#Help me plz my W okada is not working
1 messages · Page 1 of 1 (latest)
What is your PC GPU? And did you follow any tutorial or guide before this? The word "translate" likely means to convert your voice as a specified voice model in realtime rather than translating to another language.
Yes I did follow tut
Didn’t work
!howtoask
- Check Docs & Guides: Your answer may already be in the AI Hub Docs or the https://discord.com/channels/1159260121998827560/1159513888199540817 channel.
- Search the https://discord.com/channels/1159260121998827560/1192011222023950368 : Look for existing posts that solve your issue. Do not invade someone else's post.
Tell your:
- Full GPU Name: (e.g.,
NVIDIA RTX 4060 8gb vram desktop) - Operating System: (e.g.,
Windows 11) - Detailed Description: What were you trying to do and what went wrong?
- Tutorial Used: Link to the guide you were following.
- Screenshot: A picture of the full error message.
To maintain a legal, safe & ethical community, we will NOT provide help for:
- ANY illegal activities.
- NSFW/Porn.
Requests for these topics may be ignored, not helped and result in moderation action.
- Be Polite & Patient: Our helpers are volunteers. You may ping the
Helpersrole once. - English Only: Please keep all conversations in English.
- Don't Ask To Ask.
What does this mean?
Running W-Okada with only CPU would perform much slower than a dedicated NVIDIA or AMD Radeon GPU.
My pc kinda old lemme send a pic of my settings
There's alternative to the one you tried to run, and the one would be an online option one.

Wut?
It’s my settings
Nah, you wouldn't try to run any W-Okada version (even the one in your screenshot, not just Deiteris and Tg Develop forks) with only CPU.
Why?
Why tho? It's too much to answer. It's because CPU will have to process everything on W-Okada, so it will likely result in slower performance than a dedicated GPU.
Ye? But I don’t get my voice the ai voice when I speak in it that’s the proplem
I won't be helping you if you try to run the program like that. Rather, I can give you another way to run W-Okada with much better performance.
Tell me plz
Any dispute on why running with only CPU would give better performance than the one with discrete GPU?
There's other way round. Follow this guide. https://docs.aihub.gg/realtime-voice-changer/cloud/tg-develops-w-okada-fork-kaggle/
Last update: September 6, 2025
Before you ask anything, I would like to know that this will run online and won't use your PC resources like the "locally" one. 
And ok
Looks like you have never heard of Virtual Audio Cable before. Download the zip from this link. https://software.muzychenko.net/freeware/vac470lite.zip
Virtual Audio Cable and VB-Cable are two different programs made by different authors, don't always confuse against these. Double click on setup64.exe to install the program.
I’m on windows 10
So?
I didn't tell you to do this.
Well this YouTuber did
I told you to go for VAC, not VB-Cable, again.
A YouTuber told me to do it
Please.
Please listen to me, goddamnit.
I am
Tutorial videos from YouTube only telling you to download the outdated version of W-Okada, despite their upload dates.
I've already told you earlier.
I can focus on settings part, but anything you attempt to do outside the proper solution won't always be guaranteed.
Oki
You don't need to download the program to your PC. https://docs.aihub.gg/realtime-voice-changer/cloud/tg-develops-w-okada-fork-kaggle/
Last update: September 6, 2025
ohh
No, that's not what I mean.
That one literally said "cloud" which mean the program will run online. The "local" one where you run the program directly in your PC is another thing.
ohh
so you said dont run it with the thing i run it last so how else do u want me to run it?
There's more you should know, just don't do anything beyond the method.
k soo
should i call u and screen share or???
Firstly, create a Kaggle account.
After you create your Kaggle account, go to settings, and verify your "phone number" there.
k
It looks like I am telling you step-by-step on how to use W-Okada on Kaggle where the guide for it literally exists.
ye lol i did verify my phone now
look so what do i do now`?
nvm forgot i could read lemme go on kaggle for the guide

Import notebook.
The method isn't end there. Register an ngrok account.
wut?
More like the "notebook" link is loading.
k what do i do when its loaded ima say when its loaded
After you register for an ngrok account, copy your token key from your ngrok account, go in Kaggle, scroll down and paste your token key in highlighed text.
Do I need to confirm you every step? This is where you paste your token key.
no u need to tell me what to do now im lost

To start the Kaggle notebook session, go to run menu, click on Start session.
i clicked run current cell
wait nvm im stupid
alr i clicked start session
what now
This one is where you actually get W-Okada to work. Wait for a url link below, you copy it and paste it in your browser.
This is getting more and more crazy, because not only you missed some steps, but also implying that you didn't do the "ngrok" part after all.
I DID DO IT
my freaking god

Once you get a url link for your ngrok/W-Okada from Kaggle, copy and paste it in your browser, you send your screenshot to actually confirm that you successfully got the website program to work.
bruh it still havent loaded
Scroll down.
Yes.
Click "visit site".
what is this
This is Tg Develop W-Okada, and it's completely normal for this specific W-Okada to look like this.
You sure you wanna click start server without any voice model loaded?
So what is the panel on the left screen supposed to be? Click on "plus" symbol.
this ?

I CANT TELL RIGHT OR LEFT
dont blame me
@spice ginkgo can u plz tell me HOW THE HECK TO ADD MODELS
lemme have a break for a bit
If I don't launch W-Okada from my own PC to do this, you would be like desperate to ask me to do as a visual example but you still didn't know what to do and where to go, expecting me as your main guide who could absolute guide your predetermined way.
That, you wouldn't too closely follow someone's steps, but I'd suggest you should at least be initiative, think about what you wanna do and figure it out by yourself, rather than relying someone's mind at all time. A good thread won't be over 100 messages if not necessary. And if you ever feel tired doing this, you can try this later when you feel better. 

Additionally, there's another guide on how to use Tg Develop W-Okada fork, many steps are pretty much the same as online one, and the difference is where the program run. Although I didn't send this one earlier because you would still find for a download link and try to run the program locally even if I told you it was not very ideal to run with only CPU. https://docs.aihub.gg/realtime-voice-changer/local/tg-develops-w-okada-fork/
Last update: October 26, 2025
IM BACK FROM MY BREAK!!!
@spice ginkgo lady it dosent work i can hear myself in ai voice
i need realtime life not whatevr this is...
can u plz help me
Read my statements again.
wdym if i didnt launch my w-okada should i launch the thing that dosent work agian?
like lady wdym?
i just wanna roblox vc troll T_T
Scroll down, and you'll see these settings.
Here's your so much expectation settings:
Chunk: around 56 - 64 ms
Extra: 2.7 s
GPU: NVIDIA T4 (there are two T4 GPUs and the only one can work at the time)
Pitch extraction algorithm (F0): rmvpe (not rnvpe_onnx)
i still dont have NVIDIA T4
Input: your microphone
Output: Line 1 (Virtual Audio Cable)
Monitor: set this to hear the program, optional.
What the hell? You didn't do this and this steps? What were you cooking?
i did i even showed u a pic
You did not. You only installed VB-Cable, you did not install Virtual Audio Cable lite.
I did install Virtual Audio Cable lite, and VAC doesn't have input/output named "VB-Audio" or "Voice.ai".
ehh dont ask me ab the voice.ai ima uninstall it soo it aint working
k lemme do it rq
voice.ai is uninstalled now
@spice ginkgo can u send me the download link for the line 1 ?
I've already sent you the link to VAC lite earlier, and you either didn't care or being so confused with VB-Cable. https://software.muzychenko.net/freeware/vac470lite.zip
being so confused with VB-Cable. that was what i was
tf is happening?
im so confused
Try another web browser, another network connection, or else you're cooked. And that it works for me without any problem.
k aww i gotta try chrome even tho i hate chrome arhh only this time chrome
BRUH it aint wroking lemme try smth else
do you have a diffriend link?
@spice ginkgo
It looks like the access to the website is denied in your network, so it's quite hard to fix other than relying on VPN, but I don't think you should do this.
This is the zip file. Download this, quick.
ty
but i did have a vpn but this is faster
its installed i hope @spice ginkgo
yep it is
@spice ginkgo are u here?
Click "start server" to test W-Okada.
k
ehmm bad new it aint working....
this is right right?
@spice ginkgo
are u here?
less than a minute ago
Can you just send the full screenshot like mine?
then explain how it "ain't work"
i already did who are u agian?
im a stranger
COOL
Reduce your chunk number down to around 56 - 60 ms because your screenshot still has chunk 512 ms, change Processing unit to NVIDIA T4.
it's still in cpu mode
Easy, but you were making it harder to explain. 
for the 5th or 4th time i dont have T4 OR NVIDIA
T4 is the kaggle gpu, not yours locally
sorry XD
If you don't do this on Kaggle, then you wouldn't have T4 in W-Okada.
PROOF
No excuse.
wait im actually just stupid
this better?
check in the kaggle page, make sure T4 is selected as the "accelerator"
And that's like the second or third time I've been reminding you. 
if you've just changed it to such, restart the session
i told ya im stupid today
I have better memory. 
i dont
lets say im under 20 and i have bad memore and im over 14 so yall can guess if u want
sounds more like losing focus, go get hydrated
k if u say so
i got my tea
wich one`?
try the gpu 0
Watch this. I now have two W-Okadas running. This one is coming from my own Kaggle, while the other is coming from my local PC.
so 0: tesla
alr
should it work now`?
Try hear for yourself.
Till you found out.
I won't be repeating the same step, but because you have a bad memory, so that this could give you some ideas.
i have already done that
Enabling "Passthrough" would make the program to output your raw microphone audio rather than the converted one.
k
ehh hello?
You better not forcing anyone to provide an answer to you.
If I don't answer you, it can either mean I'm waiting for a better response about W-Okada rather than simple "um hello is anyone here" or I'm away from screen.
k sorry
im quitting this it shouldnt be this hard
