✨│ai-help

[low shard]

It's not as good as an actual TTS

[polar ridge]

Well but i can use sts to tts right?

trough applio

[low shard]

Yep that tutorial uses an over year old version of original wokada, and the settings he gave were completely fucked

Uninstall vb audio cable

Delete that old original wokada

Forget everything you get from those videos

Those are outdated info since a long time

never trust YouTube tuts for rvc and wokada

-realtime

[patent trellis]

💻 Local Realtime RVC

Guides for Programs that use RVC Models in Realtime for Calls/Games

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

Unavailable, the guide is outdated and the program is worse compared to the ones above, and much less updated

[low shard]

read the 1st link, wokada deiteris fork has way better quality and performance

Forget everything you got off that YouTube tutorial, it was just wasted time

[low shard]

Through applio yes, but it won't be expressive

Because it uses Microsoft edge tts

It has good quality and multi lingual but it's plain for reading text

[polar ridge]

So what do you suggest me to do if i want to clone a voice and train it?

[low shard]

If you ONLY want an expressive tts, don't use RVC, use eleven labs or fish speech or f5

RVC is not meant for expressive TTS at all

[polar ridge]

ok thank you

[low shard]

Yw

[low shard]

@lusty nest any updates?

[lusty nest]

following the guide right now, all is good so far

[low shard]

alr lmk for any issues

[lusty nest]

everything is sorted, thanks for the help!

[low shard]

on wokada, you can optionally use more advanced settings for benefits:

• Force FP32 mode: on (THIS IS OFF BY DEFAULT! Turning this on improves stability. Increases VRAM usage by 200 MB)
• Disable JIT compilation: off for faster loading speed of the program, on for slightly better performance (10-15 ms) for Nvidia only)
• Reduce the delay via: https://docs.aihub.gg/rvc-voice-changer/local/deiteris-w-okada-fork/#reduce-more-delay

do you may want me to check your settings?

[lusty nest]

Nono it's okay, I got the delay down to almost nothing and the stability is really good

[low shard]

alright, have a nice day

[formal badge]

Can anyone tell me how I can use this exact same AI model that this website is using, offline?
There is some text with two links:
This generator was made using the text-to-image-plugin.
The brain button feature uses the ai-text-plugin.

The bold part is the link and for the "brain button" it says it's a plugin that uses stable diffusion, so if I get stable diffusion offline, woudl that work the exact same way as this website?

https://perchance.org/ai-text-to-image-generator

[low shard]

it uses stable diffusion https://github.com/CompVis/stable-diffusion

this is outdated asf

dont even bother using that model locally

use flux.1-dev or sd3.5 instead

[rancid pine]

Which UVR model reduces the audio quality least? To prevent it from being lossy and turning into 4 bit lmao

[formal badge]

Appreciate the tip!! Thanks!

[rancid island]

my friends are saying that my voice is crackin up in discord

these are my settings

cant upload

[rancid island]

any fixes please?

[low shard]

!give-media-perms 1h @rancid island

show now

[low shard]

why did you set extra to 5?

setting extra to anything over 2.7 can cause cutoff issues

[rancid island]

for better voice quality

[low shard]

extra slightly controls voice quality, if a model sucks, it sucks

you should put it to 2.7, else you risk cracks and cutoff issues

put it to 2.7, check if it worked, if it didnt, check https://docs.aihub.gg/rvc-voice-changer/local/deiteris-w-okada-fork/#discord-crackle-fix

[rancid island]

alr done ...anything else that seems wrong?

[low shard]

nope, if the issue still persists you can trey the link i sent you

[rancid island]

both of em are off already

anything else?

[low shard]

does the issue still persist?

[rancid island]

yea sadly

[low shard]

try https://docs.aihub.gg/rvc-voice-changer/local/deiteris-w-okada-fork/#crackle-fix

[rancid island]

already tried this

[low shard]

can you try other models too?

https://docs.aihub.gg/rvc-voice-changer/local/deiteris-w-okada-fork/#models-to-try

[rancid island]

tried like 3 of em ..can you suggest the best ones ? for like egirl voices kind of like E-woman

[low shard]

gave you the link above

E-woman is wild 😭

why are people focused so much on e girl voices btw

[rancid island]

in what way?

[low shard]

im saying that im shocked on why many want e girl voices lol

[rancid island]

just want to troll my friends ..my normal voice is just well less pitched you can say for a female

[low shard]

anyways, is the cracking still happening?

even with the suggested models?

[rancid island]

downloaded that one...3 files init ....pth,json,index

which one should I upload

[low shard]

in rvc context:

• pth files: contain the voice
• added index files: contain the accent
• metadata.sjon file: it's just some extra info about the model download link if you downloaded it off weights.com, it's not needed and won't impact the actual model at all

[rancid island]

alright

sounds good to me ..dont have my friends online...will definetly let you know

Thansks

rmvpe is fine right?

[low shard]

yes

[low shard]

so everything else is fine right?

[rancid island]

yea

[normal oar]

Is this any different from "vcclient_win_cuda_2.0.78-beta.zip"??

Or is it the same

[low shard]

vcclient_win_cuda_2.0.78-beta.zip
you shouldn't use that one

that is original wokada

not suggested anymore

Belle uses wokada deiteris fork which is way better

@normal oar lemme guess, you used a youtube tutorial?

[normal oar]

Exactly

[low shard]

all video tutorials are outdated, delete everything you got off it

im guessing you have vb audio cable, which is known for causing issues on windows

[normal oar]

Does it contain malware

[low shard]

No, it just has way worse quality, way worse performance and much more unstable

[normal oar]

I've been using it for couple days now

[low shard]

you shouldn't

[normal oar]

I see, glad no malware 🙏

[low shard]

uninstall it

-realtime

[patent trellis]

💻 Local Realtime RVC

Guides for Programs that use RVC Models in Realtime for Calls/Games

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

Unavailable, the guide is outdated and the program is worse compared to the ones above, and much less updated

[low shard]

read up the 1st link

[normal oar]

Alrighty

[low shard]

wokada deiteris fork

there's no updated video

and prob will never be, because ai changes at sonic speed

[low shard]

what's your pc gpu btw?

[normal oar]

4060

[low shard]

alr u will be all fine

[normal oar]

Mhm

Why is fork better than the orginal??

[low shard]

it has better performance improvements

and advanced settings to have better stability and quality

just forget eevrything you see on yt tuts, as they are pretty old

they prob told you also to use crepe lol

[normal oar]

Got it, its completely safe right?

[normal oar]

Ikr, i use rmvpe

[low shard]

yes its literally FOSS, Free and Open Source software, you can check the code yourself and do whatever you want with it

plus the deiteris fork has more protection over the models

https://docs.aihub.gg/rvc-voice-changer/local/deiteris-w-okada-fork/#is-the-w-okada-deiteris-fork-safe

[normal oar]

Alright

Thanks.. if I encounter any error I'll lyk

[low shard]

alr yw

[normal oar]

Can I run it on docker?

[low shard]

Wokada isn't really meant to run it on docker and there are no guides about it

[normal oar]

Nvm then

[rancid island]

so the problem is not fixed...talked to my friends and it is just cutting off

I recorded it to show you

[low shard]

send it

!give-media-perms 2h @rancid island

[rancid island]

I am so done finding fixes tbh but I am still gonna try

lets rule out some stuffs....my mic is decent and my internet is more than decent ...and this is exactly how I sound

[noble hound]

Guys, can anyone help me make a voice for a YouTuber?

[low shard]

elaborate:

• your pc gpu
• your operative system
• what you want to do

[low shard]

wtf? show again a screenshot of ur current wokada, are u sure ur microphone is decent enough?

[rancid island]

It's exactly the same ....I can't send you now cause I am out of my home but I'll tag the last image sent

[rancid island]

I have the apple airpods for real

Also the voice is perfect when I monitor it

[noble hound]

5070
64 operative
YouTuber akvi4

[low shard]

so, it sounds perfect on monitor, but horrible when recording or using it on discord?

and you already tried all solutions?

[rancid island]

Exactly

[low shard]

Im guessing you mean windows 11 x64?

well, be aware that it will be very complex to train a model, firs tof all you need applio (rvc fork):

How to (unofficially) use Applio for RTX 50 serie cards

Follow to download it as said it in https://docs.aihub.gg/rvc/local/applio/

After you extracted the precompiled, go to the path in Windows explorer, write "CMD" and press enter, then in CMD write env\python -m pip install torch==2.7.0 torchvision==0.22.0 torchaudio==2.7.0 --upgrade --index-url https://download.pytorch.org/whl/cu128

If you get any already satisfied requirement issue, run env\python -m pip uninstall torch torchvision torchaudio then the command said above

After that, you can check: https://docs.aihub.gg/essentials/how-to-make-voice-models/

[rancid island]

Yea as far as I know

[low shard]

@noble hound I have deleted your post, you can make models yourself but NOT REQUEST them

You can search rvc ai voice models at:

• #1175430844685484042
• In #🔍│find-models , Do /find with @earnest musk
• https://weights.com/ (login required)
• https://huggingface.co/models (but watch out cus in hugging face there arent only rvc ai voice models)
• Suggested Models for Realtime Voice Changing (Wokada)
• https://voice-models.com/
• https://thevoicemodels.com/ (for Turkish Models, login required with discord and level 2 on their server)

if there isnt one, you can:

• make it yourself with our docs guides
• Be aware that we don't allow any model request, so don't fall for any "pay me 20 dollars and i will make the model for you" dm

[earnest musk]

:wave: @low shard, How can I help?

Available Commands:
• @weights find <query> or /find <query> - Search for RVC Voice Models
• /create - Create an AI Cover
• /image - Generate an Image

[noble hound]

Honestly, I would like someone to do it for me, I'm too stupid to do it myself

[low shard]

you can't, we don't allow anyone to do commissions, either do it yourself or search if someone already made it

we have removed model requests

@noble hound are you still looking to make it yourself? if not, you can check the sites above to see if anyone has made them already

[brittle wing]

hey ive been using deiteris' okada fork and ive got it well and running. but is there any tips to help increase the accuracy of the voice? sometimes what i say doesent get correctly interpreted by the model. (dont have an accent). e.g dog gets interpreted to doll by the model

[odd shale]

Maybe it's something related to your settings or the model you're using wasn't properly trained or got a short/bad dataset. Also, you can try imitating the voice itself.

[tender flint]

Ok

[dim cape]

I think I'm seeing a sign of over-training, should I keep it running?

(10 min dataset, training with MAINLINE on Kaggle)

[low shard]

@pastel oak maybe you got any idea?

[pastel oak]

Are these the settings you have rn when you recorded

[simple ore]

at 2.4k steps - fat chance, unless you runing batch 64 on a tiny set 🙂

[knotty moth]

testing on different epochs may somewhat surprise you

[knotty moth]

try another model

if you trained the model, try using longer dataset with better articulation

[tawdry lodge]

CAn I please get some help?

[low shard]

elaborate:

• your pc gpu
• what you want to do
• your operative system
• whats the issue
• what tutorial link did you follow

[sly root]

why is it so laggy

[hardy cradle]

can someone help me the client is loading but ntohing pops up

[knotty moth]

!howtoask

[patent trellis]

How To Troubleshoot

__**GIVE CONTEXT.**__ 📝

• Don't simply mention your issue, like "my rvc is not working".
• Describe the step you are on, what you're trying to do, the RVC you're using, a screenshot, etc.
• The more context, the better.

__**BE POLITE.**__ <:matsuripray:1159685390156967936>

• Don't be desperate. You can ping a Helper, but if they ignore, they aren't available/don't know the answer.
• It's okay if you're frustrated, but don't take it into this server.
• Don't DM without prior consent.

__**BE PRODUCTIVE.**__ 🤝

• Don't ask for every little instruction. Put your own effort & test things by yourself.
• Don't ask to ask.
• Check if your answer is a Google search away/on our guides website.

If you've changed your mind about your message, feel free to delete/edit it.

[knotty moth]

there are many cases of being laggy

pls explain

[sour summit]

Does anyone have any advice regarding flux gym? their isnt alot of info on the internet about how many training epochs and repeat images you should do. Im training a Lora for a realistic women.

[sly root]

like whenver is ay something it comes out like 10 seconds later and im breaking up

[knotty moth]

don't expect reasonable performance with cpu mode or too weak spec

[sly root]

i have really good specs i think its my settings or smth

[knotty moth]

gpus older than RTX 20-series and RX 580 are not good

or even igpu

[dire zinc]

I need help getting publicity

For ma Ai

[arctic sand]

How much db should I put the microphone volume in wokada?

-6db?

what is best

[tawdry lodge]

I need settings help as well..

[rose arrow]

can i run the client in the background smoothly with RTX 3060Ti?

[arctic sand]

What is the default for volume in and out in okada? 100%? Does this make the sound neither loud nor small?

@low shard helpp

[tawdry lodge]

Yes please help

[brittle wing]

how do you run the w-okada voice changer on linux, the guide tells you to how to combine and extract the two archives but what do i have to run locally

[knotty moth]

the guide should be explanatory enough

otherwise try troubleshotting by yourself or ask chatgpt

[knotty moth]

the one that will sound fine for you

anyway 100% is the safe one to avoid some distortion due to audio clipping

[glass crescent]

why does my VC client thing say wait web server... and doesnt do anything

[arctic sand]

but in 100% and out100% so loud

[knotty moth]

bruh system volume

set it to less than 100% and it will never be clipping

[hidden storm]

Hey, what is the best cost-efficient API model at the moment?

[next geyser]

/downloads

[knotty moth]

Here are the resources:

https://tenor.com/view/watame-rickroll-vtuber-hololive-gif-23716796

[normal oar]

Hey! I am getting no output

It doesn't seem to work

Getting 0 input, 0 output

[low shard]

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

We are a general ai server, not just voices

[low shard]

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

This is a general AI server, we aren't rvc focused

[low shard]

Wdym? If you're asking to advertise your videos, this isn't the right place

[low shard]

There isn't a best setting, just play with it, it doesn't affect quality

[low shard]

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[earnest musk]

:wave: @low shard, How can I help?

Available Commands:
• @weights find <query> or /find <query> - Search for RVC Voice Models
• /create - Create an AI Cover
• /image - Generate an Image

[low shard]

Wdym with client?
Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

We are a general ai server, you need to elaborate

[low shard]

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[low shard]

For what? There's literally thousands of AIs

[low shard]

Show a screenshot or screenrecording

[normal oar]

Headphones issue, solved it 👍 thanks

[dire zinc]

What videos no mate I’m trying to make a discord bot and need actual help how do I get the help I need

[willow shale]

So I have been running in cicles trying to figure out on my own, what am I supposed to do about pytorch? I'm trying to set up Okada for the first time.

I have a RTX 5070 so there seems to be a bit of a snag, but I don't know what to do about it

https://i.imgur.com/5K1GjYm.png https://i.imgur.com/Ex4lTXV.png

[knotty moth]

a decent alternative is vonovox, it already supports 50-series cards as well

https://docs.aihub.gg/rvc-voice-changer/local/vonovox/

[willow shale]

No way to use Okada?

[low shard]

-realtime

[patent trellis]

💻 Local Realtime RVC

Guides for Programs that use RVC Models in Realtime for Calls/Games

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

Unavailable, the guide is outdated and the program is worse compared to the ones above, and much less updated

[low shard]

1st link, wokada deiteris fork

you prob used some old original wokada off youtube

[willow shale]

Close enough

[low shard]

wokada deiteris fork has way better performance and quality

[low shard]

original wokada is old asf lol

[willow shale]

You sold me

[low shard]

tho, vonovox is a similar fork which may have slightly some improvements for nvidia

so it might slightly be better but im not sure, i haven't used it, most people use wokada deiteris fork

[willow shale]

So which one do I grab then, deiteris or vonovox?

[knotty moth]

unfortunately RTX 50-series doesn't have backward compatibility to the older pytorch versions used by the older voice changer as well

nor the current deiteris wokada

nor even the og 2.x beta

[willow shale]

https://i.imgur.com/0FTJXqx.png downloaded, ran setup, what do I do now?

[pulsar yarrow]

start it

[willow shale]

Doesn't go anywhere

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

[pulsar yarrow]

Try launcher.py

[willow shale]

Just had to wait actually

[pulsar yarrow]

[pseudo steppe]

okada link

[latent kettle]

-realtime

[patent trellis]

💻 Local Realtime RVC

Guides for Programs that use RVC Models in Realtime for Calls/Games

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

Unavailable, the guide is outdated and the program is worse compared to the ones above, and much less updated

[low shard]

Elaborate:

• your PC GPU
• your operative system
• what you want to do

[pseudo steppe]

rvc
rx 580
livestream

[low shard]

what's your operative system? windows?

[pseudo steppe]

11

[low shard]

rvc

livestream

RVC is not for realtime

RVC = Retrieval-based-Voice-Conversion, the best Few Shots Speech To Speech AI Models (on v2), Inferences (use models) pre-recorded audio (ai covers) and train (make) models. Technically, Mainline RVC does have a go-realtime.bat (aka RVC-GUI), but it's pretty messy and outdated so it's extremely not suggested for realtime.

Wokada = uses RVC for realtime inference. There's 2 main versions, Original made by Wok, and the most suggested one is Deiteris Fork (modified version)

Do you need RVC or Wokada?

They do 2 different things

[pseudo steppe]

wokada

[low shard]

btw, be aware that your pc gpu is bare minimum, you won't be able to use the voice changer in games like marvel rivals

and will play every game at lowest settings

-realtime

[patent trellis]

💻 Local Realtime RVC

Guides for Programs that use RVC Models in Realtime for Calls/Games

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

Unavailable, the guide is outdated and the program is worse compared to the ones above, and much less updated

[pseudo steppe]

not for games

[low shard]

1st link, read wokada deiteris fork

don't follow youtube tuts

they are all old and use an over year old version of wokada

[fallow zodiac]

does anyone has a working collab link for training RVC V2 model?

[low shard]

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using

[low shard]

what? you shouldn't go download it off github, if you used any video tutorials, they are old

delete everything you got

read the guide I sent you

missing steps could lead you to fuck up your whole audio system

[fallow zodiac]

my pc has gtx 1650 and windows 11 , I want to train a voice model on cloud

[low shard]

I mean, your pc gpu could technically train locally, but it would be slower and limited so yeah

Train (make) RVC Models on cloud:

1. Prepare the Dataset
2. Setup RVC:
   Choose a cloud way to use RVC,

• Google Colabs (max 4 hours of daily T4 16gb gpu not granted for free, not much hours for training, but easy to use, there's a paid tier):
  • Applio (UI)
  • Mainline (U, broken right now)
• Kaggles (a bit harder to use and needs phone number but gives 30 hours weekly of better gpus, either T4x2 16gb each or P100 16gb, only free):
  • Applio (UI)
  • Mainline (UI, broken right now)
• Lightning.ai (Kinda hard, needs login, no issue with web uis or anything, but only free 15 credits monthly, Free Studios run 24/7 but require restart every 4 hours. There's a paid tier):
  • Applio (UI)
  • Mainline (UI, No guide as of right now)

3. Be sure to know about the tensorboard

Google Colab = Easier but risk of getting disconnected
Kaggle = Harder but way more gpu time
If you are looking for the easiest way and for free, is using https://weights.com/ which ofc uses RVC

RVC Inference (use models) on pre-recorded audio on Cloud

You can use either:

• Weights.com: Easiest Possible Ever Automatic
• Ilaria RVC Zero: Fastest free on cloud
• Applio UI Colab: RVC Fork with some extra features like TTS
• RVC AI Cover Maker UI: Automatically Separates the Vocals and Instrumentals, converts the voice and mixes them back

[pseudo steppe]

which one would be better

[low shard]

you are on the wrong place

you got an amd gpu

@pseudo steppe get https://docs.aihub.gg/rvc-voice-changer/local/deiteris-w-okada-fork/#download-amd-intel-and-cpu-on-windows

Read carefully the guide

[latent rune]

Which microphone would you recommend me to get for ai voice? (edited)

It sounds bad on my headset ones

[fickle tide]

How do i train ai to duplicate my friend voice? (i already got his voice like 3 mins)

[low shard]

replied to you https://discord.com/channels/1159260121998827560/1387831485901574195

[low shard]

Elaborate:

• your PC GPU
• your operative system
• what you want to do

[runic void]

Hello, I have a problem with the application. I would like to prevent myself from being heard in the background. Is this possible?

[primal valley]

^ @swift ether 's screen, they are having this issue with Okada

[analog obsidian]

change sr to 48000

[runic void]

How do I remove my voice in the background when I speak please?

[swift ether]

doesn't work

[analog obsidian]

use wasapi, not asio

[primal valley]

wasapi used

[analog obsidian]

is that og wok or the fork??

[primal valley]

og

[analog obsidian]

then download the fork ez fix

[cosmic knoll]

Yo im using otaka voice changer and i tested some models but they all sound kinda chopped even with good settings. I mean like they stutter. How can i fix that typa stuff?

[primal valley]

some models are just choppy

even with good settings, models can have bad moments

[cosmic knoll]

but its basically with all of the ones ive tried

[analog obsidian]

do they sound choppy only when playing games?

[cosmic knoll]

no its alltime

[analog obsidian]

tell me ur gpu and which wokada version u have

are you using vac lite or vb audio cable?

[cosmic knoll]

vb audio

[cosmic knoll]

its like 2.0.78-beta

and i have gtx 1650 but is it only gpu dependent

[low shard]

This is a General AI Server, we won't be focused on voices anymore

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[analog obsidian]

uninstall that
install fork version

uninstall vb audio cable
install vac lite

[low shard]

-realtime

[patent trellis]

💻 Local Realtime RVC

Guides for Programs that use RVC Models in Realtime for Calls/Games

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

Unavailable, the guide is outdated and the program is worse compared to the ones above, and much less updated

[low shard]

1st link

wokada deiteris fork

don't trust video tutorials

the version you got is original wokada

[primal valley]

video tutorials telling people to use crepe_tiny

[analog obsidian]

[primal valley]

in this day and age

[low shard]

yeah just saw someone using it #1387831485901574195 message

[primal valley]

Ruby who wanted help also used it from a video tutorial

[analog obsidian]

installing the fork and replacing vb audio cable with vac lites fixes every weird error

[low shard]

yes but people are attached to original wokada

[analog obsidian]

they don't care enough to check this channel

[low shard]

ping everyone 🔥

[cosmic knoll]

do i have to change any setting in vac or js keep it how it is?

[analog obsidian]

dont change anything and be sure you're downloading vac lite, not the trial

[analog obsidian]

[cosmic knoll]

which one do i download? ( nvidia gpu)

[analog obsidian]

amd64 cuda

devs loving using super long unnecessary names for their shit

[low shard]

get https://docs.aihub.gg/rvc-voice-changer/local/deiteris-w-okada-fork/#download-nvidia-on-windows

[analog obsidian]

amd64 = x64 bit cpu
cuda = nvidia

[low shard]

you got a gtx 1650,

no need to use the github rtx 50 serie one

[analog obsidian]

here is the compiled version for nvidia https://huggingface.co/Shadicti/deiteris-Fork/blob/main/voice-changer-windows-nvidia-b2332.zip (non 50xx series)

[next tree]

!cover @BangChan [فایل یا لینک]

#✨│ai-help

[cosmic knoll]

i js downloaded the cuda thing, do i delete that again and download this or wt

[fickle tide]

gtx 1650 , window and i want to make my friend cover a meme video

[analog obsidian]

they're the same thing, the difference is that the huggingface link already merged the two cuda zips

[cosmic knoll]

oh so i need 2 cuda?

[low shard]

your pc gpu could train locally, but its not suggested and slow

Train (make) RVC Models on cloud:

1. Prepare the Dataset
2. Setup RVC:
   Choose a cloud way to use RVC,

• Google Colabs (max 4 hours of daily T4 16gb gpu not granted for free, not much hours for training, but easy to use, there's a paid tier):
  • Applio (UI)
  • Mainline (U, broken right now)
• Kaggles (a bit harder to use and needs phone number but gives 30 hours weekly of better gpus, either T4x2 16gb each or P100 16gb, only free):
  • Applio (UI)
  • Mainline (UI, broken right now)
• Lightning.ai (Kinda hard, needs login, no issue with web uis or anything, but only free 15 credits monthly, Free Studios run 24/7 but require restart every 4 hours. There's a paid tier):
  • Applio (UI)
  • Mainline (UI, No guide as of right now)

3. Be sure to know about the tensorboard

Google Colab = Easier but risk of getting disconnected
Kaggle = Harder but way more gpu time
If you are looking for the easiest way and for free, is using https://weights.com/ which ofc uses RVC

RVC Inference (use models) on pre-recorded audio on Cloud

You can use either:

• Weights.com: Easiest Possible Ever Automatic
• Ilaria RVC Zero: Fastest free on cloud
• Applio UI Colab: RVC Fork with some extra features like TTS
• RVC AI Cover Maker UI: Automatically Separates the Vocals and Instrumentals, converts the voice and mixes them back

[analog obsidian]

yeah, the cuda version are 2 zips that needs to be merged

[low shard]

or you can just get the precompiled

literally, everything is in the guide

[fickle tide]

thank u

[analog obsidian]

the github has a gif that teaches u how to merge the zips, but if u dont wanna do that just download it from huggingface

[low shard]

^

[distant turtle]

-colab

[patent trellis]

📒 Google Colab Notebooks

Google Colab is a Cloud (Remote Good PC) Service. While the Free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

• **Applio**

by IA Hispano
Google Colab

• **RVC Mainline**

by Hina
Google Colab

• **UVR5 NO UI**

by Eddy
Google Colab

• **UVR5 UI**

by Eddy
Google Colab

• **Wokada Deiteris Fork**

by Deiteris & Hina
Google Colab

• **Hina's Modified Original Wokada**

Google Colab

• **RVC-AI-Cover-Maker-WebUI**

by Shiro & Eddy
Google Colab

• **FaceFusion UI**

by Nick088
Google Colab

• **FaceFusion NO UI**

by Nick088
Google Colab

• **Music Source Separation Training (Inference)**

by Jarredou & Makidanye
Google Colab

[cosmic knoll]

which one do i have to use ( vac lite)

[low shard]

uninstall vb audio cable from windows app settings

did you perchance install voice.ai?

[cosmic knoll]

nop

[young halo]

ValueError: Unable to detect audio format for file extension: .opus

what does that mean?

i've been trying to make a cover several times and i always get an error

[low shard]

get vac lite here https://docs.aihub.gg/rvc-voice-changer/local/deiteris-w-okada-fork/#virtual-audio-cable

uninstall vb audio cable

[cosmic knoll]

i did

[low shard]

did you follow both steps?

if you miss this, you can fuck up your whole audio system

[low shard]

This is a General AI Server, we won't be focused on voices anymore

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

@cosmic knoll the file is setup64. not setup64a

[young halo]

My GPU is Intel HD Graphics 520, my operative system is windows, i want to do an ai cover, this the link:https://colab.research.google.com/github/Eddycrack864/RVC-AI-Cover-Maker-UI/blob/main/assets/RVCAICoverMakerUI.ipynb#scrollTo=sDo4MkcLcN5l

and the tunnel is gradio

[tawdry lodge]

Anyone able to help me with background noise for the voice changer?

[cosmic knoll]

which one do i select?? i can hear the ai voice but it doesnt do it ingame

[low shard]

did you upload a video with the .opus format?

[low shard]

line 1

and set game input to line 1

also its better u show full settings

[low shard]

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[upper goblet]

why i cant send screenshoot

I have a problem

for some reason my voicemod is not working

and I connected evrething right

[drowsy oak]

hi i have a voice changer and when i used to add models normally but now when i add models it says PermissionError: [WinError 5] Access is denied: 'model_dir\1' what should i do

[viral mason]

are you talkinga bout okada or voicemod

[simple ore]

did you click 'start'? 😛

[light current]

why the voice changer(okada) doesnt work in valorant, anyone knows how to fix that?

[low shard]

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[low shard]

!give-media-perms 1h @upper goblet

This is a General AI Server, we won't be focused on voices anymore

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[low shard]

This is a General AI Server, we won't be focused on voices anymore

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[viscid moss]

do not use .opus as input

[boreal flint]

how do i run rvc on mac?

[simple ore]

poorly

[boreal flint]

😭

that ik

whats the best way? i have trained an model with rvc web ui now i wanna use that with mac

[simple ore]

realtime?

or file to file?

[willow escarp]

hey, just wondering as a freelancer starting out, what the best automations are to sell to businesses. I have good knowledge of n8n, html, css, and im decent with js and make

[boreal flint]

real time

some delay is fine

[brittle wing]

Anuew dereverb or MDX23C dereverb for inference vocals

[swift thunder]

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

Help, it's not working anymore

[simple ore]

https://huggingface.co/spaces/r3gm/RVC_HFv2 ?

[shrewd badger]

for some reason on weights i cant create

[brittle wing]

Hiya

My Okada is very breathy

And

I can't for the life of me get Voicemeeter to release sound out of B channel

No matter what.

Any ideas?

Further, what about seemingly random and painful "Skipping"

I cant say "hello hi there" without getting at least 4 cut pops

on an rtx 4090

[queen quiver]

post a screenshot of your okada here

[brittle wing]

pok

[knotty moth]

this is not the voice changer version we know so far, perhaps too old or a random fork?

[brittle wing]

recommend me the best fork

i'll hop on it

[knotty moth]

https://docs.aihub.gg/rvc-voice-changer/local/deiteris-w-okada-fork/

[brittle wing]

From there, official fork?

Or Dei

[knotty moth]

just the recommended one

[brittle wing]

Says Dei is recommended.

That's the one I got.

[minor isle]

can w-okada damage gpu?

[brittle wing]

Or maybe I'm seeing smt wrong.

Let me try again, fresh.

U were correct.

I had the wrong fork somehow.

Thank u sirs.

Btw, Server or Client?

(I really don't understand why there are two of those)

And which audio driver do u all use the most?

wasapi, etc

[brittle wing]

Yes, I have been up for 40 hours. yes I should apologize.

I am sorry.

And I am grateful to u

[knotty moth]

well so you stumbled upon a strange fork I've never heard of

and ye I think you should consider the recommended fork as written in our guide here

[brittle wing]

Can I somehow send u a vc to ask

What u think

Or just hang in supp vc

Uhm I dont know what banned word I used

Let me try again

If anyone wants to spend 60 seconds in vc with a noob

To help

[low shard]

-realtime

[patent trellis]

💻 Local Realtime RVC

Guides for Programs that use RVC Models in Realtime for Calls/Games

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

Unavailable, the guide is outdated and the program is worse compared to the ones above, and much less updated

[low shard]

1st link

wokada deiteris fork

[low shard]

elaborate

[low shard]

you dont seem to have wokada deiteris fork,get it from the guide

[brittle wing]

Better than main?

[brittle wing]

oh nonono

[low shard]

lemme guess, you followed some youtube tutorial?

[brittle wing]

I learned that a few minutes ago.

I got Deit now

And it's better

[low shard]

all video tutorials are old

[brittle wing]

Still feeling breathy tho

I chalk it up to my actual hardware

[low shard]

u can also show a screenshot

of ur settings

[brittle wing]

And not the tech.

Sure, let me just try this new model.

One sec.

[low shard]

!give-media-perms 1h @brittle wing

[brittle wing]

I'm using a random voice I found

Sounds fantastic on the site

Sounds... similar when I do it but

It's missing a lot of "Depth" and reality.

Sounds like a breathy mess

[minor isle]

[brittle wing]

poke

[low shard]

set rmvpe without onnx, uninstall vb audio cable it can cause random issues on windows, get vac lite https://docs.aihub.gg/rvc-voice-changer/local/deiteris-w-okada-fork/#virtual-audio-cable

set extra to 2.7

are you using it in games? what other programs do you have in background

[brittle wing]

Basically none

And the PC is stronk.

I'll do the other stuff u said.

uno momento por favore

[low shard]

unless you manuallu overclock your gpu, which that already shortens its life span, and your pc isn't overheating 24/7 with bad airflow, no

what damages gpus is extremely hot temperatures

[low shard]

are you sure no other program is in background? you should have less perf

[brittle wing]

I mean... Discord, Telegram, a couple tabs in Chrome

Steam doing nothing.

[low shard]

you should close as many programs as possible just keep open the voice changer, discord and whatever program you're going to use wokada in

on wokada, you can optionally use more advanced settings for benefits:

• Force FP32 mode: on (THIS IS OFF BY DEFAULT! Turning this on improves stability. Increases VRAM usage by 200 MB)
• Disable JIT compilation: off for faster loading speed of the program, on for slightly better performance (10-15 ms) for Nvidia only)
• Reduce the delay via: https://docs.aihub.gg/rvc-voice-changer/local/deiteris-w-okada-fork/#reduce-more-delay

[brittle wing]

Ok, I think I'm set. Wanna talk to me for 10 seconds in support chat?

To tell me what u think

[minor isle]

which temp would be over heating for 2060

[low shard]

I can't VC, but you can set monitor to headphones to hear yourself

[brittle wing]

No I did that ofc.

It's just, u do this a lot

U can "sniff out" any artifacts.

If any.

I think it's decent tbh.

[low shard]

all the other thing you can do is just try other models, not all models are good

plus, there's no such thing as a perfect model

voice models got limitations

they can't laugh or scream well for example

so rvc is 100% detectable

[minor isle]

that seems to be case for all models

none of em can laugh

[low shard]

just dont get like over 85 celsius degrees, anyways you shouldn't have any issues unless you do overclocks, play super intensive games and have bad airflow

[low shard]

yep, @brittle wing RVC is detectable ofc, there's nothing perfect

@brittle wing do you need any other help?

[brittle wing]

No sir. Not trying to trick anyone. Just enjoying the software.

It's fun.

[minor isle]

it reaches 83 degrees when playing demanding games like rdr or fh5

[low shard]

alr do you need any other help?

[brittle wing]

No sir, thank you.

[low shard]

shouldn't be too much of an issue, @simple ore what do u think?

[simple ore]

"Maximum manfuacturers operating temperature for the RTX 2060 is 88 degrees."

I'd worry if it was constantly >80C

[minor isle]

idle or while playing demanding games

[simple ore]

during load, obviously

[minor isle]

yeah it does stay 82-83 while playing rdr

[simple ore]

this is normal-ish

HWInfo sensor mode

[minor isle]

what is memory junction temp

[knotty moth]

try using GPU-Z to check the performance metrics

or HWinfo as said above

[minor isle]

where do i download hwinfo

nvm

[minor isle]

what's the normal fan rpm for gpu's

[simple ore]

~1500RPM on load, 0 on idle

[minor isle]

mine is at 2000 rpm tho

[knotty moth]

it depends on the AIB model

go check some review spec for yours

[minor isle]

is zero rpm actually a thing

casue mine always spins

[knotty moth]

not sure but it has been since RTX 30-series

I knew a TUF gaming RTX 4070 has zero rpm and its temp stays below 60 in full load

while average gpus stay around 70

[pulsar yarrow]

Yo does anybody know if UVR or some other vocal separating software has a model that can separate a reference / similar voice vs other voices? Like a male and female talking in a podcast and isolate them. I know like resemblyzer can sorta do it kinda

[pulsar yarrow]

Nvm I found a doc that tells me some stuff

[sand bison]

I have a question, what is the minimum graphics card required to create AI models locally?

[knotty moth]

assuming for RVC, at least RTX 20-series

[desert hinge]

Im having some issues with the W-Okada RT, Using a vitrual cable to try and run it in games/discord. When I listen in on the cable trough sound manager I hear the changer but discord and games seem to either not pick it up or just go with unchanged voice.

[dusty arch]

can sm1 help?

[lucid veldt]

how can u search for rvc of some specific character voice? like aizen

[hallow thistle]

!howtoask

[patent trellis]

How To Troubleshoot

__**GIVE CONTEXT.**__ 📝

• Don't simply mention your issue, like "my rvc is not working".
• Describe the step you are on, what you're trying to do, the RVC you're using, a screenshot, etc.
• The more context, the better.

__**BE POLITE.**__ <:matsuripray:1159685390156967936>

• Don't be desperate. You can ping a Helper, but if they ignore, they aren't available/don't know the answer.
• It's okay if you're frustrated, but don't take it into this server.
• Don't DM without prior consent.

__**BE PRODUCTIVE.**__ 🤝

• Don't ask for every little instruction. Put your own effort & test things by yourself.
• Don't ask to ask.
• Check if your answer is a Google search away/on our guides website.

If you've changed your mind about your message, feel free to delete/edit it.

[desert hinge]

@lucid veldt in #1175430844685484042 or #🔍│find-models

[hallow thistle]

What search? Is it #1175430844685484042 or Weights.com?

[dusty arch]

i can hear myself in my headset

[lucid veldt]

the one for okada i cant find aizen some how

[hallow thistle]

Do you mean "can't"? Is it about W-Okada the realtime voice changer and virtual cable?

[dusty arch]

i mean i can hear myself

with the voicechanger

idk how to fix it

its kinda annoying

[hallow thistle]

It's annoying indeed as you use the "original" version of W-Okada from YouTube tutorials, which are outdated. Which W-Okada version are you using? What is your PC GPU?

[dusty arch]

wtf is W-Okada?

[hallow thistle]

W-Okada the realtime voicechanger. A program that converts your voice into AI voice in realtime.

[dusty arch]

bro idk what i use

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

saw this toturial

[hallow thistle]

Weird and crazy. You can send your "program" screenshot here so I can identify it for you.

[dusty arch]

i cant send a ss here

can i dm u?

[hallow thistle]

No.

[dusty arch]

than this is going to be kinda hard

[hallow thistle]

I thought users with green name can send an image here and in #1192011222023950368 ? Did this server changed or something?

[dusty arch]

nah bro, i just think that i have turned one of the things on

[hallow thistle]

That's not the correct answer.

[dusty arch]

It tells me to Select "none" for your monitor device.

i already did that

so yeaaahh

[hallow thistle]

If your "voicechanger" looks like this, this is "W-Okada". However, the "original" version opens its own window. I use the better version, which it opens in browser. https://cdn.discordapp.com/attachments/1159290139609137264/1386272484352856104/image.png

[dusty arch]

ill just wait for sm1 there know smt about this voice changer ig that is my only choise

[hallow thistle]

Did you understand what I say?

[low shard]

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[desert hinge]

Just managed to get it working I think, Needed different audio cable

Thanks for checking up on it

[low shard]

be sure ur using vac lite and wokada deiteris fork, not from yt tus

[desert hinge]

Got it working 😄

Just checked in a VC

[hallow thistle]

That didn't answer some of his questions, but close enough of getting W-Okada to work by yourself.

[minor isle]

my okada broke, it seems to not hear anything from my mic but instead its translating other ppls voice

[hallow thistle]

I think I remember you asking about this before, or maybe I just recalled a different individual.

I think we have talked about "index damaging the CPU" thing before, but the messages I replied long ago were deleted after I went away, thinking it gonna sound like a bad idea to ask about some basic ahh question.

[viral mason]

I don't think it does?

okada just kinda

is a voice changer

voicemod doesn't damage gpu

[hazy eagle]

which is better crepe_full or rmvpe

[hallow thistle]

Rmvpe.

[hazy eagle]

crepe_full sounds HD but gpu heavy

crash a lot

[hazy eagle]

alr imma just use rmvpe

thanks

[hallow thistle]

Crepe is an old F0 pitch extractor, used in RVC. Rmvpe is newer.

[analog obsidian]

the real answer to this is to choose the f0 estimator used to train the model

most models use rmvpe

[hazy eagle]

with crepe_full i can even sing with it without them noticing its voice changer

[viral mason]

when is rmvpe v2

[analog obsidian]

in terms of f0 estimation precision, it's pretty close to rmvpe

there are some moments where it can be better than rmvpe

but for the most part, rmvpe is better

[hallow thistle]

However, there's an exception. If a model is trained with "crepe", use crepe on W-Okada. The RVC voice models that trained with crepe are usually from 2023 and considered old.

[analog obsidian]

i recall crepe being decent for singing

[viral mason]

What

[analog obsidian]

but using crepe in a rmvpe model may be a bad idea, the model learned the sibilants and breaths using rmvpe knowledge

rmvpe zeroes the breaths and sibilants, but crepe doesn't

it gives them a value, a value rmvpe never gave to them

[hallow thistle]

I've never seen an actual rmvpe V2 being released anywhere. But I just see this in a GitHub repository.

[analog obsidian]

rmvpe v2 doesnt exist namari u cant be serious

[hallow thistle]

[analog obsidian]

the thing here, crepe in terms of precision, isn't trash, it does have its limitations

but can work

[analog obsidian]

this is the reason why u dont wanna use crepe in a rmvpe model, besides perfomance issues

[viral mason]

let's produce it

[analog obsidian]

if singing works fine for him using crepe it's because the thing isn't trash and horrible

its quite decent

[hallow thistle]

All I see is "RVC v2, rmvpe" but not "rmvpe v2" in #1175430844685484042

[analog obsidian]

for 2018 standards ofc, rmvpe is better than crepe anyway

however, f0s have nothing to do with realism, thats just placebo

and for inference, the result in sound is the same

your rmvpe model is gonna sound like a rmvpe model no matter what u use

[simple ore]

well, teeeechnically the inference f0 matters

because that is the input (in addition to phonemes from feature extractor)

[analog obsidian]

[minor isle]

I dont use index tho

[hallow thistle]

[minor isle]

[queen cargo]

Heya! 🫡
I am trying to rollback from LatentSync 1.6 to 1.5 but I need the LatentSync_v1_5.ckpt and it is nowhere to be found now 😦 anyone have a clue to where to get it please? 1.6 doesnt work anymore on my setup (3060ti) but 1.5 was fine for 8sec in 720p, I shouldnt have try to update 😦

[late flicker]

What is the best choice of voice trainer if Applio doesn't work properly

?

[simple ore]

whym does not work properly?

there's always mainline, just download the compiled version and return to 2023 glory days

[edgy compass]

i got issue

when i click the start thing it opens then closed after 0.1miliseconds

[edgy compass]

idk if i can ping but i need this to work 😭

[safe token]

can someone help me with w okawa colab

its keep on saying

ERROR: Could not find a version that satisfies the requirement faiss-gpu (from versions: none)
ERROR: No matching distribution found for faiss-gpu
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 261.0/261.0 kB 15.2 MB/s eta 0:00:00
Preparing metadata (pyproject.toml) ... done
Building wheel for pyworld (pyproject.toml) ... done

Installing dependencies from requirements.txt...
ERROR: Ignored the following versions that require a different python version: 1.21.2 Requires-Python >=3.7,<3.11; 1.21.3 Requires-Python >=3.7,<3.11; 1.21.4 Requires-Python >=3.7,<3.11; 1.21.5 Requires-Python >=3.7,<3.11; 1.21.6 Requires-Python >=3.7,<3.11
ERROR: Could not find a version that satisfies the requirement onnxruntime-gpu==1.13.1 (from versions: 1.15.0, 1.15.1, 1.16.0, 1.16.1, 1.16.2, 1.16.3, 1.17.0, 1.17.1, 1.18.0, 1.18.1, 1.19.0, 1.19.2, 1.20.0, 1.20.1, 1.20.2, 1.21.0, 1.21.1, 1.22.0)
ERROR: No matching distribution found for onnxruntime-gpu==1.13.1
Successfully installed all packages!

[quiet token]

Can anytone help me wiht okada it sounds so ahh even tho i got a 5070

[thorn flax]

@acoustic scarab was there an update becuase my rvc just dosent open anymore

it runs the command prompt but nothing happens

[acoustic scarab]

highly unlikely

[thorn flax]

whys it broken then

[thorn flax]

i think im just gonna reinstall it, do i get cuda 1, cuda 2 or dml?

i dont remember what they do differently

[acoustic scarab]

i can't say unless you provide a log or any other output that might be analyzed

[acoustic scarab]

cuda for nvidia gpu, dml for amd/intel, but honestly i've never heard of a choice between cuda 1 and cuda 2

[thorn flax]

cuda 1 and 2

[acoustic scarab]

both

[thorn flax]

oh

[acoustic scarab]

it's a divided zip archive

[thorn flax]

is the file that huge?

[acoustic scarab]

i guess so

[sour summit]

I am training a LoRA in fluxgym of a real life person, training for face and body. Using a 58 training images. Would love some advice on how many epochs and repeats I should be using, and how many total training steps I should be aiming for. Im pretty concerned about wasting my time training such a big data set for the LoRA to just come out over or under baked.

[thorn flax]

theres an issue

its not possible to use cuda 2

its not a zip file

[acoustic scarab]

what file archiver do you use

[thorn flax]

7zip

but that file is not a zip file

its a 002 file

which isnt a real file type

[acoustic scarab]

put them in the same dir

and use extract to on the first one

7zip should extract both

[thorn flax]

it still says the second one isnt a zip archive

[acoustic scarab]

what about the first one

did it extract without any issues

[thorn flax]

yes

no issues with the first one

[acoustic scarab]

the thing is 7zip should've considered all of archive parts as an entire archive so you don't have to extract the remaining parts(.002 in your case)

[thorn flax]

yea but they separated them in download

[acoustic scarab]

try the extracted app

does it work

[thorn flax]

idk what that is

but if 7zip cant do it then that definitely cant

[acoustic scarab]

dude

you just extracted an app from these files

try it

[thorn flax]

wdm

i didnt extract an app

[acoustic scarab]

what was in the archive then

[thorn flax]

just the rvc

[acoustic scarab]

try it

[thorn flax]

ohh you mean

[acoustic scarab]

yea

[thorn flax]

ok yea that dosent work

does the same thing as the vrc i just deleted

it initalizes then nothing happens

[acoustic scarab]

does the console stay open after initialization

[thorn flax]

yes

[acoustic scarab]

take a screenshot of it and send it

[thorn flax]

stops at initalized

[acoustic scarab]

where exactly did you get the archive

[thorn flax]

from the offical github

[acoustic scarab]

gimme the link

[thorn flax]

https://discord.com/channels/1159260121998827560/1301535386572427364

[acoustic scarab]

i've never used deiteris fork and can't say what could've gone wrong, you can try the default w-okada from here https://huggingface.co/wok000/vcclient000/tree/main, latest stable build https://huggingface.co/wok000/vcclient000/blob/main/MMVCServerSIO_win_onnxgpu-cuda_v.1.5.3.18a.zip

[thorn flax]

yea but isnt deiteris fork wayy better

[acoustic scarab]

idk

never tried it

[analog obsidian]

https://huggingface.co/Shadicti/deiteris-Fork/blob/main/voice-changer-windows-nvidia-b2332.zip
download this instead, it's the fork but with the two zips already merged

[thorn flax]

ah ok

[thorn flax]

this download did the same thing

just initalizes and then nothing happens

[analog obsidian]

that is indeed weird

uh

what gpu?

[thorn flax]

geforce rtx 2060

[analog obsidian]

the cmd should output a localhost link hmm

[paper veldt]

anybody have any tips on how to cancel echos from VC in Discord when I use RVCC under server mode? The only solution I have rn is to lower chat volume

[analog obsidian]

switch to client and enable echo in w-okada

and maybe enable sup2

[paper veldt]

okay i was hoping to keep server because it works a lot smoother and faster

[analog obsidian]

maybe if you wait maybe 5 minutes it'll fix by itself? weird issue, never seen that before

[analog obsidian]

you could try downloading a third party noise supressor, then using that as your input

[thorn flax]

how is waiting 5 minutes going to help

[analog obsidian]

so by doing that you can keep server mode

[analog obsidian]

it may be stuck

too much for the cpu to load maybe?

[thorn flax]

i mean it used to work fine

[analog obsidian]

is this problem also present in the original version?

[thorn flax]

no

but today i tried to run it and it just does this

no idea why

[analog obsidian]

me neither sorry, this is new to me

w-okada is so buggy sometimes

[crude flame]

switch to vonovox 👍

[analog obsidian]

(most of the time )

[analog obsidian]

good idea

[thorn flax]

whats that

[analog obsidian]

https://docs.aihub.gg/rvc-voice-changer/local/vonovox/

a new voice changer made by a known dev of the community

[thorn flax]

is it more optimized

[analog obsidian]

faster than w-okada

yeah

way way way more optimized

and doesn't run in your browser

[thorn flax]

thats like exactly what i need

[analog obsidian]

if something goes wrong you should join their discord server tho

[thorn flax]

wait where do i download it

[analog obsidian]

https://docs.aihub.gg/rvc-voice-changer/local/vonovox/

there

[thorn flax]

oh its twards the bottom

[analog obsidian]

vonovox discord server is in their github

any issues regarding the app go there and notify the dev about it

[thorn flax]

kk

[knotty moth]

deiteris wokada is still okay for beginners

but for those having RTX 50-series I'd rather suggest trying vonovox

[hybrid iris]

Does anyone know a good base fnf cromatic for RVC?

[knotty moth]

https://discord.com/channels/1159260121998827560/1163592055830880266

[cyan elk]

i have an AMD Radeon RX 7800 XT gpu, is this good enough to train a voice model on my pc and if so what should I use to do so?

[versed edge]

why does every single female rvc model sound so dookie without being merged into oblivion 🥀

[simple ore]

you can patch Applio to use Zluda emulation and run it on your card

[cyan elk]

would it be better to run it locally or on the cloud

[simple ore]

on 7800 better locally

[knotty moth]

still pretty good with 16 GB vram and zluda

[cyan elk]

ok so i may be dumb but i downloaded applio and extracted it to a folder but i'm not seeing a run-install.bat file, but i do see run-applio.bat, do you know why this may be?

[safe token]

can someone help me with w okada its really laggy

[simple ore]

open a help thread

if you're using Adrenalin driver older than 25.5.1 there's a guide https://docs.applio.org/applio/getting-started/installation#amd-gpu-support-windows

for newer driver there are changes

[simple ore]

For Adrenalin 25.5.1 driver or newer

1. Download a compiled version of Applio v3.2.9, unzip to the desired folder.
2. Download HIP SDK 6.2.4, install the components except the video driver at the bottom
3. Add C:\Program Files\AMD\ROCm\6.2\bin to your Path environment variable.
4. Using command line from the Applio folder run
   env\python -m pip uninstall torch torchvision torchaudio
env\python -m pip install torch torchvision torchaudio --upgrade --index-url https://download.pytorch.org/whl/cu118
5. download these into Applio folder
   https://github.com/IAHispano/Applio/blob/main/assets/zluda/patch-zluda-hip62.bat
   https://github.com/IAHispano/Applio/blob/main/assets/zluda/run-applio-amd.bat
6. replace rvc/lib/zluda.py with the following

import torch
if torch.cuda.is_available() and torch.cuda.get_device_name().endswith("[ZLUDA]"):
    # disabling unsupported cudnn
    torch.backends.cudnn.enabled = False
    torch.backends.cuda.enable_flash_sdp(False)
    torch.backends.cuda.enable_math_sdp(True)
    torch.backends.cuda.enable_mem_efficient_sdp(False)

7. run patch-zluda-hip62.bat
8. run run-applio-amd.bat to start Applio

[cyan elk]

i'm currently on adrenalin 24.10.34, would you recommend just following the old guide or updating real quick and doing the new version?

[simple ore]

that's not a right version, probably some non-whql beta

[cyan elk]

soo update then

[simple ore]

yeah, should be better with latest zluda

[thorn flax]

im gonna be honest this branch is pretty awful can someone just fix okada

[viscid mural]

need help

[thorn flax]

id rather not use some offbrand version of okada with no features

[thorn flax]

@knotty moth do you have any idea how to fix this?

[rancid pine]

Which UVR model is best to minimize bitrate dropping?

[minor isle]

What does merging models do

[analog obsidian]

combines the timbre of each models together

[minor isle]

what does that exactly do

[analog obsidian]

create a new voice

[minor isle]

does it work better?

[analog obsidian]

honestly i have no idea, if done right it should be ok
i'd rather not do it lol

i havent used that feature much to create new voices

it does have more uses than just creating new voices but thats a bit more advanced

like merging two different training runs

[minor isle]

where do I even merge it

wheres the option

[analog obsidian]

you use applio
or like
the voice changer app?

[minor isle]

uhno

i thought theres an option in wokada itself

[analog obsidian]

there is one in the og version, not sure if the fork kept that feature

i personally only use rvc's merging

click options and find it, should be there

the results wont be always so great

most of the time is ehh, weird

basically its all random

[rancid pine]

...Which model is Gabox's voc_fv4?

[crude flame]

80% of the time the new voice sucks/ is mid

Sometimes you can get a good one but you'll need to be lucky

[knotty moth]

by avoiding to convert to lossy formats in any process

[minor isle]

how to use windows wasapi line 1 in apps

normal line 1 doesnt work

nvm

[idle osprey]

Hi! I need help with b3223 nvidia version w-okada thats fork, and today i wanted to open the voice changer, oh noes! The voice changer doesnt open! the cmd says this "json.decoder.JSONDecodeError: Expecting value: line 1 column 1 (char 0)" and includes a screenshot. When i press enter to continue, the cmd closes and nothing happens.

I dont know why but i think it could be something about model_dir file, because recently i got a new model

Nevermind it was the model_Dir file, i deleted it and it got fixed

[ashen cairn]

Hey. Which program for generating ai videos based on text? Something free and good like Rvc Gui, please.

[simple ore]

easiest is something like comfyui + t2v model of your choice

or t2i model to get a starter image + i2v model to make a video

[ashen cairn]

Okay... Thank you Noobies. One more question I'm writing long texts, and make free audiobooks possible would be cool i mean text to cloned voice. @simple ore

[brittle wing]

What's the best pretrain for anime Japanese speaking characters?

[ashen cairn]

I know now. Google text to speech and next step using rvc gui.

[fossil cape]

Hey, anyone here with experience in LLM memory systems, fact consolidation, or contradiction detection? I'm working on a project that handles reinforced semantic memory and dynamic fact resolution.

[ember root]

what software should I use for voice cloning RTC on win pc

[pseudo steppe]

yeah same happened with me too

[low shard]

This is a General AI Server, we won't be focused on voices anymore

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[low shard]

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[low shard]

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[low shard]

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

!howtoask

[patent trellis]

How To Troubleshoot

__**GIVE CONTEXT.**__ 📝

• Don't simply mention your issue, like "my rvc is not working".
• Describe the step you are on, what you're trying to do, the RVC you're using, a screenshot, etc.
• The more context, the better.

__**BE POLITE.**__ <:matsuripray:1159685390156967936>

• Don't be desperate. You can ping a Helper, but if they ignore, they aren't available/don't know the answer.
• It's okay if you're frustrated, but don't take it into this server.
• Don't DM without prior consent.

__**BE PRODUCTIVE.**__ 🤝

• Don't ask for every little instruction. Put your own effort & test things by yourself.
• Don't ask to ask.
• Check if your answer is a Google search away/on our guides website.

If you've changed your mind about your message, feel free to delete/edit it.

[golden walrus]

Sir Nick, can i ask if i want to improve a real-time model, i should focus on the data right?

because people somehow still spot i'm using AI right away

[low shard]

RVC models are 100% detectable, there's no such thing as a perfect model, RVC has limitations for like laughing and screaming btw

[golden walrus]

[low shard]

are you trying to train your own model, or use one made by someone else?

[golden walrus]

I'm using my own model

[low shard]

i can also suggest better wokada settings if u show a screenshot

[golden walrus]

I'm using Vonovox xD

Because of Spin

[low shard]

if you're like vtubing, you shouldn't laugh or scream lol

[low shard]

be sure to get as much high quality clean data as you can, and to use the tensorboard

[golden walrus]

Aye, i guess it's time to learn how to imitate voice

I use tensorboard ah

Just my data is from podcast

They have pretty much monotone voice

Like someone is telling stories

[simple ore]

there are plenty of tts options with zero shot voice cloning

[crude flame]

Longer dataset = more realistic results

Expressive data also helps

[topaz dirge]

anyone knows how to fix my problem?
i have a model with only a model and no index. at the beginning it worked well but now the voice is cut when i try to speak. my other model with index is working

[low shard]

This is a General AI Server, we won't be focused on voices anymore

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[arctic sand]

How many db should I put the microphone volume into the voice changer? And what is the voice changer input and output volume basic without amplification?

somebody helpp

[arctic sand]

helPppp

[pastel oak]

First part of your question, input volume is subjective, theres no right or wrong, every case is different

Just make sure your normal voice going into the voice changer is not too quiet and not fucking up your ears (turn on passthru to hear your normal voice coming out the voice changer to test this)

[low shard]

elaborate all i asked

[simple ore]

it is likely some model slot .json got bricked

MMVCServerSIO\model_dir\

[pastel oak]

adding to that, in case its not clear, delete the model_dir folder like the person you quoted before that already said himself @pseudo steppe

Make sure to save the models inside model_dir in case you deleted them off your pc

[arctic sand]

I need an exact figure. I'm going to put in a microphone volume of -6db

[pastel oak]

There is no exact figure because nothing is right or wrong, its just your mic volume

As long as your mic is clear going into the voice changer you have the most optimal

[arctic sand]

But isn't the sound distorted if you turn down or turn up the input or output volume?

[pastel oak]

vol in: controls how loud your normal microphone is going into the voice changer. You can leave at default 100%

vol out controls how loud your changed voice sounds like going out the virtual cable. This you can also leave at default 100% unless you prefer a louder voice, I would go for max 150-200% personally but everyones different

Some models are trained quieter or louder than others too, that is a factor

A quick self test can be to use discord self hearing and you determine if the voice coming out is adequate to your liking or not. The main point is there is no right or wrong, so we can not suggest you a setting because its all preference

[arctic sand]

thx

i need self testing

[analog obsidian]

anything works as long the input audio is not distorted

[pseudo steppe]

actually the thing is it for working fine from months but suddenly from today only this error is coming

[pseudo steppe]

4060
windows 11
discord

[arctic sand]

So if the sound is too loud, would reducing the output volume less damage the sound quality than the input volume?

[analog obsidian]

if the volume is too loud the model may sound robotic

[low shard]

u can try deleting the model_dir folder to fix it, might be a bad model, but u will lose ur model

[pseudo steppe]

thx it is resolved now

[low shard]

yw

[arctic sand]

why??

[analog obsidian]

because by turning the volume too high you're also boosting the natural artifacting rvc has

so they become more easily audible

[arctic sand]

what is natural artifacting rvc?

[analog obsidian]

model randomly sounding robotic

[arctic sand]

I do noise treatment on the microphone and it sounds the cleanest at a high volume. Microphone noise ratio snr

[analog obsidian]

as long everything sounds fine, don't worry much about it

[arctic sand]

and i use sample rate 48000hz and bit depth 24bit is it good?

and buffer size and max latency(vb cable) which value is good?

i use 2048 buffer size and 7048 max latency

[golden walrus]

I have around 1 hour of dataset.

Let me train it again

[stiff dome]

hi i am trying to make my own voice moddle for W-okada but im to dumb to get Applio to work even though i have the guide of the logs coud someone pls help me?

the program is working but im to dumb to use it

[ember root]

I want to clone a voice

[low shard]

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[low shard]

Elaborate:

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[stiff dome]

i have a 4060
i use windows 11
i wanted to make a voice model of my frend
i am using the totorial from the Ai hub logs

[stiff dome]

like i said everything works or at least i think it doas im just to dumb to use it

[low shard]

have you tried reading https://docs.aihub.gg/essentials/how-to-make-voice-models/ ?

[stiff dome]

thx

[wraith berry]

why doesnt my output work???

like im using the cable but its not working

[short hill]

help does anyone have an ai humanizer

[simple ore]

all good?

otherwise there's another alternative

[quick terrace]

how can I add model to applio

can you help me please?

[vapid mantle]

I start the first process in RVC v2 Disconnected colab but I get an error, what should I do?

[simple ore]

outdated af, tries to download pip 22 from 3 years ago

[vapid mantle]

-colab

[patent trellis]

📒 Google Colab Notebooks

Google Colab is a Cloud (Remote Good PC) Service. While the Free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

• **Applio**

by IA Hispano
Google Colab

• **RVC Mainline**

by Hina
Google Colab

• **UVR5 NO UI**

by Eddy
Google Colab

• **UVR5 UI**

by Eddy
Google Colab

• **Wokada Deiteris Fork**

by Deiteris & Hina
Google Colab

• **Hina's Modified Original Wokada**

Google Colab

• **RVC-AI-Cover-Maker-WebUI**

by Shiro & Eddy
Google Colab

• **FaceFusion UI**

by Nick088
Google Colab

• **FaceFusion NO UI**

by Nick088
Google Colab

• **Music Source Separation Training (Inference)**

by Jarredou & Makidanye
Google Colab

[vapid mantle]

So how can I fix this problem?

[simple ore]

very unlikely to fix, since it is probably using python 3.9 code while colab is now 3.11

[vivid hornet]

What web browser works best for the RVCC Wokada Deiteris Fork voice changer? I currently use Opera GX and have some issues with it spiking and lagging when set to recommended settings for my GPU

[vapid mantle]

Okay, I solved the problem thanks to ChatGPT.

[bleak nymph]

sorry where is this

mb for replying to a decade old message

[analog obsidian]

scroll down until you find the section where you set the batch size value then click advanced settings

[bleak nymph]

ohhh ty

wait so for everything else

this is good right

[analog obsidian]

if you truncate the silence of your dataset using audacity, yes
if not, use automatic

[bleak nymph]

i truncated in audacity

[analog obsidian]

how many minutes this dataset have?

[bleak nymph]

it didn't get all of it so i made the gain threshold a lil lower and i manually deleted some

[bleak nymph]

lets see...

21 mins 17 seconds

[analog obsidian]

you're not supposed to remove every silence, i personally left 0.35 of silence and let the automatic slicer do its job

[bleak nymph]

i also got an xlr cable for my mic and my golly does it sound soo much better than usb 😭

[bleak nymph]

oh yeah not all of it is gone

i just did it to remove like

keyboard noises

[analog obsidian]

nice

[bleak nymph]

it has some silence left

[analog obsidian]

batch size 8 may be ok, but if the results come out bad, you could try batch 4

[pine star]

How do I uninstall the original version of W-Okada? I dont see a uninstall file and idk if i want to just delete the folder cuz idk what else it installed on my pc

[bleak nymph]

what does batch size do again?

[analog obsidian]

as long is not over 0.35s of silence you'll be fine

[analog obsidian]

amount of slices the model learn every step

[bleak nymph]

don't think it is

[analog obsidian]

just delete the folder

[pine star]

oki

[bleak nymph]

oh and by "bad" u mean like overtrained or like robotic or smth

also should i turn on overtraining detector

[analog obsidian]

model randomly having issues like voice glitching/breaking randomly, weird artifacting

[bleak nymph]

oh ok

i see

what about this option?

[analog obsidian]

set it to auto

auto = if dataset is over 1 hour, uses kmeans instead

kmeans is intended to be used with big datasets to speed up index creation

for anything lower than 1 hour, use faiss imo

kmeans learn less thats bc it's faster

[bleak nymph]

oh ok

[simple ore]

over 200 hours

[left ore]

I wanted to try morphing my voice for an animated short project - I set up "Retrieval-based-Voice-Conversion-WebUI" in an virtual environment, but when I try to get any of these voices working, I have no success in making them show up in the gradio GUI
https://huggingface.co/QuickWick/Music-AI-Voices/tree/main
I tried looking for a how to and asking ChatGPT, but most documentation is in Chinese and ChatGPT is suggesting nonsensical "solutions" - help would be really appreciated
I used "python tools/download_models.py" and "Inferencing voice:" in the GUI is still empty.

[analog obsidian]

oh mb, it's over 1 hour in mainline

[simple ore]

it automatically does that when you have ~2500(?) slices

the setting is for 200000 slices

[bleak nymph]

200 HOURS???

😭

that's insane

[simple ore]

so it does not matter for normal model

[bleak nymph]

12.5 years to train fr

[simple ore]

and nobody makes indexes for a pretrain

[bleak nymph]

oh

uhhh

😭

[analog obsidian]

wait so it's a different thing vs what is in mainline?

[bleak nymph]

i already clicked generate index

[analog obsidian]

dont think much about it, it's one of those complex settings that were originally hidden from mortals like us lol

[simple ore]

exact same

[bleak nymph]

😭 damn

[simple ore]

it is entirely useless for anyone

[analog obsidian]

nice, i always thought 1 hour sets used that kmeans thing

[bleak nymph]

ok so i'm gonna play spider-man 2 and let it train for a bit

hopefully my pc doesn't 💥

[simple ore]

it automatically narrows the set if you got more than ~2500 slices.. yes, it is kmeans or whatever faiss method

[simple ore]

it is automatic thing by faiss

[analog obsidian]

gotcha, so confusing but i think i got it lol

[simple ore]

well, for 200k samples it uses sklearn.cluster.MiniBatchKMeans

[bleak nymph]

oh great i got an error

[analog obsidian]

?

[simple ore]

for 2500-200k samples it uses internal faiss minibatch, same kmeans

[analog obsidian]

makes sense lol

[analog obsidian]

you forgot to extract the features

[bleak nymph]

oh 💀

[analog obsidian]

[bleak nymph]

so i extract features and start training then

[analog obsidian]

what if you hide the kmeans option? so we don't use it by accident

[analog obsidian]

ye

[bleak nymph]

what the hell

another error

much simpler

[analog obsidian]

send ss

[analog obsidian]

you forgot to slice?

[bleak nymph]

huh

wym by slice

[analog obsidian]

preprocess the dataset

the literal first step

[bleak nymph]

i already did that

i literally did that before

[simple ore]

not the features... the whole thing

pointed the input to C:\training\file.wav