✨│ai-help

[proper shale]

model download worked for me

[queen sapphire]

yeah

[proper shale]

now share the colab you're using

[queen sapphire]

https://colab.research.google.com/drive/1Gj6UTf2gicndUW_tVheVhTXIIYpFTYc7?usp=sharing#scrollTo=OVQoLQJXS7WX

[proper shale]

oh this is outdated...

[queen sapphire]

oop

[glad zealot]

training?

[queen sapphire]

saw it in a guys youtube description

[proper shale]

yeah that's been outdated for like 10 months?

[queen sapphire]

darn

[glad zealot]

-colab pretty sure its here

[azure marsh]

☁️ Google Colabs

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

How to use?

🤗 Hugginface Spaces

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

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

[queen sapphire]

what a newer one i can use

[rare goblet]

Ayo? @queen sapphire level 1 !!!

[proper shale]

i suggest using the colabs linked in https://docs.aihub.wtf

[queen sapphire]

which one

[proper shale]

https://colab.research.google.com/github/iahispano/applio/blob/master/assets/Applio.ipynb is a good one, check the guide for it

[glad zealot]

fck

[proper shale]

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

[queen sapphire]

cheers mj

[proper shale]

yw :)

[runic zephyr]

what is the chunk thing is it just quality and processing?

[proper shale]

it's the chunk that the AI is gonna process at once, if you increase it can take longer stuff better and gives the GPU a lil more breathing room

[runic zephyr]

what is like a good standing point to put it at if you want good quality real time

[rare goblet]

Ayo? @runic zephyr level 1 !!!

[runic zephyr]

i have an rtx 3080 ti with 12 gb and its OC'd

[proper shale]

112-128 is good enough ig

but ofc test and see what works well for you

[ocean holly]

Heyo MJ! How was your day? (I hope it was good) I have a question and I'm wondering if I could have some help!

[proper shale]

ehhh pretty good!

ask away

https://tenor.com/view/счастье-gif-23940530

[ocean holly]

In the past I tried to make a new voice model! I (sadly) failed! ;-; I wanted to have another go at it, and following the txt guide is quite hard for me! Its really well made, but Its quite hard for me to follow txt guides for some reason, so I was wondering if I could have some help with making a voice model and I was also wondering about making an AI cover

[rare goblet]

Ayo? @ocean holly level 3 !!!

[proper shale]

there aren't really video guides (that are accurate and updated enough)

[ocean holly]

I see, thats sad to hear

[proper shale]

so like, we have the guides here

if you have any issues while doing it, we can always clear things up for ya

[ocean holly]

Mhm! okay! I try again now! If by chance, you (or anyone knowledageble reading this and want to help) are free- I would love to hop in a vc

I know I'm asking alot Sorry-

[proper shale]

Yeah I can't really VC

[proper shale]

Naw it's ok I understand

[ocean holly]

Yeah thats okay! No worrys!

[proper shale]

You can always write stuff here whenever you're confused

[proud elbow]

just ask anything specifically here (sorry I can't also VC)

[proper shale]

Me or somebody else will clear stuff up for you

[ocean holly]

Thank you all!

I'll do my best and try again!

[proper shale]

Alrighty! Good luck

[ocean holly]

I hope you all have a wonderful day!! ❤️

[proper shale]

You too

https://tenor.com/view/dap-me-up-gif-13888687653934292805

[narrow parrot]

hey guys do yall have the link to the colab for training

[ocean holly]

Then Here comes the first question of my lil journey here! Can I over train? How long should my Dataset be?

I have around 39 min of no backround noise, no empty spaces beetween sentences of just speaking, is that a good dataset?

(Can I make it longer? I want to make the best quality I can, no matter how long it takes, as I daily drive this model rvc)

[violet heron]

Yeah a bit too much but it’ll work

I would recommend using tensorboard to check when the model overtrains (don’t ask me idk)

[ocean holly]

Alrighty!

[violet heron]

-overtrain

[azure marsh]

How do I know if my model is overtaining?

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

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

Credits: LUSBERT

Automated Overtraining Detection (AOD)

Will be available soon in #1159513888199540817

Credits: grvyscale

[narrow parrot]

aaah please where do you find the colab i'm lost

[violet heron]

-colab

[azure marsh]

Suggestions for @narrow parrot

☁️ Google Colabs

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

How to use?

🤗 Hugginface Spaces

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

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

[narrow parrot]

thanks!!

[ocean holly]

I shall give this a read! thank you timbo!

[stiff goblet]

Which graphic should i consider as the best epoch? Loss/g/mel is still decreasing. @proper shale

[rare goblet]

Ayo? @stiff goblet level 7 !!!

[glad zealot]

id say keep going

[stiff goblet]

Only loss/g/total is increasing , so im kinda confused

[narrow parrot]

-colab training

[azure marsh]

☁️ Google Colabs

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

How to use?

🤗 Hugginface Spaces

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

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

[narrow parrot]

theres only inference but I wanna create the model

[pulsar anchor]

upload file and turn my voice not in realtime

OH SHIT

WRONG PING PERSON SORRY

Is it possible to never get stuttering thingy like the video I showed you if I change input as upload audio file? @proper shale

The funnies

[maiden finch]

I think Hina's WEBUI has stopped working

Everytime I try to generate a cover, I get an error message saying I don't have a Nvidia GPU even though I do and am connected to the T4 GPU anyway

[limber gazelle]

does anyone know how can i upload files required to create voice model in Applio?

nvm i figured out

[zinc compass]

Hello, I currently downloaded the Voice Changer Client; and uh

[rare goblet]

Ayo? @zinc compass level 1 !!!

[zinc compass]

It isnt working, specifically it isnt really changing my voice

I can download the files, but it doesnt seem to affect my actual voice on discord tests

[proper shale]

keep going

[proper shale]

mmm maybe, but you have to keep chunks high

[proper shale]

Check command prompt

[zinc compass]

Okay, just a second

[zinc compass]

I got this warning screen while extracting

the file

[proper shale]

Weird...

Hope you aren't on OneDrive

[zinc compass]

I dont think so, let me check

I have onedrive on my computer, but Im unsure whether that inteferes

should I just use voiceai

[proper shale]

If you're not extracting the folder on it... it's alr

[proper shale]

no 😭

anything but that

[zinc compass]

I dont have a good software 😭

[proper shale]

W-Okada is already great

So yeah

[zinc compass]

UGHHHH, is it possible for u to help me out on stream

[proper shale]

Maybe try extracting with 7zip

[proper shale]

Ah sadly no

[zinc compass]

im not sure which one to download from the github page either

[brave garnet]

Voicechanger Download (Okada)

⠀
Download for Nvidia GPUs
Version 18a cuda

Download for AMD GPUs
Version 18a directml

Download for Intel GPUs
Version 18a directml

Download for Mac
Version 17b Mac
⠀

Made with love by Ilaria 💖

[proper shale]

^

[zinc compass]

wait so since Im on Nvidia

Ima download that, then extract it

[proper shale]

alright

[azure cape]

hello i have a amd card and i downloaded for amd gpus but it doesnt seem to register my card is there any fix?

[proud elbow]

what kind of amd gpu? RX 5000-series and newer should work

[proper shale]

either they've installed the cuda version or they just need to figure out the number

of their gpu

[azure cape]

RX 580

i dowloaded this one MMVCServerSIO_win_onnxdirectML-cuda_v.1.5.3.18a.

[rare goblet]

Ayo? @azure cape level 1 !!!

[proper shale]

alr, that's the DirectML version

you just need to select the right gpu number

check Task Manager > Performance tab for the number

[azure cape]

where should it show in the Performance tab?

[proper shale]

"GPU x"

And right below it the name of your GPU

[azure cape]

can i send you a screenshot in dm im blind

oh i can send it here

[proper shale]

gpu 0 right there

[azure cape]

oh so i need to use gpu0 in the app and thats it?

[proper shale]

yeah, you'll also need to convert any downloaded models to onnx

so it isn't extremely laggy

[azure cape]

aight thanks, so amd cards can only use crepe_tiny and rmvpe_onnx?

[pastel oak]

ye, use rmvpe

if you want to experiment and try the other voice changer (which should be better performance wise but its not as user friendly), heres a link https://rentry.co/RVCRealtimeGuide

[azure cape]

okay thanks ill give it a go

[rare goblet]

Ayo? @azure cape level 2 !!!

[snow silo]

guys how do i "convert wav to pth"

i mean i need a pth file

[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

[tough fiber]

can someone help me

[sinful quiver]

guys which one is better Ov2super or TITAN? im confused

[naive violet]

I want to use Rvc while gaming but it messes up the GPU really bad, is there any solution, like dual gpu

[ocean holly]

Oh? I never had this problem

What's your gpu?

[proper shale]

make sure you've lowered ur graphics as much as possible, also cap ur fps to 60

in w-okada, set extra to 4096 or 8192 if you have it set to anything higher

[gray plinth]

Hi, my AMD graphics doesnt seem to be detect by the RVC is there any fix?

[gusty kestrel]

@steel isle, I have found 16 results that match your search!

sztef

HOT EGIRL (LOW PITCH)

• Uploaded: <t:1701738126:d>
• Server: AI Hub
• Likes: 0
• Lang: EN

Epochs

200

Technology

RVC

Algorithm

Rmvpe

Requested by jsszxc · May 11, 2024, 6:50 PM

[proper shale]

Make sure you use the right version for your GPU - you might be using the NVIDIA version

https://rentry.org/VoiceChangerGuide <-- Read this, very important and has everything you need to know on how to set it up with AMD stuff too

[south vortex]

hello, im trying to extract a main vocal mixed with low harmonies inside. Used MDX Karaoke 2, but it wasnt being extracted. Im trying to find the pitch shift? option, but i can't find it

[rare goblet]

Ayo? @south vortex level 2 !!!

[tacit tinsel]

hi i get this " ModuleNotFoundError: No module named 'mega'" in colab

what am i doing wrong?

i just add the model from huggingface

does it only accept zip from mega?

[proper shale]

might be an outdated colab, use the ones in https://docs.aihub.wtf

[tacit tinsel]

thank you, will check it out

[rare goblet]

Ayo? @tacit tinsel level 1 !!!

[dawn yoke]

When ever i try to download model i get a dsonerror

[rare goblet]

Ayo? @dawn yoke level 1 !!!

[dawn yoke]

any fix?

[proper shale]

old colab, use the ones in https://docs.aihub.wtf

[dawn yoke]

preciate it

[sick otter]

How to make brawl stars buster Voice model

[proper shale]

https://docs.aihub.wtf

[summer creek]

my output voice is sounding a bit too robotic how do i fix that?

[flint seal]

i don't really need help with rvc, but how much models do you need to have made to get model master?

is there a minimum or is it given out to people who have been really active in making/sharing models?

because i'm really trying to become active in this community, to get model maker role and stuff like that

[proper shale]

like 10 good ones

it takes good knowledge to make 10 good models

[flint seal]

alright, thanks

[proper shale]

so yeah

[proper shale]

yw :)

[flint seal]

currently trying to master my model making skills

[proper shale]

if u have any questions ask away

[flint seal]

alright thanks

i submitted a model in #1159514067187277865 but i don't think its very good

[proper shale]

well it has some artifacting

but overall... it's ok

[flint seal]

thanks!

how do you get rid of artificating?

do you have to have a wider variety of datasets?

[proper shale]

well... clean your dataset of any background noise (mic bleed on studio sessions screw it up, personal experience), only get the best takes, and if possible denoise with RX10

[muted fiber]

Hey guys where can I download the Applio thing again? sorry I deleted mine

[proper shale]

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

[muted fiber]

ty

[flint seal]

someone told me to denoise the dataset

but that was after i submitted

gonna denoise the dataset then retrain

[night warren]

what file type do i need for the dataset? spent like a day cutting noise and isolating in audacity just want to know how to export it

mangio btw

[proper shale]

wav is best, to prevent any issues with file format conversion n stuff

also, no data loss

[night warren]

gotcha

[rare goblet]

Ayo? @night warren level 1 !!!

[gusty kestrel]

@waxen grotto, I have found 2 results that match your search!

enhayler

Latina Egirl voice model

• Uploaded: <t:1702037275:d>
• Server: AI Hub
• Likes: 0
• Lang: EN

Epochs

480

Technology

RVC

Algorithm

Rmvpe

Requested by jairangelo · May 11, 2024, 9:21 PM

[unreal thunder]

Hi

https://drive.google.com/file/d/1446Zp4PHV4UiLLeBIuAj1FZpDnpwz2zM/view?usp=drivesdk

Someone turn this voice into huenning please

Hugginface

PLEASE

[violet heron]

Make the cover yourself, we have guides #1159513888199540817

[restive basin]

is there a way to make audio outputs not suck the longer they are? the first ~5-10 sec are completely fine, but it degrades extremely fast if its anywhere over that length.

[brittle wing]

where do i get a rvc model

[restive basin]

#1175430844685484042 is a good place to search

[valid fractal]

32 # File exists, proceed with creation of creds and client

[brittle wing]

can i not build my own?

[valid fractal]

whats up with this? its causing me to be unable to make anything

[violet heron]

You can

On the “train” section of rvc

[restive basin]

is RVC a text to speech or a speech to speech model I... honestly am not sure what I am doing lol

[valid fractal]

how do i fix this

[violet heron]

Speech to Speech

[restive basin]

ah so I probably need help with tortoise, im new to this lol

[rare goblet]

Ayo? @restive basin level 1 !!!

[violet heron]

You might want to consider using GPT So Vits instead for tts

https://github.com/RVC-Boss/GPT-SoVITS

(Don’t ask me for help I don’t really know it well)

[restive basin]

ill check it out, thanks.
it did end up doing better with more samples, I just forgot to turn them up lol

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

📚 All-In-One English documentation

https://docs.aihub.wtf/

Full AI Voice Model Training Guide (Local)

Link: YouTube
Credits: Christopher Villanueva

Model training with Mainline RVC

Link: Rentry
credits: Raven (ravencutie21)

AICoverGen Colab Guide

Link: Google Docs
Credits: Eddy (Spanish Helper)

Create a model with RVC disconnected (colab)

Link: Google Docs
Credits: Angetyde

[dawn yoke]

whats the best rvc to use?

[naive violet]

What does extra do (also, I was wondering if If there's a way to not the GPU, dual gpu could also work)

[thin stump]

@alpine breach click here to see the runtime you have left

Again it doesn't mean you finished training. It's google warning you that they will kick you out

[thin stump]

2 hours left so keep on going

[alpine breach]

yeah i can wait 2 hours my data set 28 min so can i have ot point before 2 hour as ur experience

[thin stump]

yup great

[alpine breach]

what to do after ?

if i see clear ot

and i got step number

on ot

[thin stump]

If you're done training, then you just save the steps and epochs before the overtraining point

you can download multiple files to test it out if you want

then you package it in a folder with the Added_index in your google drive

[alpine breach]

im sorry because i didit understand fast my first experience

but for example

my ot is here

[thin stump]

it can keep on training

that's not overtraining

[alpine breach]

what to do now

ik

just example

to know what to do in case

because u wil no be here after 1 hour

[thin stump]

kk, here's your folder

[alpine breach]

yes

[thin stump]

click on it. Mangio fork > weights> Steps_epochs

it only appears when you stopped training

you trained the index already right?

[alpine breach]

idk realy what i did i just followed doc aihub

[thin stump]

k

so you should package your files like this. And test out different epochs if you want

[alpine breach]

yeah there no yet step epoch folder

what i understand

is when i see ot cleat

wil press stop

and go to folder

[thin stump]

not really, just keep using the runtime to see if g/total still wants to go down

[alpine breach]

then will searc for the ot step

in folder u said ?

[thin stump]

yeah, the mangio fork folder

[alpine breach]

from graph

detecet step number

ot

[thin stump]

just post your picture of your progress when you're sure g/total won't go down again

[alpine breach]

here ?

[thin stump]

ye

[alpine breach]

iwill

ty so much

[tacit tinsel]

Hi has anyone heard of Sovits4.0 ?

[alpine breach]

bruh

[thin stump]

what?

[hot lagoon]

What is going on?

[alpine breach]

ilost the work just now but i can do it tomorrow ty

but btw if u can teach me what to do

even not yet ot

[thin stump]

yeah that's why you have the monitor the usage time

google says maybe you have 1 hour left

[alpine breach]

yeah it was still 50 min

but was conexion prbleme

[proper shale]

Extra makes the overall result better - but takes a toll on your CPU.

Well you can't really make it not use the GPU (unless you want laggy stuff... ) so you could make another PC run Okada, and open the link in the PC you run the game on... that's surely an idea

[alpine breach]

his grph let say here is ot to learn how to sauvgrade epoch section

[thin stump]

is it still training or did it stop?

[alpine breach]

this from scratch

[proper shale]

Maybe you could try the built-in realtime RVC tbf. Could help on less delay: https://rentry.co/RVCRealtimeGuide

[alpine breach]

tomorow wil do the work again

just wana know wwhat to do i want do exercice on this one ik its not yet ot

[thin stump]

or use another google account if your impatient

[alpine breach]

its late here 2.35

i cant wait again iwil do it again its ok just wana learn how

to sauvgrade when i see ot

like an example or exercice

to be ready tomrrow

[thin stump]

alright. you should stop the training when you're sure it's done, or save your progress when you're sure google will kick you out. It's all in the guides

[alpine breach]

oh this gnna be nice if its on guide by picture

wicl check for that

[thin stump]

https://docs.aihub.wtf/rvc/cloud/rvc-disconnected/
https://docs.aihub.wtf/rvc/resources/epochs--tensorboard/#--monitoring

[alpine breach]

to learn it

i need download only the ^pth ?

in weight ? thats all

[floral kiln]

hey guys is rvc better than w okada for voice quality and fidelity?

[thin stump]

In the weights folder, it'll be called e__s*.pth

e stands for epochs, and s is steps

you're only looking at the steps since the graph doesn't tell you epochs

[alpine breach]

ok got it i apreciaite alot ur help

ty so much gn bruh

[pure scroll]

some says rvc some says w okada

i too wanna see

try testing them maybe...?

[tacit tinsel]

what settings should i focus on, if the inference output still sounds too much like the original?

or should i get a better base model?

[floral kiln]

i read about them and i think rvc is better but more difficult to use

[rare goblet]

Ayo? @floral kiln level 1 !!!

[floral kiln]

and w okada is more user friendly

[wintry palm]

-rvc

[azure marsh]

Documentation

📚 All-In-One English documentation

https://docs.aihub.wtf/

Full AI Voice Model Training Guide (Local)

Link: YouTube
Credits: Christopher Villanueva

Model training with Mainline RVC

Link: Rentry
credits: Raven (ravencutie21)

AICoverGen Colab Guide

Link: Google Docs
Credits: Eddy (Spanish Helper)

Create a model with RVC disconnected (colab)

Link: Google Docs
Credits: Angetyde

[barren canopy]

-rvc

[azure marsh]

Documentation

📚 All-In-One English documentation

https://docs.aihub.wtf/

Full AI Voice Model Training Guide (Local)

Link: YouTube
Credits: Christopher Villanueva

Model training with Mainline RVC

Link: Rentry
credits: Raven (ravencutie21)

AICoverGen Colab Guide

Link: Google Docs
Credits: Eddy (Spanish Helper)

Create a model with RVC disconnected (colab)

Link: Google Docs
Credits: Angetyde

[cyan spear]

i just trained this entire model

but when i clicked download it said error

[dawn yoke]

for like the pitch what should i put it to for like a woman voice on a male song

[cyan spear]

how do i get my stuff back up

[thin stump]

g/total

[cyan spear]

how do i do that

[thin stump]

you type it in

[cyan spear]

oh

now what

i dont know how this thingy works

[thin stump]

ok start reading from right here https://docs.aihub.wtf/rvc/resources/epochs--tensorboard/#-setting-up

[cyan spear]

whats a scaler

[thin stump]

-12. If it doesn't sound good then the male voice is too deep

[brittle wing]

What vocal extractor was a game changer for you guys? What came out the cleanest for you?

[fast carbon]

Anyone encountered this "utils" error before?

Traceback (most recent call last):
  File "restapi\MMVC_Rest_Fileuploader.py", line 92, in post_load_model
  File "voice_changer\VoiceChangerManager.py", line 170, in loadModel
  File "voice_changer\RVC\RVCModelSlotGenerator.py", line 37, in loadModel
  File "voice_changer\RVC\RVCModelSlotGenerator.py", line 42, in _setInfoByPytorch
  File "torch\serialization.py", line 809, in load
    return _load(opened_zipfile, map_location, pickle_module, **pickle_load_args)
  File "torch\serialization.py", line 1172, in _load
    result = unpickler.load()
  File "torch\serialization.py", line 1165, in find_class
    return super().find_class(mod_name, name)
ModuleNotFoundError: No module named 'utils'
paramDict {'voiceChangerType': 'RVC', 'slot': 5, 'isSampleMode': False, 'sampleId': None, 'files': [{'name': 'GLaDOS_e20_s1160.pth', 'kind': 'rvcModel', 'dir': ''}], 'params': {}}
RVC:: slotInfo.modelFile GLaDOS_e20_s1160.pth
[Voice Changer] post_load_model ex: No module named 'utils'
Traceback (most recent call last):
  File "restapi\MMVC_Rest_Fileuploader.py", line 92, in post_load_model
  File "voice_changer\VoiceChangerManager.py", line 170, in loadModel
  File "voice_changer\RVC\RVCModelSlotGenerator.py", line 37, in loadModel
  File "voice_changer\RVC\RVCModelSlotGenerator.py", line 42, in _setInfoByPytorch
  File "torch\serialization.py", line 809, in load
    return _load(opened_zipfile, map_location, pickle_module, **pickle_load_args)
  File "torch\serialization.py", line 1172, in _load
    result = unpickler.load()
  File "torch\serialization.py", line 1165, in find_class
    return super().find_class(mod_name, name)
ModuleNotFoundError: No module named 'utils'

Only has happened with this specific model, others have uploaded just fine.

A friend who has the exact same model and files it works perfectly for.

[sonic night]

hi, do u guys know why my rvc voice changer is SO slow? i have rtx 3050 16gb ram, ryzen 7, its slow on opening and starting/stopping voice convertion in real time... can someone help me? my rvc version is v1.5.3.17b onnxgpu-cuda

[violet heron]

Send a screenshot of your settings

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

📚 All-In-One English documentation

https://docs.aihub.wtf/

Full AI Voice Model Training Guide (Local)

Link: YouTube
Credits: Christopher Villanueva

Model training with Mainline RVC

Link: Rentry
credits: Raven (ravencutie21)

AICoverGen Colab Guide

Link: Google Docs
Credits: Eddy (Spanish Helper)

Create a model with RVC disconnected (colab)

Link: Google Docs
Credits: Angetyde

[proud elbow]

3050 laptop or desktop?
I'd suggest to download the latest stable .18a, and use the following settings

[brave garnet]

Voicechanger Settings (Okada)

⠀

Settings for Nvidia GPUs

F0 Det.: rmvpe (suggested for all series)

RTX 40-series: 80-96 chunk | +16384 extra
RTX 30-series: 96-112 chunk | +16384 extra
RTX 20-series: 112-128 chunk | +16384 extra
GTX 16-series: 128-192 chunk | +8192 extra
GTX 10-series: 128-192 chunk | +8192 extra

Advanced Settings

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

⠀

Made with love by Ilaria 💖

[sonic night]

desktop

[rare goblet]

Ayo? @sonic night level 1 !!!

[brave garnet]

Voicechanger Download (Okada)

⠀
Download for Nvidia GPUs
Version 18a cuda

Download for AMD GPUs
Version 18a directml

Download for Intel GPUs
Version 18a directml

Download for Mac
Version 17b Mac
⠀

Made with love by Ilaria 💖

[sonic night]

ill try

[old oracle]

-colab

[azure marsh]

☁️ Google Colabs

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

How to use?

🤗 Hugginface Spaces

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

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

[sonic night]

yes, i see the difference, thanks, i didnt knew that there was a new version

[sonic night]

is there any way to change voice faster? or it is normal?

[thin stump]

@mellow vector my dms are open by the way. It's good, but I want to look at the noise profile

[rare goblet]

Ayo? @thin stump level 26 !!!

[thin stump]

since im working on my own dataset rn

[sonic night]

i got rtx 3050 and ryzen 7 5700

[crude bolt]

what is the delay you have ? and what did you expect it to be ?

for a no crackled sound / always stable ... you can expect the delay to be between 2-3 seconds.

unstable i get it under 1 second.

[crude bolt]

let me know what you did if you got it 100% not crackled / no popping supersmooth stable under 1 second.

i got a 4080rtx (lets get that argument of better hardware out the way)

CPU and my GPU both are eating out its nose (< 30% use).

[wispy lodge]

With wokada?

It's pretty much stable for me at 100ms

It's not really usable at >500ms latency

[crude bolt]

well we can have a look at it then - cuz i suspect my whole problem to lay around a program i dislike (Voicemeeter)

and with ASIO i end up with blue screens / restarts or things breaking up as soon the audio rates are not the same (eg 44.1khz 16 bit , vs 48khz 24 bit etc)

so im using atm wasapi

if you got time , we can walk it through on a voicechannel. I'd appreciate it

[wispy lodge]

I'll probably be able to join in the evening my time

Traveling atm

[crude bolt]

sure np - what time is it for you currently ? my time now is 11am

[wispy lodge]

12am rn

[honest stirrup]

Hey guys, i know this gets aked a lot but, how do i fix harmonies? like i cant get a good version of solo vocals, just play with the aggression setting on UVR?

[crude bolt]

okay we will have to find a more suitable day in the weekend ;3

[crude bolt]

can you send me the song / mp3 ? some songs are just hard to clean up

[honest stirrup]

Sure, do u want the original one?

[crude bolt]

yes

i will run it through mine and see what i got , and send my results back to you

[naive violet]

visible confusion

[naive violet]

let's say i buy a 2nd GPU, can i make one do specifically gaming and one rvc?

[wispy lodge]

Yes

[tough fiber]

what is thisd

graph

weeird

[raven glade]

guys I want to make a sound I wont make you sing ill just chat, after setting up the sound can you help why the sound doesnt come out properly from the elevenlabs after the sound flow

[rare goblet]

Ayo? @raven glade level 1 !!!

[rotund prawn]

How the hell can I upgrade

[limber gazelle]

does anyone knows how to upload pth and index to google drive in applio? i pasted the link to the folders in my google drive and it keeps showing errors

[rare goblet]

Ayo? @limber gazelle level 1 !!!

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

📚 All-In-One English documentation

https://docs.aihub.wtf/

Full AI Voice Model Training Guide (Local)

Link: YouTube
Credits: Christopher Villanueva

Model training with Mainline RVC

Link: Rentry
credits: Raven (ravencutie21)

AICoverGen Colab Guide

Link: Google Docs
Credits: Eddy (Spanish Helper)

Create a model with RVC disconnected (colab)

Link: Google Docs
Credits: Angetyde

[brave escarp]

does anyone knows how to register a voice model on @wide breach? what model link should i use for it to work properly?

[rare goblet]

Ayo? @brave escarp level 1 !!!

[proper shale]

You're gonna need another PC tbh

Unless the game allows you to choose what GPU you want to use

[toxic ermine]

which rvc is active rn

[rare goblet]

Ayo? @toxic ermine level 1 !!!

[toxic ermine]

ilaria is offline i guess

[proper shale]

https://colab.research.google.com/github/iahispano/applio/blob/master/assets/Applio.ipynb should be working

[toxic ermine]

ty trying

[final lodge]

Question. I have some RVC models and in the zip there are 2 .pth files that start with D and G and also a "config" json file. Are these all V1 RVC's?

I noticed all the V2's seem to be a .pth and an index file.

(Sorry if this is a obvious question.)

[proper shale]

that's a SoVITS-SVC model

probably

RVC always had .pth and .index

[proper shale]

It isn't, dw bout it

You'd have to be around since April last year to know

[toxic ermine]

@proper shale yo its really complicated to use. how do i upload a voice model? i got link

from train section?

[proper shale]

there's a guide for it

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

[final lodge]

No I'm relatively new I stumbled upon a ton of generative AI stuf all at the same time still trying to catch up! Really appreciate it. I will check and see if they are indeed SoVITS-SVC's. How do they compare in quality?

[proper shale]

RVC is just faster and better tbh. Most SVC models aren't really good since people didn't really know how to make models

[toxic ermine]

ty

[proper shale]

yw :)

[final lodge]

BREAKING MY HEARRTTT... why did I grab so many of these.. ffkk

Guessing they can't properly be converted either. One last question... What is the best source for bulk download of quality v2 models. Just trying to build a library and don't want to download one by one.

[rare goblet]

Ayo? @final lodge level 1 !!!

[proper shale]

Dunno if bulk download is possible... but there's a couple of good resources for V2 models

#1175430844685484042
https://weights.gg
ooor you can search Huggingface repos... lots of models uploaded there

[wispy lodge]

Nah, it's actually a bit easier. Games always run on a GPU that has the monitor plugged in. So you can just choose a GPU in wokada (and maybe in RVC realtime too)

[barren canopy]

I don't fully understand mode collapses, but from what I understand my D is having a ton while G is fine. Is this normal?

[rare goblet]

Ayo? @barren canopy level 1 !!!

[barren canopy]

Ofc it's with outlier data & smoothing set to 0 as the local guide in the server wanted.

[opal pumice]

Are you using a pretrain? I don't think I've seen a starting point that high before

[barren canopy]

No my Mangio fork didn't come with a pretrain and when I tried loading one from huggingface I got size mismatch errors for the tensors

[proud elbow]

training from scratch will never be as good as using a pretrain. if ur looking for it, go to #1235952130855010365

[barren canopy]

Thank you!

[proper shale]

might be fine, check the model itself

[proper shale]

ah nvm

[opal pumice]

Full of static, they posted a clip before

[barren canopy]

Was just static when I tried inference with it, gonna try with a pretrain to see if it works

[proper shale]

ah... without pretrain it's screwed up ye

[barren canopy]

Ah I've now screwed the pooch:

start preprocess
['trainset_preprocess_pipeline_print.py', 'C:\\Users\\ashto\\OneDrive\\Desktop\\Mangio-RVC-v23.7.0_INFER\\Mangio-RVC-v23.7.0\\datasets', '32000', '8', 'C:\\Users\\ashto\\OneDrive\\Desktop\\Mangio-RVC-v23.7.0_INFER\\Mangio-RVC-v23.7.0/logs/harleyQuinn', 'False']
C:\Users\ashto\OneDrive\Desktop\Mangio-RVC-v23.7.0_INFER\Mangio-RVC-v23.7.0\datasets/dataset.wav->Traceback (most recent call last):
  File "C:\Users\ashto\OneDrive\Desktop\Mangio-RVC-v23.7.0_INFER\Mangio-RVC-v23.7.0\my_utils.py", line 117, in load_audio
    ffmpeg.input(
  File "C:\Users\ashto\OneDrive\Desktop\Mangio-RVC-v23.7.0_INFER\Mangio-RVC-v23.7.0\runtime\lib\site-packages\ffmpeg\_run.py", line 325, in run
    raise Error('ffmpeg', out, err)
ffmpeg._run.Error: ffmpeg error (see stderr output for detail)

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "C:\Users\ashto\OneDrive\Desktop\Mangio-RVC-v23.7.0_INFER\Mangio-RVC-v23.7.0\trainset_preprocess_pipeline_print.py", line 84, in pipeline
    audio = load_audio(path, self.sr, DoFormant, Quefrency, Timbre)
  File "C:\Users\ashto\OneDrive\Desktop\Mangio-RVC-v23.7.0_INFER\Mangio-RVC-v23.7.0\my_utils.py", line 141, in load_audio
    raise RuntimeError(f"Failed to load audio: {e}")
RuntimeError: Failed to load audio: ffmpeg error (see stderr output for detail)

end preprocess

I read earlier in the server when this happened, it was due to the file variable pointing to the .wav instead of the folder, but I have it pointing to the path with the .wav inside (or having spaces in file name):

C:\Users\ashto\OneDrive\Desktop\Mangio-RVC-v23.7.0_INFER\Mangio-RVC-v23.7.0\datasets

[rare goblet]

Ayo? @barren canopy level 2 !!!

[opal pumice]

No expert, maybe someone else can tell, could OneDrive be causing an issue?

[barren canopy]

I suspected that, so I moved onto different harddrive completely:

D:\harleyAI\vocals and still get the same error of failed to load audio

[opal pumice]

Well damn

[proud elbow]

a dataset folder with wav files not being corrupted should work

[cerulean hedge]

ok

[barren canopy]

So you think the wav file might just be corrupted? I can open and run it in media players. The only file in the harleyAI folder is dataset.wav

[proud elbow]

and same format? (e.g. 44.1khz 32-bit float)

[barren canopy]

As long as I'm understanding, yes

[tame mica]

try using ilaria's audio analyzer

[green shell]

yo guys which model is the one that can make ai songs?

[tame mica]

suno.ai

[green shell]

rvc and okada can't do that right?

[proud elbow]

and redownload ffmpeg.exe, or the entire mangio perhaps?

[tame mica]

nope

[green shell]

kk

[barren canopy]

Will do after utilizing Ilaria's audio analyzer

[proud elbow]

or alternatively mainline RVC or Applio 3.1.1

[barren canopy]

the output for reference

[barren canopy]

I might try, which do you suggest for intuitiveness?

[proud elbow]

both are good, but Applio has crepe with hop length, which is same as mangio crepe

[tame mica]

hmm looks about right .. don't know what the problem is

[barren canopy]

It still remains if I change file path, moving it around even onto another drive D:

Very unfortunate

[mystic marsh]

-colab

[azure marsh]

☁️ Google Colabs

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

How to use?

🤗 Hugginface Spaces

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

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

[pulsar anchor]

How do I... select voice model?

[tame mica]

have you installed any? :3

[pulsar anchor]

I have downloaded them

but

there is no folder for me to place model in

[restive basin]

How does RVC work is there like a chart or documentation in English, I want to learn abt the technology

[finite galleon]

Click refresh and on the dropdown if you have downloaded a model already

[tame mica]

iirc the pth file goes to the weight folder and the index goes to the logs folder

[pulsar anchor]

ooh wait

ah I found out now

I must set these things in here

the damn website

Remind me of AUTOMATIC1111

Now how do I set these things for singing?

whhaa

oke

[proud elbow]

leave anything default but the pitch shift according to what kind of voice of the sample and model

[pulsar anchor]

hmmmmm

Where can I find more voice model?

oke then

Ig Applio is a way to go for make AI Covers

ty a lot

I'll chose this

When I try to record

my microphone volume automatically keep decreasing

this is so weird

How to solve this?

What....

yeah

[barren canopy]

Update for anyone using search later on and running across this error: Just use applio please. Fixed the error.

[pulsar anchor]

Are there even good software for singing tbh?

All seem mid...

Appelio seem mid rn

I try to find an alternative that look like kits.ai

this fucking website is an embodiment of fucking Greed

Mfs want money

urgh

I'll shove my burnt money up their ass if I fucking can

[rare goblet]

Ayo? @pulsar anchor level 4 !!!

[ancient needle]

why do I keep getting (PaErrorCode -9993) nonstop

[pulsar anchor]

This automatic microphone volume changing...

A problem which CANNOT be solved

Somebody please help me, this video will explain the problem

[ancient needle]

and it's even giving me that same error code for the voices you get whenever you have the whole thing installed (something to do with illegal voices or something)

[sacred bane]

guys i need method to use voice changer in call of duty any help?

[pulsar anchor]

Do you have alternative to Appelio?

this shit suck now

and I am pissed

[brisk salmon]

what should i put here ?, in extracting feature ?

it saw error

i can't use rvc cause there no gpu

[golden zenith]

Hi everyone, sorry for my English, but I'm trying to make a cover with Google Colab and when I try to start step 2 by putting the vocal model, I get this error:

JSONDecodeError Traceback (most recent call last)
<ipython-input-8-fb90c280b89e> 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)

Do you know how to solve it?

[marble forum]

Please can anyone recommend how I can get movie clips/dialogue from movies?

Say I want clips of Optimus prime talking. How can I get those ?

[azure marsh]

Suggestions for @golden zenith

☁️ Google Colabs

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

How to use?

🤗 Hugginface Spaces

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

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

[golden zenith]

Thank you so much ❤️

[rare goblet]

Ayo? @golden zenith level 1 !!!

[tame mica]

you cannot train a model without a gpu

[brisk salmon]

a shi-

okay

i'll wait again colab have back the gpu

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

📚 All-In-One English documentation

https://docs.aihub.wtf/

Full AI Voice Model Training Guide (Local)

Link: YouTube
Credits: Christopher Villanueva

Model training with Mainline RVC

Link: Rentry
credits: Raven (ravencutie21)

AICoverGen Colab Guide

Link: Google Docs
Credits: Eddy (Spanish Helper)

Create a model with RVC disconnected (colab)

Link: Google Docs
Credits: Angetyde

[delicate skiff]

yo watah el

nbete ar

helps

easyGUI is woriking?

[visual flint]

Using Applio RVC. Im using the text to speech and if a message is so long then it creates a pause mid-sentence as shown in the screenshot, the problem is that this occurs mid-sentence rather than on a full stop so its annoying having to split everything up. is there a way around this?? thanks

[barren canopy]

Would the noise be due to dataset quality or not being trained long enough?

[proper shale]

outdated colab, use the ones linked in https://docs.aihub.wtf/

[delicate skiff]

okeoprf

[proper shale]

Use a different TTS tool - maybe that'll fix it

[delicate skiff]

im =ussing applio

[visual flint]

Suggestions? Applio seems the best with TTS from what I've seen

[proper shale]

Applio uses edgeTTS which is alright tbh

Maybe try ElevenLabs

[visual flint]

Not worth paying for what im using it for lol

[proper shale]

Unless you're going to use it that much you don't really need to pay - 10k characters is enough tbf

[visual flint]

Is that a 10k character limit permanently or per use? I assume i can do TTS with a voice model for free?

[wide breach]

google drive or huggingface

[brave escarp]

oh youre so gentle! thanks!

[wide breach]

np, incase you encounter any error feel free to message me ill look into it

[dusty mortar]

which program do you use for a 32k pretraining model? i tried mangio but it doesnt have the 32k option

[pastel zealot]

What time is it for y’all

[low shard]

9 pm

[sage bramble]

use havrest, they will use your cpu

[fluid matrix]

please dont use harvest...

[fluid matrix]

why would you want to train on your cpu.

[brisk salmon]

Colab can't use GPU rn

[rare goblet]

Ayo? @brisk salmon level 2 !!!

[brisk salmon]

And I have a AMD CPU and GPU

So I cant use it in my pc

[sage bramble]

Because I use a VPS to train my model

[fluid matrix]

uh

how long does each epoch take for you

[brisk salmon]

Or is it possible to use AMD GPU and CPU to train ?

I have a Rx 6800 xt and a Ryzen 5 5600x

[fluid matrix]

and how did that cpu not explode yet

[sage bramble]

25 sec

[fluid matrix]

does not sound right tbh

[brisk salmon]

[fluid matrix]

cuda should be an requirement

[sage bramble]

Watch

[fluid matrix]

if you use windows you cant really train on amd

[brisk salmon]

1000 epoch, god damn

A shit

[fluid matrix]

2k steps and 1k epochs

oh my god

[brisk salmon]

That's huge

[sage bramble]

They took me 7 hours

[fluid matrix]

no thats the thing

this is nothing

[barren canopy]

At a certain point you're overfitting I almost guarantee it @sage bramble

[rare goblet]

Ayo? @barren canopy level 3 !!!

[fluid matrix]

did it even train????

[barren canopy]

There's 0 ways 1000 epochs doesn't overfit no matter the size of the dataset

[fluid matrix]

@red kayak check this out

[sage bramble]

yes

[brisk salmon]

Okay I'm way too confused, I just wanted to train model, but I'll just wait colab work then ;_;

[fluid matrix]

which one are you using

rvc disconnected?

[brisk salmon]

Nope

Easy gui

For train

[fluid matrix]

https://docs.aihub.wtf/rvc/cloud/rvc-disconnected/

try this one

[brisk salmon]

Okay I will try tomorrow then

I turn off my pc

[fluid matrix]

the easygui you used might be outdated

[brisk salmon]

Oh okay

But Alot of times GPU are overused in colab

So we can't use it freely

[sage bramble]

Huge ?

[rare goblet]

Ayo? @sage bramble level 2 !!!

[barren canopy]

yes

[brisk salmon]

Godamn

[barren canopy]

That's about the size of a Call of Duty modern game

If anyone ever sends me that I would shatter their monitor

[brisk salmon]

[barren canopy]

That's practically a zipbomb

[red kayak]

Wha

I don't mnw

No context

[fluid matrix]

just scroll up a tiny bit

[barren canopy]

The context for you @red kayak

[fluid matrix]

@red kayak

[barren canopy]

oh shit

diablo was talking abt something else

[fluid matrix]

nah scroll from there

[barren canopy]

Ohhh

ait

[red kayak]

Harvest is the least of my concerns

Why are u training for so long

[fluid matrix]

this doesnt even sound right

[red kayak]

And with no tensorboard

[barren canopy]

C:\Users\ashto\Downloads\ApplioV3.1.1\logs\LinManuelMiranda | epoch=2 | step=58 | time=15:40:55 | training_speed=0:00:12 | lowest_value=24.555490493774414 (epoch 1 and step 26)

GOAT

[fluid matrix]

no way this is using a cpu

i think he is just confused

and even if everything worked

2k steps and 1k epochs oml

[barren canopy]

I don't know, pytorch can be on a CPU and burn the shit up

I once accidentally loaded a small GPT-2 onto my CPU and tried fine tuning it

[fluid matrix]

1 epoch would take ages first of all

[barren canopy]

Worst mistake of my life

[red kayak]

Lowest point doesn't mean ideal epoch

It's probably a mode collapse

Meaning ur model is stuck In local minima

And has overtrained

:p

[barren canopy]

Well I doubt the overtraining part bc its the second epoch but I get you

That's why we have tensor

Is Applio's OT detection worth turning on? Unsure of how it works under hood.

[red kayak]

No if it's a mode collapse that means the model is having a tough time producing newer results

[barren canopy]

Don't mode collapses stabilize after? That was my understanding from the tutorial video linked in the rvc embed. It only becomes an issue with an excessive amount of mode collapses (in the model).

Can anyone help explain what I'm looking at? I'm unsure what I'm meant to actually be looking for.

[red kayak]

Not how it works no

[red kayak]

G/total is the generator and d/total is your discriminator graph

[barren canopy]

What are the features that I'd like to see? When it pans out (1200 steps on the discriminator graph) is that an indication of overfitting, because I know when it happens in training NLP models and you see a plateu that looks like it's about to go back to increasing loss, it's an indication of an overfit model.

[red kayak]

Sorta yeah but ur graphs are noisy here and usually it doesn't always end up with a plateau it may also start decreasing

[barren canopy]

How might I decrease the noise? Perhaps turning off Applio's OF detection to let it run for longer (because it auto-stopped itself due to the detection module)

[red kayak]

Ohh no, In TensorBoard, a "noisy graph" could refer to visualizations of data that contain irregularities or fluctuations. This might be represented in graphs showing training loss or accuracy over time, where noise could indicate variations in performance due to factors like randomness in data or fluctuations in training conditions.

[barren canopy]

Ohhhh!

[acoustic lion]

Could someone recommend me settings for Mangio settings(Realtime voice changer)? I tried looking for guides that spoke about it but I found nothing

[brittle wing]

?

do i just copy the zip file into here as well??

im just gonna do that and hope it works out

[acoustic lion]

Also, I am having trouble with EasierGUI, no clue on what leads to this problem

[pastel oak]

https://rentry.co/RVCRealtimeGuide#gui-settings-explanation

[acoustic lion]

Isn't this for W-Okada?

[pastel oak]

Yours will look different and less information because youre using an old mangio version but will still be roughly the same

if you open the link you will clearly see a picture that looks like the same as yours

[acoustic lion]

Eh? Old Mangio? I just downloaded it

[pastel oak]

...that is not the point. you still downloaded an older version

if you downloaded the binary release, that was uploaded on july 27 2023

[acoustic lion]

Is there a way to update it by chance?

[pastel oak]

scroll up on the guide

theres download links

[acoustic lion]

RVCNvidiaLatest, correcT?

[pastel oak]

if you have nvidia gpu then yes

[mild osprey]

hi, I downloaded the program, I have a graphics card, the AMD 6600, but when I am in the program the card does not appear, it only says cpu

i need help : (

[rare goblet]

Ayo? @mild osprey level 1 !!!

[median heath]

Hi i just danllowed this again and when i turn my miic on i see this

Voice Changer] Waiting generate pipeline...
Voice Changer] Pipeline is not initialized.

[pastel oak]

i assume you downloaded wokada. you downloaded the wrong version

Follow the guide
https://rentry.co/VoiceChangerGuide

@mild osprey i personally suggest since you have an amd gpu you try out the other voice changer which runs better
https://rentry.co/RVCRealtimeGuide

[pastel oak]

delete pretrainand model_dir and launch start-http again

it will redownload everything cause something went wrong

[mild osprey]

I'll try it, thank you very much

[brazen skiff]

Does anyone have a guide on how the tensorboard works?

[pastel oak]

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

[late gate]

How to post own model on HF?

[thin stump]

@alpine breach I wrote this guide so you can't be confused on exporting and resuming https://rentry.org/rvc-disconnected-saving-and-resuming

I lost my edit code though

[tacit tinsel]

what settings should i focus on, if the inference output still sounds too much like the original?
or should i get a better base model?

[vernal relic]

heyo!

hi

so these two or anyone else reading

I've got a similar issue and I'm hoping someone can help

I've downloaded applio

and I try to run the program but the terminal stays empty

as you can see, run applio is on there, and it's the one i use

but there is no "install" batch file

when i run applio as admin

it wants me to run that install file

anyone able to help?

nvm!

i got it to work I

am dumb

[proper shale]

10k monthly

and no tts models for free sadly, but i normally just convert with RVC anyway

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

📚 All-In-One English documentation

https://docs.aihub.wtf/

Full AI Voice Model Training Guide (Local)

Link: YouTube
Credits: Christopher Villanueva

Model training with Mainline RVC

Link: Rentry
credits: Raven (ravencutie21)

AICoverGen Colab Guide

Link: Google Docs
Credits: Eddy (Spanish Helper)

Create a model with RVC disconnected (colab)

Link: Google Docs
Credits: Angetyde

[leaden lintel]

Im on the easy gui google collab and it seems the twilight links (all of the two ones on here at least) have this JSON decode error idk is says JSONDecodeError: "expecting value: line 1 column 1 (char0) im not sure how to fix it or if those ones just dont work

[naive violet]

I see

[proper shale]

1st - What settings are you running? (F0 Det, Chunks, Extra, maybe even Index)

2nd - What's your GPU?

Sorry if I didn't make myself clear on the first question (English is not my native tongue - I still have some issues with it even though I've been learning for soooo long), I was asking how you set those up, just so I could see if you were running chunks too low or something, which could cause the stuttering n stuff.

Anyways, the 1050 may have some issues running the voice changer... I'd say keep extra low, chunks high, and use either harvest or crepe_tiny as your f0.

It'll have high delay sadly, so might not be all too worth it. Perhaps you could use Colab, which runs the stuff in the cloud: https://colab.research.google.com/github/hinabl/voice-changer-colab/blob/master/Hina_Modified_Realtime_Voice_Changer_on_Colab.ipynb

[wind cove]

dunno if this is the right channel for this but ive run out of "compute units" on google colab is there anything else i can use

[thin stump]

Use a different account to train. Since you didn't export your model checkpoint, all your progress is gone unfortunately #✨│ai-help message

[wind cove]

thanks ig but i already tried that with my other accounts and i dont want to make a new one or use a vpn i was asking if there were any alternatives sorry

[thin stump]

Local training needs a good GPU. That's the only alternative

[wind cove]

oh wait no im not training models i just want to make funni covers and stuff with them

[proper shale]

weights.gg is always there

[wind cove]

im asking if there are any alternatives for that

k thanks

[thin stump]

oh my bad, I personally use tryreplay.io. Same devs as weights

[vivid saffron]

I need help

[proper shale]

Explain your issue and someone can try to help

(I might sleep soon, sorry 😭)

[alpine breach]

guys what to do now google collab wil kil me

kick me

[rich rivet]

What delays are normal? Mine is always over 1 second using the lowest settings possible with https://github.com/w-okada/voice-changer (onnx directML) using AMD RX6900 XT.

I'll test with RVC Realtime tomorrow to see if it performs better.

[alpine breach]

wi need stop run before collab kick me out ?

[rare goblet]

Ayo? @alpine breach level 8 !!!

[thin stump]

https://rentry.org/rvc-disconnected-saving-and-resuming

please read it

it's so short too

[alpine breach]

will

but its soon wil out of time i need do something fast ?

[thin stump]

we don't what is "soon" so save it already

[alpine breach]

this what i they said after more than 2 .15 hours of raining

trainign

i need stop the stop runing now ?

[thin stump]

did you read the guide

[alpine breach]

i do now

just im confused

[thin stump]

save the latest checkpoint? It's not even 1 page long. Honestly skill issue

[alpine breach]

i nees stop run time before downlding bevcause i dont find index in log

[thin stump]

your added index is already in your google drive

[alpine breach]

so i download only last epoch step

?

is that all ?

[thin stump]

the latest one that isn't overtrained

yes

[alpine breach]

this one

so i need stop traing ?

[thin stump]

e_155_s11005

stop the training

and export https://rentry.org/rvc-disconnected-saving-and-resuming

[alpine breach]

yeah

i pressed download from the folder

is that all ?

ok ty so much

wil stop it finaly

[thin stump]

Did you export it? You're not being clear

downloading it doesn't mean you can just resume

[alpine breach]

i want resume later

wil export now

[thin stump]

Ok, great. You can continue training after that or leave colab

[alpine breach]

ty so much i found the file add u right

[thin stump]

yes, the added index

[alpine breach]

yes is was there

for the export

i dont need to change noting ?

i aleardy did it like that

i see now

[thin stump]

I believe there's a language barrier here. But it wouldn't make sense to export something that doesn't exist, right? Export e_155_s11005

[alpine breach]

ty so much i was confused and afraid to click somthing wrong after 3 hours

[thin stump]

epochs 155, steps 11005

don't thank me yet until you completed it

[alpine breach]

i need put this values in place of 0000 ?

[thin stump]

yes

[alpine breach]

if they kick me out i can downleded files that aleardy exist ?

[thin stump]

I don't know if you exported it or not so don't come whining in general-chat

[thin stump]

good, run that

[hollow loom]

i have a small issue where my ai breaks so bad when making the lyrics to a ai cover

[hollow loom]

i'm using rvc gui to make the lyrics

[alpine breach]

this is my folder i see in file they put last one

is normal ?

just need changed it with 155

?

thats it ?

[thin stump]

I think it's saved in your D_2333333.pth. Please read the guide here on resuming. It's not even a half page long. So again, skill issue if you don't understand https://docs.aihub.wtf/rvc/cloud/rvc-disconnected/#section-17

[alpine breach]

i read it em all u send me im sorry to be skill issue mb

[thin stump]

It's not exclusive to rvc gui. Some Ai models weren't trained on particular words or high notes. You have to remove the reverbs in songs too

for the most parts, the lyrics can cover up the bad parts but its noticable

[hollow loom]

dang so i need to just find songs that don't have reverb in them

i was trying to do a cover of it's my life by bon jovi and it broke so bad

[thin stump]

dms? I could try

[hollow loom]

sure hang on

[rare goblet]

Ayo? @hollow loom level 1 !!!

[alpine breach]

for now im done ? when i want resuming wil follow the guid resuming thats all ?

or now dont need to change nothing idi it good ?

[thin stump]

yeah, you can resume tommorow

[alpine breach]

ty so much u saved my night lol

gn all

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

📚 All-In-One English documentation

https://docs.aihub.wtf/

Full AI Voice Model Training Guide (Local)

Link: YouTube
Credits: Christopher Villanueva

Model training with Mainline RVC

Link: Rentry
credits: Raven (ravencutie21)

AICoverGen Colab Guide

Link: Google Docs
Credits: Eddy (Spanish Helper)

Create a model with RVC disconnected (colab)

Link: Google Docs
Credits: Angetyde

[proud elbow]

nah reverb is quite easy to remove using foxjoy Reverb HQ (or UVR De-echo-de-reverb alternatively).
the lead/back vocal separation might be tougher thing to do.

[thin stump]

Yup definitely a backing vocals problem. I had to cheat a lot when running hp_5, hp_6, and BVE models for my AI covers. Basically mixing and matching to find the right one @hollow loom

[stoic basin]

Guys what's the best girl voice for rvc in okada

[errant plume]

@thin stump hi does the audio file for conversion must also be clear and recorded in a professional mic? cuz the dataset is good quality but the the audio for conversion was only recorded in mac's inbuilt mic

[proper shale]

not necessary to have hq mic for conversion

anything that's "clear" enough, with no bg noise is alr fine

[errant plume]

ok thanks

[rare goblet]

Ayo? @errant plume level 1 !!!

[errant plume]

@proper shale hey when resuming the training is it possible to add more dataset files?

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

📚 All-In-One English documentation

https://docs.aihub.wtf/

Full AI Voice Model Training Guide (Local)

Link: YouTube
Credits: Christopher Villanueva

Model training with Mainline RVC

Link: Rentry
credits: Raven (ravencutie21)

AICoverGen Colab Guide

Link: Google Docs
Credits: Eddy (Spanish Helper)

Create a model with RVC disconnected (colab)

Link: Google Docs
Credits: Angetyde

[proper shale]

Ya ran out of GPU time... wait for a day or so for it to come back

ooor use an alt account

btw, last night i got a ping from u, about the upload fails n stuff

make sure to not upload the index, seems to break uploading models

[rare goblet]

Ayo? @brittle wing level 2 !!!

[proper shale]

oh, right

it gives (for free users) 4 hours of GPU time per session, eventually you run out of time and have to wait 24hrs for it to come back

nope, only computational credits, which you get like 100 for $10

but those do run out

yup

yeah

[strange mirage]

anyone know how to fix harsh "P" sounds with RVC? like whenever i say a word that starts with a P or has a P in it the P is harsh as hell

[shrewd vine]

Does RVC V2 have any limitations on audio creation?

Maybe 1 account have only 5 hours of audio duration?

[rare goblet]

Ayo? @shrewd vine level 1 !!!

[pulsar anchor]

I should upgrade??

[tame mica]

if it aint broke don't fix it /hj

[proper shale]

Tbh idk if it is worth it.
You could also pay as you go, which gets you the computational credits whenever you need them instead of monthly paying, which might be worth it, but frankly I'm not sure how those would last

[proper shale]

just open http://127.0.0.1:6969 on ur browser

[pulsar anchor]

Well it did for me

it did broke before I am gonna upgrade it

[proper shale]

Might be a model issue

[proper shale]

It did not break.

[pulsar anchor]

Then explain how to fix this problem please

[proper shale]

Not really, the limitation is more on the hardware

[pulsar anchor]

If you think it's not broken

[proper shale]

again, open http://127.0.0.1:6969 on ur browser

its where applio is being hosted, in ur local network

[pulsar anchor]

sigh

Man

[shrewd vine]

Oh... Fine then. I'll try to install this AI

[proper shale]

Here is a collection of guides that will help ya along your way: https://docs.aihub.wtf/

[pulsar anchor]

I know

and what next?

[proper shale]

now just... use the thing

[pulsar anchor]

What?...

[strange mirage]

it didn't do it on w-okada

just rvc

[proper shale]

might be ur audio then...

[pulsar anchor]

Is this the part of the software?

Where it auto-adjust my microphone volume?

but

why?

I don't get it

[proper shale]

wh

[shrewd vine]

And what is the best thing?

[spare shale]

1

[proper shale]

depends on your use case, easier to use is applio

[spare shale]

Applio best

[proper shale]

not necessarily

[spare shale]

But I think it's better

But I use colab anyway x3

[shrewd vine]

What is "colab"?

[spare shale]

ilaria is also a good option

[spare shale]

Google PC

https://colab.research.google.com/

[shrewd vine]

Mm... OK

[proper shale]

Might not be

For some person

[spare shale]

smile

[shrewd vine]

What cases can there be?

[proper shale]

Look, @shrewd vine, if you have a good gpu, go for Applio or Mainline - hecj even both if you want to

[pulsar anchor]

Collab is for those who have shit pc but want to use Appelio

If I remember

[proper shale]

Can't go wrong without that

[proper shale]

Or any other RVC thing*

[pulsar anchor]

Just like Stable Diffusion ahh thing

[spare shale]

colab Yes.. If you have a bad PC you can use colab

[shrewd vine]

So, good GPU - Applio or Mainline. I'll choose Applio then

[proper shale]

Right. Let us know how that goes

[shrewd vine]

Is zip recommended?

[shrewd vine]

Yes, sir!😉

Already

[proper shale]

Honestly, not a lot that can be done.

It will have a substantial amount of delay sadly

[proud elbow]

it is same

[pulsar anchor]

hmmm

I think I know the reason why Appelio auto-adjust my volume

if my voice too loud, the result of using ai to change my voice will be really bad?

I guess AI Singer don't know how to screamfully sing a song

or even be quiet

[proper shale]

mmmm yeah kinda

[pulsar anchor]

How to make AI Voice scream?

[proper shale]

...try to scream

most models can't, though

[pulsar anchor]

shiet.

[potent root]

why is it erroring out

[tough fiber]

bro please can someone tell me wtf is this graph

[bright wren]

Hello i have a question

I was training a vm in RVC for 1000 epochs

around 442 i get this error

unforunately i am too new to know what it is or how to fix it

[tough fiber]

@proper shale mj can u check this -*-

[proper shale]

not enough space on the hard drive, so it stopped writing

or something like that

[proper shale]

nah wtf

[bright wren]

yyaaasssss thank you for the help

most likely what happened since i only had like 50gb of space left on my hd

[tough fiber]

so i didnt get what should i do

[proper shale]

switch batch size to anything else

thats very very weird

[molten dirge]

Is there an android version?

[rare goblet]

Ayo? @molten dirge level 1 !!!

[proper shale]

There ain't a specific app for RVC, but you could use weights.gg for conversion, or Colab

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

[orchid acorn]

gojo saturo

[dull orchid]

Is there anything you can do about voice cracks when inferinen

Ahh does it matter when you cant sing perfect, like not hit notes very precise,

Or do you need to have very good vocals that sound spot on, or can it also be not perfect

[idle briar]

Is this software full online?

is there possible to run it without internet connection?

im new here, sorry

[scenic path]

hi guys

[grim bay]

What is the best Separation type and Vocal model type on MVSEP for AI cover?

[tame mica]

bs roformer > viperx then denoise and dereverb first for separating vocals :3

[proper shale]

BSRoformer

[finite galleon]

MVSEP Bsroformer ver 2024.4 i think?

Just pick the bsroformer version that has the higher sdr i forgot the exact model name

[brittle wing]

rvc disconnected

anyone know how to fix this?

[quasi dagger]

You need to put the dataset in a folder named rvcDisconnected in your Google drive and then specify the file name of the zip in this colab cell (after you mounted the drive of course)

[brittle wing]

exactly what i did

as you can see it extracted them

[dusky whale]

Hi, this is a bit of an unorthodox question. I am writing a bachelor's work about voice conversion, particularly about its realtime usage. There is plenty of documentation and papers about VITS and TTS, but I've yet to find good sources about RVC specifically (any searches are filled with samey tutorials). Can you share any sources/leads/ideas of where to look? Much appreciated!

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

📚 All-In-One English documentation

https://docs.aihub.wtf/

Full AI Voice Model Training Guide (Local)

Link: YouTube
Credits: Christopher Villanueva

Model training with Mainline RVC

Link: Rentry
credits: Raven (ravencutie21)

AICoverGen Colab Guide

Link: Google Docs
Credits: Eddy (Spanish Helper)

Create a model with RVC disconnected (colab)

Link: Google Docs
Credits: Angetyde

[brittle wing]

so thats not the issue

[quasi dagger]

Indeed, is the WAV files in a subfolder in the zip? Idk why it said multispeaker

[brittle wing]

ive tried formatting it in that way now

doesnt work either

strange

does easygui work just as well

might just switch to that idk

[quasi dagger]

Maybe that "/D:" partition in the file structure causing the problem

Does this happen with other datasets?

[brittle wing]

happened to me months ago i think

i really havent trained since then

because it put me off

but i thought id have another crack at it

its annoying bc i know this model could be fucking amazing 😭
i did a test train on weights.gg to see how good it sounds and its really clean lmao

and i want to do it with the titan pretrain now

[quasi dagger]

Let me try RVC disconnected here, I'm not sure about that "/D:" in the file path

[brittle wing]

ight

i just got rid of the else statement and its giving me this

[rare goblet]

Ayo? @brittle wing level 8 !!!

[brittle wing]

so i think that fixed it?

the else statement was:

else:
#If anything else, we're dealing with multispeaker.
#Move all folders to /dataset/ indiscriminately.
print("Dataset Type: Multispeaker")
for fi in directories:
shutil.move(fi, final_directory)

do you think im safe to continue here

[errant plume]

@azure marsh hey bro if you're okay can i ask you some qs in regard to rvc in dm?

[azure marsh]

Wassup I'm Automaze!

• My prefix in this server is - (customizable with -prefix)
• Interested in how I'm built? I'm actually open source!
• Forgot a specific command? Try /help or -help

[brittle wing]

go ahead dm a bot

[errant plume]

lmao

[brittle wing]

im assuming you meant kalomaze

[errant plume]

fri thought its a real user💀

[brittle wing]

😭

[errant plume]

sed life

[split frigate]

how can i switch to gpu-cuda version
i got nvidia

[brittle wing]

i continued and i dont think i was safe to lmao

[quasi dagger]

Seems like it's how the file was zipped, notice it should show a partition such as C: or D: in the file path

*Shouldn't

[brittle wing]

hm

ah shit i see now

i zipped it using something called breezip

think i downloaded it from the microsoft store years ago

ill do the "send to" thing on windows and see if it makes a difference

well shit

@quasi dagger do you think its possible for you to zip it for me and see if it makes any difference whatsoever

i can send the zipped folder via google drive and then you unzip it and rezip it then send it back

small favour

if it doesnt work then ill just try a different method lmao

[errant plume]

@brittle wing yo!

[quasi dagger]

Does it show a dataset folder when you click on the folder icon on the left?

[brittle wing]

yep

[tame mica]

make an ngrok account and then copy the ngrok token

[quasi dagger]

Weird!

[brittle wing]

yeah 😔

[quasi dagger]

Try without the subfolder and put all WAV files in the zip, then delete the datasets folder on colab and try to run the cell again. In my example I put the WAV file without a subfolder

Like

Jablesdataset.zip

• babyone.wav
• ...

Before it wasn't working because of that partition (/D:), now it should I guess

[brittle wing]

ill try in a sec ive just disconnected and deleted the runtime rq

[quasi dagger]

Good

[brittle wing]

wait so delete the folder but keep the dataset file?

[quasi dagger]

When you delete the runtime it also deletes the dataset folder on colab (not your drive)

[brittle wing]

yeah

[quasi dagger]

Try now

[brittle wing]

wait wtf

its just worked fine now

but i didnt change anything

[errant plume]

ok its working now thx

[quasi dagger]

You changed the zip

[rare goblet]

Ayo? @errant plume level 2 !!!

[brittle wing]

oh yeah

[quasi dagger]

@brittle wing Good that it's working now 👏

[brittle wing]

its working now then