✨│ai-help

[tranquil cliff]

Hi

[burnt thorn]

Are these the real .pth files?

What should be the normal size of a .pth file? Of about 250 epochs

[modern oxide]

no

your final model is in folder called weights

in mangio fork it's literally inside the root folder in other forks it's inside a folder called assets

[old hawk]

i rlly need help with this

i cant fix it at all

or seem to find the issue

[modern oxide]

in colab?

[ripe scaffold]

yo can anyone help me rq

[old hawk]

i just open rvc and added vocals and model where it said

[rare goblet]

Ayo? @old hawk level 1 !!!

[old hawk]

i cant coonvert

[burnt thorn]

Then what is this?

[old hawk]

conver*

my keyboard is broken

[modern oxide]

it's unconverted models

[modern oxide]

locally or in colab? i can't help you if i don't know what you use

[burnt thorn]

But on colab it said 250 epoch model saved in d_233333.pth

[rare goblet]

Ayo? @burnt thorn level 3 !!!

[modern oxide]

the name of the rvc you use

[old hawk]

let me check

[modern oxide]

yeah it's just logs, don't mind it, just take the model from weights folder

[old hawk]

can you tell me how to check

which one im using

[burnt thorn]

Well fck cus i selected 250 epochs and there is only the one with 100 epochs

[modern oxide]

you read the name of the rvc

[old hawk]

i jus downloaded it from a yt tut

oh

ok

[modern oxide]

has it finished?

[burnt thorn]

Yeah

Its said done

[modern oxide]

can you send the screenshot of weights folder and your model's logs folder

[burnt thorn]

Well i deleted it after it said done

Nvm then i will just retrain

[old hawk]

i legit cant find it

[modern oxide]

do you use rvc v2 disconncted?

[modern oxide]

ah

you use the old one

[old hawk]

oh

https://github.com/Tiger14n/RVC-GUI/releases/ i downloaded it from here

[modern oxide]

https://huggingface.co/MangioRVC/Mangio-RVC-Huggingface/resolve/main/Mangio-RVC-v23.7.0_INFER.7z?download=true download this

[old hawk]

ok

will that possibly fix the issue

[modern oxide]

possibly yes

[old hawk]

ok tthanks

ill download it

[burnt thorn]

I used this one

[modern oxide]

nice

dunno why it didn't save the final model tho

[burnt thorn]

Becaus i am dumb and my google drive storage was full

[old hawk]

what file do i need to run

[modern oxide]

go-web.bat

[old hawk]

oohh nvm

[copper pebble]

use rvc disconnected

[old hawk]

is there any tutorial vid or doc on how to use the site

[rare goblet]

Ayo? @old hawk level 2 !!!

[modern oxide]

nah easygui is better because it automatically saves everything to you drive

[modern oxide]

#1159513888199540817

[old hawk]

thx

[copper pebble]

it has banned codes

[modern oxide]

what's codea

[copper pebble]

codes**

im in school rn

[modern oxide]

ah

i dunno never had problem with that

[silver raft]

https://garticphone.com/ru/?c=1250ad57e4 - ребята давайте поиграем в гартик фон

[rare goblet]

Ayo? @silver raft level 1 !!!

[modern oxide]

чувак это канал для помощи а не для поиска игр, иди в #🌏│русский

[ocean furnace]

Guys, a quick question, any other alternatives for rvc which can be operated or google colab or something?

[ocean furnace]

If any of you really just got your hands on some other handy tools, it'll be cool if you can share

[modern oxide]

#1159949278270193734

[modern oxide]

#1163883017446629508 message

[ocean furnace]

Is there a tutorial for easygui one?

[modern oxide]

good question

hmm

i don't see any

[viscid moss]

that colab is banned from Free Colab 💀

[modern oxide]

it's not

i use it

i mean the ui is banned for free users but the colab itself works

[viscid moss]

ah, the another, Without GUI

[ocean furnace]

Mind to guide me for a bit?

[rare goblet]

Ayo? @ocean furnace level 1 !!!

[modern oxide]

well you just link it to ur drive, load in the dataset and press every cell step by step that's all

[ocean furnace]

Didn't really do that stuff when was working at grading tbh, gotta find a way to do the same in mobile

[modern oxide]

i guess it's kinda the same, you'll just have to find a way to clean your dataset in your phone

i guess using another colab woth uvr in it

[brittle wing]

What is a good epoch to use in Applio for tranining rvc models? I have mine set to 1K, but idk if that is too high or too low. I mean, the max it can go is 9999, so I would assume its fine

[rare goblet]

Ayo? @brittle wing level 1 !!!

[brittle wing]

on 15 mins worth of datasets overall

[modern oxide]

for such long dataset usually 300-400 is enough

[obtuse stream]

whats the different rmvpe or rmvpe_gpu? wich is better?

[modern oxide]

they're the same, the gpu one just uses gpu to do it faster

[toxic snow]

how do i upload my own ai model to wordband?

/colab

[quiet slate]

Hi

I have a macbook air M1 and when i tried to run the voice changer it doesn’t work as fast as it should be it works but its too bad to even run so is there any option to fix it

Like the voice just cuts off when I’m trying to talk into it or it comes out be like a robot

[proper shale]

use Dio as your f0

[misty cosmos]

im using rc on runpod with a rtx 4090 and its wort it way more than colab... look at the time hwo fast it is.....

[proper shale]

how big is ur dataset

[misty cosmos]

i took me so long to find out how to use it now i can adnits AMAZING

[misty cosmos]

very very small for testing, but still... its fast

[proper shale]

ah man

gotta test it out with the big stuff

and with 8 batch_size

[misty cosmos]

i have a dataset that is ig almost 2h should i try that later?

[proper shale]

Sure

[regal mirage]

Who Can Instala Things On Google Colab

Yes, Im New On This And I Don't Know?

[rare goblet]

Ayo? @regal mirage level 1 !!!

[proper shale]

Install what

[regal mirage]

AicoverGen mod by ardha27 and hina

[proper shale]

You just gotta run the steps

[copper pebble]

theres an epochs tab in pinned messages

[proper shale]

uhhh

that isnt quite accurate

[cyan prairie]

I followed the steps stated in tutorial. However, is it normal for the output of the audio to be loaded longer? Sometime it loaded until it state error. I failed to understand this new method than last time.

[rare goblet]

Ayo? @cyan prairie level 1 !!!

[cyan prairie]

Like this one here. I kept waiting tho.

[regal mirage]

What Can I Do If This Happens?

This

[modern oxide]

5-10 minutes is enough for a good model

[misty cosmos]

mine is 7 min like the testing one but i have the with 2h ... its old, back then when you still needed a lot of voices for a good model

[regal mirage]

@viscid moss

[rare goblet]

Ayo? @regal mirage level 2 !!!

[viscid moss]

no puedo leer nada 😅

[regal mirage]

Espera

[viscid moss]

Pero a ver.... cual descargaste?

Para saber desde el inicio todo el contexto

[regal mirage]

Descargue Rvc De La Página De Aihub.wtf

[proper shale]

I hate to be that type of guy (or gal) but I don't think you can run RVC if you have a weak PC

[regal mirage]

Oh Well

[proper shale]

Yeah sorry :(

You can use Colab though

[regal mirage]

I Gonna Try

[viscid moss]

follow this #1159513888199540817 and this #1159949278270193734

[regal mirage]

I Watch A Spanish Tutorial

[slate lagoon]

guys the player after conversion just stays stuck at 00:00

[viscid moss]

En los mensajes fijados de #🌍│español tenemos guias y cosas de ayuda

[empty dust]

Hello everybody

Yesterday I trained the model on 20 minutes of data, with batch size 8 and everything was fine. The epochs ran for 40-50 seconds. But today, having started training the model also on 20 minutes of data, the epochs started to last 2 minutes at first. And then, when I started the process again, the epoch time increased to 3 minutes

This is a very strange problem, I hope you can help me with it

PC Info

OS: Windows 10 22H2
CPU: Ryzen 5 5600X
GPU: NVIDIA RTX 3060Ti
RAM: 32 GB x 3600 Mhz

[empty dust]

Solved: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/840#issuecomment-1650953404

[half cove]

[glad zealot]

damn every 2 sec

i wish

[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

[twin shoal]

-rvc

[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

[hot lagoon]

in uvr5 there is vip modles? Wich vip models are the best?

[still spruce]

wtf does "index -1 out of bounds for axis with size 0" mean

[still crystal]

how do i fix this i ran the feature extraction

[rare goblet]

Ayo? @still crystal level 3 !!!

[viscid moss]

restarting the space

[still crystal]

restart runtime?

[viscid moss]

ya

[still crystal]

ok thank u

[still crystal]

now im getting an error here

[viscid moss]

try with this option

Disconnect and... idk

[still crystal]

ok

[still crystal]

it still doesnt work

[sonic garden]

-colab

[azure marsh]

All working colabs / spaces ☁️

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

Need some help? 🤔

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

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

[void yoke]

Quick question. More epochs and a bigger database helps with the training, right?

[misty cosmos]

the model gets better ( voice )

[half cove]

more epochs or a longer dataset wont always make it better

its more about having a good dataset

[native relic]

no spaces,just checked and changed my folder name

stil dosent work

[sand iris]

What's the error now

Also that rvc GUI thing isn't really used anymore

[gloomy dew]

@viscid moss in general it started like this

[rare goblet]

Ayo? @gloomy dew level 5 !!!

[gloomy dew]

and this is zoomed, check the notes

[viscid moss]

cuantos epochs?

No es necesario usar inglés

español está bien

[gloomy dew]

en el primer punto, 12:50pm estuvo a 260 epochs, y ahora mismo a 360 epochs. SIgue en esa recta

[viscid moss]

El mas bajo de todos fue en 260?

El dataset de cuanto era?

Yo creo que ya hasta ahi

no esta mejorando nada

esa gráfica va plana

https://tenor.com/view/giaochu01-anime-gif-4807421

like this xD

[gloomy dew]

Acabo de ver mas, y no, incluso de antes
Epoch: 208 [2023-11-10 12:14:28] llego a 29 smooth

[viscid moss]

yo creo con eso eso ya, no está mejorando

[gloomy dew]

esa era mi sospecha xd

dataset de 30 minutos, 300 mb en wav

Graciasss

[viscid moss]

Pues..... la gráfica dice que ya

de todas formas puedes probarlo

y reentrenar

[gloomy dew]

Como sabes es la primera vez que uso el tensorboard asi que aqui gracias bro xd

[gloomy dew]

a verdad. Al inicio le coloque 300 epochs limite y cuando terminó, dije voy a probar un poco mas a ver que tal, le puse nuevamente click a entrenar pero a 500 epochs. No tiene nada que ver con eso que ya no suba la grafica no?

[viscid moss]

no, eso no tiene nada que ver

[gloomy dew]

Listo. Queda. Cerrado. Comprendido

[real phoenix]

Hi, I'm new to using AI. I'm having trouble getting it to work properly. The voice comes out either very choppy or completely buggy. I've tried other models, but they didn't work either. I have an AMD RX 6600 GPU. Is there any specific configuration I need to make it work? I downloaded the MMVCServerSIO_win_onnxdirectML-cuda_v.1.5.3.16a

[gloomy dew]

garcias por tode

[gloomy dew]

have you tried with another song? maybe the vocals are not too good? not very well isolated or with noises

[real phoenix]

Claro, aqui está a tradução:

Yes, I have tried with some other models and my location is very quiet. The problem seems to be in the voice processing.

[viscid moss]

https://rentry.co/VoiceChangerGuide
Check this guide

[gloomy dew]

the very choppy buggy voice you're talking about... is about a voice you're training or a voice you're converting (like for a song)???

[viscid moss]

Remerber that for AMD u need to convert the model to onnx

[real phoenix]

I didn't know that. Does the guide have how to convert?

[rare goblet]

Ayo? @real phoenix level 1 !!!

[real phoenix]

just for using on discord for fun

[viscid moss]

go to the correct channel first #🔍│help-w-okada

[real phoenix]

okay thanks

[quartz sphinx]

@viscid moss como veo que hablas esp, me gustaria preguntarte acerca de esto, me ha estado desconectando constantemente al punto en el que no puedo durar ni 5 minutos (literalmente) conectado sin que me desconecte y em salga este mensaje, he intentado (denuevo, sin exagerar) como 15 veces y me sigue apareciendo este mensaje y desconectandome despues de un lapso de tiempo muy corto, que me recomiendas? y hola por cierto xD

[viscid moss]

emmmmm hablas de la version para colab no?

Pero del realtime o del otro?

[trim ice]

Question, is there a "best" tool/thingy for training RVC models with locally? My voice samples are directly ripped - no background noise or anything like that

[quartz sphinx]

Desgraciadamente di

[quartz sphinx]

Osea estoy usando el rvc en colab

Realtime

[viscid moss]

ese está baneado, hay una version modificada que está en los mensajes fijados en #🔍│help-w-okada

[worldly moth]

hi this is the firsty time in this server so idk if this is the correct channel but does the voice changer work on mac os?

[viscid moss]

pero tambien se desconecta, lo unico es usarlo localmente

[viscid moss]

Use RVC, mangio or applio

[trim ice]

What's the difference between the 3?

[rare goblet]

Ayo? @trim ice level 1 !!!

[static gyro]

where do i ask about UVR things?

[trim ice]

immediately blocks the level bot

[viscid moss]

go to #🔍│help-w-okada

[viscid moss]

Applio have TTS and is more visual attractive than Mangio

but it's still the same

[quartz sphinx]

Me imagino que sera este: #🔍│help-w-okada message (supongo que me tendre que hacer una multicuenta de google)

Como puedo usarlo localmente?

[viscid moss]

for train o infer

[rare goblet]

Ayo? @quartz sphinx level 1 !!!

[viscid moss]

Si. Instalando el programa, requiere GPU

[viscid moss]

mmmmm maybe #📑│making-models and #1159290066074607626

[trim ice]

Ah, okey. Another question: crepe, mangio-crepe, or rmvpe?

[viscid moss]

rmvpe is the best one

[trim ice]

aight

any settings that are worth tweaking to optimize the model for speech rather than singing?

[static gyro]

thank you

[viscid moss]

check #1159513888199540817 for that

[trim ice]

thanks

[trim ice]

Ah, any specific guide?

[viscid moss]

https://docs.aihub.wtf/

[quartz sphinx]

ese es el problema, que al no tener gpu mi unica alternativa (al parecer) es colab, aunque segun lo que vi en el mensaje fijado de help w okada hay una version modificada, pero que modificaron? y porque hay un riesgo de ban?

[viscid moss]

Las modificaciones solo fueron para ofuscar el código y evitar que te desconecte por el ban. Y hay riesgo de ban porque a google no le gusta que lo usen para eso y gratis.

[quartz sphinx]

bueno pues, tienen algun porcentaje o probabilidad de ban? para saber en ese caso que riesgo estaria corriendo

[viscid moss]

Luego de usarlo mucho mucho y que te desconecte mucho mucho, no podras usar google colab con esa cuenta

[quartz sphinx]

Hay un numero de veces promedio? Cada cuanto usualmente te desconecta?

[viscid moss]

nop, es aleatorio

[quartz sphinx]

bueno, gracias de todas formas, si me ocurre algo mas aviso

[viscid moss]

https://tenor.com/view/anya-forger-anya-spy-x-family-anime-salute-salute-gif-okey-dokey-gif-25700022

[viral rain]

Hey, anyone here to help?

I'm on CoverGen_No_UI, how do i pick the artist to which the song converts to?

[viscid moss]

just paste the model of that artist

[viral rain]

In the model download function? Or the rvc_dirname?

[rare goblet]

Ayo? @viral rain level 1 !!!

[viscid moss]

Model link, name, YT link and name again

[viral rain]

Got it, thanks!

[viral rain]

"python3: can't open file '/content/src/main.py' : [Errno 2] No such file or directory

[viscid moss]

Did you execute all the previous cells?

[sand heart]

https://FXtwitter.com/genkki_/status/1722751930089242877 CAN YOU GUYS VOTE FFOR ME IN THIS POLL

[rare goblet]

Ayo? @sand heart level 2 !!!

[sand heart]

haha you dont have to if you dont want too

[viral rain]

I don't use Twitter

[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

[wide parcel]

a lot of my rvc pth files are similar, is this because of the base being consistient? if so would it be possible to reduce the files by 50mb just by making a custom compressor/decompressor?

[glad zealot]

Not sure why but models are supposed to only be 60mb below

50mb up and 60mb bellow

[abstract wharf]

i've followed a couple guides and created my own custom model, now i'm trying to convert some audio in my tensorboard but it keeps throwing an error

logs show AttributeError: 'NoneType' object has no attribute 'tobytes'

anybody have any tips?

[modern oxide]

to convert some auido in tensorboard? you mean in rvc?

[abstract wharf]

yes sorry that's what i meant

[rare goblet]

Ayo? @abstract wharf level 1 !!!

[abstract wharf]

im just using the tensorboard ui

big nooob here

[abstract wharf]

i got it to work a couple of times, now it is throwing errors again.

[trim ice]

I have both an RTX 3080 and an Intel Arc A380. Is it possible to train using both at the same time for more performance?

[teal snow]

how do i make a ai song for free without downloading something

[molten pecan]

That happens to me sometimes

I think that's because I disconnect from the server for a moment

You could try rebooting the computer if you haven't tried yet

[molten pecan]

I think you can't. I think

[trim ice]

sad

any way to get the most out of training performance on my RTX 3080, then?

[abstract wharf]

yeah, i restart the runtime in the colab and I get another 1-2 succesful conversions but always bugs when I start changing pitch and other settings... have any other fixes?

[molten pecan]

Hm maybe try unloading and loading the audio you want to convert

Or refresh the model list

[potent tangle]

im trying to download Serj Tankian model to RVC v2 Huggingface version and AICoverGen and all i get are errors

i tried upoading zip file to AICoverGen but it just says uploading... forever

[violet heron]

What’s the link you used on the HuggingFace spaces?

[potent tangle]

https://huggingface.co/robob90/serj-tankian-toxicity-era-rvc2/resolve/main/serj-tankian-toxicity-scream-250%2B50e.zip

[violet heron]

The model is only 20 bytes big

It’s not your fault

It’s who ever uploaded it

Try using a different model

[potent tangle]

i cant find the one with the url, and when i upload it to the drive it doesnt work

[potent tangle]

https://www.weights.gg/es/models/clm72x86r101zcctcujopcx3h i fount his one but when i upload it it uploads forever

[violet heron]

If your on HuggingFace you can’t use drive models

[violet heron]

Wrong link, and seems the download is also hosted on weights

So even if you got the right link it wouldn’t work

And there’s 4 models in that 1 model

[potent tangle]

https://www.weights.gg/models/clm72ubxw0t38cctccmhwtmvc this one?

[brittle wing]

cannot import name 'TypeAlias' from 'typing' (/home/studio-lab-user/.conda/envs/default/lib/python3.9/typing.py) help tyring to use real time voice chnager on sagamaker amazon thing

[rare goblet]

Ayo? @brittle wing level 1 !!!

[low bay]

Hi guys, I've downloaded some voices and it says about epochs, do these impact what everything sounds like? If so, how do you change them?

[viscid moss]

same

[violet heron]

Epochs basically means training cycles

Basically it just tells people how much it’s been trained

More/less does not mean better

[low bay]

Ahhhh okay thank you!

[rare goblet]

Ayo? @low bay level 1 !!!

[viscid moss]

I think the problem is the version of python, you need 3.10.12

[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

[silent prism]

error is showing up

[molten pecan]

Yeah that's really common

Are you logged in?

[silent prism]

no

[molten pecan]

You have to be

Otherwise it will give you a bunch of errors

[silent prism]

how do i log in?

[molten pecan]

Oh you just need to make an account

[silent prism]

how do i do that?

[molten pecan]

Are you using huggingface?

[silent prism]

yes

[molten pecan]

Oh then just Google "huggingface create account"

It's kinda obvious

After that just make sure you are logged in when you are using Ilaria RVC

[silent prism]

alright created

ok

still showing errors

@molten pecan

[molten pecan]

Have you refreshed the page?

[silent prism]

yes

[molten pecan]

Click on the upper right icon

And see if your username appears there

[silent prism]

it does

[molten pecan]

Are you using an adblock?

[silent prism]

yes

[molten pecan]

Try disabling it

And refresh

[silent prism]

still not working

.3.

[molten pecan]

Dang

Try it in another web browser

[onyx forum]

does anyone know how to resume the training on the easy colab from 10/25

[harsh hamlet]

@silent prism did you get it working? I'm logged in without adblocker and still getting errors

[potent root]

anyone know what's the lowest point here?

here's a screencap of it with max smoothing

[flat turtle]

If you press the left most button, turn on the middle one, and press the right most one, and lower smoothing to like .86-.90 it might be easier to see

[onyx forum]

which of these would i have to press to continue training where i left off on the colab

[harsh hamlet]

I switched the Advanced Options from rmvpe to harvest and I was able to get Ilaria RVC to work. Don't know what's going on tonight haha

[rare goblet]

Ayo? @harsh hamlet level 1 !!!

[brittle wing]

I uploaded a sound I created for Mangio-RVC-v23.7.0, what should I do now? @karmic oxide

Where can I find where this model is registered?

@wheat copper

[wheat copper]

is this on paperspace

local?

[brittle wing]

methinks

I uploaded the sound here, but I couldn't find where it came from.

[viscid moss]

pth into "weights" folder, index into logs/asdasd/ADDED.index

[brittle wing]

Where can I access this on my computer?

[viscid moss]

go into the Mangio main folder

[viscid moss]

and then......

[brittle wing]

I think these are but there are too many

[viscid moss]

the first one

[brittle wing]

What will I do with this file?

[viscid moss]

If you are going to try it just go to the inference tab

[silent prism]

no i didnt

@harsh hamlet

i switched it to harvest too

[viscid moss]

guys, try with this one
https://huggingface.co/spaces/Eddycrack864/Applio-Inference

[harsh hamlet]

maybe try switching the advanced options from rmvpe to another one in the Advanced Options section. Best of luck!

[silent prism]

it said theres a issue trying to download

trying to download this

https://huggingface.co/chilombo/rvcmodels/resolve/main/drakesinging_made_by_%40minajjang.zip

[viscid moss]

checking.....

[silent prism]

ok ty

[brittle wing]

Which should I upload these to, Mangio or mmvc?

[viscid moss]

What do you want to use the model for?

[viscid moss]

model loaded

[brittle wing]

real time voice changer for twitch streams

[viscid moss]

The problem: The name had special characters

[viscid moss]

ok so....

[viscid moss]

u need the first one file

[viscid moss]

and the first one file too

[viscid moss]

why re u using the e210?

[brittle wing]

The first file and index are correct?

[rare goblet]

Ayo? @brittle wing level 3 !!!

[viscid moss]

correct

[brittle wing]

Which one should I upload them to to get the best experience?

[viscid moss]

both

[brittle wing]

no no no wait 1 minute

Will I upload it here?

[viscid moss]

ya, press upload

select RVC

and upload the pth and index file

[silent prism]

its taking forever to convert

[viscid moss]

because it uses CPU, it is the same as Ilaria's. Just a visually different version.

[silent prism]

oh

[rare goblet]

Ayo? @silent prism level 4 !!!

[silent prism]

whats the max time it will take?

[viscid moss]

depends on the time of the audio

[viscid moss]

it's ready

[silent prism]

damn its pretty good

[silent prism]

how do i make it less gittery?

and less roboticy

[viscid moss]

That depends on the quality of the acapella and the quality of the model. The Hugging Face space has rmvpe+ by default which is the best extraction method there is

[silent prism]

you know any better model of drake?

[viscid moss]

idk bro

try with another model, just for testing

[silent prism]

where do i find models

[viscid moss]

go to #🔍│find-models and use the ModelAI bot. Using "/search"

[brittle wing]

I did it but there is a lot of buzzing, how can I fix it?

[viscid moss]

try the Raven's settings

-rt

[azure marsh]

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

[viscid moss]

2nd link

[torn bay]

One message removed from a suspended account.

One message removed from a suspended account.

[brittle wing]

rvc tts right ?

or okada ?

[viscid moss]

Okada

[viscid moss]

Ilaria's RVC?

It's normal

[torn bay]

One message removed from a suspended account.

[brittle wing]

-rt

[azure marsh]

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

[viscid moss]

all RVC's the same. But locally is faster.

Installing the program

[torn bay]

One message removed from a suspended account.

One message removed from a suspended account.

[rare goblet]

Ayo? @torn bay level 1 !!!

[viscid moss]

u've 2 options Mangio or Applio

[torn bay]

One message removed from a suspended account.

[viscid moss]

Oh really? I didn't know

[torn bay]

One message removed from a suspended account.

One message removed from a suspended account.

One message removed from a suspended account.

[brittle wing]

What will I download here?

[torn bay]

One message removed from a suspended account.

One message removed from a suspended account.

[viscid moss]

just search for settings to ur Voice Changer

[torn bay]

One message removed from a suspended account.

[viscid moss]

Interesting....

[torn bay]

One message removed from a suspended account.

[silent prism]

i cant get links

it gives me download

[viscid moss]

right click and copy link

[silent prism]

alr

[molten pecan]

Either the Google Colab or the Huggingface spaces

I don't know any others

[silent prism]

i need to get a new pc this shi is soo sloowww lmao

[molten pecan]

I personally prefer the huggingface spaces since they are straight to the point and works well on mobile

[molten pecan]

Are you making a model?

[silent prism]

no converting

[molten pecan]

Oh okay

[silent prism]

why

[molten pecan]

How long is the audio and how long is it taking so far?

[silent prism]

its like 3 min long

its taking like 5 min

[torn bay]

One message removed from a suspended account.

[molten pecan]

Oh okay

Why don't you try using Ilaria RVC?

[torn bay]

One message removed from a suspended account.

[molten pecan]

Yeah but far more convenient if you have a bad PC

And it works on the phone too

[silent prism]

it didnt work errors

[silent prism]

i got a pretty good phone, but i do business on there, dont want to ruin it

ya dig

[molten pecan]

Are you logged in?

You need to be logged in huggingface to use it

[silent prism]

yea

we just talked like a hr ago

[molten pecan]

Oh you don't need a good phone, since it's cloud based

[silent prism]

oh

[molten pecan]

It doesn't take the resources of your phone. At least not in theory

[silent prism]

well my internet is just slow

[molten pecan]

Ah okay

[silent prism]

i thought it took cpu?

according to the other guy

[molten pecan]

https://huggingface.co/spaces/r3gm/Aesthetic_RVC_Inference_HF

Try using this one

If it's still giving you error, duplicate the space with the purple button

[silent prism]

im using apollio at the monet, or wtv the name is

Applio

[molten pecan]

If you use it locally then yeah

Apollio 😭

[silent prism]

i didnt know the name lmao

yea it says it uses cpu

and what is duplicating?

[molten pecan]

Basically that link I sent you is a public space and app, which means people are using it

At the same time as you

Therefore it might give you some errors and it will be slower

By duplicating it you copy it and keep it to yourself

(make sure you set it as Private)

[silent prism]

how do i duplicate it

and set it to private

[molten pecan]

With the purple button that says

Duplicate this space

[molten pecan]

It will give you a popup

You can pick the option there

[silent prism]

oh i see it, and once i click on it will it be faster?

[molten pecan]

Yeah

[silent prism]

cuz it talking forever

[molten pecan]

Yeah it's loading

It will take a few minutes

To come back to it you just go to your profile and your list of spaces will be there

Keep in mind that it will go on hibernation mode if you don't use it for 24 hours, I think

So you'll have to wait until it loads again

Like now

[silent prism]

im using opera gx for the first time, and in the taskbar, its glowing red, is that like a error, or does it always do that?

cuz idk if its a error for RVC

or what

[molten pecan]

Can you send a screenshot?

[silent prism]

nvm it disapperd

[molten pecan]

Ah

[silent prism]

disappeared*

OHHH MY GAWWWD it sounds so good

this is magic

[molten pecan]

Oh it worked?

Congrats!

[silent prism]

yeeee

its .wav

[molten pecan]

If you want to get the best quality possible lmk

[silent prism]

i do

[molten pecan]

I know what (almost) each setting does

Oh okay

[silent prism]

its kinda delayed

i can send it to you

[molten pecan]

Remind me tomorrow since I gotta leave now

[silent prism]

alright

[molten pecan]

Ping me or I'll forget

[silent prism]

@molten pecan

[molten pecan]

Sure

Hi

Morning

[silent prism]

sent a friend request

[rare goblet]

Ayo? @silent prism level 5 !!!

[molten pecan]

Accepted it

See you tomorrow

[brittle wing]

Where do I put the amount of metadata

[onyx forum]

which of these would i have to press to continue training where i left off on the colab

where would my dataset be stored in my drive since its no longer saved in the colab

or do i not need it

but do i have to click any of the other cells besides "4. train model" because i get this error

thats when i just run the train model, but i still have the same name

ok let me try again

[rare goblet]

Ayo? @onyx forum level 2 !!!

[daring raft]

hey, so I used Kits.AI when they first started, and I liked the thing that you can sing falseto for example and some models would sing it like yelling style. Now with replay i can’t get that result even with increased index ratio, it still sounds too much like a persone who gave original vocal.

Why that heppend before and how can i train and convert it like that? Is it because of f0 method??

Please i need help with this cant find solution

[onyx forum]

i would think it would depend on what types of singing were in the dataset you gave, for example if you just had yelling in the dataset then it would transform falsetto or any high enough notes into yelling

ok now my model wont even train past 420 epochs even though i always set it to 1000

it finishes the train model cell right after it loads up tensorboard and then doesnt do anything

[daring raft]

yeah but also, i have bad english accent, and before, when converted to AI model, it sounded native af, now it’s super realistic, i can’t get that accent like before with Replay

[rare goblet]

Ayo? @daring raft level 3 !!!

[thin stump]

@daring raft The index ratio just changes the amounts of artifacts you can hear in their singing. Too much of it and you can notice it messing up words

[daring raft]

What hey used before then? I used Kits.AI like 5 months ago, i even made model there and it was always the energy of the AI model singer. Now with Replay it’s more energy of real person who sings

[thin stump]

They probably have their own niche? For me, the F0 pitch on Replay is more about clarity than accent. Mangio Crepe is my go-to and RVMPE is for speech

Crepe is like the middle between the two and could be inconsistent

[modest oriole]

hi! can anyone that knows a bit abt rvc help me? -- the default models vs the models i downloaded use 2x my cpu, and RVC only uses my CPU even though i have GPU selected

[rare goblet]

Ayo? @modest oriole level 1 !!!

[modest oriole]

even w/ 5 sec it'll cut out from time to time, but it wont cut out at all with default models even on like, 192 delay :c

[twin shoal]

-rvc

[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

[rare goblet]

Ayo? @twin shoal level 1 !!!

[solemn bough]

Is there any ratio of dataset's length and epoches?

[silent prism]

my shit aint working again

[twin shoal]

why it stop here?

[civic onyx]

how do i install applio

on NPM

its showing me how but it doesnt work

https://github.com/IAHispano/Applio-API#installation

[silent prism]

how to i go to my list of dupplications?

[civic onyx]

thats it im pinging an admin

[rare goblet]

Ayo? @civic onyx level 1 !!!

[civic onyx]

its gonna take 5 years

for one to show up

[silent prism]

fr

[civic onyx]

so i might as well ping one to save both of us time

[silent prism]

ping a helper

[civic onyx]

@karmic oxide

we need help

[silent prism]

@karmic oxide

lmao

any you reccomend?

[civic onyx]

so your simply not going to help me

[silent prism]

recommend

[civic onyx]

or even try

https://github.com/IAHispano/Applio-API#installation

this is what im working with ^

[silent prism]

they all keep dying on me

[gray geyser]

yall i need help

im using the easy gui

in colab

[silent prism]

we all do

[gray geyser]

how do i put

a weights.gg model

like

the link

and download it

cuz i cant find a frank ocean model

anywhere other than weights.gg

[civic onyx]

💀 what a great helper you are

[gray geyser]

💀

[civic onyx]

didnt even bother to look at the site

[silent prism]

bruh chill

[twin shoal]

what should i choose first?

[gray geyser]

my laptop doesnt have a very good gpu at all so i can only use colab

[silent prism]

if brodie doesent know he doesent know

raven

can yo rec me a rvc

send link?

alr

ty

i cant understand that lmao

and i have integrated graphics

well raven

can you at least tell me where i can find my dupplications?

[civic onyx]

well if you cant help me with anything i need, is there some kind of collab open right now that i can use that you can help with or at least reccomend

[twin shoal]

what that :v

[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.

[twin shoal]

iam training model

[silent prism]

who is the right person for the job?

[mighty parcel]

chief, when raven is saying, is she does not use applio, has never used it, and has no experience in helping troubleshoot any issues specific to applio

you can throw the word to applio, to most people

and most will say its a buggy mess

as for why its blatantly recommended, IDK

[silent prism]

thanks

appreciate it

[twin shoal]

bruh

oh

should i use harvest instead of rmvpe?

[civic onyx]

well its fine now, i didn't know there were new collabs out since the google colab banning thing

[twin shoal]

ok

[brittle wing]

-rt

[azure marsh]

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

[twin shoal]

oh, there is two rmvpe

is rmvpe_gpu better than rmvpe?

[rare goblet]

Ayo? @twin shoal level 2 !!!

[brittle wing]

How much gain should I use? pls help

[twin shoal]

yeah

[silent prism]

yes

[brittle wing]

How many should these be?

[twin shoal]

ok

[solemn bough]

How many epoches should I use on 15 min~ dataset?

[twin shoal]

i have seen this error second time

i start train model

and its stop with that

[twin shoal]

i think 1k

idk

[solemn bough]

Ain't this too much?

I rarely see models with 500 epoches

[twin shoal]

hmm

i think those model overtraining at 500 epoches

[silent prism]

how can i fix the quality after?

[torn bay]

One message removed from a suspended account.

One message removed from a suspended account.

One message removed from a suspended account.

One message removed from a suspended account.

[silent prism]

its doing that to me too

[civic onyx]

how do i use the collabs on hugging face, they have no collab button, im not using covergen becuse it sucks

keeps disconnecting me

and its other collab straight up didnt work

had an error loading code

ah their offline

guess ill ping someone else

🤷

@pliant crescent

[rare goblet]

Ayo? @civic onyx level 2 !!!

[civic onyx]

yo

[pliant crescent]

You mean the spaces on huggingface
Well I suggest follow the Ilaria RVC guide in #1159513888199540817

[civic onyx]

do i have to do something local to use them

becuse if i need to locally download them and use them then thats not really what i want tbh

[pliant crescent]

No it's online

[pliant crescent]

Do you even have a good GPU

[civic onyx]

hell no

thats why i avoid local at all costs

[civic onyx]

okay so it didnt really give me a guide to use the hugging face colabs

it kinda just gave me a colab

unrelated

[pliant crescent]

Huh?

[civic onyx]

https://docs.google.com/document/d/186_a8d32iYxr1UAEm1LAFQvLu2w3huVIg-KTlTHshg8/edit

[pliant crescent]

Yes that's the guide

[civic onyx]

????

so do i paste the huggingface link of the colab into models??

[pliant crescent]

Are you not seeing this

[civic onyx]

yeah no i read that

but like

it didnt give me any info

on actually using the colabs

[pliant crescent]

Just follow these steps

[civic onyx]

on huggingface

[civic onyx]

i did

https://huggingface.co/spaces/TheStinger/Ilaria_RVC

[pliant crescent]

Yes
You have to duplicate that space into your account

[civic onyx]

i have no clue how to do that

doesnt tell you how

[pliant crescent]

Use this link
https://huggingface.co/spaces/TheStinger/Ilaria_RVC?duplicate=true

Make sure to set it to private

It will take 2-3 minutes to load

But this is only for the first time

[civic onyx]

well how do i use it multiple times

[pliant crescent]

Just open it
It will be in your Huggingface account

[civic onyx]

Jesus christ.

16 GB's.

[pliant crescent]

Btw it takes like 100 seconds to convert a 1 minute audio

[pliant crescent]

Don't worry it's all in Huggingface

[civic onyx]

oh good

[civic onyx]

its telling me my email needs to be confirmed, guess it must be in my gmail

[pliant crescent]

Just confirm your email
There should be an email sent by huggingface with a link

[civic onyx]

so i imagine ill need to seperate the vocals myself, it doesnt have its own seperating thing?

[pliant crescent]

Well you could use the MDX Colab

[civic onyx]

am i able to use it within ilaria?

[rare goblet]

Ayo? @civic onyx level 3 !!!

[civic onyx]

i mean its supposed to use other hugging face colabs right

unless im dumb and im wrong

which is a high probability

okay nevermind MDX confuses me

[brittle wing]

What should we pay attention to when recording a custom real-time sound?

[tawdry ruin]

wdym by a custom real-time sound?

[brittle wing]

i want a custom voice

[rare goblet]

Ayo? @brittle wing level 4 !!!

[brittle wing]

I want to do this with voice recording, but what should I pay attention to?

[pliant crescent]

Make sure there's no background noise and reverb

Also this tool can make your recording better
https://podcast.adobe.com/enhance

[brittle wing]

If I send you a voice recording, would you have a chance to review it?

[pliant crescent]

Sure

[remote maple]

hey, having a weird issue myself... tensorboard all of a sudden just isnt loading up

its showing a blank screen

anyone have this issue too?

[solar arrow]

Hi, quick question, I haven't trained a model in months, probably last one was in June. I usually use the Colab but Im interested in how is the process today? Is RVC3 out yet? Is there a new preferred method?

[rare goblet]

Ayo? @solar arrow level 1 !!!

[cobalt oyster]

I need tutorial for this pls help AI_Auto_Cover_V1.ipynb

[rare goblet]

Ayo? @cobalt oyster level 5 !!!

[cobalt oyster]

@astral jungle

[brittle wing]

/help

[harsh scarab]

​No Category:
  help Shows this message

Type /help command for more info on a command.
You can also type /help category for more info on a category.

[brittle wing]

/help send upload guides

[harsh scarab]

No command called "send" found.

[brave garnet]

Tutorial in italiano

⠀
Scegli dai pulsanti qui sotto che tipo di guida seguire.
Se hai problemi, scrivi tranquillamente nella chat italiana.

Gli autori delle guide sono creditati nella parte superiore.
⠀

Made with love by Ilaria 💖

[cobalt oyster]

how to open this guide

@brittle wing could u help me

[brave garnet]

Tutorial in italiano

⠀
Scegli dai pulsanti qui sotto che tipo di guida seguire.
Se hai problemi, scrivi tranquillamente nella chat italiana.

Gli autori delle guide sono creditati nella parte superiore.
⠀

Made with love by Ilaria 💖

[cobalt oyster]

@tawdry ruin help me pls

pls help

[tawdry ruin]

what do you need?

[cobalt oyster]

I need tutorial for AI_Auto_Cover_V1.ipynb

btw thanks for responding

[cobalt oyster]

the above link does not work

[tawdry ruin]

AI_Auto_Cover_V1?

I've personnally never heard of it

[half cove]

same

[tawdry ruin]

do you have the link of the Colab?

[cobalt oyster]

https://colab.research.google.com/github/self-destruction/AiAutoCover/blob/main/AI_Auto_Cover_V1.ipynb

yes

[half cove]

I gotta test this

[cobalt oyster]

its in some other language so I use translator

u can create cover using yt link it automatcally removes vocals it worked yesterday but now it doesnt

[half cove]

So like aicovergen?

[cobalt oyster]

I dont know about aicover gen

[glad zealot]

what this do?

[cobalt oyster]

create ai covers

[glad zealot]

why not just use covergen? it has a gui and auto separate background from audio

i mean this have uvr too but no gui :D

[half cove]

seems like the same concept as ai cover gen

[glad zealot]

yup

just not as automatic as covergen

imma try it tho

[half cove]

@cobalt oyster https://colab.research.google.com/github/hinabl/AICoverGen-Colab/blob/main/Hina_Mod_AICoverGen_colab.ipynb
i think this still works

[cobalt oyster]

I will try thanks for help

[glad zealot]

oh its using mangio as base

[cobalt oyster]

does it work for u

Transformation does not work for me

[glad zealot]

testomg

testing atm

[cobalt oyster]

ok after testing could u help me if possible

[twin shoal]

hello

[glad zealot]

theres a part that i think is confusing so imma test it

[twin shoal]

i have a question

did it overtrain?

[cobalt oyster]

which part I am using it for like few days I can help u if possible

[glad zealot]

you have smoothjing enabled?

[twin shoal]

no

[proper shale]

is KL still going down

and also give us d/total

[proper shale]

use 0.8 or 9

[glad zealot]

set it to 0.999

[proper shale]

no

[twin shoal]

0.8

[glad zealot]

nope thats good

keep going

[twin shoal]

ok

oh what if

its 1000 epochs

and doesnt overtrain?

[proper shale]

nahhhh it might be overtraining actually

d total was uhm

stagnating a bit

I'd say keep going rn

[twin shoal]

ok

[proper shale]

But just to see when it becomes blatantly obvious it's overtraining

[glad zealot]

when it doesnt want to go down thats basically overtraining

[proper shale]

Then I'd pick a point at around 7-8k

[twin shoal]

ok

[cobalt oyster]

@glad zealot testing completed

?

sorry for asking many times if completed could u msg me

[glad zealot]

almost i think?

[cobalt oyster]

which step r u in

[glad zealot]

im at the transformation part

[cobalt oyster]

I am stuck in that part

[glad zealot]

i just pressed it

and it worked

[cobalt oyster]

it shows me error

[glad zealot]

my setting

it actually sounds good

i might use this next time i make a cover

[cobalt oyster]

yes

[glad zealot]

it gets the reverb if the original have

[cobalt oyster]

this is my error

[cobalt oyster]

yeah I got this colab from a yt video

[rare goblet]

Ayo? @cobalt oyster level 6 !!!

[glad zealot]

you run the cell above it right?

[cobalt oyster]

yes

[glad zealot]

not sure

imma try again with another model

is your audio a youtube link?

[cobalt oyster]

yes

[glad zealot]

no idea

[cobalt oyster]

it extracts vocals very good too

[glad zealot]

ye

i like that i keeps the cho

ehco

[glad zealot]

your error is saying that theres no audio file somehow

[astral jungle]

Huh

[astral jungle]

give me the colab link

[cobalt oyster]

https://colab.research.google.com/github/self-destruction/AiAutoCover/blob/main/AI_Auto_Cover_V1.ipynb

@astral jungle

[cobalt oyster]

yes

[brave garnet]

Colab di Google

⠀

• Colab per usare modelli già esistenti e trainare

Link

• Colab di backup nel caso la precedente dia problemi

Link

⠀

Made with love by Ilaria 💖

[cobalt oyster]

this one

[astral jungle]

Thank you, I'll try it later

[cobalt oyster]

thanks

check it out when ur free

[half cove]

why does it link me to brazil server

when its in russian

😭

[astral jungle]

Okay

[cobalt oyster]

dont know

[cobalt oyster]

have u tested

[valid sail]

I need help with something

[rare goblet]

Ayo? @valid sail level 1 !!!

[half cove]

@cobalt oyster idk if this will help you but i put it through google translate

[cobalt oyster]

Ok

Thanks

Btw how did u get this

[half cove]

google translate

[cobalt oyster]

extension is much easier

do u have google translate extension

btw do u have any kind of guide to this

[half cove]

🤷‍♂️

[cobalt oyster]

did u get stuck at transfomation

I got stuck at transformation

could u help me

[astral jungle]

It works @cobalt oyster

[cobalt oyster]

what did u do?

I cannot geet pass transformation step

[astral jungle]

I just run from top to bottom

[cobalt oyster]

did u use yt link

[astral jungle]

No

[astral jungle]

I used the link you sent me

[cobalt oyster]

oh ok let me try again I will let u know once I tried

[astral jungle]

Okay

[copper pebble]

Tensorboard error on RVC Disconnected??

• refresh it till it works
  (like 4/5 times)

• change browser
  (if the refresh button's tip doesnt work switch to a non Chromium browser)

[cobalt oyster]

it somehow works now @astral jungle

[leaden glen]

What folder do u need to put models in in applio?

[modern sedge]

is collab working now?

[copper pebble]

yh

[copper pebble]

https://discord.com/channels/1159260121998827560/1160926856548200531 voice training guide:

[astral jungle]

Yup

[peak mantle]

hey, can someone help me? i want to add my model to the applio, but it doesn't add model index... what i need to do?

[lament eagle]

Common FireFox W

[celest latch]

how to get the huggingface download link?, i cant find any button

[copper pebble]

kurama dripping rn

[copper pebble]

files and version> right click on the red box> copy link

[celest latch]

thanks

[brittle wing]

C:\Windows\System32>runtime\python.exe rvcgui.py --pycmd runtime\python.exe
System nie może odnaleźć określonej ścieżki.

[abstract wharf]

if my converted audio sounds a bit tinny and robotic — is that due to the quality of my audio samples used for training? or is there another setting I mess around with in rvc

[narrow robin]

how do you make an voice model

[copper pebble]

https://discord.com/channels/1159260121998827560/1160926856548200531

[stiff goblet]

does anyone uses applio rvc ? i got an error in "split audio" option in inference. It says [Errno 2] No such file or directory:'assets/audios\myVocals_timestamps.txt'

[rare goblet]

Ayo? @stiff goblet level 1 !!!

[stiff goblet]

I figured out what causes that error. Just don't use space character in your input audio filename. For example change it "my vocal.wav" to** "my_vocal.wav"**. It works now 👍

[molten pecan]

It can be both things

You could try tweaking the settings but if your audio or your model don't sound good, then you can't do much about it

[glad zealot]

use ngrok :D

[obtuse stream]

if the sentences are so far apart, is that good?

[proper shale]

trim those silences

thats vad

bad*

[obtuse stream]

completely?

now good?

[nimble zealot]

Hey

So I’ve been making ai covers through the aicovergen colab page this week

[proper shale]

yup

[nimble zealot]

But I just got on it again and they changed the settings to only have like 6 or 7 pitch options for vocals

So what do you guys use

[proper shale]

It's up for you to decide honestly

[nimble zealot]

The other one I found burns through all my colab gpu mad fast

Well I’m asking like what is a relatively simple web ui I could use

Or a colab link that works well

I’m about to get colab pro but I feel like the one I found that I like is gonna burn right through it

[obtuse stream]

its good ? the sample lenght is enough? it's okay that there's not much pause between sentences, right?

[proper shale]

seems to be good enough

[obtuse stream]

and 300 epoch is enough?

[proper shale]

not really?

there isn't a set value for epochs in every minute

go for 500, if it's not overtraining yet, 750, if not then 1000

[rare goblet]

Ayo? @obtuse stream level 3 !!!

[proper shale]

Not exactly accurate

[molten pecan]

Do you have a more accurate graph?

I need it so badly

[proper shale]

There isn't one

[molten pecan]

😭

[proper shale]

And there won't be one

[molten pecan]

Dang

[proper shale]

Because every training session is different.

Example: For one person, 500 epochs is just right, but for the other, it's extremely undertrained/overtrained (overfitting)

Their datasets are different, their training sessions are different

There is no such thing as an accurate epoch per minute graph

[obtuse stream]

and then what is the correct epoch number based on

how do we find out?

[proper shale]

...you don't

you just train until its overtraining

That's the fun (and honestly quite bad) part

[obtuse stream]

and how can I see that?

[wet cliff]

i have a question and dont really know where to ask but what are epochs? 😭

[proper shale]

tensorboard and manually testing ur models

[obtuse stream]

these, how do I know?

[proper shale]

-overtrain

[azure marsh]

Hey, @wet cliff!

👇 Here are some resources to help you identify if your model is overtraining

How do I know if my model is overtaining?

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

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

Credits: LUSBERT

Automated Overtraining Detection (AOD)

Will be available soon in #1159513888199540817

Credits: grvyscale

[proper shale]

oh no wrong person

[rare goblet]

Ayo? @proper shale level 42 !!!

[proper shale]

An epoch is basically how many times your dataset has been trained fully.

E.g. 200 epochs
Means the dataset has been trained fully 200 times.

[obtuse stream]

i used this

[wet cliff]

so more the merrier?

[stiff goblet]

Is there any pitch setting for your model TTS in applio fork ? I can't find it

[obtuse stream]

? what do you mean

[proud elbow]

epoch = training cycle, like a lap in a race

[stiff goblet]

I'm trying to use TTS . text to speech. But there is no pitch setting in that tab. Did you tried to use tts before in Applio ?

[obtuse stream]

no

[toxic edge]

How to avoid delay?

[obtuse stream]

how to check the overtrainb?

[molten pecan]

Are you using a wireless mic?

[neon valve]

What is the appropriate amount of epochs for a 6 minute dataset? 200?

[rare goblet]

Ayo? @neon valve level 3 !!!

[molten pecan]

Try plugging it if so

And make sure you are using rmvpe

[toxic edge]

yes

[rare goblet]

Ayo? @toxic edge level 1 !!!

[molten pecan]

Try plugging it

Or use a wired mic

[brittle wing]

https://i.imgur.com/g2HX5Vx.png

what's with this one? trying to download pytorch

[neon valve]

Anyone?

[molten pecan]

Try asking on #📑│making-models

[nimble zealot]

Can someone please help me out with a google colab for ai song covers

[rare goblet]

Ayo? @nimble zealot level 1 !!!

[nimble zealot]

I’m trying to make some for my cousins and just bouncing around between trying random ones if anyone has a colab link that works well please help

[visual niche]

Hello, first of all, due to my lack of proper English, I am contacting you by translating from the translation site, I apologise if there is any mistake. If we move on to my problem, since I am interested in AI cover, I found a simple way of making it by researching on the internet, but I can only use the ones with clean vocals by separating them from the music and vocal shape, I can never use the opposite because the sounds are distorted. My request from you is to teach me how to cover the music I want. (As I said in the first sentence, my English is not enough, so if you offer any solution, I would be happy if you tag me).

And I should add that if there is one on the server, I don't know which one will work for me, so I would appreciate it if you don't blame me for that.

[molten pecan]

I can help you!

So first, to get clean vocals from a song use the software called Ultimate Vocal Remover 5

It's for PC only

[frozen sandal]

thanks

[stiff goblet]

You are welcome @frozen sandal

[brittle wing]

do i need to use line 1 when i'm using okada's colab voice changer?