✨│ai-help

[viscid moss]

like Mainline, Mangio, Applio

[brittle wing]

Mangio

[viscid moss]

I mean... install the app on ur PC not in colab

https://docs.aihub.wtf/rvc/local/mangio/

Checj guides for that pls

[tawdry edge]

does anyone know how to get a hatsune miku voice working on voice.ai... like the "make yr own voice" type thing

[viscid moss]

we don't support voice.ai sorry

[tawdry edge]

damn

a guy said to check here for voices

mb

[viscid moss]

U can found voice on #1175430844685484042

but idk if that models works on voice.ai

we use w okada voice changer

[tawdry edge]

does it work on games?

[rare goblet]

Ayo? @tawdry edge level 1 !!!

[tawdry edge]

yuhh

and is it free?

[viscid moss]

yep

[brittle wing]

Will this voice sound nice?

[viscid moss]

of course

[brittle wing]

not eddy

[viscid moss]

maybe

it depends on ur audio files that u use to train the model

[tawdry edge]

can u send me a download link pretty please i HAVE to be hatsune miku

[brittle wing]

.flac

[viscid moss]

https://rentry.co/VoiceChangerGuide

[tawdry edge]

thx

[viscid moss]

i mean the quality of the audio

but u need to test

so u can hear ur model

[brittle wing]

high choose

I chose a good one when I made an mp3 converter

and I could have done the best too late

I also had the option of the best

kbps

[tawdry edge]

oo that is a lot of words im a little bad at following directions do u think u can help a bit with that too or😭

[brittle wing]

wait

helps me

[tawdry edge]

i js rlly wanna be hatsune miku

oh okay

lmk when you guys r done

[brittle wing]

okay

not eddy already

[viscid moss]

okke

[brittle wing]

open web.bat?

[viscid moss]

let's talk in #🔍│help-w-okada

[viscid moss]

yep

[brittle wing]

wait

what next?

not eddy

[viscid moss]

https://docs.aihub.wtf/rvc/local/mangio/#inference-
pls check the guide

index on logs folder and pth in weights folder

[brittle wing]

not working

I can't find the tab where the vocals are sent

[viscid moss]

the converted audios?

[brittle wing]

vocals

I don't have such a specifier what I write there, there is no such thing

[viscid moss]

u need to search fot the folder "audios" in ur RVC folfer

[viscid moss]

there

[brittle wing]

put vocal?

folder audios?

And where do I choose the vocal once I've put it in "Audios"?

?

[viscid moss]

u need to change the path

or press auto detect and choose the audio file

[brittle wing]

Which one exactly?

[viscid moss]

The name of the audio you uploaded in audios folder

[rare goblet]

Ayo? @brittle wing level 13 !!!

[viscid moss]

yep, that

[brittle wing]

and convert?

yes?

[viscid moss]

yep

[brittle wing]

mp3 flac?

[viscid moss]

The format you want

[brittle wing]

How long will it take?

[viscid moss]

depends on ur pc

[brittle wing]

1h??

[viscid moss]

do u have a Nvidia GPU?

[brittle wing]

no

inte

l

[viscid moss]

only CPU?

oh ok

how long is your audio file?

[viscid moss]

.

[brittle wing]

10 min

[viscid moss]

I guess it will take about 10 -15 min.

it depends as I said

[brittle wing]

and can I send it to my phone and play with TTS as soon as I send myself.pth and index?

yes?

huggingface public

[viscid moss]

ya

[brittle wing]

And a lot of people see it on hugging face afterwards?

Can you hide it for me after uploading?

[viscid moss]

The model will be public but people will only be able to see it if they enter your HF profile or if they find your model by the name you give it.

Like a social network

[brittle wing]

maybe it's wrong that I took this flac instead of mp3 it will take longer

format

How do I make colab save my progress to my drive? If colab disconnects IT REMOVES ALL FILES IN THE WHOLE NOTEBOOK, like wtf?

Even if I refresh browser it delets everything

[viscid moss]

I think there is not a colab with autobackup (I think)}

-colab

[azure marsh]

☁️ Google Colabs

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

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[brittle wing]

wtf

[viscid moss]

I think it is better for you to use Colab 'cause GPU

[brittle wing]

okay

lets go

why?

[sturdy charm]

what do I download if I run off cpu for windows

actually I have no idea what im doing

I dont even know if I have the right app download

[brittle wing]

so you just downloaded something random

[sturdy charm]

nah I followed the tutorial from Jarods Journey

I just dont have a GPU so Im not exactly understanding what to do

[sturdy charm]

what apps do I need to use the app if I have no gpu?

[analog obsidian]

-colab

[azure marsh]

Suggestions for @sturdy charm

☁️ Google Colabs

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

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[sturdy charm]

oh uh

[rare goblet]

Ayo? @sturdy charm level 1 !!!

[sturdy charm]

which one should I use

[analog obsidian]

i personally prefer applio

[sturdy charm]

I see

[sturdy charm]

do I need any apps to run it?

like python or anything

[analog obsidian]

nothing, they run in your browser

log into your google account and you can use applio's colab, tho it has a time limit

[brittle wing]

good?

[brittle wing]

Applio and Paperspace bad imo

[rare goblet]

Ayo? @brittle wing level 3 !!!

[sturdy charm]

well I have no idea what im doing regardless rn

[brittle wing]

Have you tried reading the docs?

[analog obsidian]

http://docs.aihub.wtf/rvc/cloud/applio-colab/

please read this

[brittle wing]

^colab applio sucks

but their local app is good

[sturdy charm]

this is way too much work to catfish people

gotta make content somehow

looks like Ill actually have to work

RIP 🙏

[brittle wing]

laziness in catfishing does not get you anywhere

you need to be passionate about it

[sturdy charm]

ehhhh

its not that im lazy

its that im poor

I can NOT afford a gpu in this market

because I am infact using a work computer

and would need a whole pc

[brittle wing]

Why does my model speak with a trained voice for a while, and talk like a robot in longer words

[near kiln]

Im trying to find a realistic egirl voice model

[brittle wing]

?

[near kiln]

I have a pretty good one, but my rvc makes the voice kinda sound glitchy at times and dont know any work arounds

how dependent is the voice changer on your pc's hardware?

[cobalt pilot]

Uhh what happened to ilaria's link on https://docs.aihub.wtf/ ?

[brittle wing]

how do I train a model?

[rare goblet]

Ayo? @brittle wing level 1 !!!

[limpid flicker]

how do i get the .pth & .index files on huggingface (i had a model made for me and i'm pretty confused since i don't have any clue how all this works)

[rare goblet]

Ayo? @brittle wing level 1 !!!

[polar wyvern]

IS applio faster than mainline rvc?

[hard frost]

ХУЙ

[hoary gyro]

what do i put in the voice model box on the applio interface

[rare goblet]

Ayo? @hoary gyro level 1 !!!

[pliant relic]

JSONDecodeError Traceback (most recent call last)
<ipython-input-6-30c00ab7158a> in <cell line: 31>()
31 if os.path.exists(config_path):
32 # File exists, proceed with creation of creds and client
---> 33 creds = Credentials.from_service_account_file(config_path, scopes=scope)
34 client = gspread.authorize(creds)
35 else:

5 frames
/usr/lib/python3.10/json/decoder.py in raw_decode(self, s, idx)
353 obj, end = self.scan_once(s, idx)
354 except StopIteration as err:
--> 355 raise JSONDecodeError("Expecting value", s, err.value) from None
356 return obj, end

JSONDecodeError: Expecting value: line 1 column 1 (char 0)

[hoary gyro]

got the same error

[pliant relic]

what doing here?

[hoary gyro]

i think its just outdated

try using applio

-rvc

[azure marsh]

Documentation

AI HUB Docs

https://docs.aihub.wtf/

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

• Full AI Voice Model Training Guide (local) by Christopher Villanueva
• Model training with Mainline RVC by Raven (ravencutie21)
• AICoverGen Colab Guide by Eddy (Spanish Helper)
• Create a model with RVC disconnected (colab) by Angetyde

[hoary gyro]

wrong thing

-colab

[azure marsh]

☁️ Google Colabs

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

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[polar wyvern]

Use ilaria on huggingfacr

It is cloud

Fast

Easygui

If It gives a erro like 500

Create a account, clone the space And select cpu free

Then you Will have a personal cloud rvc

[pliant relic]

how to use applio rvc?

[proper shale]

outdated colab, use Applio instead:

https://docs.aihub.wtf/rvc/cloud/applio-colab/

[woeful nest]

@proper shale how can i use the voice changer

[proper shale]

https://rentry.co/VoiceChangerGuide

should help you with that

[woeful nest]

Why is the client not working? I only get a white screen

[proper shale]

what's command prompt saying

if you just opened MMVCServerSIO.exe yeah it wont work

[woeful nest]

i opened voice changer native client

but all i get is white screen

can i send a pic?

[proper shale]

did you open with start_http though

[scarlet wedge]

how can i post voice models on #1175430844685484042 ?

sorry, my english is bad :v

[proper shale]

#1159514067187277865 - upload the model with QCBot, then a QC will review it and, once you get approved, just post on #1175430844685484042

[scarlet wedge]

thanks, but i type /submit and i type what more? is there a template of the information that i need to type?

[proper shale]

the thing has all the options you need to fill out

[scarlet wedge]

oh, thanks!

[proper shale]

like this

[scarlet wedge]

yeah, i see it

[grizzled tundra]

@proper shale The voice when speaking is not understood or there is confusion

how i can fix it

[rare goblet]

Ayo? @grizzled tundra level 1 !!!

[grizzled tundra]

@red kayak

[red kayak]

I don't think I understand what you are trying to say

Could you please provide a demo

[grizzled tundra]

The sound is not clear , How can I solve it

and not real

@red kayak

[tame mica]

does rvc support flacs? honestly i forgor so im just asking this

[glad zealot]

yes

[brittle wing]

-colab

[azure marsh]

☁️ Google Colabs

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

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[brittle wing]

-hf

[azure marsh]

<:huggingface:1179800228946268270> Hugginface Spaces

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

More commands: -audio, -colabs, -kaggle, -local, -overtraining, -realtime, -rvc, -help

[brittle wing]

-colabs

[azure marsh]

☁️ Google Colabs

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

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[toxic patrol]

I have this problem where when i talk the word i said is being said multiple times how can i fix that?

[vague sage]

Rvc okada?

[toxic patrol]

Yes...?

[rare goblet]

Ayo? @toxic patrol level 1 !!!

[vague sage]

Rvc real-time, okada or inference?

[toxic patrol]

Okada i think

[vague sage]

Can you send a screenshot?

[proper shale]

Enable Sup2 and the Echo options, it should mitigate the echoing issue

[toxic patrol]

Thank you for the help appreciate it. Lars I couldn't send a screenshot, because i was just offline sorry.

[proper shale]

You're welcome! :)

[vague sage]

Ah ok np

[brittle wing]

How do I get the ngrok token value?

[fading drum]

how install tensorboard

i mean, i do pip install tensorboard, but tensorboard no work, no commands or anything

[analog obsidian]

http://docs.aihub.wtf/rvc/resources/epochs--tensorboard/#tensorboard

[fading drum]

its also in my environment variables

[fading drum]

but it doesnt work

it spits out
tensorboard_venv already exists, skipping creation and activation...
Launching TensorBoard...
2024-06-15 01:33:32.328918: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable TF_ENABLE_ONEDNN_OPTS=0.
2024-06-15 01:33:33.597228: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable TF_ENABLE_ONEDNN_OPTS=0.

wait... whats an environment variable

[fading drum]

it says this 3 times then closes

[analog obsidian]

What is your gpu? Which one you have?

[fading drum]

nvidia 1660 ti SUPER

[analog obsidian]

Oh that explains it

[fading drum]

i reinstalled windows a week ago and it was working then, but now it doesnt

[analog obsidian]

1660 gpu cant do half precision

[fading drum]

i mean the error changed, when i did the system environment variable change

it is now:

[analog obsidian]

Rvc is set to work using fp16

Ur gpu cant do that

[fading drum]

rvc works fine on my pc, ive trained multiple models, its just tensorboard will not

[analog obsidian]

Oh hmm

[fading drum]

im using rvc1006nvidia

[analog obsidian]

Sorry i have no idea, but i know is something related to the 1660 itself

Tensorboard may need some special parameters

Wait a sec

[fading drum]

glances back to stackexchange

yyyyeahhh nothing useful there

well, useful, but unintelligible/stuff ive already done

[analog obsidian]

i think rvc automatically changes half precision to full precision if it detects a 1660 gpu

But tensorboard doesnt

[fading drum]

do you know how i can tell tensor to use full precision?

man i cant wait for my nvidia tesla k80 to arrive

[analog obsidian]

Ill ask someone that may know a solution

[fading drum]

thx

[analog obsidian]

It seems tensorflow doesnt support 1660

[fading drum]

but... how did it before??!?!?

[analog obsidian]

It worked before?

[fading drum]

unless it was something strikingly similar to tensorboard

[low shard]

are you using a colab/kaggle that requires you a ngrok auth token? u can get it at https://dashboard.ngrok.com/tunnels/authtokens/new

[fading drum]

waiiit, i think it was in a google colab

[analog obsidian]

Oh that explains it

I was also searching in google and also got the response 1660 dont work in tensorflow

[fading drum]

damn, is there a tensor alternative i can use to see training patterns do you know? or am i just screwed

i mean, ill probably need to re-code a bunch of stuff, probably easier to use excel

[analog obsidian]

Yea this is a bit advanced, i have no idea how to make a 1660 gpu to work in tensorboard

Srry

[brittle wing]

Oh nice I was using RVC Mainline

[fading drum]

alg thanks for helping tho

[low shard]

like the rvc mainline kaggle?

[brittle wing]

This

The girl on the photo on the corner is me.

[low shard]

ohh lol

[brittle wing]

E che puoi dirmi?

Come me la cavo?

Come faccio a fare il Colab funzionare?

[low shard]

?

[brittle wing]

Puoi aiutarmi?

[low shard]

in cosa

[fading drum]

I DID IT, I GOT IT WORKING

[analog obsidian]

[fading drum]

you gotta uninstall cuda and tensorboard, re-install python, and reinstall tensorboard, then cuda

apparently cuda likes to fight with tensor, but tensor doesnt fight with cuda

[analog obsidian]

lols so it wasnt the 1660 at all

[fading drum]

apparently not

1 random dude on stackexchange under like 20 comments and im like "imma do THAT one"

[analog obsidian]

lmao amazing

[glacial cargo]

is illaria rvc down? if so, you got any other rvc sites?

[rare goblet]

Ayo? @glacial cargo level 2 !!!

[fading drum]

im feeling so sugoi rn

[wheat plover]

if my audio is in a sample rate of 44100khz, should I use a 48k or a 40k sample rate to train?

[azure marsh]

Suggestions for @glacial cargo

☁️ Google Colabs

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

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[glacial cargo]

ah ok ty

[soft tiger]

what software can i use to convert audio files into a ai model

not make a ai model

use my ai model to sound like this vocal file

[wheat plover]

you mean make an ai cover?

well, you can use applio

-applio

[azure marsh]

🖥️ Local stuff

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

Credits to Faze Masta and Antasma for compiling these links.

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

[proper shale]

Can you link to Mainline real quick

[proven hill]

message got deleted lmfao

[proper shale]

ah

wait ila

[proven hill]

[proper shale]

are u also trying to fix the colab

dumb question ik

but just so i can write somethin

[proven hill]

now we are just moving to the better mainline one

[proper shale]

ah

[proven hill]

no reason to fix the old unsopported one

but we are working on HF

as talked to modz too

[proper shale]

yeye

[tall depot]

I can't wait to see Applio with a GPU

[rare goblet]

Ayo? @tall depot level 8 !!!

[proven hill]

[proper shale]

The original Ilaria RVC on Colab is now unsupported.

What do I use instead?
https://colab.research.google.com/drive/1mHKTGH5e3SAyDSBss1KtiYRbDdQzwSMs - Ilaria RVC Mainline, the most up-to-date fork by Ilaria.

What about Huggingface Spaces?
The original Ilaria RVC on HF will be fixed soon. You should use Ilaria RVC Mainline (linked above.)

[proven hill]

I am Ilaria and i support this message

[proper shale]

Are you really Ilaria, the italian girl who made the wonderful "Ilaria RVC"???

[tall depot]

[proven hill]

once we fix the huggingface, everyone will have a FREE A100 80GB for inference

[peak tusk]

asked one, but zero gpu crashes idk why https://huggingface.co/spaces/IAHispano/Applio/discussions/4

[neon gazelle]

whenever i press start it just plays back my own voice

[pastel oak]

make sure that PASSTHRU next to start - stop is GREEN

[neon gazelle]

thank you

[quasi ether]

Is this good to train ai models on computer?

[pastel oak]

if its the 6gb vram itll be okayish, i personally prefer 8gb vram but i am not a pro so others might say something else

im not sure if batch size 8 is possible comfortably but if you use titan pretrain and 4 batch size on a 10-20 minute dataset it shouuuuld work well

[quasi ether]

Idk where to find Vram so have my entire task manager

[ornate summit]

I don't know coding so I'm a bit confused. I'm trying to run the RVC and instead of opening it says this. What did I break?

Traceback (most recent call last):
File "C:\Users\sludg\Downloads\RVC1006Nvidia\gui_v1.py", line 874, in <module>
gui = GUI()
File "C:\Users\sludg\Downloads\RVC1006Nvidia\gui_v1.py", line 110, in init
self.config = Config()
File "C:\Users\sludg\Downloads\RVC1006Nvidia\configs\config.py", line 35, in wrapper
wrapper.instance = func(*args, **kwargs)
File "C:\Users\sludg\Downloads\RVC1006Nvidia\configs\config.py", line 50, in init
self.json_config = self.load_config_json()
File "C:\Users\sludg\Downloads\RVC1006Nvidia\configs\config.py", line 68, in load_config_json
d[config_file] = json.load(f)
File "json_init_.py", line 293, in load
File "json_init_.py", line 346, in loads
File "json\decoder.py", line 337, in decode
File "json\decoder.py", line 355, in raw_decode
json.decoder.JSONDecodeError: Expecting value: line 1 column 1 (char 0)

[low shard]

i feel like that community grant discussion will do the same end of https://huggingface.co/spaces/TheStinger/Ilaria_RVC/discussions/3 ngl

prolly they focusing on zero gpu as they spent millions on it

[peak tusk]

nah I got in touch with a friend at huggingface and I will get it on monday

[low shard]

lmfao

does anyone here even know abt https://huggingface.co/spaces/r3gm/rvc_zero

[brittle wing]

hello

[rare goblet]

Ayo? @brittle wing level 1 !!!

[brittle wing]

Can anyone help me? I want to use the voice of a character and it doesn't sound like it is

[limpid flicker]

is there like a video tutorial? i am soo confused and not sure how it works

[wheat plover]

read our documentation at https://docs.aihub.wtf/

-help

[azure marsh]

Command has expired

[wheat plover]

well, i need a documentation myself on these bot commands

[earnest pier]

Please can someone dm like a updated tutorial of how to use rvc I keep trying upload voices but then I don’t know how to use them

[brittle wing]

-colab

[azure marsh]

☁️ Google Colabs

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

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[quasi ether]

-RVC

[azure marsh]

Documentation

AI HUB Docs

https://docs.aihub.wtf/

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

• Full AI Voice Model Training Guide (local) by Christopher Villanueva
• Model training with Mainline RVC by Raven (ravencutie21)
• AICoverGen Colab Guide by Eddy (Spanish Helper)
• Create a model with RVC disconnected (colab) by Angetyde

[quasi ether]

-help

[azure marsh]

Command has expired

[quasi ether]

-links

[azure marsh]

🖥️ Local stuff

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

Credits to Faze Masta and Antasma for compiling these links.

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

[cedar surge]

Hello!

[rare goblet]

Ayo? @cedar surge level 1 !!!

[cedar surge]

I have one question

When I run the voice changer with GPU options applied, my CPU strains, but my GPU is almost like it's doing nothing

Is this normal?

I have a 12th Gen Intel(R) Core(TM) i5-12400 2.50 GHz processor

[analog obsidian]

try rmvpe (no onnx)
but yes realtime* is a bit cpu intensive too, your cpu only has 6 cores

make sure rvc quality is set to low instead of high (there's barely any difference)

also #🔍│help-w-okada use that* channel for wokada help

[snow hazel]

I'm using mvsep to remove hamonies but I don't know what settings to use

[tall depot]

it run very fast with a T4 (the same as Colab's GPU) https://huggingface.co/spaces/mealss/rvc_zero

[snow hazel]

Is there another way instead of mvsep to do this?

[low shard]

Yea, but rvc zero was meant to run on ZeroGPU (A100, like more than 11 times better), that's why its called rvc zero and its only on hf spaces

its not meant to be a local fork like ilaria rvc or applio

[flint seal]

I'm training on Applio RVC Colab right now, and my .pth file generated successfully, but my index is nowhere to be found. I clicked 'Generate Index,' but it still isn't there

When I finished training, I also received this error.
Training has been successfully completed with 250 epoch, 250 steps and 4.31 loss gen.
Lowest generator loss: 3.200925827026367 at epoch 212, step 211
Saved model '/content/program_ml/logs/2022_250e_250s.pth' (epoch 250 and step 250)
Failed to train index: need at least one array to concatenate
If you are running this code in a virtual environment, make sure you have enough GPU available to generate the Index file.
/usr/lib/python3.10/multiprocessing/resource_tracker.py:224: UserWarning: resource_tracker: There appear to be 20 leaked semaphore objects to clean up at shutdown
warnings.warn('resource_tracker: There appear to be %d '
Failed to train index: need at least one array to concatenate
If you are running this code in a virtual environment, make sure you have enough GPU available to generate the Index file.

[cedar surge]

Thank you!

[snow hazel]

I need help. I'm trying to use mvsep to remove hamonies but I don't know what settings to use.

[flint seal]

where do i put my dataset path?

never mind, I see now.

[flint seal]

I changed my path to that, and retrained my model but now I get the same error.

Training has been successfully completed with 100 epoch, 100 steps and 6.995 loss gen.
Lowest generator loss: 3.4762566089630127 at epoch 67, step 66
Saved model '/content/program_ml/logs/the2022project_100e_100s.pth' (epoch 100 and step 100)
/usr/lib/python3.10/multiprocessing/resource_tracker.py:224: UserWarning: resource_tracker: There appear to be 20 leaked semaphore objects to clean up at shutdown
warnings.warn('resource_tracker: There appear to be %d '
Failed to train index: need at least one array to concatenate
If you are running this code in a virtual environment, make sure you have enough GPU available to generate the Index file.
Failed to train index: need at least one array to concatenate
If you are running this code in a virtual environment, make sure you have enough GPU available to generate the Index file.

[snow hazel]

What is the best way to get rid of harmonies?

[flint seal]

yeah

i ran the mount drive cell

i will retrain it, as it trains pretty fast

yup, python 3, t4 gpu

yup

inside /content/drive/MyDrive/dataset

the .zip with my audio files

oh crap

alright

[flint seal]

hey, sorry for the late reply, i was training my model

but it worked!

so thanks

[undone steppe]

whats the best gui app?

[proper shale]

Depends on what you wanna do:

https://docs.aihub.wtf/rvc/local/applio/ - Applio for audio conversion
https://rentry.org/VoiceChangerGuide - W-Okada for realtime conversion

[fallow marlin]

whats the best free app to make ai song covers?

[brittle wing]

Is the Multinotebook RVC kaggle notebook broken?

I executed all the cells in order

nvm

i found another notebook

[lofty musk]

https://media.discordapp.net/attachments/1234675157747105853/1251359688205926420/image.png?ex=666e4b06&is=666cf986&hm=6f2557712cd6d9d27cf73ff077446d0fb7a1b3d956ec1d4b22bb40c647e1aec7&=&format=webp&quality=lossless&width=1920&height=226

anyone know why this is happening

[snow hazel]

What is the best way to get rid of harmonies? I tired doing mvsep but it didn't work.

Please help me

[dawn cairn]

guys is there any particular model or settings which u recommend for optimal use for vc changer?

for real time voice conversion

rtx 4060

[fading drum]

how do i know if a model is overfitting or underfitting when cloning, using tensorboard with rvc1006nvidia

[brittle wing]

Is there any upcoming pitch extraction algorithm?

Or they'll stay the same forever?

A more realistic one

[analog obsidian]

http://docs.aihub.wtf/rvc/resources/epochs--tensorboard/#--monitoring

undertrained models do not sound like the original voice and have low quality
overtrained models do sound like the voice but since the model lost their ability to generalize, they will struggle to generate new stuff and their quality is also degraded, (most of the time overtrained models sound robotic or glitched)

most common overtrain happens when you reach the lowest point in your graph, then its gonna raise forever

[stiff goblet]

Titan or Original pretrain for multi languages?

[late beacon]

Where and how do I download RVC? I watched some youtube videos but can't seem to find where to download it or how

[glad zealot]

yes

[astral perch]

What is Titan pertrained model and should I add it to my local RVC

[glacial pollen]

There's few ways. In essence, there is original ( official ) RVC available on github along with it's forks ( custom variants - some are just qol-changes based, some have some bonuses and so on

Easiest approach would be probs going for applio or my fork for easier installation

[late beacon]

Do you by any chance have the applio or fork link?

[rare goblet]

Ayo? @late beacon level 1 !!!

[glacial pollen]

@late beacon Gimme a bit of time. Gonna check what's my fork missing compared to the newest mainline ( official ) and link it
~ Was away for some time

[pastel oak]

Local if your gpu is good (for inferencing + creating model)
https://docs.aihub.wtf/rvc/local/mainline/

on cloud if no gpu:
https://docs.aihub.wtf/rvc/cloud/applio-colab/

if you want to train a voice model on cloud
https://docs.aihub.wtf/rvc/cloud/rvc-disconnected/

actually if you download local

i have a more recent version prepackaged if you dont want to manually install. do you have nvidia or amd

[late beacon]

NVIDIA

[pastel oak]

https://huggingface.co/Shadicti/RVCLatest/blob/main/RVCNvidiaLatest.zip

[late beacon]

Btw

[pastel oak]

this is recent version local

[late beacon]

What does voice model training mean?

[pastel oak]

RVC has essentially 3 core things

• inferencing (ai covers, you put an audio in there and it convrts with whatever voice model you chose)
• voice model training (you create your own voice model with a dataset)
• realtime voice changer

[late beacon]

Alright

[tribal saffron]

What is the best collab for android

[proper shale]

https://colab.research.google.com/drive/1mHKTGH5e3SAyDSBss1KtiYRbDdQzwSMs - Ilaria RVC Mainline, easy to use and pretty nice UI

[tribal saffron]

Can you train the models?

[rare goblet]

Ayo? @tribal saffron level 1 !!!

[proper shale]

yeah, you can

though i'd recommend doing that on a pc instead

[tribal saffron]

Ight

[astral perch]

Wait this is the latest RVC local file?

[late beacon]

I've downloaded it, what do I need to start to use it?

[pastel oak]

its from april 24, 2024

the one on github is from like 2023 so atm thats the latest prepackaged yea

[pastel oak]

go-web.bat to start

read up on everything else here https://docs.aihub.wtf/rvc/local/mainline/

[late beacon]

Thx

[astral perch]

Their github has one from last week

[odd shale]

It will work on Paperspace?

[pastel oak]

Oh thats amazing, I didnt know they finally uploaded it

Do you got a link?

[astral perch]

Im guessing that the on github already has the titan pre trained installed?

https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/releases

[pastel oak]

I am not sure how paperspace works, but if the older one used before works, then it should work too

[pastel oak]

Im not sure I can see it, the latest one it shows is RVC1006 which is the one from october 2023

i might be github noob

[astral perch]

Wait it says the same thing but it also says last week

[pastel oak]

The note saying last week is most likely because it was updated, and its like a note lol

since theres been new uploads on the repo

[astral perch]

Oh I see

So the one that you send is the latest one then im guessing

[pastel oak]

Yes

[astral perch]

Alright thanks!

[pastel oak]

You can technically manually install the latest version, but I prepackaged it for those that dont know how to

[astral perch]

Is that the one with the newest pre trained models?

[pastel oak]

Well - in this case, someone else did. I prepcakged the AMD version

[odd shale]

Altho i would have to download the entire Nvidia package and throw it on a GPU environment, right?

[pastel oak]

No it doesnt have them included, but I could do that and upload

However, you can also manually download the newest pretrains. theyre on this server

[astral perch]

Can you please do that, it would help a lot!

Oh

Where?

[pastel oak]

For titan:
#1236445948834611260 message

[odd shale]

Ig you could make a prepackaged version with Titan and KLM pretrains.

[pastel oak]

Most likely yes. I am not sure I understood the question fully

[pastel oak]

Good shout. Are those the two most used from the new ones?

[odd shale]

Well, Titan and KLM (mostly the 32k versions) are the best ones.

[pastel oak]

Ok ill do that

I think it would be great to include the links on the docs guide for rvc mainline

[odd shale]

Does the most latest RVC versions have crepe for pitch extraction?

[pastel oak]

Since in this version, you can also use the merging tab

[pastel oak]

It has everything that the previous one also had

Let me run it up quickly

[astral perch]

wait @pastel oak are you gonna make a prepackaged version?

[pastel oak]

Oh, it doesnt

[pastel oak]

Leo suggested to upload a new prepackaged version with the Titan and KLM pretrains included

So yea

[odd shale]

And there's no way to add mangio crepe? (the same crepe Applio has)

[astral perch]

Okay then I'll just wait until you have it all set

[pastel oak]

It would be best to ask in RVC Developers server. This is beyond me

[upper granite]

hello guys, I'm new to this ,I'm using a radeon rx 5700xt,ryzen 5 3600 computer ,is it possible for me to create my own voice model/

if so is there any simple tutorial? cause most thing I found so far is complicate as hell, even if I manage to set those up there are many error and things that cause mental distress to my soul

[odd shale]

Ehh.. We don't suggest you watching tutorials on YT since these tend to be outdated.

Instead, ask my buddy @pastel oak for help.

[pastel oak]

afaik you can not create voice models with AMD gpu unless youre on Linux with ROCM

[upper granite]

ok tks brother

[rare goblet]

Ayo? @upper granite level 1 !!!

[pastel oak]

you have a few options :

Ilaria RVC Mainline:
https://colab.research.google.com/drive/1mHKTGH5e3SAyDSBss1KtiYRbDdQzwSMs

RVC Disconnected (with quick tutorial aswell)
https://docs.aihub.wtf/rvc/cloud/rvc-disconnected/

[late beacon]

Is there a way to use RVC in real time?

[pastel oak]

go-realtime-gui.bat

[upper granite]

tks you brother

[pastel oak]

I am so cooked

[late beacon]

I tried that, but it somehow won't work in realtime, like I can't seem to be able to use the voices in a call

[pastel oak]

Check out my guide on how to set it up. You probably havent installed a virtual cable

https://rentry.co/rvcrealtimeguide

[upper granite]

I just want to create FIlian cover of Filthy Frank songs and end up 2 days wandering youtube now

[late beacon]

Thx

[upper granite]

just download some ram my dude

and memory

=))

[crystal cairn]

hi guys im new to this type of stuff. Which program i sht ebest to make some tts with voice model from this discord? Are there any tutorials maybe. i was trying to find it myself but its pretty confusing

[astral perch]

Oh damn

[crystal cairn]

thanks ill try it oout

[rough vale]

have anybody tried using zluda with ilaria rvc?

[dusk barn]

im getting an error in the part where im supposed to paste the link

json decode error

[azure marsh]

Suggestions for @dusk barn

☁️ Google Colabs

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

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[dusk barn]

wheres the new one

can i use easygui cause its the one im using

is that updated

ok thx

[rare goblet]

Ayo? @dusk barn level 1 !!!

[upper granite]

but I did put the zip file in

[crystal cairn]

i tried to use it but the output never sound the same as the model i choose

[rare goblet]

Ayo? @crystal cairn level 1 !!!

[crystal cairn]

how do i make it sound the same

for example i tried to make mr boss from skmiling friends model found on this discord server

tried it still doesnt work

why do i have to pick a tts model for that i kinda dont get it

okay okay

but i see that evertytime i chose different tts model the output is different in quality

okay

thankls

do u know how to change language in applio gui

bcuz it translates to my native language poorlyu

thanks

i think i got everything now the output is satisfying at least

thanks bro

bw does applio use my gpu or cpu

[rare goblet]

Ayo? @crystal cairn level 2 !!!

[crystal cairn]

bcuz it runs pretty smoothly with decoding

okay, ye ai have an nvidia

why doesnt it use amd?

okay

[analog obsidian]

@shell shoal http://docs.aihub.wtf/

here it explains all

[shell shoal]

thanks!

[rare goblet]

Ayo? @shell shoal level 1 !!!

[cyan maple]

Need some guru wisdom : P
Training a model, mel and kl are still dropping, but g/total is plateauing, continue to train or no?

[bold magnet]

hey

can somebody help me with the file index not generating?

i already trained the model

but the index doesn't generate

ight

thx

[proper shale]

keep going

[bold magnet]

applio

yeah i already did that

ight

[gusty nimbus]

hi there i’ve tried to update the rvc_hfv2 hugging face collab, and i keep getting a runtime error when trying to upload it?

this is what it brings up each time

[proper shale]

I'm uh, pretty sure that's outdated atp

If you wanna keep using Huggingface there's always https://huggingface.co/spaces/TheStinger/Ilaria_RVC

[gusty nimbus]

i’m so confused, when you load it up it says it’s using an old version of gradio so you need to update it, you update it and it completely breaks it, never seen it before

thankyou i’ll try that one out!

[proper shale]

alr, let me know how it goes!

[gusty nimbus]

it’s a little bit different from the other one, when converting the acapella using a model, how do you know on the inference or am i looking at the wrong bit?

[proper shale]

You just have to set the model in the same tab as you download it

If I understood you correctly

[gusty nimbus]

yeah you did i put it in the part of model loader, but on the inference bit it doesn’t give you a choice, is that the right part?

[proper shale]

yeah

it has the model selected already

[gusty nimbus]

perfect, i’ll let you know if once it’s downloaded if it did it mate, but thankyou!

[low shard]

zerogpu 🔥

[proper shale]

fr

[low shard]

its gonna be a 1v1 between ilaria rvc on A100 and applio on a T4 hf spaces lol

[brittle wing]

guys

i installed this

MMVCServerSIO_win_onnxgpu-cuda_v.1.5.3.18a

how do i add #1175430844685484042

the one i downloaded, idk how to add

it asks for json on website but the one i have has pth file

[proper shale]

probably selected the wrong model architecture, switch back to RVC and try uploading the .pth again

[tall depot]

It doesn't matter 'cause I'm always the first

[low shard]

Huh?

[glossy coral]

guys , i didnt use any model for so long , what colab should i use? the one i got is outdated

[proper shale]

https://colab.research.google.com/drive/1mHKTGH5e3SAyDSBss1KtiYRbDdQzwSMs - Ilaria RVC Mainline, pretty easy to use

https://docs.aihub.wtf/rvc/cloud/applio-colab - guide for Applio colab

I'd suggest trying both and see what works best for you

[glossy coral]

thx!

[brittle wing]

could u help me better? maybe through dms

idk much about this, not really technically qualified

[proper shale]

if you can provide more detail that would help a ton, tbh

i don't really do helping on DMs so uh, yeah

[proper shale]

Yw, let me know how that goes

[brittle wing]

alright I'll lyk by tomorrow since I'm omw to school so can't access the laptop

[rare goblet]

Ayo? @brittle wing level 1 !!!

[proper shale]

Alrighty

[brittle wing]

what all do u need?

[proper shale]

Screenshots only, probably

[brittle wing]

like the laptop specifications of what I downloaded?

or*

[proper shale]

Don't think that's needed, unless you downloaded the wrong version... but we'll find that out later

[brittle wing]

kk, I'll see if I can access it tonight, otherwise tomorrow morning it is

[proper shale]

alrighty! ping me once you're there

[vast berry]

i have a model i used a while back for rvc and i ran it on gradio,, cant seem to run any google drive links that give me a gradio public link so i can use the model

i also cannot upload my .zip which includes a .pth and .index to any gradio link i can get

[azure marsh]

Suggestions for @vast berry

☁️ Google Colabs

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

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[rain prism]

i need help

@proper shale

[proper shale]

switch browsers, id say use chrome

[wheat plover]

is there any way to download the tensor board graphics in RVC Disconnected?

[proper shale]

if that isnt a browser issue idk what else to do

[proper shale]

automatically in your drive

once you export the model

[rain prism]

i am tho

@proper shale btw im using the colab

[proper shale]

yeah

[wheat plover]

but if I detect overtraining, how can I export an older version of the model?

[rain prism]

@proper shale what do i do?

[proper shale]

i believe you can probably do that by knowing the steps per epoch and just exporting the lowest value of the thing

you can set epoch and step count in export model

and then just enable "save weights only"

[proper shale]

idk tbh

[molten root]

hi, where can i ask question about TensorBoard ?

[proper shale]

you can do this here

[wheat plover]

thank you, I'll test it out later.

[rain prism]

btw if i wanna rec the audio do i make it on client or server? @proper shale

[proper shale]

both can record, anyway

[molten root]

ok, i followed the ai hub doc to dl tensorBoard, and it opened my console window. but then, it says that it will automatically open a browser tab with tensor board and if not, i have to copy paste the localhost link with the port number.

but, the thing is, it didn't opened anything and copy paste my localhost link doesn't do anything

[wheat plover]

also, is there any way to modify the default values of Applio's batch inference files?
I have to change the folder with the audios and the export folder every damn time I open it to test a model

[proper shale]

just click on the REC thing in there

[rain prism]

no like record voice only @proper shale

[proper shale]

just click on the REC thing in there

[rain prism]

where

@proper shale

[proper shale]

switch back to client audio

[rain prism]

everytime i rec it doesnt work

see its 0 seconds

but i recorded and talked

[proper shale]

you haven't even selected the model, nor the inputs/outputs

select those

[rain prism]

wdym

[proper shale]

i got no time rn

mb

but

select the stuff

[rain prism]

oh

i kept clicking it tho

it only worked now

💀

ok thx

the recordings are still 0 seconds tho :/

[wheat plover]

figured it out, modified two lines of code inside \tabs\inference\inference.py
line 403 value=os.path.join(now_dir, "assets", "audios", "inputDemo"),
line 412 value=os.path.join(now_dir, "assets", "audios", "outputDemo"),
then I created a folder inside assets\audio\ for inputDemo and outputDemo

[rose arrow]

why does it sound muffled sometimes?

[rain prism]

@proper shale the ai voice is mocking me and keeps repeating the same word 😭

[rare goblet]

Ayo? @rain prism level 4 !!!

[proper shale]

honestly... im not even sure why this is happening

id reinstall tensorboard

[proper shale]

yeah it comes in a delay

[rain prism]

no like it keeps repeating the same word forever

[proper shale]

ah

[rain prism]

like if i said hi

the ai voice would keep saying hi forever

[proper shale]

enable echo + sup2, and increase s thresh till its over

[rain prism]

uhhhh

[proper shale]

in what?

[rain prism]

my s thresh is 0.0003

[rose arrow]

the ai conversion

[rare goblet]

Ayo? @rose arrow level 1 !!!

[wheat plover]

yes, thank you.
skip_models and manual_save did exactly that

[rose arrow]

does it have something to do with my mic, the model, or the client settings?

[proper shale]

might be the mic or the model tbh

sometimes when the mic is at a distance it gets pronunciation wrong, thats what happened to me at least

once i got a headset that was mitigated

[rose arrow]

ok thanks, i'll try using a diff mic

[amber perch]

How can i use voice models from mobile

[proven hill]

wheres a tutorial for ilaria rvc

[proper shale]

b-but you made it

[proven hill]

sorry i menta llaria rvc

[proper shale]

go in model loader, load model
go to inference tab, upload audio, convert

if it doesn't sound right pitch wise > settings, mess with pitch level

[proven hill]

thank you youre my hero

[proper shale]

no youre my hero

[proven hill]

youre my egirl voice model🗣️

[molten root]

since it already installed everything on my pc and idk where it did it, when i re dl and want open tensorBoard, it says that : tensorboard' is not recognized as an internal or external command
or external command, an executable program or a command file.

[rare goblet]

Ayo? @molten root level 3 !!!

[proper shale]

you're my mj voice model 🗣️ :

[proper shale]

it's all in "tensorboard_venv" folder, so just delete that and run the bat again

also, that "TB is not recognized" issue, maybe you got spaces in your folder? or something like that

[pallid wasp]

What is the best to use if i want to make good ai covers

[molten root]

still doesn't work and no i've no weird characters in my path

[crimson umbra]

how do I make people hear my voice changer? (RVC)

[brittle wing]

can someone link me to docs

[proper shale]

you need a virtual audio cable

https://rentry.org/VoiceChangerGuide#virtual-audio-cables-mandatory

[proper shale]

https://docs.aihub.wtf

[brittle wing]

thx

[proper shale]

yw! :)

[brittle wing]

wehats the name of the thing in audacity to select just audio clips i forgot

[proper shale]

[brittle wing]

ive cleaned the audio w noise gates and stuff

[proper shale]

labels?

[brittle wing]

yeah how do i do it i forgot

[pallid wasp]

How do i make good ai covers

[rare goblet]

Ayo? @pallid wasp level 3 !!!

[brittle wing]

it looks like this

[brittle wing]

lowkey when u record the vocals urself its clearer thats how i do it now

[sullen aurora]

Hello guys, i have a serious problem with my RVC and i want to know if it's normal or not : When i speak the time it takes to translate is really too huge. And the models don't sounds like they should be. Can you help me please ?

[brittle wing]

change the chunk

the lower ur chunk is the more it lags in games tho

[proper shale]

i never used labels tbh

[sullen aurora]

I have 960 as chunk

[analog obsidian]

#1203830360878743612 message

[proper shale]

ah good one lyery

ty

[brittle wing]

is there a way to stop it from selecting these little noises

[analog obsidian]

delete them

[brittle wing]

or do i manually delete them

its a 1 hour recording

[analog obsidian]

1 hour for a rvc model is a bit overkill xd

[brittle wing]

alot of it is just silence tho

Lol

[analog obsidian]

lols idk i just remove those small noise manually if noise gate didnt remove them

[brittle wing]

i basically just recorded my friend thru a discord call so theres like random noises

[sullen aurora]

maybe it's because i don't have an rtx card ?

[rare goblet]

Ayo? @sullen aurora level 1 !!!

[brittle wing]

best use of ai is making your friends say gay shit

[analog obsidian]

rmvpe can handle random noises quite well

[brittle wing]

yea but its better if its just like cleaner u kno

[analog obsidian]

tho is safer to remove them

[brittle wing]

should i dele something that sounds like this or is it fine

[analog obsidian]

distortion + voice getting cut

uhh rvc doesnt like that at all

rx can fix distortion i suppose

audacity should too tho not sure how to do it

avoid very loud audios

volume in a dataset has to be consistent

otherwise that volume discrepancy is gonna be present in the model

and voice getting cut is gonna cause collapse

[brittle wing]

should i just normaliza

for the volume

most of it is like same volume

[analog obsidian]

you should always normalize anyways
to -3db

[pallid wasp]

Can someone send me the link to the newest RVC

[analog obsidian]

https://huggingface.co/lj1995/VoiceConversionWebUI/resolve/main/RVC1006Nvidia.7z latest mainline

[pallid wasp]

Thanks

[analog obsidian]

no prob, the version i sent is for nvidia only

[grave river]

I’m not sure if this is the right chat, but how do I make or use a voice mod?

[molten root]

when i try this command line, i get an error, i don't understand what's wrong

[glacial pollen]

@molten root Using Mangio's ver of rvc effectively disqualifies you from tensorboard

[rare goblet]

Ayo? @glacial pollen level 5 !!!

[glacial pollen]

There's absolutely no point in checking board for mangio produced tensorf data due to steps misalignment etc etc

[proper shale]

https://rentry.org/VoiceChangerGuide should help

if you wanna make your own voice model for Okada, https://docs.aihub.wtf to know the basics

[molten root]

but i think i've something,
I manually installed tensorboard en tenserflow with console window with pip install
i gave only the logs directory to tensorboard command ( see picture)
then it worked, it opened a windows with some boards that refresh in realtime, i suppose it's my current training...🤔

[glacial pollen]

It really doesn't matter

Mangio rvc is not suitable for " precise " trainings

Effectively, usage of tensorboard in your case is a waste of your time

[molten root]

ok

thks for reply

[glacial pollen]

Well, I am saying it to save your time

and frustrations

But whatever the case is. Good luck

[molten root]

so much effort for nothing then 🥲

[glacial pollen]

Well, you can always train again

using different rvc / repo

You want most vanilla rvc with mangio essentials + up to date + some neat thingies?
I have a solution for that

There;
https://github.com/codename0og/codename-rvc-fork

[astral perch]

Should i put the titan pre trained models in pretrained v2 folder?

[rare goblet]

Ayo? @astral perch level 10 !!!

[glacial pollen]

It does not have faulty " auto syncing " og mangio does which causes issues I mentioned

as for tensorboard

aside of having done:

pip install tensorboard

you use that batch

just paste in the path to your model's log folder ( or rather, a folder containing your tfevents file - per model )

naturally you can paste the direct path to your at-the-time trained model ( it's log ) and have it auto refresh, that will do the trick

And if you're curious on what's exactly the issue with mangio's fork, it is:

The way the fork's handling the training is, it tries to calculate and predict steps count per model ( which can't be 100% accurately predicted ) and that is causing, in the long run, steps misalignment. That means, whenever you check the tensorboard later on, for the model, it is not representing the actual / real training points ( in this case, saved epochs )
It is unfortunately hardcoded in the fork so that's that. No other way around but using a different rvc release

[molten root]

oki thks

[glacial pollen]

Np, hope it goes well on your side

[analog obsidian]

i get this error when i run nvidia_standard.bat
RVC ( NVIDIA-RTX30XX )- CUDA_118.bat does not fix this and same error is seen

[glacial pollen]

You have rtx 30 series I understand @analog obsidian

[analog obsidian]

4060

[glacial pollen]

Have you used any other rvc before? if so, what release was it

[analog obsidian]

both mainline, applio work fine no issues

[glacial pollen]

mainline or something else

That must mean both of those moved their torch compatibility up or you're using em in venv

[analog obsidian]

deleting all python packages + python itself causes the same error

[glacial pollen]

If you wanna use my fork, you need torch 2.0.1

That's the list:

joblib==1.3.2
numba==0.56.4
numpy==1.23.5
scipy==1.11.1
librosa==0.9.1
llvmlite==0.39.0
fairseq==0.12.2
faiss-cpu==1.7.3

required for compatibility reasons

gradio==3.34.0
gradio_client==0.2.9
urllib3==1.26.12
pydantic==1.10.11
pydantic_core==2.1.2
typing_extensions==4.7.1

required for compatibility reasons

Cython==3.0.4
pydub==0.25.1
soundfile==0.12.1
ffmpeg-python==0.2.0
Jinja2==3.1.2
json5==0.9.14
Markdown==3.4.4
matplotlib==3.7.2
matplotlib-inline==0.1.6
praat-parselmouth==0.4.2
Pillow==10.0.0
resampy==0.4.2
scikit-learn==1.3.0
tensorboard==2.15.0
tqdm==4.65.0
tornado== 6.3.3
Werkzeug==3.0.0
uc-micro-py==1.0.2
sympy==1.12
tabulate==0.9.0
PyYAML==6.0.1
pyasn1==0.5.0
pyasn1-modules==0.3.0
fsspec==2023.10.0
absl-py==1.4.0
audioread==3.0.1
uvicorn==0.23.2
colorama==0.4.6
pyworld==0.3.2
httpx==0.23.0
onnxruntime==1.16.3
onnxruntime-directml==1.16.3
onnxruntime-gpu==1.16.3
torchcrepe==0.0.20
fastapi==0.88.0
torchfcpe
ffmpy==0.3.1
python-dotenv==1.0.0
av==10.0.0
sox==1.4.1
soxr==0.3.7

and confirmed / recommended python is 3.10.6

[analog obsidian]

if i downgrade torch is gonna work?

[analog obsidian]

yes i even tried that and still had the same error

i managed to fix it by manually downgrading torch last time

but training didnt worked and fork didnt had ffmpeg

inference worked tho

[soft tiger]

could someone help me with this

[glacial pollen]

All I can say is that my fork assumes you're having a clean ( other custom rvc forks' requirements wise ) environment ready

torch 2.0.1 is a must

anything above will fail

[soft tiger]

@analog obsidian 🙏

[analog obsidian]

i mean i deleted every trace of python and didnt work

so idk

[glacial pollen]

do:

pip show torch

[analog obsidian]

run install.bat
then run-applio.bat

[glacial pollen]

and that is your issue

you gotta have 2.0.1

as for cuda, whatever is supported for 4090

[analog obsidian]

gonna try dowgrading it 2.0.1

[glacial pollen]

I have 30 series so I am following the cu 117 / 118 path

as for 20xx and 40xx series, it should be fine with just standard non guided torch installation

meaning:

pip install torch torchvision torchaudio

[analog obsidian]

oh

[glacial pollen]

not sure why some folks are getting higher ver of torch

[analog obsidian]

so first step for 40xx is first downgrading to 2.0.1 then do that?

[glacial pollen]

in general, you gotta have 2.0.1

that's how it is for my fork and used to be for mainline, at least before

maybe there were some major changes

just as switching to python 3.11 or 3.12

and so, torch changes

hard to tell as I was away from rvc for a bit now

Will have to verify it but, for now, downgrade to 2.0.1

[analog obsidian]

can you tell me how to do it? i did last time but probably i did it wrong

[glacial pollen]

Gimme a second, gotta check what applio has for requirements

[analog obsidian]

K, i found pytorch site telling how to downgrade it

[glacial pollen]

Ah yea

applio has it at 2.1.1

[analog obsidian]

that explains why i have 2.1.2 probably

any should work?

[glacial pollen]

Hmmmm.. try 11.8

I for sure know ampere, so, 30xx cards require 117 so, 1.17

so if 118 doesn't do the justice for you, try 11.7

[analog obsidian]

ok its downgrading now, after that i should run nvidia_standard again?

[glacial pollen]

haven't had 4090 testers so, can't say for sure

[analog obsidian]

nvm u told me to install it manually

[glacial pollen]

I'd actually purge all you have related to pip

[analog obsidian]

okei i'll delete pip stuff

[glacial pollen]

just do the

pip list

and then

pip freeze list.txt

and from wherever you save the list, open up the cmd and remove the packages using it ( the list

pip uninstall -r ur_list.txt -y

[proven hill]

you have a fork?

[glacial pollen]

Yea

for a while now

[proven hill]

send gh

[glacial pollen]

mh? gh?

[analog obsidian]

where is this saved? i run terminal in desktop and cant see a txt file

[proven hill]

github

[glacial pollen]

oh, lol
https://github.com/codename0og/codename-rvc-fork

[glacial pollen]

check in what location you had ur commandline when you did the freeze

by default it'll be your user's root

C:\Users\xxxxx\here

[analog obsidian]

weird, is not here lol

[glacial pollen]

weird weird.
Run the cmd from any folder then

and attempt it there

like so

[analog obsidian]

[glacial pollen]

used the freeze properly?

[analog obsidian]

idk i just used
pip freeze list.txt

after
pip list

[glacial pollen]

huh, could it be another w11 bullshittery? 🤔

[analog obsidian]

[glacial pollen]

maybe try this:

pip freeze | xargs pip uninstall -y

[proven hill]

"Proper fusion for 32k isn't supported" why?

[glacial pollen]

Because that's how rvc was made

[rare goblet]

Ayo? @glacial pollen level 6 !!!

[glacial pollen]

Not my gimmick really

but directly related to the code

[proven hill]

but you can fuse 32k models

[wispy lodge]

Pip freeze does not output to file. You have to redirect the output like pip freeze > requirements.txt

[glacial pollen]

You can but it's " reliability " leaves a lot of unknowns

might look and souind fine on the outside

[wispy lodge]

And don't butcher global environment. Use venv

[glacial pollen]

but we'll never know how it handles network distribution in 32k models

[analog obsidian]

translation: xargs command not found

[proven hill]

never had this kind of problems tbf

[glacial pollen]

That's why I mention that it might work fine, but officially it is not supported

because sure, you have the 40 and 48 handling on 32k but it is not 100% accurate

[analog obsidian]

oh

[glacial pollen]

They have different distributions

oh, you didn't add the ext ? @analog obsidian 🤔

[analog obsidian]

i just copied and pasted those commands

[glacial pollen]

ye, and in this case ^ it should properly output you the file

[elder whale]

guys why my discord dont see chenger micro?

[analog obsidian]

isnt there

[glacial pollen]

try to open the file directly with the address bar

to confirm if it's legit not there or invisible for whatever heck ( win11 is bs ) reason

[elder whale]

its to me?

[glacial pollen]

no

[wispy lodge]

Just use pip freeze > list.txt...

[elder whale]

u can help me pls

[glacial pollen]

That's what they did my man

I suspect w11 being trash

[analog obsidian]

it

worked

lmao

[glacial pollen]

hm

[analog obsidian]

[glacial pollen]

so we going powershell now lol

[wispy lodge]

I don't see the arrow neither in your command nor their

[glacial pollen]

One more reason to evade w11

oh, lmao nevermind. didn't notice I forgot to slap it in there

Hawk eye

[analog obsidian]

xD

that took comically long

[elder whale]

guys?:(

[rare goblet]

Ayo? @elder whale level 1 !!!

[glacial pollen]

must be tired ig, either way

get the packages right and all should be working with no issues

[analog obsidian]

install virtual audio cable and restart pc after that

[elder whale]

I do this

programm work

but ds dont see

[analog obsidian]

i see, so run nvidia_standard?

[elder whale]

i am put virtual cable and its dont work

but programm work

[glacial pollen]

yea

[elder whale]

what i need to do

[elder whale]

bro?

[analog obsidian]

ok doing it rn

[glacial pollen]

if I get you right

it ain't there?

[violet heron]

What is your inputs/output on Okada

Send a screenshot of your settings on it

[analog obsidian]

i think he means he can't see vac on discord

[glacial pollen]

must mean that A) vac isn't properly installed / configured
or B ) the output in the voice changer isn't set to line

[analog obsidian]

probably installed wrong version?

[glacial pollen]

also, it must be exactly:
https://vac.muzychenko.net/en/

vb-audio nah

[violet heron]

Even then it might be good to check it so the person doesn’t complain about it being laggy

[violet heron]

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

[analog obsidian]

true

[violet heron]

Person might accidentally click the trial version instead of lite

So provide the direct download link instead

[analog obsidian]

codename i think i didnt get the error this time

nothing red showed

[glacial pollen]

means you're good to go

[glacial pollen]

I, with all respect, expect people who fiddle with AI to be competent enough to read : )

[analog obsidian]

damn so weird applio caused this issue

[proven hill]

https://tenor.com/view/wrong-drumpf-trump-stupid-gif-6220235

[violet heron]

have you seen general?

[glacial pollen]

Because they moved to higher ver of some packages

[glacial pollen]

I've been a member since the only fork available and commonly known was mangio + maintained only by kalo and mangio

so I naturally did seen enough

well enough to bother with indepth explanations

[analog obsidian]

[glacial pollen]

so that'd be cuda version

hold up

try the 118 version

[analog obsidian]

ok

[glacial pollen]

and if that won't work, Imma check up on what's favored by 40xx

[analog obsidian]

damn i can't choose cuda version if i want to install this with pip lol
i dont have conda

[glacial pollen]

so, torch 2.0.1 should click with 117 and 118

yes you can

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

[analog obsidian]

yeah im so stupid lfmao

[glacial pollen]

dw dw

And, for whatever reason, it still won't work for you, here's my list

compatible with my fork and my voice changer's fork ( keep in mind, I have 3060 )

[analog obsidian]

i was about to say it didnt work lol

[glacial pollen]

🤔

[analog obsidian]

without cu188 link it does the same thing