✨│ai-help

[analog obsidian]

its the max amount 2s or something?

[viral mason]

yea 0.2

[analog obsidian]

ah ok, set it to 0.1s

[viral mason]

alr

[analog obsidian]

longer crossfade length gives results comparable to rvc but increases delay

[viral mason]

oh ok

this place was wild when that weirdo was in here a second ago

https://cdn.discordapp.com/attachments/1177719453744644247/1362873370110201906/Bigbaby.gif

[untold marten]

I see that they migrated to web browser now

tho it gives error when trying to set the name of the model. also it only accepts RVC? can't upload GPT sovits models then?

[low shard]

I was eating lmao wtf happened

[misty wigeon]

So I don't remember what batch size and other settings I used to make a particular model, is any of that info stored in the logs folder?

[low shard]

tho it gives error when trying to set the name of the model
You should not rename the model in the UI, it's bugged

you should rename the model files then re-upload the slot

can't upload GPT sovits models then?
RVC is Speech To Speech
GPT-So-VITS is only Text To Speech

Those are 2 different AI types

[viral mason]

what were you eating?

[viral mason]

chaos

[low shard]

smt similar to focaccia

[untold marten]

ooh okie, gotcha. Thanks for explanations ^^

[low shard]

do you need any other help ?

[untold marten]

As of right now as far as I can remember, not yet but i will come with any questions/problems as I explore this web version if thats fine

[viral mason]

got it working again

[low shard]

alright

[untold marten]

guess it doesn't have as well an app version that can be added to the system tray right?

[low shard]

it's still a program running locally on your pc hardware

both original wokada and wokada deiteris fork use a Web User Interface coded in Javascript & TypeScript

Web User Interfaces are used by the great majority of AI programs because:

• easy to edit
• can be used for cloud

the original wokada just made it's own 'browser' to open the local hosted url webui, and iirc that was removed in the deiteris fork for performance issues

[untold marten]

I see. Thankyu ^^

[low shard]

You're welcome

[karmic egret]

why does my voice sound like a broken Roberter

I don't understand it, I have everything like in a Yt video

[low shard]

Yt video
That's one of the issues, all video tutorials contain outdated info

AI progresses at sonic speed

you can't expect an over year old video tutorial to set you up

[karmic egret]

soo that video was yesterday

[low shard]

share a link to the video you used so i can check if they used updated software

and also tell what's your pc gpu

[karmic egret]

Is it because of the voice model?

[karmic egret]

an Gefroce 4090

[low shard]

great, share a link to the video you used

[karmic egret]

idk i watched that on tv haha

wait

[low shard]

share a screenshot of the program you used

it could be both a bad model and outdated software/ wrong settings

!give-media-perms 1h @karmic egret

[karmic egret]

Should I send a screenshot of the voice changer?

[low shard]

an entire screenshot of it yes

or the youtube tutorial since you did the same thing, but you said u forgot which

[low shard]

yep, awful settings and extremely outdated software lmfao

what video did you even watch

[karmic egret]

idk

[low shard]

that's an old version of original wokada

[karmic egret]

a yt video haha

[low shard]

vb audio cable also gives random issues on windows

uninstall both

[karmic egret]

okay

[low shard]

and also the settings were pretty fucked, crepe tiny sucks in quality

[karmic egret]

wait a sec

i have

uninstall both

[low shard]

great

[karmic egret]

so what is the new vc?

[low shard]

wokada deiteris fork

there's 2 main versions of wokada:

• original made by wok
• deiteris fork (modified version) made by deiteris

the one you're using is the old version of original wokada which has worse performance and quality

read the wokada deiteris fork guide: https://rentry.co/forkvoicechangerguide

all video tutorials are outdated, only the written ones are up to date

[karmic egret]

Another supporter sent this to me too, I have no idea where the VC download link is

[low shard]

you have to read the guide, it's all in it

it's exactly in the windows nvidia part https://rentry.co/forkvoicechangerguide#download-nvidia-on-windows which redirects you to https://huggingface.co/Shadicti/deiteris-Fork/blob/main/voice-changer-windows-nvidia-b2332.zip

it's crucial to read all the guide and not miss steps

also be sure to not skip step 3 and get vac lite

then do the other steps

[karmic egret]

okay i will try

thanks

[low shard]

you're welcome and let me know

[karmic egret]

what should I choose here I can't speak English well haha

[low shard]

you should just set your usual microphone as the default one

seems like you did that already

[karmic egret]

alr thanks Download VoiceChanger now

[kind drum]

I am gong to change my Asian voice to European voice. Which model is fast and efficient?

[untold marten]

it's nrcessarily to uninstall virtual cable that i installed some time ago or can install vc lite along?

[karmic egret]

what is that 😦

[kind drum]

Are you all using realtime voice changer client?

[low shard]

let me know

[untold marten]

i use server most of the time

as i heard it's better

[karmic egret]

idk what to do know

[low shard]

not necessary but would be better you uninstall vb audio cable

[untold marten]

while i do not know what are those scripts are exactly, u don;''t need to run them as the app is inside the folder

[karmic egret]

okayy ty

[untold marten]

called MMVCServerSIO.exe

[low shard]

The force gpu clocks, Forces ur gpu to be constantly active, this helps if ur gpu goes idle state

it's usually not needed tho

you can just go inside the folder and run the .exe

[karmic egret]

okay thanks

[low shard]

shortly:

Server = can have less delay with wasapi and asio

Client = can use noise suppression and echo reduction

that's the difference

[karmic egret]

why im on an website

[untold marten]

thats the interface

it's basically the same as old okada

just better in performance and easier to edit

[low shard]

it's not a website, it stills runs locally on your pc gpu, just it's a web user interface

please check https://rentry.co/forkvoicechangerguide#why-does-it-run-in-a-browser-and-not-its-own-window for futher details

[karmic egret]

okay

I fixed it but I'm still a broken roberter

[untold marten]

broken in what way? did u try the crackle fix ?

[karmic egret]

yue can you go in a call with me for an sec to see if it sounds weird to you too

[untold marten]

sure, can try

[karmic egret]

im a cracky roberter

wait

[untold marten]

hello?

[karmic egret]

wait

[untold marten]

was sounding fine i believe, i mean no crackle or robotic voice

[karmic egret]

i cant talk discord dont accsp my cable

[karmic egret]

okay thanks

[untold marten]

np

[karmic egret]

haha

yue accsp my dm

[low shard]

in discord, set input as line 1 and output as headphones

[karmic egret]

yes

ty

[low shard]

and ofc remember to set the input as microphone and output as line 1 in wokada

do u want to share a screenshot of ur settings?

i can help u set them up

[tropic garden]

I am getting this error: RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory in Applio Google Collab. Any suggestions?

[low shard]

Elaborate:

• your PC GPU
• the google colab link
• what you want to do and did
• the model download link

[waxen pike]

what do I have to do to change my tone of voice?

[untold marten]

change the pitch slider

[waxen pike]

the robot's voice is not natural.

[tropic garden]

1. Not PC 2. Google Collab Link from AI Hub Guide 3. What you did? Nothing. 4. What am I doing? Waiting. 5. What did I do? Pressed Start Training. 6. What was I doing before that? Placed KLM pretrain from AI Hub Link 7. The model download link from KLM Hifigan 32 K available through AI Hub.

What I think was going on? Maybe that KLM download is corrupted.

What do I do next?

[low shard]

those settings are wrong

gpu: rtx 4060
f0: rmvpe without onnx
extra: 2.7

about tone, change the pitch, it depends by your voice and the model

did you download vac lite?

[low shard]

Not PC
Are you on mobile? Was asking if you're on a desktop/laptop pc since you could have a good one and do it locally instead

Google Collab Link from AI Hub Guide
yeah but there's applio no ui and ui collab

The model download link from KLM Hifigan 32 K available through AI Hub.
Which?

[waxen pike]

I don't understand why the old one is worse, and I can't find anything about it.

[low shard]

The old one has worse performance, the newer one, deiteris fork, has better optimization and also settings that can help the quality

for example, if you turn force fp32 mode in advanced settings, you will have better quality and a more stable model at the cost of a bit of performance, as of my test on my 4060 ti 16gb, turning that on gave a 20ms delay but it did help with quality

[karmic egret]

yes

[low shard]

!give-media-perms 1h @karmic egret

[karmic egret]

here

[low shard]

@waxen pike try the settings I told you, and be sure you downloaded vac lite

[low shard]

the entire wokada settings page

that's just the start

[karmic egret]

I don't know how to set everything up correctly. When I opened it, it was like this.

[low shard]

scroll down

[tropic garden]

Applio UI Collab and this is the 32Khz
G-https://huggingface.co/SeoulStreamingStation/KLM49_HFG/resolve/main/G_KLM_HFG_32k.pth?download=true
D-https://huggingface.co/SeoulStreamingStation/KLM49_HFG/resolve/main/D_KLM_HFG_32k.pth?download=true
I'm going to use another link and see if that is the issue.

[karmic egret]

the rest works perfectly

[low shard]

you already setted the extra, f0, and everything else?

because you shouldn't leave them on default

on default it uses your CPU

[karmic egret]

I have to come later, my parents want me to come down

[low shard]

alright, let me know

[karmic egret]

I have everything perfect When I turn it on everything sounds good but I think I have to use a different voice haha

[low shard]

seems like the pth wasn't downloaded fully #✨│ai-help message

maybe retry

[untold marten]

I have few small questions regarding onnix files/settings. Basically based on the guide for deiteris fork, onnix isn't needed anymore for NVIDIA users but can help AMD/Intel users right?

And regarding chunk, iirc on old okawa, if i was giving more, there was the delay but was giving better quality, and this remains the same here right? tried 85ms (hard to hit 72 but that was recommended for 3090 as was highest card with recommended settings) and i believe it gives good results but if i want better, having on idk.. 100-200 should help the quality right? altho delays

and F0 est can be changed from rmvpe_onnix toi simple one as no onnix file is used i guess

[analog obsidian]

amd users werent able to run .pth before, so they were forced to use onnx
nowadays the fork allows the usage of .pth in amd/intel

you can use rmvpe_onnx in a non onnx model but why? its a downgraded version of rmvpe that is meant to be used with an onnx model

you're just increasing your cpu usage for no reason if you use onnx f0 with a .pth model

onnx is there for compatibility, not because its better

[worthy vale]

hey guys what is this? Please open the following URL in your browser.
http://<IP>:<PORT>/

like I downloaded 1 model and tried to reopen start http

[untold marten]

guess i can't post the link to the guide

but basically i followed the settings in the guide from rentry link and set the settings there (+ advanced) and helped the voice to sound more smoothly.

[analog obsidian]

w-okada does not give better results than the fork

they're the same thing..

[vast elm]

can you provide further log? theres not much to make from that single line, perhaps the errors/outputs before it?

[analog obsidian]

actually the fork should give better results because fp32 inference is way better than fp16

but that depends in the model

a trash model is gonna sound trash no matter what u do

[untold marten]

i think fp32 helped more ? not sure but anyway using as well now vc lite

[analog obsidian]

its not a quality setting

is a stability setting

fp32 calculations are more precise than fp16

in simple words, that translates in the voice not randomly glitching for no reason

[untold marten]

it said in guide that greenm means stable right? can it be misleading? i mean be a bit unstable but still showing green?

bcs it was having some weird cuts first time

[worthy vale]

I message you

check out

screen

[untold marten]

then prob thats why it wasn't working that well

[analog obsidian]

original w-okada has only fp16 inference

so things randomly explode there

random artifacting, voice completely glitching, etc

[untold marten]

was on this one from deiteris at begging when i said that it was having random cuts

[analog obsidian]

a thats a you problem

try chunk 150ms, 2,7s extra

.pth

don't use onnx

1s crossfade length

[untold marten]

did that after looking more

wanted to show final advanced settings pic but my perms expired i think

[analog obsidian]

changing browsers also help, i've heard people have issues while using operagx

its weird idk

[untold marten]

I have operagx :))

[analog obsidian]

that explains

try chrome or edge

[untold marten]

tho i think i managed to make it work

prob bcs i have a good gpu? 4070 ti super

[analog obsidian]

i know that random cuts happen when the extra is too low or when you're lagging

or also when chunk size is too low

onnx puts more stress in your cpu so if u were using onnx that also might explain why the voice was lagging

[untold marten]

but anyway, i have:

protocol: sio
crossfade 0.15
silence on
fp32 on
disable jit on
false onnx
protect 0.5 (left as it is as idk what to set exactly...)
skip pass through conf no
chunk 85
extra 2.7s

i have a model with pth file no onnx

and no index as i heard it's not needed anymore

anything that needs change?

using rmvpe as f0det

-90 db in. sens

[simple ore]

it means the file it tries to open is damaged/not downloaded properly

[untold marten]

and at audio i set the SR to 48k

[analog obsidian]

i would try:
close the voice changer > install chrome > set default chrome as default browser > open the voice changer > change chunk to 150ms, protocol rest

[untold marten]

will try as i have chrome as well but not using it unless smth doesn't work on opera or firefox :))

[analog obsidian]

okay good luck fixing ur thing

i guess

[untold marten]

well it's more or less ok. just asking if those settings are fine as they are or needed small changes but gonna try now chrome

[analog obsidian]

chunk 85 is fine as long you're not gaming

for games increase chunk to 150ms ~ (well that depends how gpu heavy the game is)

its random so i cant tell ya

[untold marten]

gotcha!

[analog obsidian]

every game has their unique chunk value lets say lol

[untold marten]

iirc, i think it said somewhere to increase if i use a game

[analog obsidian]

yeah

just play with the chunk value, you know u did it well when the voice is stable

[untold marten]

thanks!

uhhh, wanted to try another pth model with index file and got this error...

wait, i stopped this and started again and it works now?!?

[low shard]

lemme guess, are you using a youtube tutorial for a realtime voice changer?

[worthy vale]

ye

[low shard]

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

what's your pc gpu?

[worthy vale]

3050

rtx

[worthy vale]

I reloaded everything but now its even dont opening

[untold marten]

download the fork made by deiteris. that should work after setting up and reading the guide from rentry website

[low shard]

good enough

[worthy vale]

ty

[low shard]

just uninstall the version you got off youtube and vb audio cable

vb audio cable gives random errors on windows

[worthy vale]

I mean I installed this of website

[low shard]

the version you got is an old original wokada

[untold marten]

https://rentry.co/forkvoicechangerguide#windows

[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

[worthy vale]

from the git hub

[low shard]

did u follow the 1st link guide?

[worthy vale]

I can send link

[untold marten]

follow the guide here for deiteris fork

[low shard]

the updated link is the one above

the wokada deiteris fork

Wokada has 2 main versions:

• Original made by Wok
• Deiteris fork (modified version) made by Deiteris

each version has it's own updates

the wokada deiteris fork is more suggested

[untold marten]

also contains the link for virtual cable lite

[worthy vale]

I just reloaded it

now all works

[low shard]

it's better you don't use the version you got off youtube

it's outdated, it got worse performance and quality

[worthy vale]

ty guys'

nah I got it from website from github

I just watched how to setup this

[low shard]

if you're talking about https://github.com/w-okada/voice-changer , this is the version i'm talking about

you should not use it

it's the original wokada, it's not as good as the wokada deiteris fork

[low shard]

it's better you just forget what you got off youtube and follow the deiteris fork guide: https://rentry.co/forkvoicechangerguide

video tutorials aren't up to date

[fallow river]

Hello, is there a way to do some TTS client that automatically applies an RVC model onto the TTS?

[crystal gull]

Currently bashing my head against my desk rn, I'm trying to get llama py to properly use my gpu with cuda to generate llm text. and no matter how I install the pip packages, no matter how i set my build flags. It keeps saying it's assigning layers to my cpu, any help?

I've tried pip install llama-cpp-python --upgrade --force-reinstall --no-cache-dir --compile --extra-index-url=https://pypi.nvidia.com, set CMAKE_ARGS=-DLLAMA_CUBLAS=on before building, CMAKE_ARGS="-DLLAMA_CUBLAS=on" pip install llama-cpp-python --force-reinstall --no-binary llama-cpp-python, $env:CMAKE_ARGS = "-DGGML_CUDA=ON". etc etc

Python 3.12.6
RTX 4070
Toolkit 12.8

[kind drum]

May I know the effect of Pitch, Index, F0EST, N.GATe parametsrs in Voice Changer Client?

[simple ore]

python 3.11

I do have build tools installed though

[crystal gull]

Yeah I've followed through with all that in the past on 3.12.6 and it just uses a cpu version, not sure if that different py version will make a difference

Unless I have an issue with my py

llm = Llama(model_path="./models/ana-v1-m7.Q4_K_M.gguf", n_ctx=2048, n_gpu_layers=32)

I do specify ctx and gpu layers

[simple ore]

make sure you uninstall CPU one, then remove the pip cache

then pip install llama-cpp-python --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cu126

although that failed for me

[crystal gull]

sure ill clear my venv make a new one and use that and see if it makes any difference

[simple ore]

not venv, the pip cache

[crystal gull]

well pip is in the venv no?

clearing the venv would clear the pip cache unless im mistaken

[simple ore]

no, it is un global repo

[crystal gull]

huh okay

[simple ore]

pip cache purge

[crystal gull]

pip install llama-cpp-python --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cu126
states requirements already satisfied for everything

Oh 1s

[crystal gull]

Okay I've cleared everything out including cache, install llama-cpp-python using that.
When I start my python script it still appears to not raise my gpu usage or my vram, and it states cpu layers in the terminal.
load_tensors: layer 0 assigned to device CPU

[simple ore]

anyway, I did not bother much with llama

i have lm studio installed

[crystal gull]

Does LM Studio have an api?

Huh well I finally found this error after trying to compile my own with cuda

        No CUDA toolset found.```

-- CUDA Toolkit found

So I have cuda toolkit but not toolset.. No clue what that is and can only find toolkit downloads

[simple ore]

there's just way too much wizardry required for installing things on windows

LM studio provides a standard API for the model

send a payload request, get whatever back

I run orpheus TTS with it

[karmic egret]

I sound like a stoner, at the beginning everything was fine and after 30minutes of use I sound like I've smoked weed

[sick fog]

How do i run this through vrchat without the thing echoing there voice

[low shard]

share a screenshot of ur wokada

!give-media-perms 1h @sick fog

[sick fog]

The frick is wokoda

downloaded this do not know if this si the one

the github is a mess to traverse

like it just will not work on vrchat even when i got vb audio connected

or smtimes the oputput of my actual game goes to my vb audio which is so annoying

i got these settings if i have to I'll swapout my 4060 for my rtx 4090

or just go afk ig

[trim sparrow]

0%| | 0/56 [00:00<?, ?it/s]/kaggle/tmp/.venv/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: Grad strides do not match bucket view strides. This may indicate grad was not created according to the gradient layout contract, or that the param's strides changed since DDP was constructed. This is not an error, but may impair performance.
grad.sizes() = [64, 1, 4], strides() = [4, 1, 1]
bucket_view.sizes() = [64, 1, 4], strides() = [4, 4, 1] (Triggered internally at ../torch/csrc/distributed/c10d/reducer.cpp:325.)
return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
/kaggle/tmp/.venv/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: Grad strides do not match bucket view strides. This may indicate grad was not created according to the gradient layout contract, or that the param's strides changed since DDP was constructed. This is not an error, but may impair performance.
grad.sizes() = [64, 1, 4], strides() = [4, 1, 1]
bucket_view.sizes() = [64, 1, 4], strides() = [4, 4, 1] (Triggered internally at ../torch/csrc/distributed/c10d/reducer.cpp:325.)
return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass

i keep getting this error

with applio

[fast scarab]

Hey guys, I’m having an issue with Codename-RVC-Fork (v3.1.0-rev1).
When I start training, it crashes while loading the filelist.txt due to an encoding error.
Anyone know how to fix this or had the same problem before?

[proud hinge]

Im having troubles with the Real time voice changer, it keeps buffering and not play the full audio just parts of it how can I Fix this

[knotty moth]

it is a mess because there is a better one here https://rentry.co/forkvoicechangerguide

[sick fog]

i just got that one

took a long time searching through old support messages tho

it seems fine tested it through discord idk bout vrchat thoo

[analog obsidian]

vrchat in vr mode or desktop

?

coz the 4060 might not be strong enough to handle vr mode + the voice changer

but i mean the only way to know is to try it

just disable badly optimized avatars and reduce every gpu intensive task in the game

[sick fog]

want me to swap to my 4090?

also i have all avatars off cus I'm friends with alot of crashers n they join me n crash so its annoying

[analog obsidian]

ah you've got 2 gpus in one pc

[sick fog]

Nah

my 4090 is in my other pc behind my desk

I'm upgrading the case but I'm too lazy to swap the parts but if i need to I'll do it rn

[analog obsidian]

first try if the 4060 can handle vr mode, iirc the 3060 can't, and the 4060 isnt that different from a 3060 sooo

i think its like 10% faster?

[sick fog]

I've made my own vrc client

it is like a optomiser

[trim sparrow]

Can anyone help me fix this?

[simple ore]

there's something wrong with your dataset

[trim sparrow]

it sounds perfectly fine and is at the right sample rate

[simple ore]

i did not say how it sounds, I mean the files are fked

did you resume training with a wrong sample rate?

[trim sparrow]

no everything at 40k

[simple ore]

double check

the error is because it tries to cut a piece of the file and gets past the end of the file

[trim sparrow]

yeah im sure i tried deleting the problematic dataset but even then its still not working even with the previous datasets that worked

[knotty moth]

the preprocessed files might have fked up
if not sure, start over from preprocessing

[trim sparrow]

redoing it rn

yeah still doing it

[trim sparrow]

not sure what to do now

i mean i started a new voice model with the same ones as before and i mea nits working now..? only issue is now i gotta wait for 1000 epochs sadly but it is what it is i guess if anyone knows how to fix the other one please lmk

[coarse pier]

yo
quick question
does mic quality in okada matter?
cause my mic is fucking ass

[sick fog]

What settings should i use for a pretty quick ai voice but for it to sound realistic

[knotty moth]

define being realistic

if you want it sound less robotic, try another model

[lost dove]

Hello, I am looking to get information or an analysis on a specific Ai instagram model / influencer.

This influencer uses a unique approach to video creation. And I cant pinpoint exactly how they are creating their videos

I am willing to pay someone to do research for me or tell me how they are doing this, how they set it up, and How I can recreate it.

Here is the information I know:

This is the creator I want to analyze: https://www.instagram.com/marina_hasegawa_/

I suspect that they are using real reference videos taken from other sources and replacing the characters / scenery with their own. I'm not sure what software they are using to do it though

I think I know how they created the character(s) in their videos. But I don't know:

• how they applied them to the video format mentioned above
• How they are able to maintain photorealistic quality
• How they are able to do all of the above and keep the custom Text and Jersey number 28 in all of the videos
• How they change the outfit for the characters.

---

[hallow thistle]

Stop.

[lost dove]

Sorry

Can I not post links?

[hallow peak]

Is Okada colab working?

[hallow thistle]

No where in this server is allowed to promote your own things.

[lost dove]

Im not promoting?

I'm requesting, Trying to learn how this ai influencer makes their videos

[hallow peak]

??

[hallow thistle]

What is your PC GPU? Most of W-Okada Colab notebooks are broken currently.

[hallow peak]

I see. I don't have a pc as of now. I plan on getting set up with one later this December.

Thank you for letting me know. I'll be patient.

[hallow thistle]

The what? You don't have one? That's bad.

W-Okada won't gonna work with mobile devices since they lack the feature to change input/output audio devices like the one in PC.

[knotty moth]

sorry but that sounds like another "I will pay you after I get some money from that"

[hallow thistle]

W-Okada can run locally. If you're looking to buy a PC, I'd say to buy one that has a decent Intel/AMD CPU and NVIDIA GeForce RTX 30-40 series GPU.

[knotty moth]

afaik it may still have issues #📰│dev-updates message

you can try this kaggle notebook https://www.kaggle.com/code/suneku/voice-changer-public

[craggy bough]

i replied in #🧬│ai-chat, that isnt an ai influencer, its a real person

[knotty moth]

still they could have been using some AI tools for certain tasks

[gritty zinc]

is that all it is to make it sound less robotic is different models?

[elder coral]

How do you make a model with applio when im in the training path it says the parameters of the model do not match

[simple ore]

start from the start

select correct sampling rate that matches the pretrain you're going to use

[brittle wing]

umm guys i was trying the ai on a game

and it kept capturing the voices of the ppl playing asw

idk how to fix this can some1 helpp ??

[simple ore]

select the actual mic, make sure you playing in headphones and they dont leak audio

[brittle wing]

yes i put my actual mic as input and virtual cable as output

and i have echo sup1 and sup2 turned on

ughh i think i need to buy new headphones

[knotty moth]

use headphones that leak audio and turn down its volume to listen

[brittle wing]

wdym

like i should turn my headphones volume down to listen to what

[low shard]

that is wokada, the original wokada

you got an old version

wokada is the program meant for doing RVC realtime inference (use models)

never use video tutorial, delete that wokada along with vb audio cable

since you got a 4060, you're good enough

read the updated wokada deiteris fork (modified version): https://rentry.co/forkvoicechangerguide

[low shard]

elaborate:

• your pc gpu
• what you want to do
• what tutorial link did you use
• a screenshot of the program and error

[vagrant gale]

best rvc fork for gtx 1660 I wanna train my own voice model

• if anyone knows any slovenian pretrain models

[simple ore]

best gas for 1975 pinto? I wanna race F1

[vagrant gale]

or slavic languages

[low shard]

original wokada hina mod colab is broken

wokada deiteris fork colab works, but you need a pro subscription because google detected it uses a web ui, which isn't allowed in free ui

tell your pc gpu and what you want to do

[vagrant gale]

I already made 2 with this gpu

but I forgot which rvc I used cuz it was like 2 years ago

yes it took 2 days of my pc running non stop but it was worth it

[low shard]

elaborate:

• what's your pc gpu
• what you want to do
• what tutorial link did you use
• a screenshot of the program

[vagrant gale]

?

[low shard]

simply that gpu is cooked

[vagrant gale]

.

[low shard]

you could even train but it would be slow asf and limited by vram

it's not worth to train locally when cloud exists

[vagrant gale]

you can train with cloud?!?

wow

[simple ore]

kaggle, google colab

will be faster than 1660

[vagrant gale]

alr thx I'll check it out

[low shard]

Train (make) RVC Models on cloud:

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

• Google Colabs (4 hours of daily gpu for free, not much hours, but easy to use):
  • Applio (ui)
  • Mainline (UI)
• Kaggles (a bit harder to use and needs phone number but gives 30 hours weekly of better gpus):
  • Mainline (UI)
  • Applio by Vidal (UI)
• Lightning.ai (Kinda hard, needs login, no issue with web uis or anything, but only free 15 credits monthly):
  • 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

Mainline is broken though

[vagrant gale]

thx 🙏🙏

[low shard]

it's just better to use applio kaggle or colab

[vagrant gale]

?

or any slavic pretrains

[low shard]

yw, if u really want to use local, applio is good https://docs.aihub.gg/rvc/local/applio/ but it would be just way slower

[low shard]

don't think so

maybe check #1235952130855010365

but last time i checked i don't remember seeing one

[vagrant gale]

ye that's why im asking haha

[vagrant gale]

didn't know I can train on cloud

I'll check that out first

[low shard]

cloud means you're using using a remote good pc btw

[vagrant gale]

ikik

just thought that g collab method was broken last time I checked

[simple ore]

I would not recommend those claiming to have some, they may have the audio of specific language, but the content so variable it sounds shit

[low shard]

cloud isn't as much stable as local ofc, it does happen it breaks easily, but currently applio colab and kaggle are working, just mainline is broken but anyways applio is better

[simple ore]

regular default pretrain + 30-60 min set produces good output

[vagrant gale]

so what pretrain should I use?

[simple ore]

with default pretrain, no issues with the language

[vagrant gale]

alr

[vestal helm]

Is there any reason why

there isn’t a 44k option in rvc / pre trains

[odd shale]

Noobies tried to train a 44k pretrain for refinegan but it didn't work out well

[vestal helm]

Interesting

[fast scarab]

Hi,
I haven’t been able to find any Kaggle notebook that works with Codename-RVC-Fork 3.
Has anyone here managed to adapt it or get it running on Kaggle?

[gusty finch]

i dowlaod The VoiceModulation but how do i open it

[low shard]

lets talk in https://discord.com/channels/1159260121998827560/1363521952358072430

[untold marten]

Heya, since it's possible to use it on multi pc setups, i have an older pc with 1660 3gb vram. i believe that won't be enough even if i run just the app on it, right?

[hallow thistle]

Explain about the "app" you're trying to run over LAN network.

[untold marten]

altho i know that i must run at these settings for 16xx series:

256 ms chunk + 2.7s extra

[untold marten]

the deiteris fork

looked on this https://rentry.co/forkvoicechangerguide#for-nvidia-on-windows

but just wanted to ask if it may work decently

since it has just 3gb vram...

[hallow thistle]

With NVIDIA GeForce GTX 10/16 GPU in a host PC, it will still work fine, especially Dertis' fork W-Okada. Don't expect it to run W-Okada with a game on host PC that well.

[untold marten]

sorry if it doesn't make sense or u don't understand. please ask and i'll try to explain better.

[untold marten]

Yeah, ive tried with the old one 1y ago+ and didn't like it at all :))

and wasn't deiteris fork on that time too :))

[vagrant gale]

tried to use kaggle applio and got error when starting : ERROR: Exception in ASGI application

when I try to use the ui link ngrok says : dial tcp 127.0.0.1:6969: connect: connection refused

[vagrant gale]

TypeError: argument of type 'bool' is not iterable
An error occurred launching Gradio: When localhost is not accessible, a shareable link must be created. Please set share=True or check your proxy settings to allow access to localhost.

[untold marten]

then thanks for the fast reply ^^

[simple ore]

local install?

[vagrant gale]

cloud

kaggle

applio

[simple ore]

for kaggle import the latest notebook from github

[vagrant gale]

alr

what github tho?

I did tho

https://www.kaggle.com/code/deiant/applio

this right?

[simple ore]

create a notebook, import notebook,

[vagrant gale]

alr

it wont let me run the ngrok cell?

only the first install one

i'll run the install one first ig

works now 👍

[low shard]

sooo, did you try the settings I told you?

btw how's it going

[long oyster]

can i have some guide to use?

[simple ore]

just install it

[low shard]

https://rentry.co/forkvoicechangerguide#virtual-audio-cable

wait

what's your pc gpu?

aren't you using wokada deiteris fork already?

[simple ore]

input: real mic, output: line 1 - out of the virtual cable

[low shard]

yeah, uninstall vb audio cable, get vac lite at step 3, it's said in the same guide you followed to get the program

[simple ore]

discord input: line 1 - in of the virtual cable

[low shard]

don't skip steps

[long oyster]

okay

can i get some helps, please?

@low shard is this right?

[lost dove]

No, I will pay you to get me the information about it.

[lost dove]

Check the Request, Its not a real person.

[low shard]

did u run setup64.exe

and did u uninstall vb audio cable before that

[lost dove]

Read my responce to you

It is Ai generatred, and I'll prove it to you.

Check your dms

[vagrant gale]

I tried using aicovermaker keggle but it gives me erorr : python: can't open file '/kaggle/working/main.py': [Errno 2] No such file or directory

and ngrok when i go to link : dial tcp 127.0.0.1:7755: connect: connection refused

[low shard]

@viscid moss your ai covermaker ui is broken

[vagrant gale]

copied from here btw

should I import it from somewhere else?

[viscid moss]

hmmm lemme try rq

it was working

[vagrant gale]

what is the github for the kaggle version

[vagrant gale]

,

[viscid moss]

https://github.com/Eddycrack864/RVC-AI-Cover-Maker-UI

[vagrant gale]

aaa lm try rq

[viscid moss]

local installation isn't working too

so just colab ig

[vagrant gale]

haha rip

[viscid moss]

lemme try

[vagrant gale]

I tried importing it from github

it's installing rn

nope

it's broken

[viscid moss]

are u sure that u run the right notebook?

Version 6 btw

it's working for me

[vagrant gale]

can you send me the link?

[viscid moss]

on Kaggle and Colab

[vagrant gale]

I also tried collab rn and this doesn't work either

or am I just stupid

[low shard]

the run.bat?

[vagrant gale]

nvm collab works

just needed some time 👍

[viscid moss]

https://www.kaggle.com/code/eddycrack864/rvc-ai-cover-maker-ui

[vagrant gale]

i'll try it later

collab works tho

[viscid moss]

Ok, make sure to run version 6

[viscid moss]

ye

not the run.bat, it has dependencies conflicts

ppl said that precompiled version isn't working too

[low shard]

the run.bat seems to be working so far on my end, cloning main branch

let's see

[eternal aurora]

@low shard

I have an igpu

[low shard]

😭

you don't have any dedicated gpus?

[eternal aurora]

My cpu is amd Ryzen 7 5700g

[low shard]

did you check in task manager, performance tab ?

[eternal aurora]

Unfortunately not

[eternal aurora]

That is what it says

[low shard]

GPU is one of the most important things in intensive tasks like AI, running it on cpu is useless since you're gonna have unstable ping

you can run it on cpu but it's not worth at all

[eternal aurora]

Ohhh fairs right I’ll leave it then

[low shard]

so yeah your pc would be too weak for a decent local experience

[low shard]

you can use cloud

[eternal aurora]

I’ll just get a 1080 then

[low shard]

which is remote good pc

[eternal aurora]

The google drive thing?

[eternal aurora]

Is it free?

[low shard]

i think you meant google colab

you can use google colab for 4 hours daily max for free but the gpu time is kinda random

the issue is that hina mod original wokada is broken, and that the wokada deiteris fork colab gets detected for using a web user interface which isn't allowed in the free tier, so you'd need a pro subscription

but, you could use Kaggle for free, with 30 hours of weekly gpu, it's just a bit harder to use and needs a phone number verification (it's owned by google)

this is the wokada deiteris fork kaggle: https://www.kaggle.com/code/suneku/voice-changer-public

[low shard]

I mean that's kind of decent, but I advise you atleast an RTX card, since that card is barely higher than the bare minimum

are you going to use this in discord vc or games too?

[eternal aurora]

Thank you appreciate it I’ll check it out

[eternal aurora]

Dc vc

[eternal aurora]

Is a rtx 3050 good?

[low shard]

with a gtx 1080 it's decent only for discord vc, like it works but not max quality and low delay

[low shard]

btw are you going to buy a laptop?

[eternal aurora]

Naaa I’ll just add it to my current pc

[eternal aurora]

Ohhh I see

[low shard]

I would suggest an rtx 3060, btw you're not really forced to buy it since you got cloud

[eternal aurora]

yeahh its for the future possibly

[low shard]

yeah, if you want even an rtx 3050 would be good but an rtx 3060 might be better for higher vram, tbh depends also on what you're going to use the PC for other than this program

[eternal aurora]

Yeah that’s also true I would be prolly be using it for gaming

I’m just setting up my acc rn

[low shard]

Then yeah if you want a budget GPU an rtx 3060 might be better

Or you could get an AMD card since they are cheaper, it would be good for gaming

Just keep in mind that ai support on AMD isn't as good as Nvidia, but fortunately Wokada deiteris fork improved amd performance

[eternal aurora]

Oh is it? That works quite well ngl

I’m just confused rn with the installation

[low shard]

Yeah, you could have issues on AMD for other ai programs though

[low shard]

What's wrong ?

[eternal aurora]

And that’s true also

[eternal aurora]

i cant find the notebook option

[low shard]

Click edit my copy, then check session options

Kaggle recently changed the name of notebook options to session options

[eternal aurora]

okay im on session options what should i do?

oh wait

nvm

i just need to put the gpu

how do i do the 2nd part

like this?

[low shard]

That's just the name of the cell basically, just run what's below

[eternal aurora]

okay

[eternal aurora]

so like this?

just paste it into an empty line?

and for this do i copy the whole thibf?

[low shard]

Nope

All you gotta do is run the cell

There's a play button at the top left of the cell

[eternal aurora]

that?>

[low shard]

Yup

It will execute the code

[eternal aurora]

yeahh like this?

i pressed this command but im unsure if its doing anything

[gusty cloak]

Is that ok if Time per epoch is 0:00:01 while training model?

[long oyster]

yyes

[eternal aurora]

i got it working

[low shard]

yup

[low shard]

yup it did as u see in the output

[low shard]

Nice, be sure to set https://rentry.co/forkvoicechangerguide#virtual-audio-cable only this part on your pc

A VAC (Virtual Audio Cable) makes a fake audio device, used to re-route the audio of different programs

In Wokada context, it's used to get the output of wokada as the input in other programs

[low shard]

• what's your pc gpu
• what tutorial link are u using
• what's your dataset size and training settings

1 seconds per epoch seems weird

[eternal aurora]

yhyh i got all of that

thank you tho

[low shard]

u got vac lite, not vb audio cable, right?

bc vb audio cable gives random issues on windows

[eternal aurora]

nooo i got vb

[low shard]

[eternal aurora]

i got it yesterday

[low shard]

uninstall it

[eternal aurora]

ill change it dw

[low shard]

get vac lite from that link i sent, unless u want more issues lol

[eternal aurora]

na no more issues pls

[low shard]

also share a screenshot of ur wokada, so i can suggest settings

[eternal aurora]

how do i uninstall?

[eternal aurora]

okay ill do that right after i get vac lite

[low shard]

go in the windows settings, app > installed aps > find it > 3 dots at its right and click uninstall

[eternal aurora]

yep i just done that

[low shard]

🔥

[eternal aurora]

what do i download from the link?

[low shard]

vac lite

[eternal aurora]

yh i just done that]

[low shard]

@viscid moss

[low shard]

be sure to check the guide installation too below the blue installation link

[eternal aurora]

yh will do

is this what u want?

[low shard]

set gpu to the kaggle gpu you set before in session options

set f0 to rmvpe without onnx

set extra to 2.7

set input to your microphone, and output to line 1

[eternal aurora]

theres 2 options

[low shard]

select the first

[eternal aurora]

okay

looking good?

ill set my mic in a bit

and for the characters shall i just import them?

[gusty cloak]

1. RTX 3060TI
2. First tutorial - https://youtu.be/Hx2IHzt5tAc , Second - https://docs.applio.org/applio/getting-started/training
3. Dataset Size - 5 minutes.
   Settings:
   RVC V2, Sampling Rate 32K, HIFI-GAN, rmvpe, contentvec, Batch size 8, Total Epochs 500, Save only latest, Save every weight, Pretrained, Custom pretrained (G_SnowieV3.1_32k.pth and D_SnowieV3.1_32k.pth), Index alghoritm - Auto

[low shard]

you can just click edit and upload the models

[low shard]

you need to set the input to microphone and output to line 1

[eternal aurora]

yeahyh ill do that later as i havent got my mic rn

[eternal aurora]

oh yh i forgot

[low shard]

the first tutorial seems outdated

the settings look fine, are you sure it's doing 1 seconds per epoch?

could you send a screenshot?

[gusty cloak]

I can't send screenshots

Mel Spectrogram Similarity: 99.84%
Time per epoch: 0:00:01

[low shard]

!give-media-perms 1h @gusty cloak

[low shard]

okay wtf 😭

[gusty cloak]

idk 💀

[low shard]

something surely went wrong in the training

[gusty cloak]

I'm using Codename-RVC-Fork v3.1.0, but i have same results in Applion

[viscid moss]

any fix needed?

[low shard]

it's the first time I see such thing, have you also tried retraining it? maybe changing the pretrain?

@simple ore have you ever seen this before?

[low shard]

nope, check #💬│engineer-chat

[gusty cloak]

I had time per epoch 10 seconds before i started using pretrained

[simple ore]

@gusty cloak training a model on mute files because they forgot to slice stuff

[eternal aurora]

kinda random

but would u know the name of this song?

[gusty cloak]

Should I change Pretrained or what?

[analog obsidian]

do the preprocessing again

[low shard]

I haven't heard the name of this song, though it's written to be named "Phantom"?

[eternal aurora]

yeah, i tried looking for it but i cant find it oh well

[last robin]

does anyone know why my thing doesnt work when i try use my gpu for okada

[gusty cloak]

Checked all checkmarks in preprocessing and it works fine thx

[low shard]

elaborate:

• what's your pc gpu
• what you want to do
• what tutorial link did u use
• a screenshot of ur wokada

!give-media-perms 1h @last robin

[midnight sable]

Hey I used some google colab back when I used to make covers and I found it here but now it does not seem to work so like... How do I make covers lmao

[last robin]

i have a RTX 3060 im trying to use a voice changer i dont know what you mean by tutorial link but i have been using it for the last couple days and has been work fine but noe it says invaild or unsupported data type ComplexFloat on the command thing

the wokada stays the same but just says failed to fetch when i try switch

voices

it works with everything else but my gpu even tho it had been working on it for the last couple days

[eternal aurora]

how do i select a voice model?

@low shard

[low shard]

click edit, select a slot, upload model (.pth) and index (.index)

then just click the slot

and start

[eternal aurora]

like this?

[low shard]

what's your pc gpu first of all

[low shard]

yes, i suggest you to rename the model file to smt else though, not just model

[eternal aurora]

okay do i just press edit?>

[midnight sable]

Geforce RTX 2070

[low shard]

can you share a screenshot of your wokada?

[low shard]

no don't edit directly by the ui, it's broken, you need to edit the model files name, and if you already uploaded the slot, you can just re-reupload it

[eternal aurora]

ah okay

[last robin]

is that fine

[eternal aurora]

this?

[low shard]

lmfao you didn't even need to use google colab

google colab is a cloud method, meaning it lets you use a remote good pc, and it's meant for people with a bad pc, it has limited gpu time in free tier

As you got a good PC, you can use RVC locally, you can choose between:

• Applio: A fork of RVC with some extra features like Applio TTS, kinda faster and simpler but same quality tho
• Mainline: The original RVC

locally means you will able to utilize your own gpu and have no limits

or do you still wish to use cloud?

[low shard]

yes

[eternal aurora]

okay

[low shard]

lemme guess, you used a youtube video tutorial?

[last robin]

yes i did is that wrong

[low shard]

also weird it's not showing all settings for some reason

[low shard]

yup, all video tutorials are outdated

you're using original wokada which is more bugged, and also you are using vb audio cable which gives random issues on windows

[low shard]

so yeah, using this it's normal that it will stop working randomly lol

you can uninstall everything you got off youtube

[eternal aurora]

i uploaded it

[last robin]

where would i find the updated version

[low shard]

there's 2 wokada main versions:

• original made by wok
• deiteris fork made by deiteris

each has it's own unrelated updates

check https://rentry.co/forkvoicechangerguide for the latest wokada deiteris fork

be sure to uninstall the original wokada + vb audio cable

[low shard]

alright, are you having any issues?

[last robin]

to uninstall t is there a certain thing or do i just delete the files

[eternal aurora]

yeahh i cant seem to choose it

[low shard]

show a screenshot of what u see and what browser do u use

[eternal aurora]

opera gx

the place where the models are for some reason is collapsed

[low shard]

wait it shows you didn't upload anything, did you click upload ?

[eternal aurora]

yeah i did

[midnight sable]

Oh LMAFO ty bro I just watched a yt vid like a year back that said to use the colab so I did

[low shard]

so did you do everything as said as in https://rentry.co/forkvoicechangerguide#uploading-models ?

[low shard]

youtube video tutorials are outdated asf and don't give all info for rvc/wokada lol, yw and lmk

[eternal aurora]

yep i did

if it helps

the person that uploaded it says that

you cant use a gpt sovit can you on here

[low shard]

😭

it says it explicitly you can't use it on the voice changer

[eternal aurora]

yeah i just saw

im so blind

[low shard]

RVC and GPT-So-VITS are 2 completely different AIs

RVC is meant for Speech to Speech

GPT-So-VITS is meant for Text To Speech

not all voice models are RVC

[eternal aurora]

ohhhh

riht

right

i see

makes sense

with index

[low shard]

yup

[eternal aurora]

does it

change accent?

[low shard]

in rvc context:

• pth = the voice
• index = the accent (be sure it's the added_index and not trained one, anyways most models should only include just the pt and added index)

[eternal aurora]

i see makes sense now

[gusty cloak]

I just started making models and i'd like to know if everything going fine

[somber kraken]

hi guys

how do i set my output for the voice changer as mic input?

[last robin]

hi again im very confused where i actully down load the new wokada

[low shard]

elaborate:

• what's your pc gpu
• what tutorial link did u use
• what you want to do
• a screenshot of ur wokada

!give-media-perms 1h @somber kraken

[low shard]

did you read the guide? it's all inside the guide, get the windows nvidia version

[last robin]

ohhhh sorry im very dumb i had misread the guide

[low shard]

it's fine lol, lmk

[somber kraken]

so pretty much

im running a 4060

i was searching for a voice changer

to troll my friends

[low shard]

good enough

[somber kraken]

i remembers watching this video

when i was little

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

[low shard]

I hope not a youtube video tutorial

nvm

[somber kraken]

and that it had

a voice changer

so i searched it back u[

it works but

i have allot of output media

[trim sparrow]

does anyone know how to get more gpu usage from applio

[somber kraken]

and like a lot

i can output it to my headphones

works perfectly

[low shard]

this shit is an old version of original wokada

it's over a year old lmfao

[somber kraken]

o

well it works

[low shard]

simply you wasted time watching that youtube video

[eternal aurora]

how do i hear myself

[somber kraken]

i just dont know which one to choose

[low shard]

no, it's bugged and has worse performance and quality

[somber kraken]

oh

[low shard]

forget everything of that youtube video

all video tutorials are outdated

only written guides are up to date

[somber kraken]

hmm alr

[eternal aurora]

@low shard how do i hearmyself?

[somber kraken]

but which one would i theoretically have to pick to stream it to my mic

[low shard]

uninstall everything

wokada has 2 main versions:

• original made by wok
• deiteris fork made by deiteris

each has it's own updates

latest wokada deiteris fork is currently the best to use, read https://rentry.co/forkvoicechangerguide

it's just better you forget that guide at all

[somber kraken]

ik

but for now

this works for me

[low shard]

it's bugged, just don't use it bro

[somber kraken]

i just need to know which output device is

[low shard]

even the vb audio cable you used

it gives random issues on windows

it's not worth to use those things at all

[somber kraken]

i dont have an audio cable

[low shard]

you're just going to waste time trying to fix something that is not meant to work

[low shard]

A VAC (Virtual Audio Cable) makes a fake audio device, used to re-route the audio of different programs

In Wokada context, it's used to get the output of wokada as the input in other programs

[somber kraken]

i have a cheap usb mic

[low shard]

please

just uninstall the version you got

and read the wokada deiteris fork guide

[somber kraken]

i will

[low shard]

it has everything you need in it, the updated version is much better, and it got the vac lite

[somber kraken]

but its 10:30 pm rn

[low shard]

without a vac, wokada can't work at all, in all versions

so yeah you just kinda wasted time lol

[somber kraken]

oh nvm

i

i found one

ty for helping me though

ill update it tomorro

[low shard]

set the output as your headphones in wokada

[eternal aurora]

okay

[low shard]

i mean the vac is at the 3rd step of the wokada deiteris fork guide, and that should work also with the version you got rn

but anyways let me know

[low shard]

are there any other issues?

[eternal aurora]

no but forsome reason i cant hear myself

[low shard]

can you share a screenshot of ur wokada?

[somber kraken]

thanks man it works

[low shard]

set output to line 1

and monitor to your headphones

-# sorry for my confusion lol i was helping 2 people at the same time

after doing that, you should be just able to select any model, start, and hear yourself

[trim sparrow]

does anyone know anything similar to kaggle i can use to run applio for training

[low shard]

wdym?

[eternal aurora]

ohh thats fine

[trim sparrow]

i ran out of gpu usage and i cant find like a paid option or anything to get more

[low shard]

also be sure that the chunk has a bit higher value than the perf value at the top left

the rest seems fine

-# let me know ofc

[low shard]

ran out of gpu usage?

could you show a screenshot and explain more?

did you seriously use already the 30 hours weekly of kaggle?

[trim sparrow]

yeah😭

[eternal aurora]

okay done that its still not working

[low shard]

how many models are u cooking 😭

30 hours is already alot for free tier

[trim sparrow]

legit just 1 but all the errors ive had i had to restart the model from scratch completely like 6 times

[low shard]

kaggle doesn't have a paid tier to increase gpu usage

[trim sparrow]

n this time it was going fine

at 525 epoch rn

n then 30 hour limit hit😭

[trim sparrow]

what if i just export the log folder and make a new account

[low shard]

well, google colab gives max 4 hours of gpu daily (which can be random), but there's a paid option there

[low shard]

you need to verify with new acc, so new email and new phone number

if it's the same email or phone number, it won't work

[trim sparrow]

yeah thats fine ill probably js use one of my familys numbers

[low shard]

what's not working exactly?

[trim sparrow]

so if i export the log folder for the model and put it in a new one can i just resume the training like normal or is it more complicated

[somber kraken]

nvm it doesnt work

i still dont know which one to select

[low shard]

which to select? the output? you should uninstall vb audio, and then select line 1 as the output in wokada

and then select the input in the game/discord vc as line 1

[eternal aurora]

i believe my mic is working i just cant hearmyself

should i try a new broswer/?

[somber kraken]

huh

[low shard]

be sure the site has microphone permissions, also try chrome or firefox since operagx usually has issues with wokada has reported to other users

[low shard]

can you elaborate your help request?

[somber kraken]

im not supposed to

wait

lemme try and formulate

[eternal aurora]

yh will do

[low shard]

well without info it's hard to help

[low shard]

alr

[low shard]

🔥

[eternal aurora]

i just refreshed and it works

[somber kraken]

which vb audio should i select, there are multiple options. i dont know which one to picka

or should i not use vb audio

[somber kraken]

cuz i'd rather stay on the old version for now

[eternal aurora]

do i change it

[low shard]

You shouldn't, yep

[somber kraken]

ye but for now

[low shard]

Vb audio cable is bugged on its own for windows

The vac lite from the 3rd step of the wokada deiteris fork guide will also work on the old version

So you'd just need to follow that single step of the new guide to make it work on the old version

https://rentry.co/forkvoicechangerguide#virtual-audio-cable specifically only this

If you want to update it, you will just need to run the other steps, but for now, you can just get the newer vac lite to make the old version work

[low shard]

I could just suggest you the force fp32 mode, this will use a more stable and higher quality inference (use models) mode, but will higher up a bit the delay

[eternal aurora]

i want to use this on ig would u know how to do thar?

[eternal aurora]

oh no it works now

[low shard]

IG? Instagram?

[eternal aurora]

yeahh

[low shard]

It's an advanced option to get higher quality at the cost of some delay

It's optional lol

[eternal aurora]

ill try it

[eternal aurora]

i mean my delay is not that bad

[low shard]

I don't use it, but You'd just need to find the voice settings, set input as line 1, and then set output as headphones

[eternal aurora]

okay ill see

[low shard]

Yeah bc it runs on cloud lol, the only thing that could kinda nerf the delay is your internet speed in this case

[eternal aurora]

yhh thats true

so if i wanna use it on dc or any other app

how would i do that?

[low shard]

you'd need to do the same process I said in #✨│ai-help message

it works for every app

[eternal aurora]

where is that?

[low shard]

you'd need to check the voice settings inside the settings of the app

in discord just click the gear, go to voice & audio tab, and here you can see the input and output

[eternal aurora]

ohh okay

[low shard]

that's for wokada

for other apps (like use wokada in disocrd):
input: line 1
output: headphones

[eternal aurora]

ohh okay

i found it on ig

[thorn flax]

how do you delete a model that youve uploaded

[low shard]

elaborate:

• your pc gpu
• what program are you talking about (example: Applio, Original Wokada, Wokada Deiteris Fork, etc)

[low shard]

alright

[thorn flax]

idk what any of that is

[low shard]

what tutorial link did u use?

[eternal aurora]

YH i think i fixed it ngl

but thanks a lot ngl

[low shard]

I hope you didn't use a random youtube video tutorial lol

[low shard]

so, everything is fine now right?

[thorn flax]

i just kinda did it all by hand i didnt need a tutorial

[low shard]

where did you download the program?

[thorn flax]

i have 3 models i was testing but i dont really need the bottom 2 anymore

[eternal aurora]

yhh should be however my voice doesnt sound liek the sample one

[thorn flax]

idk hwo to remove them off my list though

[low shard]

either reupload the slot, or remove the model files in the model_dir folder inside the program

also, be sure you're using the b2332 version

[thorn flax]

wheres the folder

[eternal aurora]

the accent of it is not there

[thorn flax]

and whats b2332

[low shard]

can you share a screenshot of your program folder?

[thorn flax]

where is it

[thorn flax]

this program is in a web browser idk where its folder would be

[low shard]

you can higher up the index value, which will make your accent similar to the trained accent, but could sound like using autotune

[low shard]

could you send a screenshot of whatever you're using to opening the program ?

[eternal aurora]

thats true yeah ill do it a bit

[low shard]

there's thousands of AI programs

[thorn flax]

this

i was told this is the most optimsed one

[low shard]

that looks like a shortcut,
right click it > properties > check the absolute path

you will get the path of where the program is

[thorn flax]

ah ok

this

[low shard]

I feel like you're using wokada deiteris fork latest version (so b3223)

yep

[thorn flax]

is it the latest version?

[low shard]

check inside model_dir, here you will find the model files

[low shard]

yup

[thorn flax]

last i updated it was october

oh

[low shard]

you can also check the version by opening the program itself, you should see b2332 at the top left

[thorn flax]

ok

[low shard]

it's wokada deiteris fork, not sure if it's the latest though

[thorn flax]

it says B2332

[low shard]

yup latest

[thorn flax]

guess its abandonware

[low shard]

here you can delete the model files of that slot

[thorn flax]

ok

thx

[eternal aurora]

sorry i asked again

but lets say i use it on dc

how do i stop hearing myself

[low shard]

welp nope, the b2332 version is actually from 2024 december

[low shard]

set monitor to none

[eternal aurora]

okay ty

[low shard]

-# which is why it's weird you said you got it from october

[low shard]

any other issues ?

[eternal aurora]

not that i know of

it sounds a bit weird in voice messages

[low shard]

there's thousands of AI models, maybe try using other models

[eternal aurora]

yhh i might do that

do yk any girl ones that are good

[low shard]

also it doesn't automatically add emotions, it depends on how you talk too

[low shard]

https://rentry.co/forkvoicechangerguide#voice-models-to-try-out

[eternal aurora]

ty

[low shard]

or

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)
• 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:

• #1159289738314919936
• #1191429836321849435
• make it yourself with our docs guides https://docs.aihub.gg/essentials/how-to-make-voice-models/

[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

[eternal aurora]

okay will do

[somber kraken]

nick

i want to update now :D

for sum reason

when i start a discordcall the model stops properly outputting

[low shard]

yeah, I would have guessed that since the version you used works on hopes and dreams

please, start by reading https://rentry.co/forkvoicechangerguide and let me know for any issues/errors/questions

[eternal aurora]

If I wanna use it again what do I do?

[low shard]

you will just need to run the last cell of the kaggle since you already installed it on kaggle and the files remain

[eternal aurora]

okkk

[trim sparrow]

@low shard i just throw the folder inside of the logs and continue training thru applio?

or do i need anything else

[queen horizon]

I got a 4060ti which should i download

I cant find latest w 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

[brittle wing]

i desperately need help on how to even get the voice changer

[fervent trench]

it was like a year ago when I installed okada, I got a windows new pc but can't find the download link

[low shard]

what's your pc gpu

[low shard]

what's your pc gpu

[fervent trench]

GTX 1660 TI

[low shard]

good enough

[fervent trench]

my CPU is a Ryzen 9900x, maybe its more powerful

[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

[fervent trench]

why are people using a fork?

[low shard]

video tutorials are all outdated btw

[fervent trench]

what happened to the original

[low shard]

fork means modified version in IT term

[fervent trench]

I am a programmer, I know

the fork seems dated

december last year

[low shard]

the wokada deiteris fork has improved performance, and has advanced settings that can help with better quality

[fervent trench]

well alright, I'll see if it works