]

You might have a .venv folder right in your user's folder

Because of the wrong folder earlier

might want to delete that to claim back space

[paper nymph]

no I don’t have

[forest sluice]

Still, from the error, you should install aria2

[paper nymph]

brew install aria2 this one right?

[forest sluice]

yeah

[sand iris]

Not sure why they're using aria2 and not curl or wget

[paper nymph]

okay something different

ı think this is what it supposed to do

[forest sluice]

faster dl, ig

Holy cow, the script gave me a heart attack https://github.com/rvc-project/retrieval-based-voice-conversion-webui/blob/main/tools/dlmodels.sh

[paper nymph]

how much time does it will take and storage space

[forest sluice]

Prob ~3GB

[sand iris]

(btw this was actually in the applio codebase )

[paper nymph]

what now?

[velvet crown]

how come when i do inference in hugging face it sounds good but when i do locally it mumbles all the song?

[paper nymph]

what should I do

[sand iris]

Plan B : source .venv/bin/activate && pip install pyworld

[paper nymph]

downloaded

next?

[sand iris]

Give it another shot with run.sh

[paper nymph]

bruh

[rare goblet]

Ayo? @paper nymph level 3 !!!

[sand iris]

Wait what does the shell/terminal look like right now ?

[sand iris]

does source .venv/bin/activate not add a (.venv) at the start of the terminal line ?

[paper nymph]

nothing changed

[sand iris]

pip install pyworld yet again

gotta make sure it's doing that in the right place

[worn stag]

what is best vocal remover ?

[paper nymph]

did it and still same

[forest sluice]

uvr5

[sand iris]

What's it outputting?

[worn stag]

ty

[forest sluice]

pip show pyworld?

[paper nymph]

I just downloaded it

where did it goes

[sand iris]

What's the output of pip install pyworld in that same terminal? Maybe it'll give us a clue

[forest sluice]

OK, maybe .venv/bin/python3 -m pip install pyworld?

[sand iris]

scripts dir is windows exclusive IIRC

[forest sluice]

then maybe look for python3 somewhere in .venv, then replace it with the path found

[sand iris]

should be .venv/bin/python

[paper nymph]

it shouldn’t be too hard to setup something

[earnest jasper]

what target sample rate shall i set for this dataset?

https://prnt.sc/OAvdXwknFP5P

also its around 4 mins of audio, how much epochs shall i set? 🤔

[earnest jasper]

didn't have the image perms that's why upload the image, its just the spectrogram graph

@modern oxide sorry for the ping but can ya help 🥹

[modern oxide]

sup

[modern oxide]

is it only 6k???

[earnest jasper]

yup : (

[modern oxide]

how, whose voice are you making?

[earnest jasper]

a TTS

american english voice

[modern oxide]

from the google translator?

[earnest jasper]

no it's of microsoft's TTS

[modern oxide]

damn

well, 32k is the lowest option possible rn

so choose it

[earnest jasper]

ok, what will happen if i set 40k or 48k

i want to know xD

[modern oxide]

as for epoch count put 600 should be good enought

[modern oxide]

basically nothing, it'll be the same, but the training will slowdown

[earnest jasper]

oh

[earnest jasper]

also i'm using the RVC disconnected colab

anymore settings or changes i should do?

batch size?

[modern oxide]

well, batch size 4 should be good

[paper nymph]

can you help me my man I really want to do this

[sand iris]

This requires the macos development tools installed via xcode-select --install

Ya we're getting there lol

[earnest jasper]

what really do the batch size does? like it fasten up the training or something else?

[rare goblet]

Ayo? @earnest jasper level 4 !!!

[paper nymph]

so what should I do now

[modern oxide]

it spkits samples per epoch basically

[earnest jasper]

ok, so 4 should be my go to value while training?

also what about the 'save frequency' thing?

[modern oxide]

yeah, the dataset if only 4 minutes and if it's TTS it should be pretty clean

[modern oxide]

save frequency you can put 20, it'll save each 20 epochs of your model

[sand iris]

You did the xcode thing ?

[earnest jasper]

alright, thanks for the help man! 🤗

[paper nymph]

how much storage does it takes

[sand iris]

Unsure

[paper nymph]

I am not sure I have enough space for it

[forest sluice]

https://github.com/JeremyCCHsu/Python-Wrapper-for-World-Vocoder/issues/87#issuecomment-1639504850

Ahh, I know the problem

Refer to the link above, there's a workaround

[sand iris]

nah pyworld builds fine on 0.3.4

We're just missing the c/c++ compiler atm

Messed with this enough to know lol 0.3.3 had that posixpath serialization issue, 0.3.4 builds fine OOB

[paper nymph]

what???

it says 120 hours

[sand iris]

Yeah the time estimation is fucked up

does that on macos

It'll go down

[paper nymph]

ı have to keep you waiting for downloading

like 15 minutes

or something like that

[modern oxide]

no problem :3

[paper nymph]

fuck rvc

I'm tired of this errors

pfff

[sand iris]

We're getting somewhere!

pip install -r requirements.txt and that should be it

OH WAIT LMFAO

Guess what

It's running on python 3.11, I missed that earlier

As unfortunate as this may be, we'll have to re-do part of the things

rm -r .venv && brew install python@3.10

Edit run.sh so that the line with python -m venv becomes python3.10 -m venv

Re run and that should be it

You can edit with any text editor

[sand iris]

line 33

replace python3 with python 3.10

[paper nymph]

did it

[sand iris]

great

[paper nymph]

saved

executing run.sh again?

[rare goblet]

Ayo? @paper nymph level 4 !!!

[sand iris]

Time to run run.sh for the last time !

(hopefully this time )

[paper nymph]

downloading different things

let's see if it will work this time

[sand iris]

It should now

it's got everything it needs

[storm ermine]

So RVC GUI is outdated? It seems like it was easier to use than the RVC WebUI

[paper nymph]

YEAHAHHHA

WE DİD İT

[paper nymph]

NO YOU DID IT

THANK YOU SO MUCH MAN

now ı have some question

what should ı do to close localhost

[sand iris]

Close localhost?

[paper nymph]

and should I do all these thing from at the beginning every time

how can i import voicemodels from huggingface links

[sand iris]

You could create a script for it

#!/bin/sh

cd ~/Desktop/Retrieval-based-Voice-Conversion-WebUI/

source .venv/bin/activate

python infer-web.py

Something like this

Make it executable and it can be used from anywhere by double clicking

[sand iris]

For that you get the .pth and .index files from the model, you put the .pth under assets/weights/ and you out the .index under logs/ in the RVC folder

[visual pumice]

Note: you can still call torch.view_as_real on the complex output to recover the old return format. (Triggered internally at C:\actions-runner\_work\pytorch\pytorch\builder\windows\pytorch\aten\src\ATen\native\SpectralOps.cpp:867.)
  return _VF.stft(input, n_fft, hop_length, win_length, window,  # type: ignore[attr-defined]
INFO:torch.nn.parallel.distributed:Reducer buckets have been rebuilt in this iteration.```

its been like this for like 10 minutes

[queen plaza]

hi uhm..

[sand iris]

What's your GPU and how much time worth of audio files did you give it ?

[queen plaza]

im trying to have like a very grumpy old man voice

[visual pumice]

gtx 1650 and 1:59 dataset

[queen plaza]

but my voice always cracks

😭 is there any fix?

[sand iris]

I think the 1650 is just too weak to train

[queen plaza]

i have my settings on like input thingy to like 500+ and chunk to the highest

I have a 3090ti and a good cpu

[sand iris]

it doesn't have enough VRAM for it, it should crash, but some drivers fallback to system RAM

Which is really slow

[visual pumice]

oh well

[sand iris]

Glad it worked after all of this lol, sorry this took a while

[sand iris]

#🔍│help-w-okada

[rare goblet]

Ayo? @sand iris level 59 !!!

[sand iris]

Does that depending to microphone and model quality

[queen plaza]

ah thank you

oh so uhm

i have a logitech g7333 or something

is that okay

i heard its bad

oh no

my 100$

[paper nymph]

I have one more question

how can ı change the language

turkish translation is very bad

[sand iris]

can you run echo $LANG in a new terminal?

[paper nymph]

in rvc folder?

[sand iris]

Don't know how macos handles languages but just trying to verify it's the same as linux

anywhere

[paper nymph]

forget about it

ı'm gonna learn how to use it

[sand iris]

LANG=en_US.UTF-8 python infer-web.py

In the start script could probably set it to English

But I'm not 100% sure

that or straight up remove the i18n/locale/<TUSKISH>.json could reset it to English

not sure about the filename

[zealous shuttle]

is there any inference colab that fixed the cuda issue?

[sand iris]

Yeah

-rvc

Wrong command

-colab

[azure marsh]

☁️ Google Colabs

• 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
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab

🤗 Hugginface 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

[sand iris]

AIcovergen-WebUI

has it fixed, but because the nvidia driver is stinky it takes a while to fix on launch

[zealous shuttle]

is it safe from banned?

[sand iris]

No*

It hasn't been banned yet though

[visual pumice]

can someone link me the google collab to train a simple dataset

my 1650 is actually ass

[sand iris]

RVC_Disconnected right above in the bot message!

[zealous shuttle]

so it run slowly?

[sand iris]

It takes a long time to launch

but it doesn't have the cuda error

[visual pumice]

im actually fuming

[sand iris]

And infers much faster. Unless you meant another cuda error I'm not aware of

[visual pumice]

my 1650 was lightning bolt like in 2019 but now its weak sauce in 2024

[sand iris]

I guess you can game on it still

it's just that gtx 1000 cards have issues with compute for some reason

[visual pumice]

its even worse because if i want to upgrade my gpu i need to upgrade my cpu or else it creates a bottleneck

[sand iris]

And that card happens to have a low amount of VRAM by today's standards

[visual pumice]

this was set on 2 on default, is it okay to set it to 12 cores?

[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

[visual pumice]

thats what my cpu can do so google should do the same?

[sand iris]

keep it on 2

The google machines are kinda underpowered on the CPU side

[visual pumice]

Google must have good gpus because wtf is a Tesla T4

[rare goblet]

Ayo? @visual pumice level 5 !!!

[sand iris]

big compute GPU

[visual pumice]

crazy GPU

why was i torturing myself wtf, colab is so much better than forcing my grandpa GPU from doing all the work

[paper nymph]

uvr5

where can i see error logs

[forest sluice]

What do you see in the rvc folder?

[paper nymph]

I don't think it is relevant with rvc

[harsh umbra]

hey guys, is this the right channel to get help from regarding putting voices onto songs?

[forest sluice]

Oh, you're running UVR5 as a separate app? Sorry about that, I thought you were attempting to run UVR on RVC

[forest sluice]

Yes, also next time just straight out tell the issue

[harsh umbra]

thank you, i wasnt sure. so i would like to put either my voice or my friends voice onto a song, the youtube guide didnt say anything regarding that

[coral turret]

how do i use the voice models in #1175430844685484042

i have rvc installed

[queen plaza]

hi uhm

is there a good mic

for this software

[old oracle]

-colab

[azure marsh]

☁️ Google Colabs

• 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
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab

🤗 Hugginface 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

[forest sluice]

Hmm, if you mean training voice model, just follow the guide

-guide @coral turret

[azure marsh]

Suggestion for @coral turret

Documentation

📚 All-In-One English documentation

https://docs.aihub.wtf/

❔ English FAQ

https://docs.aihub.wtf/faq

Credits: AI HUB FRANCE

✨ More guides

Full AI Voice Model Training Guide (Local)

Link: YouTube
Credits: Christopher Villanueva

Model training with Mainline RVC

Link: Rentry
credits: Raven (ravencutie21)

AICoverGen Colab Guide

Link: Google Docs
Credits: Eddy (Spanish Helper)

Create a model with RVC disconnected (colab)

Link: Google Docs
Credits: Angetyde

How To Make an AI Cover With Ilaria RVC

Link: Rentry
Credits: 👽 Julia (ailen2091)

[pine grail]

okay i havent been active in a while but im trying to find that colab without webGUI or whatever but i can only find the official banned rvc and the disconnected one

[rare goblet]

Ayo? @pine grail level 3 !!!

[pine grail]

the one that makes covers

[paper nymph]

why it gives error within 1 second

[sand iris]

What's logged in the terminal next to it ?

[sand iris]

Oh for fucks sake

That's one of the weird errors

Happens sometimes for no reason

Then re works randomly

[paper nymph]

fuck rvc again until morning

[paper nymph]

.

[sand iris]

Happened to me as well at some point

[harsh umbra]

how can i make my own model of a friends voice for personal use? couldnt find anything on yt

[rare goblet]

Ayo? @harsh umbra level 1 !!!

[sand iris]

Then just repaired itself randomly

[paper nymph]

so I have to wait for heal itselfs

[paper nymph]

thanks

[paper nymph]

I restart the computer still same 😦

[sand iris]

¯_(ツ)_/¯

If I ever get the issue again I'll try finding the cause

[visual pumice]

Yall.... how do i reset the colab

like how do i refresh it all

because its doing this when it didnt do this before

[queen plaza]

wait so

techincally

if you train your own model

to recognize ur voice and like

satuff

cant u sound like

100% real

like anothjer person

or soiemtihbngn?

[visual pumice]

im a tiny bit confused with RVC V2 disconnected colab, how do i make it forget about the previous stuff i was doing and whatever, like i wanna start a training process from scratch

[nocturne otter]

Does anyone know what to do it says pipeline failed whenever I go on any voice

[final palm]

erm

im using disconnected

and i got this

oh nvm i got it lol

[visual pumice]

how to i restart my colab

it still thinks im on the dataset i trained earlier but i have no idea

[gritty vigil]

Which pre trained checkpoint(s) should i use to train german voice models?

[visual pumice]

ive disconnected and deleted my runtime, i refreshed the page but it still is continueing with my previous session

and now everything is giving me errors because stuff already exists

[sage rampart]

what do i use for real time

[hollow hinge]

Sorry not sure where else to ask, but is there any replacement for UVR? It seems to have a lot of dependency issues right now.

[molten pecan]

mvsep.com

[hollow hinge]

Anything local?

[molten pecan]

Ask @odd shale

[spiral osprey]

Pls help me.

[odd shale]

Download UVR.

[hollow hinge]

I have. It's python dependencies have a lot of version issues. I could probably try and de-spaghettify it but I remember last time I did that it taken a while. Was hopeful there was some kind of replacement

[proud elbow]

UVR is the only good local one, and it only needs python 3.9 (recommended)

[hollow hinge]

I'll keep at it then. If I can get it working I'll save the requirements file somewhere lol

[proud elbow]

the UVR through installer and patch shouldn't be hard to set up, or alternatively through git repo download/git clone, just make venv and then pip install requirements within

[hollow hinge]

The installer's for windows right? I'll probably go through git and use a venv.

[rare goblet]

Ayo? @hollow hinge level 1 !!!

[hollow hinge]

I think part of my issue is I'm on linux and trying to use the ROCm version which doesn't really have many docs written about it. (Which I accept is a me problem lol)

[forest sluice]

Oh Linux

[hollow hinge]

I'm used to solving issues, it's just python is a particular pain in the ass a lot of the time

[proud elbow]

UVR through installer supports directML tho, unlike the manual install within venv which needs pytorch CUDA in order not to have issues

[hollow hinge]

There is no Linux installer, only Windows/MacOS

[proud elbow]

probably ask someone else using AMD GPU which manages to do RVC training on Linux & ROCm (yeah someone said it possible with some workarounds)

[forest sluice]

What's your Distro btw?

[hollow hinge]

I run Fedora Kinoite, I was trying to get RVC working in an archlinux container

I've had it work in a container before, either Ubuntu or Arch, don't remember.

Same for RVC

[forest sluice]

And you've read the guide, yes?

[hollow hinge]

For UVR? Yeah

[forest sluice]

And you did this in your Arch container? sudo rm /usr/lib/python3.11/EXTERNALLY-MANAGED?

[hollow hinge]

Yup

[forest sluice]

Plus, have you tried enabling venvs?

[hollow hinge]

That's what I'm trying now

[rare goblet]

Ayo? @hollow hinge level 2 !!!

[hollow hinge]

Also going to try the non-ROCm version

[proper shale]

¯_(ツ)_/¯

[proper shale]

go for 500

and check the graphs

for every model

[karmic heron]

i left rvc v2 disconnected bcuz of that, now i use rvc v2 disconnected but you use huggingface for it (you dont waste your google drive storage)

[proper shale]

is it the Brazil version

[karmic heron]

yes

[proper shale]

you can't retrain on it

[hollow hinge]

Finally getting somewhere

[proper shale]

i dont mean to hate on it but normal one is better

[karmic heron]

just change the link and model name

[proper shale]

imma put the matters in my own hands

translate the current disconnected colab to portuguese

[karmic heron]

just change title and markdown

[proper shale]

[hollow hinge]

Seems to work, might see if replacing torch makes ROCm work

[sand iris]

boy

do not

plus fairseq breaks on python 3.11+

[sand iris]

• get python310 from the AUR
• get aria2 from the arch repos
• setup a venv python3.10 -m venv /wherever/you/want
• Activate it source /path/to/venv/bin/activate
• clone the rvc code
• cd to it
• install the right torch build according to pytorch.org (you can omit torchvision)
• edit requirements.txt, remove mentions of torch, torchaudio and onnxruntime-gpu (will cause issues & pull in unneeded nvidia cuda packages)
• add onnxruntime to the reqs, by default it only pulls it on macos, but we want it here
• pip install -r requirements.txt
• ./tools/dlmodels.sh to download the pretrains
• HSA_OVERRIDE_GFX_VERSION=10.3.0 python infer-web.py note: the HSA override might be different depending on your GPU

[hollow hinge]

I did, seems to be working though so 🤷

[hollow hinge]

I was getting UVR working, but thank you this was next on my list 🙂

[sand iris]

Ah sorry

Might be a similar story for UVR

[hollow hinge]

UVR's problem was basically the ROCm tree sucks

[sand iris]

Does it use onnx ?

Also you don't really need ROCm on the system side for pytorch, surprisingly

[hollow hinge]

I think so, but either way the requirements list is wildly outdated

[sand iris]

Might be stuck on CPU-only for the onnx stuff

buidling onnxruntime on ROCm is a big big pain

[hollow hinge]

CPU is fine but GPU would be preferable

[mental sierra]

guys i need a little help

basically im trying to make an ai cover with juice wrld

but his voice doesn't sound right at 0 pitch change, but then when I change it -4 where it sounds better, it's out of tune. how can I fix this?

[proper shale]

pitch the instrumental to -4 too

-12 would be an octave deeper, but might be too deep

[mental sierra]

i tried that too

[proper shale]

with -12, -24 and so on it would be in tune w/ the original

else you can just pitch down the instrumental, should be fine

[mental sierra]

yeah just theres another voice of the original singer and they don't sound good together.

[proper shale]

unless you got background vocals 😭

[mental sierra]

sorry i'm not really sure how to explain it 😭

[proper shale]

it's ok

again, if you pitch the instrumental down 4 seminotes you should be fine

and it wont be off tune

[mental sierra]

yeah but what about the original voice?

[proper shale]

you just gotta make sure both the vocals and the instrumentals have been pitched down by the same number

so if you use the -4 vocals on the -4 instrumental it'll work out

[mental sierra]

oh alr i'll try that

is this the only way to fix it?

[proper shale]

...yeah

to keep everything in tune yeah

[mental sierra]

yeah but now the original voice is too deep 😭

[proper shale]

try doing everything in -2

[mental sierra]

alr ill see

is there anyway i can do it without changing pitches?

[proper shale]

don't think there is

[mental sierra]

damn then how do people make those covers that are perfect?

pitch wise

[proper shale]

it all depends on the type of voice really

[wet fog]

how do i use the voice model

[proper shale]

and the song

[proper shale]

some songs will sound weird with some artists on normal pitch

[proper shale]

#1159513888199540817

[mental sierra]

yeah the original is a lot higher than juice wrld so it sounds a little weird

damn it still sounds weird no matter what way i do it

[coral frigate]

Hello, does anyone have any ways to boost the quality of the voice changer? Like any additional audio software to make the rvc sound more crisp?

[proper shale]

just make sure you're using a good quality model, it should be fine

the mic also is kinda important, but not as much ngl

[narrow dove]

Can anyone tell me why my Applio file looks so different from what it should be after installation?

[proper shale]

spaces on folder name

remove spaces and try installing again

[narrow dove]

ah im a dumbass hahaha

thank you

[proper shale]

noooo you're not

rvc stuff just takes some time to get used to

[proper shale]

you're welcome :)

[narrow dove]

wait but it doesnt matter if my User name on my pc has a space in it does it?

[proper shale]

I don't think it does

[pulsar tartan]

hello guys
how can i use rvc models?
i want to make my own custom voice lines for valorant and i dont know how to use thecan someone help me please

[rare goblet]

Ayo? @pulsar tartan level 1 !!!

[proper shale]

https://rentry.org/ilarvc_inf_guide

[final palm]

would it even be possible to run rvc on a 2015 mac air locally

obviously i won't be training locally but inferencing could work

maybe

[nocturne otter]

on the rvc voice changer it says pipeline failed whenever I go on any voice and press start and when I speak I don't here anything

[rare goblet]

Ayo? @nocturne otter level 1 !!!

[mental sierra]

guys ive been trying to make my own voice model i've tried literally everything can someone please help me

tysm

[autumn hill]

TY IT WORKED

[half cove]

let goo

[proper shale]

delete pretrain and model_dir folders and restart w-okada

[mental sierra]

mj can you help me please

[rare goblet]

Ayo? @mental sierra level 3 !!!

[proper shale]

what's up

[mental sierra]

ive been having trouble making my own voice model

i have the audio i made to train it with

i just dont really know how to do it

[proper shale]

https://docs.google.com/document/d/1XuxQYiqEhYrdYeCZRRLrmV_ciMKo0bV-jTCGHu_-5Cc/edit?usp=drivesdk

[mental sierra]

ive watched lots of youtube videos on collab and other stuff i just cannot get a working one

[proper shale]

this should help

[mental sierra]

tysm

[proper shale]

and rvc disconnected works actually

[mental sierra]

ill let you know how it goes

[proper shale]

-colab

[azure marsh]

☁️ Google Colabs

• 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
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab

🤗 Hugginface 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

[proper shale]

heres the link for it too

[mental sierra]

thanks

[proper shale]

you're welcome🙏 good luck on the model

[mental sierra]

👍🏿

[mental sierra]

where is the link to the actual collab it self

[mental sierra]

is it this? rvc disconnected?

nvm i got it

mj so far so good, im a little confused as what to put for total epochs?

[proper shale]

keep at 500, you should be fine

[mental sierra]

alright thanks

[proper shale]

this doesn't mean you gotta train to that in one session though

you can stop after 1 or 2 hours

and export the model

[edgy tangle]

Hey guys, I just got a slightly strange error, and that is that I was training an AI voice with a 30 minute dataset, but it doesn't sound even close to the original voice.

[proper shale]

then you can retrain

[proper shale]

huh weird

[edgy tangle]

I have an AI voice that has a shorter dataset, but with the same voice, and I feel that it is much more similar to the original voice

[rare goblet]

Ayo? @edgy tangle level 3 !!!

[proper shale]

then you should probably use that then

see the thing is, adding more dataset doesn't mean the model will be necessarily more accurate

[edgy tangle]

ouh

[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

[edgy tangle]

I said this because I made the new dataset with a higher audio quality and a little longer to improve pronunciation and things like that, but it did the opposite

[vague sage]

pls help, someone that uses voicemeeter, call me dm

[slender scarab]

could anyone help me with this

python trainset_preprocess_pipeline_print.py "/content/dataset/jesse" 40000 2 "/content/Mangio-RVC-Fork/logs/jesse" 1
start preprocess
['trainset_preprocess_pipeline_print.py', '/content/dataset/jesse', '40000', '2', '/content/Mangio-RVC-Fork/logs/jesse', '1']
thread:0: 0% 0/1 [00:00<?, ?it/s]
thread:1: 0it [00:00, ?it/s]
end preprocess

[proper shale]

it's fine

[slender scarab]

ok

[rare goblet]

Ayo? @slender scarab level 1 !!!

[slender scarab]

i might need more help cause feature extraction never works

[slender scarab]

feature extraction keeps saying no-f0-todo. i have my dataset set up correctly idk whats wrong

[proper shale]

what are you using btw

local or via colab?

[slender scarab]

colab

[proper shale]

rvc Disconnected i suppose?

[slender scarab]

yes

[proper shale]

okay so

1. has to be a .zip, not a .rar
2. the folder you zipped must have the same name as the model's name
3. make sure the .wav file has been exported properly, id recommend re-exporting from audacity if that isnt the case yet

[verbal marsh]

how do i download

[paper nymph]

yep

an error again

what should I do to fix that

[paper nymph]

solved*

[solemn gust]

hi

[sand iris]

Is everything miraculously working now ?

[solemn gust]

i have an issue

[rare goblet]

Ayo? @solemn gust level 1 !!!

[paper nymph]

yeah somehow

[sand iris]

[paper nymph]

now I must learn how to use it

[sand iris]

Told ya it's random

[paper nymph]

voices are coming like mosquito

[sand iris]

If I remember correctly rmvpe_gpu is broken

[paper nymph]

not it'S not

export PYTORCH_ENABLE_MPS_FALLBACK=1

[sand iris]

OH

[paper nymph]

ı wrote this comment before running it

[sand iris]

OF COURSE

Yeah really good find

I swear they made it default somewhat recently

Probably RVC using an older version

[paper nymph]

and this is my double click run shortcut

I think I'm good for who doesn’t know anything about terminal

[rare goblet]

Ayo? @paper nymph level 5 !!!

[paper nymph]

I am very grateful you for your help and patience

english is not my first language

[solemn gust]

why my voice doesn´t change?

[violet heron]

First off

Your using crepe and not Rvmpe

Set CPU to your GPU

[proper shale]

check cmd

[solemn gust]

i don´t hace gpu

[proper shale]

oh i think they might be on the wrong version for their gpu

[proper shale]

oh.

[violet heron]

What’s your GPU?

[proper shale]

prolly integrated

[solemn gust]

i have an APU

[proper shale]

yeah it won't sound that good

[solemn gust]

I have integrated graphics

[proper shale]

delay wise

[violet heron]

You can try the colab

[proper shale]

dedicated graphics are needed for a good experience

[chilly jasper]

is there an easy way to batch infer a few hundred files

[proper shale]

^

[solemn gust]

what is that

[proper shale]

isn't there a batch inference section in most local forks

[violet heron]

https://colab.research.google.com/github/hinabl/voice-changer-colab/blob/master/Hina_Modified_Realtime_Voice_Changer_on_Colab.ipynb

[proper shale]

you should try using that

[solemn gust]

what i need to do with that?

[violet heron]

It’s a google colab

So you run it online

[chilly jasper]

ah I'll have to check deeper, i was using Mangio rvc and saw that inference batcher was in planned to be added so I didnt look much further, thanks!

[solemn gust]

idk how to use it

[violet heron]

It walks you through

[solemn gust]

explainme

i don´t understand anything KJGJFKDKJGD

[half cove]

dw there is a guide i think

[half cove]

your using hina colab?

https://docs.google.com/document/d/e/2PACX-1vTIceEcBfS6Zqolv_QEysrFfI_EJikPxozWptP_EjkpLVl-l-gdo-ijBonQMTviAHEYm5emmd9k9TdC/pub

[edgy tangle]

I already managed to fix the problem, now I only have to wait 10 minutes for each epoch

[half cove]

how long is your dataset

[edgy tangle]

30 min

[half cove]

are you using it on cpu?

[edgy tangle]

gpu

[half cove]

what gpu

[edgy tangle]

rtx 3050 with 4gb vram

but

im using laptop

[half cove]

try plugging it in ig 😭

that usually makes laptops faster

[edgy tangle]

?

[half cove]

like in charge

[edgy tangle]

ooh

its already plugged

[half cove]

[edgy tangle]

sorry im not good at english

[half cove]

dw

[edgy tangle]

[half cove]

its fine

[patent pasture]

or use colab, it is way faster

[edgy tangle]

almost every time that i try to use collab i cant use gpu

that's because i use local training

[tulip pine]

i use w-okada for real time voice chaning. but it doesent change anything in discord?

[violet heron]

Did you install the virtual audio cable?

[half cove]

what you mean?

[tulip pine]

uuuhhm what audio cable😅

[violet heron]

https://software.muzychenko.net/freeware/vac470lite.zip

[edgy tangle]

i get this message

[violet heron]

Download and extract, run the setup64 file

[patent pasture]

it's weird and concerning, didn't you try to use colab on other google accounts?

[edgy tangle]

i tried

[violet heron]

Then on Okada set your output to the cable

[half cove]

you prob used all the gpu limit for the day

[violet heron]

And on discord set your input to the cable

[cinder bobcat]

Hi, im training a dataset (12 mins) and when i checked the 10 epoch version it sounds very different to the desired voice. Im using local RVC - should I maybe wait?

[rare goblet]

Ayo? @cinder bobcat level 2 !!!

[half cove]

when did you get the message

[tulip pine]

setup, setup64 or setup64a?

nvm

nvm

[edgy tangle]

when i try to run collab

[patent pasture]

uh, try vpn

[edgy tangle]

sometimes i can use it

[patent pasture]

maybe here is some location isues

[half cove]

like right after or while training

[edgy tangle]

but i prefer local training

because i have more space

[patent pasture]

okay

[edgy tangle]

yeah

[cinder bobcat]

Should the songs I want to train be split, or is it enough if I leave it as it is? (WAV files)

(Talking about local training)

[half cove]

if you mean split it into multiple audios rvc does that already

[cinder bobcat]

Oh okay, I created a folder in "dataset" and copied the path. I wonder why the voice sounds so different when I tested the 10 epoch file 🤧

[half cove]

that cuz its only 10 epoch

[patent pasture]

@half cove, are you good in analyzing graphic?

[cinder bobcat]

yee ik, but the voice didn't go in that direction at all so I was really wondering xd

[half cove]

what do you mean by that

[patent pasture]

why does it look unusual?

[half cove]

turn smoothing up

[patent pasture]

I turned 0,999

[half cove]

how many epochs have you trained?

[patent pasture]

4 steps = 1 epoche

[half cove]

mfer making me do math

[patent pasture]

Ok, I used Mailine RVC

is it normal?

@half cove, do you train on local PC?

[half cove]

yea but im a bit busy rn gime a sec

[patent pasture]

oh, sorry then

[edgy tangle]

I found the problem

When I use the GPU in the pitch extraction everything is done much faster, but it happens that the voice does not look anything like that of the samples

But if I use the CPU, everything is much slower, however the voice has greater fidelity to the samples

[polar siren]

Need help- tried just ignoring this because I thought it was just weird but then I tried to train the feature index and nothing happened, how do I fix this?

[edgy tangle]

RuntimeError: Expected tensor for argument #1 'input' to have the same type as tensor for argument #2 'weight'; but type torch.cuda.FloatTensor does not equal torch.cuda.HalfTensor (while checking arguments for cudnn_batch_norm)

This error appears several times when I do pitch extraction with rmvpe_gpu

There is a solution?

[polar siren]

looking back on here I think I found the issue I'll try the whole thing again but this time with the mp3 being a wav instead

[rare goblet]

Ayo? @polar siren level 2 !!!

[lusty iris]

Hi all! I am completely new to making voice models. I installed Mangio RVC and tried to make my first model, using one single 40-minute .wav file (clean audio, no background noise), trained with 300 epochs. When I try to change my voice using that model, the result is very noisy (it sounds like in an old FM radio, lot of white noise or so). When I'm using a downloaded model, the quality is good. Could someone give me a few hints on what to do to improve? Thank you in advance!

[rare goblet]

Ayo? @lusty iris level 1 !!!

[edgy tangle]

Someone? (btw im using applio)

[spiral osprey]

anybody know?

[cinder bobcat]

i created a 50 epochs model and it doesnt even sound 1% similar to what i wanted (dataset was like 5 minutes) I dont need quality but this is really weird. I selected the rmvpe algorithm. Someone knows what can I do?

My colab training was always succesful but local training is really frustrating for me

[candid lichen]

So can someone who actually isn't a dick tell me what I need to fix with my model in order for it to get "approved"

There's the pth and index boom

"Zipped Incorrectly" my ass

[half cove]

-upload

[azure marsh]

Suggestions for @candid lichen

huggingface.co upload Guides

English guide

Link: https://rentry.org/fdg_guide_newer
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

[candid lichen]

Bro I literally did that

[half cove]

seems like you didnt

[candid lichen]

Yeah it's the index and pth

What's the issue

[half cove]

get them both and put it in a zip

then upload to huggingface

[candid lichen]

It was in a zip

[half cove]

18 items

inside the zip file should look like

[candid lichen]

Well thank you for not being a duck and actually telling me what I need to do

Dick

[half cove]

who was like that?

[half cove]

oh

[candid lichen]

Yeah called me bozo for asking what's wrong then not giving a explanation

So yeah pretty annoying but anyways thank you I'll do that

[edgy tangle]

Aaah

[rare goblet]

Ayo? @edgy tangle level 4 !!!

[edgy tangle]

I still can't solve that error

[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

[woeful pond]

Heya

I was wondering whats the most simple way to update rvc?

Just a complete reinstall or is there another way?

Im talking about the one by okada if thats still the one to use

[bold canyon]

whenever i try to download a model on the rvc it says theres been an error?? does anyone know how to fix??

[half cove]

-howtoask

[azure marsh]

How to troubleshoot

• Give context: Throwing your error/issue alone will not help. Tell what step are you on, what were you doing before issue, your settings, your terminal, et cetera.
• Be relevant: You come to ask for help, not to complain. You can't ask for RVC help in voice changer help and vice versa.
• Be polite: DONT DEMAND FOR HELP OR I WILL FUCKING MURDER YOU. Ask politely, the server was not made for you.
• Don't be an attention whore: If people are ignoring your help, kindly ping them. If they are ignoring you, either they don't know how to solve it, they are busy, or you are an asshole.

WITH ALL DUE RESPECT, DO NOT BE AN ASSHOLE AND HAVE COMMON SENSE.

[native copper]

when i set my output to cable input its not outputting any sound with the voice changer?

[brittle wing]

Does anyone have experience with Stable Diffusion AI, I use Fooocus and have a question

[violet heron]

#🔍│help-ai-art

[slender scarab]

can anyone help with this? steps i was shown earlier didnt work

do i have to resample the dataset to my target sample?

[chilly jasper]

Feature extraction at 48k sample with rvc v2 on around 12 minutes of training data should take a while right? (Using 3070 and rmvpe)

[half cove]

How long per epoch

[chilly jasper]

How would I check this?

[half cove]

On the cmd

If your already training a model

Are you asking before your training a model?

[chilly jasper]

This step, cmd has been updating very slowly

[half cove]

Like how slowly

[chilly jasper]

every few minutes, some threads havent moved in the last ~30 min though

[chilly jasper]

thread 0 actually hasnt updated for like 47 min now

[proven hill]

E-GIRL VOICE MODEL SALES!!

DM ME FOR DISCOUNT PRICES!

E-Girl, Mommy, V-Tuber, Anime girl and much more!

Wondering what models?
You find them here: https://ko-fi.com/ilariaowo/shop

INCLUDING FREE ONES!! :swaygirl:

I also do commissions if you want aspecific voice!

I accept the following payment methods:

• Ko-Fi
• PayPal
• Steam Gift Cards
• Amazon Gift Cards
• Throne

Grab yours today!!!

[half cove]

[molten pecan]

YIPEEE

[molten pecan]

Which RVC version exactly? Can you send a screenshot?

[orchid aurora]

Hola

Testing

Kewl

Come hago tags aqui?

@molten pecan

Got it

[molten pecan]

@half cove how to colab with kits

[orchid aurora]

@molten pecan este es la versión que tengo

[orchid aurora]

☺️

Every
Single
Time

[molten pecan]

Podrías nombrar el modelo como "test"?

A veces da ese error cuando subes 2 modelos con el mismo nombre

[orchid aurora]

Ok

Error

[molten pecan]

Revisaste si el modelo tiene el archivo .pth e .index?

[orchid aurora]

Deja reviso

[molten pecan]

Fijate dentro del zip

Eso quise decir

[orchid aurora]

Ok

Si en dos folders separados

[rare goblet]

Ayo? @orchid aurora level 4 !!!

[molten pecan]

Que raro

[orchid aurora]

Los tengo que subir unzipped?

[molten pecan]

Deberían de estar en una misma carpeta

[molten pecan]

Subilos en una misma carpeta, que no estén separados

[orchid aurora]

Ok

Agrégame para no molestar a gente aquí?

[molten pecan]

No hace falta. Este es el canal de ayuda

[orchid aurora]

Ok

[orchid aurora]

Unzipped?

[molten pecan]

En un .zip

[orchid aurora]

Puedo subirlos en un folder o tienen que estar en un zip ?

[molten pecan]

Tienen que estar en un zip

[orchid aurora]

Todos estos en un zip? Bear with me…. 😩

[molten pecan]

Solamente el .pth y el .index llamado "added_"

[orchid aurora]

Please do t hate me

[molten pecan]

El resto no van

[molten pecan]

No te disculpes. Estoy para ayudar

[orchid aurora]

Gracias ☺️

[opal depot]

anything I need to alter here?

using rvc disconnected collab

[molten pecan]

Nuh uh. Unless you are a more advanced user

You are good to go :3

[opal depot]

fym nuh uh /j
Kk

[rare goblet]

Ayo? @opal depot level 3 !!!

[opal depot]

I'll probably be asking a bunch of questions

Cause i'm new

[molten pecan]

That's alright. We all start somewhere

[violet heron]

Sample rate

Change it to be lower

[opal depot]

I got told to not change it 💀

[violet heron]

You should change it

[opal depot]

To?

[violet heron]

Most of your time your model does not follow the target sample rate

[violet heron]

32k is your best bet

[opal depot]

My files are wav but they were originally ogg

Idk the sample rate for them

[violet heron]

You try a bit higher

[molten pecan]

According to the rvc disconnected guide you should leave it at 40k