✨│ai-help

[knotty moth]

both files on yours seem at the Downloads, so let them both there, and then do extract

[nocturne marsh]

I did that

and my files page doesn't look like that for sure

[hallow thistle]

[nocturne marsh]

I only extracted 001 tho

and they weren't in the same folder

[hallow thistle]

Why you still not changing your sample rate to 48000 Hz?

[acoustic bluff]

i set to it 48000 hz

when I click start server

it goes back to 41000 hz

[knotty moth]

tg-develop fork may have a bug to show back to 44.1k for the server mode

[nocturne marsh]

oh wait I'm in

it took a few minutes to load

[knotty moth]

no idea but just set it again to 48k

[acoustic bluff]

i did

it doesnt work

tried everything

same error

[candid crater]

what does it mean select audio inut

input

[nocturne marsh]

is adding an index important or only thr path file?

[hallow thistle]

When more than 2 people asking for voice changer in here at the same time, it makes things messy for me to provide an answer.

[candid crater]

Sorry.

?

[hallow thistle]

Close or Ctrl + C the voice changer's terminal, go to your MMVCServerSIO folder, open stored_setting.json file with Notepad, edit every sample rate to "48000", click save, and relaunch the program again.

[hallow thistle]

When you said "last question", this one is not a last question.

[acoustic bluff]

still getting the same error

sample rate is set and staying on 48000hz now tho

does the same thing on both

[hallow thistle]

Try another web browser.

[acoustic bluff]

if i do passthrough on it works for both 44100hz and 48000hz

[candid crater]

My turn?

[acoustic bluff]

tried edge and vivaldi browsers

I'll get chrome and try on chrome now but I don't think it's a browser issue

am I missing files?

    result, vol = self.process_audio(audio_in)
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "voice_changer\VoiceChangerV2.py", line 133, in process_audio
    audio, vol = self.vcmodel.inference(audio_in)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "voice_changer\RVC\RVCr2.py", line 226, in inference
    raise PipelineNotInitializedException()
Exceptions.PipelineNotInitializedException: 'Pipeline is not initialized.'```

[nocturne marsh]

well I'm sorry I'm tryna figure this out

[acoustic bluff]

I got no voice_changer folder

[nocturne marsh]

dude I did let Toxic go and then asked when you told me to

I'll ask someone else

while you're busy

the voice sounds pretty laggy, how can I fix it to become more stable?

[candid crater]

fish

[nocturne marsh]

then imma wait

[candid crater]

What do i do here

[nocturne marsh]

okay so you select a microphone

[acoustic bluff]

can you send me a screenshot of your folder?

[nocturne marsh]

look at the input device tab

[acoustic bluff]

like this

[nocturne marsh]

on the right side

[nocturne marsh]

you are not missing files

I have the same

[candid crater]

there

[nocturne marsh]

I'm assuming your default laptop one

then you should be good

[hallow thistle]

I observed for myself, the program worked for me even my laptop is old and has no GPU, of course the perf number is over 30000 ms which is crazy. I don't see any "pipeline not initialized" error so far.

[nocturne marsh]

@acoustic bluff

go to _internal file

[nocturne marsh]

and scroll down to where you see a list of folders

thats where it should be

[nocturne marsh]

yes correct

[acoustic bluff]

weird

[nocturne marsh]

is this for me mate?

[nocturne marsh]

oh it's because you're doing client

I had the same thing so I just did server

[candid crater]

server is not avaliable..

[nocturne marsh]

oh shoot

yeah I had the same issue with client

so we would be on the same page

[hallow thistle]

The "pipeline not initialized" error sometimes has to happen with a non-RVC voice model (like a Beatrice voice model) being loaded to W-Okada voice changer fork (which only supports RVC voice model), your GPU's driver is outdated or something else.

[candid crater]

😭

dam

[nocturne marsh]

also

[acoustic bluff]

send me a voice model that should work I'll try it

[nocturne marsh]

when you get an answer, tell me

[nocturne marsh]

lowkey missing out on my favourite show setting this up 😭

[hallow thistle]

The question is: why the program works for me the first try while y'all keep struggling?

[hallow thistle]

#1175430844685484042

[candid crater]

it feels like im doing damage on my windows

[acoustic bluff]

I got models from that channel

already

[nocturne marsh]

you're using neither client or server?

[nocturne marsh]

nah it won't mate, don't sweat

[candid crater]

alr

[hallow thistle]

With "server" audio mode:

[candid crater]

idk

[nocturne marsh]

oh yeah I see why it looks unchecked now

it happens when you start server

I was using RMVPE ONX

do I not?

[candid crater]

it dosent work i switch still

[nocturne marsh]

ask chatgpt maybe

if I knew I would help man

you sure you can't use server

[candid crater]

Yeah

[hallow thistle]

Here's your settings:
Chunk: around 130 ms
Extra: 2.7 s
Pitch extraction: rmvpe (not rmvpe_onnx)

[nocturne marsh]

do you know why?

[candid crater]

No

[hallow thistle]

Refresh your tab.

[nocturne marsh]

done

[candid crater]

still wont budge

[nocturne marsh]

the hell why isn't the settings saving

nevermind it was just the sample rate

keeps going back to the default after I change it to 48000

weirdo

[hallow thistle]

Try switch input device from "microphone" to "Line 1" and then "microphone".

[nocturne marsh]

sounds laggy a little and robotic

[grim ravine]

hello. i did everything correctly i think but i hear my changed voice how do i close it

[hallow thistle]

[candid crater]

Dosen't work..

[hallow thistle]

On W-Okada, set "monitor" device to none.

[hallow thistle]

Check if your browser gives microphone permission to your current tab (voice changer).

[nocturne marsh]

ah I changed the pitch, a little better now

[acoustic bluff]

do you think getting spin helps

idk rip i'll just use main outdated one

this ain't working

[nocturne marsh]

sounds so robotic

HUH

bro I refreshed my page

and I can't use server anymore

the hell

oh

tf

it's normal now

[candid crater]

/:

[nocturne marsh]

why can't I enable noise suppresiuon?

[candid crater]

idk what to do anymore bro

[twin rain]

i cannot find my gpu on processing unit tap

how i can fix it ?

[nocturne marsh]

my voice sounds so robotic, a little choppy and so weird, any way to fix it to make it more realistic?

[grim ravine]

its not letting me use gpu i can only use cpu why?

[celest violet]

after that? just unzip normal??

.

or before unzip i must do any? @hallow thistle

[hallow thistle]

Use WinRAR or 7-Zip to open or extract the .zip.001 one.

[hallow thistle]

What is your PC GPU? And are you following Tg Develop's W-Okada fork?

[celest violet]

im using winrar then use click Extract to voice changer windows amd64 cuda, then 1 folder appear

in that folder, appear MMVCServerSIO + 2 file Force GPU Clocks

after that?

[hallow thistle]

Go inside MMVCServerSIO folder, there's a program of the same name, double click the program to launch the voice changer.

[celest violet]

with blue mic icon?

ok

rmvpe auto download?

[knotty moth]

show the spec in task manager, if it is integrated intel or AMD radeon graphics, it falls back to cpu mode

[grim ravine]

it is am d radeon

awmd

amd

[celest violet]

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'set_sampling_rate'

why this appear on black window

[merry forge]

everytime i upload a voice

it dont sound right

it sounds nothing like the audio samples people post

so idk whats going on

[viscid scarab]

Help me , I have voices in my head

[royal vapor]

I'm trying to get the best quality and least delay using RVC Client. I've seen the guide showing recommended settings for xx50/xx60/xx60TI/xx70 cards etc.. Does it also matter which series these are? 2060, 3060, 4060? Could I get a 2060 card and get the same result as a 3060 or 4060? The card is not being used for anything else

[serene swallow]

Im using 1.7 and it doesnt have the settings like in ur pic

Could it be that the author of Vonovox removed features instead of adding in the latest update?

[proud wagon]

any help to humanize ai text

[lunar trout]

does anyone know how to fix the thing where when your using your mic
all of your audio sounds like your under water

[brave axle]

it seems to work but it doesnt detect my mic at all even though i put my correct microphone, and every single app hears me besides okada

[knotty moth]

"humanizers" are just plain BS

[proud wagon]

i kinda know but what do i do

[knotty moth]

what purpose is it exactly? to outsmart the AI detectors or even real humans' common sense being savvy enough?

[proud wagon]

to outsmart the detectors

[knotty moth]

so it could help you do the school/uni projects or pass the job interview despite your real skill incompetence?

sorry but we couldn't support for such abusive purposes

[proud wagon]

its just literature so it doesnt require any skill, its not that big lol

ur talking like im a med student

[knotty moth]

in your head, zombie

[daring hatch]

Real professional from you

Repxic: can someone help
seiso💖: you're incompetent

[knotty moth]

I saw you seem like a friend of that bob guy, could you tell me what's wrong with me so he repeatedly insulted me as a clown?

[inland field]

ModuleNotFoundError: No module named 'pyngrok'
""COLAB""

[nocturne mural]

Please provide more details. That error line doesn’t specify anything and it won’t help to deduce what your solution might be. For example, you could tell us which Colab you are using, what you did that led to that error in that Colab, and for what purpose you went to that Colab.

https://media.discordapp.net/stickers/1442670901828124802.png?size=160&name=Sandia

[inland field]

Colab link = https://github.com/w-okada/voice-changer/blob/master/Realtime_Voice_Changer_on_Colab.ipynb
Error happened when I ran it
I went to that Colab because My gpu is amd "weak one" r7 200 series (2gb)

[knotty moth]

the way you make that invalid conclusion makes you look nothing but a random immature person I've ever seen across average discord servers

[nocturne mural]

Have you executed the cells in order? Because the error you describe means that the cell may not have been executed or may have failed during installation.

[inland field]

Yes
full error
/content/voice-changer/server

ModuleNotFoundError Traceback (most recent call last)
/tmp/ipython-input-341872215.py in <cell line: 0>()
22 get_ipython().run_line_magic('cd', '/content/voice-changer/server')
23
---> 24 from pyngrok import conf, ngrok
25 MyConfig = conf.PyngrokConfig()
26 MyConfig.auth_token = Token

ModuleNotFoundError: No module named 'pyngrok'

---

NOTE: If your import is failing due to a missing package, you can
manually install dependencies using either !pip or !apt.

To view examples of installing some common dependencies, click the
"Open Examples" button below.

I think solution may be by adding !pip install pyngrok
to notebook

[nocturne mural]

umm

[nocturne mural]

I think you should switch to another notebook. I don’t usually check #✨│ai-help much, but I’ve seen that they no longer use the original w-okada, since it’s no longer being updated. Instead, they use its forks, which you can see with the -realtime command

-realtime

[patent trellis]

💻 Local Realtime RVC

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

• Vonovox

A Realtime Voice Changer with similar performance to Wokada Tg-Develop Fork, with extra features, but it supports only Nvidia GPUs on Windows 10/11 unlike other options that have wider support. and without cloud options GUIDE

• Wokada Tg-Develop Fork

A personal fork (modified version) of Wokada Deiteris Fork, it just adds some Quality of Life improvements to it like supporting Spin Embedder and Audio Effects. Don't expect too much about it since the creator made it originally as a personal project. GUIDE

• Wokada Deiteris Fork

Deiteris' fork (modified version) of wokada that doesn't get updates anymore. GUIDE

⚔️ Wokada Tg-Develop Fork vs Vonovox

For Windows Nvidia, Both Wokada Tg-Develop fork and Vonovox have similar performance & quality. Users should read the pros and cons for both and choose based on their differences, such as UI and Vonovox's paid effects.
Read Wokada Tg-Develop Fork Pros&Cons & Vonovox Pros&Cons

⛔ Outdated/Discouraged

These options are not recommended for use.

• Original Wokada

Not suggested, older versions in youtube tuts are even way worse. GUIDE

• RVC GUI Mainline Realtime

The program is worse compared to the ones above, and much less updated. GUIDE

[inland field]

Could I run it locally with amd r7 200?

[nocturne mural]

AMD is not very good for running RVC properly, but I don’t think you’ll get much out of only 2 GB of vram.

https://colab.research.google.com/github/deiteris/voice-changer/blob/master-custom/Colab_RealtimeVoiceChanger.ipynb

Try that notebook; it might work for you since it keeps the same interface as the original w-okada, just with some changes.

[hallow thistle]

Do you mean AMD Ryzen 7 CPU? While W-Okada voice changer could work with CPU only, it's not always recommended for most scenario. For W-Okada Colab notebook, try this one. https://colab.research.google.com/github/tg-develop/voice-changer/blob/master-custom/Colab_RealtimeVoiceChanger.ipynb

[hallow thistle]

!howtoask

[patent trellis]

❓ How to Ask for Help

✅ Before You Ask!

1. Check Docs & Guides: Your answer may already be in the AI Hub Docs or the https://discord.com/channels/1159260121998827560/1159513888199540817 channel.
2. Search the https://discord.com/channels/1159260121998827560/1192011222023950368 : Look for existing posts that solve your issue. Do not invade someone else's post.

📝 How to ask?

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.

🚫 Prohibited Topics (We Will NOT Help With These)

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.

<:matsuripray:1159685390156967936> Community Expectations

• Be Polite & Patient: Our helpers are volunteers. You may ping the Helpers role once.
• English Only: Please keep all conversations in English.
• Don't Ask To Ask.

[hallow thistle]

Try downgrade Vonovox to v.1.6.9, the latest stable version. v.1.7 is a beta/pre-release version, so a bug or glitch is kind of expected.

[hallow thistle]

GPUs do matter for W-Okada voice changer. The performance roughly depends on the generation, GPU architecture (Ada Lovelace, Blackwell), VRAM (8GB, 16GB), the specs of a GeForce RTX GPU (like CUDA cores), it's not always the number naming (xx60, xx90) itself.

While the guide of Deiteris' W-Okada lists several recommended chunk and extra settings for known GPUs there, these "max settings" sound more of "known average" settings rather than the absolute one, and while most use "extra" value at 2.7 s, the actual chunk value may be raised over supposed chunk value in a real scenario if run along with a graphic-demanding game.

[white shadow]

anyone else having issues with hugging face

nvm we good ig

[rapid sable]

hey i keep getting an startup error about web server

any fix?

[warm oriole]

Any good Bangla (bangali) TTS out there?

???

[rocky isle]

Do you have to do this for Vonovox as well or is it meant to be done only for W-Okada?

“1 - Open Task Manager, click "Details"

2 - Right-click audiodg.exe and set the priority to "High"

3 - Right-click again and choose "Set affinity" then select only CPU 2.”

[lunar trout]

does anyone know how to fix the thing where when your using your mic
all of your audio sounds like your under water

[lavish maple]

Runway Gen-4.5

just donwload that on ur computer and run it

its the best one on the market apperenrtly so u will have to tweak and overclock ur systems if u want max of it

i was wrong

its licesned so its not open source anymore or it wasnt ever

so jsut use their website to us eit on cloud

ur gpu is well enough tbh

[viral mason]

You shouldn't have to do this for anything, it should just run normally after u download it

[viral mason]

I don't understand what u mean

[rocky isle]

I have screenshots showing the process actually had 1.9 GB of data in the Working Set (RAM) initially. Then it dropped to 300 MB.

Since the app didn't crash, that missing 1.6 GB of data had to go somewhere. It wasn't just 'empty reserved' space because it was literally sitting in RAM a few minutes prior. Windows definitely PAGED it out.

The thing is, audio still sounds the same and delay hasn’t changed so I don’t know what’s going on here.

[viral mason]

you must be using an old version, did u get it from a tutorial on youtube?

[rocky isle]

Yes, but the same thing also happens with vonovox and Deiteris W Okada

[viral mason]

that's odd

[rocky isle]

with all versions basically lol

[viral mason]

you should use either vonovox or wokada tg fork, deiteris is ok but anything else is outdated

but that's weird that they all do that

[rocky isle]

yeah i’m using vonovox. Shouldn’t audio be impacted when some of it is running on your ssd instead of your ram though? cause this is happening to me and it still sounds the same lol

[viral mason]

I wouldn't know as I don't know the impacts of using it on different stuff

I keep whatever voice changer I use in my downloads

[limber tangle]

What settings do you recommend for RVC? My dataset is studio quality, but the resulting product has excessive metallic artifacts in the letter “s.” (t de esser)

[viral mason]

have you tried using legacy core 1.5, it's a really good pretrain

[limber tangle]

ı use klm 4 pre train 48khz

[viral mason]

personally using the 32k version has gotten good results every time

klm 4 is kinda old if I am remembering correctly

[limber tangle]

hmm

[limber tangle]

u have download link ?

[viral mason]

the pretrain can be found here
https://discord.com/channels/1159260121998827560/1453062294446805244

it has many different versions, 32k, 40k, and 48k, as well as spinv2 I think and also Refinegan

[limber tangle]

Does it work properly for Turkish too?

[viral mason]

I wouldn't know as I speak english but you can try ^^

[limber tangle]

Well thank you for your help I'll give it a try

[viral mason]

you're welcome! hope the results are better for you

[alpine lotus]

so uhhh gradio just killswitches itself when you try to open the public url on colab

idk if it's a me issue but this didn't happen when i was about to train a model on colab earlier

[robust mirage]

guys what voice changer should i take? i have rtx 4060 laptop gpu, should i get local or cloud instead?

[alpine lotus]

speaking from my experience the 4050 laptop ran the voice changer just fine so if you have storage root for local but if you don't have enough storage go for cloud

[brave axle]

i just followed the guide and it still cant hear me at all

[robust mirage]

aight thankss

[elder kelp]

im tryna use the L voice, but how I configure the settings and get the best ones?

[viral mason]

did u get ur voice changer off a youtube tutorial?

[elder kelp]

no?

[viral mason]

what version are u using, it'll be easier to help if I know what one u have

the only 3 that tare recommended that are up to date is wokada deiteris, wokada tg fork, and the best currently vonovox

[crimson depot]

@viscid moss

[tame oracle]

-rt

[patent trellis]

💻 Local Realtime RVC

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

• Vonovox

A Realtime Voice Changer with similar performance to Wokada Tg-Develop Fork, with extra features, but it supports only Nvidia GPUs on Windows 10/11 unlike other options that have wider support. and without cloud options GUIDE

• Wokada Tg-Develop Fork

A personal fork (modified version) of Wokada Deiteris Fork, it just adds some Quality of Life improvements to it like supporting Spin Embedder and Audio Effects. Don't expect too much about it since the creator made it originally as a personal project. GUIDE

• Wokada Deiteris Fork

Deiteris' fork (modified version) of wokada that doesn't get updates anymore. GUIDE

⚔️ Wokada Tg-Develop Fork vs Vonovox

For Windows Nvidia, Both Wokada Tg-Develop fork and Vonovox have similar performance & quality. Users should read the pros and cons for both and choose based on their differences, such as UI and Vonovox's paid effects.
Read Wokada Tg-Develop Fork Pros&Cons & Vonovox Pros&Cons

⛔ Outdated/Discouraged

These options are not recommended for use.

• Original Wokada

Not suggested, older versions in youtube tuts are even way worse. GUIDE

• RVC GUI Mainline Realtime

The program is worse compared to the ones above, and much less updated. GUIDE

[hallow thistle]

Still looking for UVR5/MSST or something? You have been pinging Eddy for the same topic three times this year.

[hallow thistle]

W-Okada voice changer (b2397) fork and Vonovox will work with NVIDIA GeForce RTX 4060 Mobile GPU, locally. The storage size doesn't directly impact the performance of the voice changer, though a drive with highly free space (> 256 GB) is preferred.

Also, your account "pronouns" part is questionable, by the way.

[hallow thistle]

Do you mean W-Okada voice changer? And what is your PC GPU?

[rapid sable]

i fixed it by reinstalling so no worries but i got another question

im using RVC and im having trouble selecting models

[hallow thistle]

No, I'm more of investigating deeper issue.

[rapid sable]

okay then im using Gtx 1080

[hallow thistle]

W-Okada realtime voice changer or Applio RVC (retrieval-based voice conversion)? The initials RVC doesn't always mean realtime voice changer.

[rapid sable]

how can i know which one it is?

[hallow thistle]

[frank juniper]

hello everyone, can someone please connect me with an Ai training job or any given programming task I would really appreciate 🥺

[crimson depot]

I already DM

[wind orbit]

Does someone have problems with odaka aswell? Yesterday it was perfect and now those high pitches and cut offs are there and I changed nothing

[hallow thistle]

If it's a general knowledge, you can talk in here. It doesn't have always be a single moderator who knows about UVR5, unless you were trying to contact Eddy for his UVR5 fork bug fixes.

[crimson depot]

I'll just wait for the reply, It's a personal matter and I need a little help with it.

[hallow thistle]

Most people here for sure. What is your PC GPU? And did you follow any tutorial before?

[hallow thistle]

[wind orbit]

CPU: Ryzen 5 3600x gpu: rx 5500xt, I always have high temp and yesterday it was PERFECT, but today it’s just messy and whenever I play a game as soon as I speak it keeps micro freezing etc. Yesterday none of that happened. + the voice cuts off and it glitches like in those micro high pitch stutters mid speaking

[hallow thistle]

Ok, but which W-Okada version you're using?

[wind orbit]

The std one, only one working for my amd

[hallow thistle]

Try this W-Okada version. https://github.com/tg-develop/voice-changer/releases/download/b2397/voice-changer-windows-amd64-dml.zip

[wind orbit]

The dml one doesn’t work to me somehow

I tried it from huggingface and it just opens the cmd and closes right after

It says something abt windows pkg and then writes a huge chunk of commands and then just closes

[hallow thistle]

What I sent to you is a different (b2397) version to what you're using (likely being v.1.5.3.18a).

[wind orbit]

leemme try

[hallow thistle]

The issue with v.1.5.3.18a W-Okada, especially its DirectML variant has to happen with its outdated and buggy code, where the program only uses CPU even if your AMD Radeon RX GPU is present.

[wind orbit]

ohh and that explains the high pitches?

because with the voice im using there shouldnt be any, its one of the best teached ones

[hallow thistle]

No idea which "high pitches" refer to, but if you mean those background noises coming to your microphone it's likely the cause, or when the program is so laggy that the audio is pitching up and down unexpectedly.

[wind orbit]

no i mean, when i talk normally it sometimes gives like those high beeps between the speaking of the voice, plus it stutters or cuts off completely sometimes

[hallow thistle]

Still, though.

[sweet osprey]

where is the AI HUB FR

@hallow thistle

[wind orbit]

this one is way better

the one u sent me

can u send me the settings for noise suppression? i cant change it somehow

like the noise 1 and 2 doesnt wanna check

[hallow thistle]

AI Hub FR or AI Hub France no longer exists. What you're here now is "AI Hub by Weights" Discord server, which is the second iteration of previous AI Hub that was taken down in 2023.

[hallow thistle]

These three "noise supression" settings only work when you set audio mode to "client". If you set audio mode to "server", both noise and echo choices will be grey out. It's a quirk in every W-Okada version.

[wind orbit]

okay tysm

it works so good

[hallow thistle]

You're welcome. Make sure to set extra number to 2.7 s for more audio quality, though always check performance number at top right screen if there any delay.

[robust mirage]

aahh okkay i'll use vonovox then, i'm afraid that my 8 gb vram can't handle it at first, thanks

[robust mirage]

uuhhm... just ignore it

[viscid scarab]

Help me

I need some help

[hallow thistle]

What is your query again?

[wild kindle]

what program can be used to generate the images that are listed in the ai-images section?

[dry turret]

I have a question, it's my first time using w-okada and with each voice that I use, and listen back to the recording of my voice, the voice is always so choppy and cuts out mid word. Any idea how this can be fixed?

[hallow thistle]

Stable Diffusion.

[tulip onyx]

I need someone to tell me if there's an alternative to Hugging Face to remove noise, echoes, etc., because the page isn't working when I try to access it.

[spare vale]

I think this is the right place to ask. Any models I upload myself have super high res, never gets below like 3000 ms, and when I first use the models i uploaded, they start off at like 100ms and work just fine until it quickly skyrockets in ms. i've uploaded even the most downloaded models in this server and I never get anywhere close to lower, even using the quickest settings. Using the premade models I usually get around res 200ms. Please help. (i'd send pictures but i cant)

[hallow thistle]

More like UVR5.

[hallow thistle]

What is your PC GPU? And did you follow any tutorial or guide before?

[hallow thistle]

Yes, this is the right channel for query about voice changer. What is your PC GPU? Did you follow any tutorial before?

[spare vale]

9070xt, and yes I followed the tutorial, down to every setting.
Also I apologize if I don't respond, I have work in a few hours so I'm laying down, but I shall be back here soon to figure out my issue!

Tutorial I used:
https://youtu.be/SxdnGxicJOg?si=OUXQbwaUpZp54DYy

[hallow thistle]

For better voice changer version, you might like to try this version. https://github.com/tg-develop/voice-changer/releases/download/b2397/voice-changer-windows-amd64-dml.zip Simply, the v.1.5.3.18a W-Okada has long been outdated, its DirectML variant is very buggy as well.

[viral cradle]

u got one for mac os?

hello

?

[hallow thistle]

Apple Silicon (M2, M3) or Intel Mac?

[viral cradle]

macBook air

[hallow thistle]

Yes, that's the Macbook name. But what about which CPU/chip does your Macbook Air has?

To check your Mac's CPU/chip, click Apple icon at top left corner screen, navigate "About This Mac".

[viral cradle]

its a applke m1

*apple m1

[hallow thistle]

Try this W-Okada version https://github.com/tg-develop/voice-changer/releases/download/b2364/voice-changer-macos-arm64-cpu.tar.gz, and make sure to follow this part of the guide as well. https://docs.aihub.gg/realtime-voice-changer/local/tg-develops-w-okada-fork/#download-for-apple-silicon-mac

[viral cradle]

wont work

tells me this may harm device when i try terminal

[hallow thistle]

https://cdn.discordapp.com/attachments/1159290139609137264/1459543412461273168/image.png?ex=696a40ba&is=6968ef3a&hm=c3b4ff9125645d1a89891f1e2b575c420b8431fff527ea1c2f93fc69ec19383d&

[runic cape]

who can help me to get the Server URL

[sturdy nimbus]

how to fix the delay

[flint hazel]

Can someone help uh my gpu is rtx 2060 super so my chunk size is not going lower than 2400 but guides say it should atleast go down to 128 this is my version vcclient_win_cuda_2.0.78-beta

[viral mason]

That version is very old

[flint hazel]

💔

[viral mason]

You should switch to wokada tg fork

[flint hazel]

I thought it was the latest and they discontinued it feck chatgpt

[viral mason]

Lemme get u the links

[flint hazel]

👁️👄👁️

[viral mason]

U can't trust a robot lol

[flint hazel]

true true

tho is it the reason why the Chunk count is so high?

[viral mason]

You'll need these, the first one is a virtual audio cable and the other two are for the voice changer
https://software.muzychenko.net/freeware/vac470lite.zip

https://github.com/tg-develop/voice-changer/releases/download/b2397/voice-changer-windows-amd64-cuda.zip.001

https://github.com/tg-develop/voice-changer/releases/download/b2397/voice-changer-windows-amd64-cuda.zip.002

[viral mason]

Possibly just limitations in the code or it being weird

[flint hazel]

ya sure it will run on my gpu..?

[viral mason]

You can try 🤷‍♀️

[flint hazel]

💔💔

[viral mason]

Btw to use this one extract the 001zip and then place 002zip in it, since 002 can't be extracted

Idk why it's needed but it is

And for the virtual audio cable just extract and run setup64

Then to run wokada tg fork run mmvcserversio

It's an exe file

[flint hazel]

so ya mean after extracting the 001.zip and put the 002 zip in the 001.

[flint hazel]

Got that already

[flint hazel]

are those in 001?

[viral mason]

Wokada tg fork is the 001 and 002 zip thing

[flint hazel]

aa..I see..

[viral mason]

Yup

[flint hazel]

ty for the help

[viral mason]

You're welcome! If u need more help just @ me here or if u need to send pictures and can't here just ask to dm me ^^

[flint hazel]

oki

[tame oracle]

!howtoask

[patent trellis]

❓ How to Ask for Help

✅ Before You Ask!

1. Check Docs & Guides: Your answer may already be in the AI Hub Docs or the https://discord.com/channels/1159260121998827560/1159513888199540817 channel.
2. Search the https://discord.com/channels/1159260121998827560/1192011222023950368 : Look for existing posts that solve your issue. Do not invade someone else's post.

📝 How to ask?

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.

🚫 Prohibited Topics (We Will NOT Help With These)

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.

<:matsuripray:1159685390156967936> Community Expectations

• Be Polite & Patient: Our helpers are volunteers. You may ping the Helpers role once.
• English Only: Please keep all conversations in English.
• Don't Ask To Ask.

[wide perch]

Using RVC with virtual audio cable and the output isn't producing my voice at all.

What can I do to fix this?

[viral mason]

Wdym?

Do you mean the real-time voice changer

[wide perch]

yert

When I start it, the vb doesn't work at all

Don't hear my voice even though I have Listen to Device enabled

[viral mason]

What input and output settings do u have set?

<@&1159293140440723499> kill him

[viral mason]

Uh oh

[wide perch]

In RVC, output is vb input and my input is my normal microphone. Then in windows settings, I have the vb output selected as the mic

[viral mason]

I hope he doesn't turn me into an ice statue..

[manic parrot]

which graph should i follow?

[wide perch]

Did you see

[viral mason]

For training a model don't follow the graphs, I learned just listening to each epoch is better, start around 100-200 epochs

[viral mason]

I did yea, you have them set up correctly but idk why it isn't working

[wide perch]

the vb output mic just doesn't work, dont know what to do

[manic parrot]

Okay, then I'll test it every 50 epochs

[viral mason]

Oki doki

[wide perch]

this hasn't happened to anyone else before?

[viral mason]

You could try switching to vac lite

I'll get the link

https://software.muzychenko.net/freeware/vac470lite.zip

To set this up just download it, extract the zip file and run setup64

Not as admin btw just regular

[wide perch]

Doesn't work, now I don't know if it's the vb cable or rvc or my computer preventing them from working for some stupid reason

the vb mic doesn't even show the green bars when I talk, so it's literately not picking up my voice at all

[viral mason]

damn, yea idk what could be the issue did u get the voice changer off a youtube tutorial possibly?

if so it's outdated

[wide perch]

I'm just using normal RVC from github

is there a new rvc that I'm not aware about?

[viral mason]

which github page?

[wide perch]

https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/releases

[viral mason]

there are 3 currently yes, wokada deiteris fork, wokada tg fork, and vonovox

[wide perch]

do all of them work

[viral mason]

they worked for me so I would think so

[wide perch]

So I shouldn't be using the standard one then?

[wide perch]

this one

[viral mason]

I looked at the link u sent and I've never seen this one before so most likely it's outdated

what gpu do u have?

[wide perch]

rtx 5060 ti

[viral mason]

ooh nice

[wide perch]

thanks dude

So out of the 3 which do you recommend?

[viral mason]

that one is peak, I have the 5070 ti

[wide perch]

siick

[viral mason]

I'd say prob wokada tg fork as that's what I use, tho vonovox is better tg fork is easy to learn and has fast model swapping times

vonovox also has a ton of features that might confuse u

[wide perch]

alr thanks ill give it a try

[viral mason]

let me get u the link so u don't download the wrong one

[wide perch]

thanks

[viral mason]

https://github.com/tg-develop/voice-changer/releases/download/b2397/voice-changer-windows-amd64-cuda.zip.001

https://github.com/tg-develop/voice-changer/releases/download/b2397/voice-changer-windows-amd64-cuda.zip.002

all u gotta do is just download both, extract the 001 zip and put the 002 zip into the folder for the 001 zip since 002 cannot be extracted but is needed for whatever reason

[manic parrot]

I have google colab pro, can i train 2 models at once

l4

[viral mason]

kaggle is better tbh, 30 hours free

can't train more than one model at a time but still, for free 30 hours, and colab at most will give u maybe 4 for free

per account

[wide perch]

is it only web

never used web ones before

[viral mason]

the interface shows up on web browser yes

but it's still on local

runs from your pc gpu and the task manager

[manic parrot]

Can I open a second Google Colab workspace while the first one is open?

[wide perch]

IT WORKS

LETS GOOO

Thank you so much

[viral mason]

yea but u need to use a second google account to train the other model with it

[viral mason]

you're welcome! do u have any questions abt it?

like how to use it or anything

[wide perch]

nah I figured it out thanks

[viral mason]

alr, hope it works well for u in games and stuff

[wide perch]

hope so too

[wide perch]

I have a question actually.

I want to switch the processing unit to use my GPU instead of my CPU but it doesn't seem to work when I click Start Server.

Just wondering if it is possible to use my GPU since it would run better

[wide perch]

No yeah I did change it but clicking Start Server does nothing

The performance stats don't move

[viral mason]

odd

did u change from client to server?

client does not work

[wide perch]

ooh

thats probably why lol

its working

[lament prawn]

i know this is yesterday but um this is happening to me

[neat patio]

Can I get some proformence tips (specs rx 580 8gb
I3 10100)

[lethal dirge]

Hi everyone, could you help me download the program to create voices for my characters, please?

[torpid spire]

ive got a 4060 gpu and when i open start http bat it doesnt do anything js opens then closes

[hallow thistle]

!howtoask

[patent trellis]

❓ How to Ask for Help

✅ Before You Ask!

1. Check Docs & Guides: Your answer may already be in the AI Hub Docs or the https://discord.com/channels/1159260121998827560/1159513888199540817 channel.
2. Search the https://discord.com/channels/1159260121998827560/1192011222023950368 : Look for existing posts that solve your issue. Do not invade someone else's post.

📝 How to ask?

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.

🚫 Prohibited Topics (We Will NOT Help With These)

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.

<:matsuripray:1159685390156967936> Community Expectations

• Be Polite & Patient: Our helpers are volunteers. You may ping the Helpers role once.
• English Only: Please keep all conversations in English.
• Don't Ask To Ask.

[hallow thistle]

Try W-Okada fork. https://docs.aihub.gg/realtime-voice-changer/local/tg-develops-w-okada-fork/ https://github.com/tg-develop/voice-changer/releases/tag/b2397

[torpid spire]

what link theres 2?

thats amd @hallow thistle im nvida

[lethal dirge]

Could you please send me the download link for my PC?

please

[hallow thistle]

Download https://github.com/tg-develop/voice-changer/releases/download/b2397/voice-changer-windows-amd64-cuda.zip.001 and https://github.com/tg-develop/voice-changer/releases/download/b2397/voice-changer-windows-amd64-cuda.zip.002. AMD64 is another CPU architecture name for x86-64 or x64, used in most Intel and AMD CPUs, and it doesn't refer to AMD Radeon GPU that way. The term "CUDA" mostly refers to NVIDIA GPU.

[hallow thistle]

What is your PC GPU?

[lethal dirge]

Or rather, what processor or what version of Windows do I use?

[hallow thistle]

No joke question. To check your PC CPU and GPU, open Task Manager, go to Performance tab, see CPU, GPU 0 or GPU 1 on the left panel.

[lethal dirge]

Could you please send me the download link for my PC?

[hallow thistle]

Because you keep asking for it but never actually answer my questions. https://huggingface.co/IAHispano/Applio/resolve/main/Compiled/Windows/ApplioV3.6.0.zip

[lethal dirge]

Is that the download link?

[hallow thistle]

Why should you ask more? That's a download link.

[lethal dirge]

But is this the download link for the program?

[hallow thistle]

It's not that hard to click that download link.

[viral mason]

yes it is

[hallow thistle]

W-Okada voice changer or Applio RVC (non-realtime)? While the voice changer could work with AMD Radeon RX 5xx GPU, it would likely struggle when with higher settings (Extra 2.7 s on bXXX W-Okada fork). Applio RVC (retrieval-based voice conversion) can work with AMD GPU but with some tweaks.

[hallow thistle]

The user has Google Colab Pro tier, so the runtime limit now depends on the user's Colab compute units and not always limited to 2 - 4 hours within 24 hours, which usually happens with non-Pro free users. Kaggle also has an option to link Colab account (that has Pro tier enabled) for more Kaggle GPU quota, by the way.

[viral mason]

interesting

[hallow thistle]

With Colab Pro, while this tier allows you to run few more notebook instances on the same account, training 2 models at the time (if they on separate instances and use same GPU) sounds impractical or impossible, because resources would be shared and distributed across notebook instances, leading to slower performance for both and potentially drain compute units faster. For full performance, running only one instance is always better.

[manic parrot]

thanks for info

[candid urchin]

What is the most stable real-time software at the moment?

I’m currently using Applio v3.6.0 to train a voice model. Could anyone recommend the best configuration settings for achieving optimal results?

[manic parrot]

i got this error while starting training in applio

[unborn oracle]

hello everybody, I have never ever did anything AI related other than using AI, I wanna get into it and learn about it and models and stuff, I dont know where to begin or how to, all I know is that you need a decent pc, I have a 4080 super and a ryzen 7 7800x3d, anyone who has tips please share and thanks (;<->;)
I use windows 11

[hollow matrix]

guys can someone help , when i hit the start.http nothinf happpens....

[hallow thistle]

Tg Develop's W-Okada voice changer. I was about to tell you this 7 days earlier, but you were away.

[hallow thistle]

What is your PC GPU?

!howtoask

[patent trellis]

❓ How to Ask for Help

✅ Before You Ask!

1. Check Docs & Guides: Your answer may already be in the AI Hub Docs or the https://discord.com/channels/1159260121998827560/1159513888199540817 channel.
2. Search the https://discord.com/channels/1159260121998827560/1192011222023950368 : Look for existing posts that solve your issue. Do not invade someone else's post.

📝 How to ask?

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.

🚫 Prohibited Topics (We Will NOT Help With These)

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.

<:matsuripray:1159685390156967936> Community Expectations

• Be Polite & Patient: Our helpers are volunteers. You may ping the Helpers role once.
• English Only: Please keep all conversations in English.
• Don't Ask To Ask.

[cosmic epoch]

which Ai source is this?

[hallow thistle]

This is Tg Develop's W-Okada voice changer fork. Basically, it's an implementation of W-Okada (MMVCServerSIO). https://cdn.discordapp.com/attachments/1159290139609137264/1457950706500763809/image.png?ex=696b0ce7&is=6969bb67&hm=8bd91ae54b7db47d22da490dd6d6c5c145640cd42f5a235d0e217b8018939f9b&

[cosmic epoch]

is this AI for free? and can you train models in it?

[hallow thistle]

W-Okada can't train a voice model, only inference, which means converting your vocal in realtime using RVC voice model.

[cosmic epoch]

how do you use this W-Okada? and is it for free?

[hallow thistle]

W-Okada is a free and open source program, much like RVC (retrieval-based voice conversion; like Applio RVC fork).

[hollow matrix]

Nvidia

[hallow thistle]

Which NVIDIA GeForce RTX?

[cosmic epoch]

is it something you use online or do you need to download it?

[hallow thistle]

How am I supposed to answer that?

[cosmic epoch]

give me a link?

[hallow thistle]

What? Since your PC has zero NVIDIA GPU (assuming you have no PC GPU), and you wanted to try it, this is Colab notebook link for that one W-Okada. https://colab.research.google.com/github/tg-develop/voice-changer/blob/master-custom/Colab_RealtimeVoiceChanger.ipynb

[cosmic epoch]

okay thanks

[hallow thistle]

Generally, I do not like to tell little trivias (like your queries) whose common information can be accessed either in #1159513888199540817 or Google, and if you do research and understand a bit on why W-Okada voice changer is like that, it would be great.

[hollow matrix]

Yeah ti 3050

[cosmic epoch]

@hallow thistle Also, is it impossible to to upload models to weights? because when i click on "train model" the bottom that says "upload model" dissappeared.

[hallow thistle]

Download and try Tg Develop's W-Okada fork. https://docs.aihub.gg/realtime-voice-changer/local/tg-develops-w-okada-fork/ https://github.com/tg-develop/voice-changer/releases/tag/b2397

[hollow matrix]

Okay!! Thanks so much

[hallow thistle]

[cosmic epoch]

also, is there any other way to train models besides the applio colab and weights? like, is there any other colab for training models that WORKS nowdays?

[hallow thistle]

Why should you want more? Does Applio RVC on Colab not work?

[cosmic epoch]

it does work, but... i dunno... once there were more colabs that allow you to train voices. is this the ONLY colab that does it now?

[hallow thistle]

Applio RVC is the only best known working RVC fork. Some other RVC implementations are either outdated, no longer maintained.

[thick knot]

w-okada don't work, why

@pseudo depot pro help

[hallow thistle]

Try Tg Develop's W-Okada voice changer. https://docs.aihub.gg/realtime-voice-changer/local/tg-develops-w-okada-fork/ https://github.com/tg-develop/voice-changer/releases/tag/b2397 The one you use is an original and outdated W-Okada.

[thick knot]

bro I want something useful

[hallow thistle]

Yes, that one is more useful, just that when you were overlooking it.

[hallow thistle]

Download both of them.

[thick knot]

why

[hallow thistle]

Simply, these are split zips of a single zip file.

Why why? You use WinRAR or 7-Zip to open the .zip.001 one.

[thick knot]

7-Zip

[pseudo depot]

why me ?? I don't know anything about this topic

[thick knot]

I felt you were Arab

and for

[pseudo depot]

I am

but I don't know anything about the AI models that are created here, I am specialized in another area of software engineering

[thick knot]

عربي اخوي

[hallow thistle]

That doesn't mean the person knows about voice changer.

[pseudo depot]

I'm a girl, and yes I am arab

[viral mason]

Weights removed the option because they're greedy and horrible

[cosmic epoch]

True. This site is falling apart completely. It's like they're trying to keep everyone away from this site

[viral mason]

It's very sad to see them remove useful features, like downloading the outputs of ai vocals and now uploading good models

[serene swallow]

Deprecated. Use vonovox. That one does actually work if you have an nvidia gpu

[wild bobcat]

i was wondering that too

ugh

this sucks

[royal vapor]

I'm trying to download tg-devlop's version but i get that the 002 rar file is damaged... I've tried downloading it several times. Anyone experienced this?

[viral mason]

Just place the 002 zip into the folder of the 001 zip after extraction

And run mmvcserversio

You can't unzip it but it's still needed

[royal vapor]

so directly into the MMVCServerSIO folder?

[viral mason]

Just drag and drop into the folder of 001, then u can run mmvcserversio so wokada tg fork starts

[robust halo]

PyInstaller\loader\pyimod02_importers.py:378: UserWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html. The pkg_resources package is slated for removal as early as 2025-11-30. Refrain from using this package or pin to Setuptools<81.

[robust halo]

☹️

[viral mason]

<@&1159293204038955078>

[fair crag]

Hello

[lethal basin]

hi

your local worm

[viral mason]

that increases the index, the index basically just controls the model's "accent" it's not needed to be moved past 0 tho ^^

[livid idol]

same thing happening here

[nocturne patrol]

is there a tutorial? I'm on PC windows and i have a GTX 1070?

[nocturne patrol]

To download the voice changer

[alpine lotus]

and this is why from now on i'll be moving models to jammable

[alpine lotus]

this happens on the main acc but not in secondary ones

i wonder why...

[nova abyss]

what chunk size can i use? rtx 5050

[viral mason]

what version of the voice changer are u using?

I use a 5070ti and I can get pretty low, all the way to 122.7 for chunk

[nova abyss]

tg

[viral mason]

try chunk at 122.7 and extra 2.7

[nova abyss]

i feel like its coming out too fast

[viral mason]

what?

[nova abyss]

isnt that what chunk does?

the lower the chunk the faster it comes out but the higher the chunk the slower it comes out?

[worthy helm]

why does it sound like there is a delay\

[silent stratus]

anyone know whats goin on with the applio notebooks rn?

this what i just got in kaggle:

Save the link for later, this will take a while...
Traceback (most recent call last):
File "/kaggle/working/program_ml/app.py", line 30, in <module>
from tabs.inference.inference import inference_tab
File "/kaggle/working/program_ml/tabs/inference/inference.py", line 9, in <module>
from core import (
File "/kaggle/working/program_ml/core.py", line 20, in <module>
from rvc.lib.tools.model_download import model_download_pipeline
File "/kaggle/working/program_ml/rvc/lib/tools/model_download.py", line 14, in <module>
from rvc.lib.utils import format_title
File "/kaggle/working/program_ml/rvc/lib/utils.py", line 9, in <module>
import wget
ModuleNotFoundError: No module named 'wget'

[viral mason]

Not sure, I was just using the kaggle notebook for applio and it worked fine

[silent stratus]

hmm

[hallow thistle]

That's close. The lower chunk value means less delay, while higher chunk value means higher quality.

@viral mason By the way, aside from sending these people links, how much do you know about the voice changer?

[hallow thistle]

What is this Python program/environment supposed to be? And is your PC GPU still AMD Radeon RX 570?

[nova abyss]

what chunk size can i use if i want higher quality?

[hallow thistle]

I have so some mixed thoughts about whether if Weights.com still considered good or no, as much as people here especially Local_Worm keep shitting on the site as if the website itself is some political material.

As much as I felt guilty for didn't actually pay for the site whereas I once won a prize for a free Weights subscription, the Weights and related websites (Voyages) are only good for "draft" creations, especially AI covers, inference and some model trainings. The most parts could be done through separate softwares like UVR5 and Applio RVC, but this workflow is more complex than the Weights website itself, to which I just think some lazy people wouldn't even do this.

[ashen minnow]

i cant add models form #1175430844685484042 to vcclient

[hallow thistle]

What is your PC GPU? And did you follow any tutorial before?

[ashen minnow]

when i try to upload any model it's says it's missing toml file

[hallow thistle]

!howtoask

[patent trellis]

❓ How to Ask for Help

✅ Before You Ask!

1. Check Docs & Guides: Your answer may already be in the AI Hub Docs or the https://discord.com/channels/1159260121998827560/1159513888199540817 channel.
2. Search the https://discord.com/channels/1159260121998827560/1192011222023950368 : Look for existing posts that solve your issue. Do not invade someone else's post.

📝 How to ask?

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.

🚫 Prohibited Topics (We Will NOT Help With These)

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.

<:matsuripray:1159685390156967936> Community Expectations

• Be Polite & Patient: Our helpers are volunteers. You may ping the Helpers role once.
• English Only: Please keep all conversations in English.
• Don't Ask To Ask.

[ashen minnow]

every model what i dowland it's only have 2 files and default from vcc have a lot more

[hallow thistle]

You still have not directly answer my questions.

[ashen minnow]

intel core i3-9100F and yes i used tutorial

[hallow thistle]

That's an Intel CPU, not a GPU (graphics processing unit).

[ashen minnow]

Nvidia Geforce Gtx 1650

[hallow thistle]

Try Tg Develop's W-Okada fork. https://docs.aihub.gg/realtime-voice-changer/local/tg-develops-w-okada-fork/ https://github.com/tg-develop/voice-changer/releases/tag/b2397

[ashen minnow]

it's keep showing cannot install device driver (1072 - the specified service has been marked for deletion)

[hallow thistle]

Try extract MMVCServerSIO folder to somewhere else like "D:\MMVCServerSIO", shouldn't be on your desktop folder, check your GPU driver.

[languid turtle]

@hallow thistle

what is the best version

of tg okada

the most faster

and realistic

[hallow thistle]

Download https://github.com/tg-develop/voice-changer/releases/download/b2397/voice-changer-windows-amd64-cuda.zip.001 and https://github.com/tg-develop/voice-changer/releases/download/b2397/voice-changer-windows-amd64-cuda.zip.002

[languid turtle]

alr

[hallow thistle]

While the voice changer might work with NVIDIA GeForce GTX 980 Ti, the entire GTX 9xx GPU series is slightly older than GTX 10xx and the more recommended RTX 20xx, so you don't always expect it to work fast whereas the program might struggle with that GPU.

[languid turtle]

Is there a more optimized version?

[hallow thistle]

Oh. There's Vonovox, a complete voice changer alternative, but this specific software is known to work best with RTX 20xx, so even if you get GTX 980 Ti to work it would likely struggle either way. So, "W-Okada voice changer fork" is your only hope to run locally. For full performance otherwise, consider for a more recent PC with better GPU (like GeForce RTX 3060) or an online website instead.

[languid turtle]

is it realistic?

[languid turtle]

i wanted to say if its sound like human and not robotic

[hallow thistle]

The audio quality has to happen with chunk/extra settings and the RVC voice model itself, not just specific versions.

[distant hamlet]

Weights was good until they removed model uploading

[swift thunder]

Is Gradio down?

[wind yacht]

Yo

How do you fix the res ms thing skyrocketting when you're using the voice changer in a VC

Like when I'm not speaking it's like res 14 ms, and when I do speak it launches up to 500ms and sometimes even above 1k, making my voice cut badly

I'm running a gtx 1650

[viral mason]

I know that extra should be set at 2.7 not any higher as it'll cause cutoff issues, chunk is dependent on the person's GPU and pc as not all can handle lower values and does change in delay when lowering or raising the value.

I don't know anything about fixing errors caused by the person's computer or if they're missing anything like python, I also cannot help at all with the cloud versions because I know nothing about setting them up

But besides that stuff I mainly know that if they're not using wokada deiteris fork, wokada tg fork, or Vonovox they are using an outdated software

Probably stuff I'm missing I just woke up

[cobalt prairie]

guys

How can i fix the constant cutting in my voice

with the chat bot

i mean voice bot

[viral mason]

???

[cobalt prairie]

Like the voice is cutting

i verified and no external sound exept my voice is

like hearing

[viral mason]

I'm very confused, could you explain what you mean by voice bot

[outer valley]

@distant hamlet i post it here because in general i cant post images.
This is my tensorboard of my first tries to train. With different settings (because me dum and a bit confused about applio settings).

And i have the feeling i get something wrong.

Thank you a lot.
Maybe this is a bit dumb to ask. But what applio settings would you use for a dataset of 1h 13 min that is with a sample rate of 48k and very diverse with normal talk and many emotional talk and noises.

I used BS-Roformer-Viperx-1297 in UVR5 to remove the background music. And in RX 10 i used this module chain (screenshot) to go over all the files. They raw audio data had background music and other noises and personally i hear a lot of this room/reverb in it a lot. So i wonder if that workflow i did is a good way to go. @distant hamlet

[distant hamlet]

I guess since your dataset is so large, it works out. Maybe batch size of 10-12?

[distant hamlet]

chain looks fine. how bad is the original audio tho? And are they all pieced audio from different sources? I personally prioritize finding best-quality audio to begin with > trying so hard to repair mid/bad-quality audio. unless it’s reallly hard to find audio then that’s an exception

[outer valley]

The model of the orange line that was trained for 8h with batch size of 8. Was bad but better then the other ones with batch sizes of 16-22 that only trained for around 2h.

Epochs are something i wondering even after reading about it what it really means when you lets say train for 400-1000 epochs. More is not better i got that because of overtraining. But for that there is tensorboard and even in applio a setting to stop training after a xx number of not doing and better process.
But doesnt that not mean i could like always go for more epochs and just need to find and pick the best epoch or does a high epoch setting alone affects quality? No right?

[distant hamlet]

I usually find de-reverbing from MVSEP (or other AI models) to be sufficient. RX’s deverbing might not be necessary. and wait, what’s loudness control? lol

[distant hamlet]

gotcha. so how many epochs was the bs8 model at 90k steps? and how did it sound to you? that line looks like it was potentially continuing to trend downwards or just about plateau, but just evaluate from how it sounds

[outer valley]

I think if you fine with a little wall of text....

I have a voice model that i created with merging 2 models in Okada.
I really like that model and that sounds really good but it is old. Like from 2023 and 2024 and it has this high pitch and ss/zz/hiss problem that it makes this weird ai voice sounds.

Since back then i wanted to fix that model but i used already existing ones. Sadly there was some sort of wipe back then? Not sure i was never much active in AIhub.
So i was not able to find the orginal uploader of the 2 models.
And since then try to find better base models. When there was still "commission section" i got some models made from the "Master model makers" but... to be honest they not reached the quality of this one random model i got before the wipe or what ever happened to that time.

They had to much hall/reverb or to muddy or had the same hiss/zz problem.

So tried to merge around with different models i thought could maybe help to what i want. I only could get a small upgrade from that.

In short now in 2026 i hoped there is better ways to go for my goal with like "pretrains" and using my 4090 to just do it myself now.

Back then i got told trying to make a model with emotional data is not a good idea and not works out well.
Now at least "gemini" kind of tried to tell me that changed or would be possible.

So i got raw data from youtube from the voice i wanted with like many different emotional sounds and talking and a lot of normal talking gemini said a 13 min of emotional stuff and around 30-40 min of talking would work.

So i used UVR5 and RX 10 with was i showed even manual cut A LOT of stuff UVR5 missed in RX10 and cut to long silent parts between talking. But only were i felt the pause was too long.
So the dataset is a big mix of different sources of the same voice.

[viscid scarab]

help me

[outer valley]

So gemini was telling me it would be good to "normalize" audio to bring it all to the same lvl ?
Like laughing or scream would otherwise be to loud and whisper to quiet for training. Because i was very unhappy with the first try of training and checking the dataset i felt like... it still had to much of strong reverb and this noises of for example a PC being near the mic.
So the 2.0 cleaning gemini suggested "loudness control"

[outer valley]

that would be around 250 epochs from what i see

[dreamy seal]

a bug on https://colab.research.google.com/github/iahispano/applio/blob/master/assets/Applio.ipynb#scrollTo=nAlXiNYnFH9F

if i visit the website the execution instantly got terminated

would be just like this after it finished lanching and it got terminated

[distant hamlet]

gotcha, I figured it was some sort of compression or normalization. which is good

[distant hamlet]

looks perfectly normal then. have you tested to hear how it sounds at 250e?

[outer valley]

So the sound was... a bit hard to explain...hmmm
There was glitchy noises (hmmm sounds for example) and had the hiss problem.

But i could hear the emotional talk quality and so on. Sorry not sure how to explain it the best.
But the biggest problem was :

• glitchy voice sounding
• hiss problem
• metallic/reverbing?

[distant hamlet]

any inference audio sample you can show? you can DM me if you want

[dreamy seal]

why is this paused now? how do i use it again?

[outer valley]

the 2.0 cleaned dataset i trained with 16 batch size.
Should i try go down to 8 and normal training settings?

Was my idea of a lot of emotional audio in the dataset maybe a wrong idea? Emotional talk i mean like :

• funny weird mouth noises that for example some vtuber do.
• whispering
• burst of laughing or other noises

i manual cleaned and removed what i felt could be to bad from quality or maybe not work well. But maybe it is still contain to problematic sounds?

There is a mix of talk from like 2022-2023 and from 2025.
So there is a little difference from for example accent of the voice. But my goal is a own voice with 2 different voices so i not care for accent.

It only would be a problem if that means using audio from different times of the person voices with different accent and maybe a small change in audio from maybe different mic setup hurt the training a lot.

[outer valley]

i send you one one moment

[worn lodge]

sup guys, so, Im trying to run qwen image with the upscaler fine tune. I got a rtx 3060 12gb + 32gb of ddr4 ram. Now, I tried both the quant4 and quant8, nothing, im always hitting OMM, like Im probably doing something wrong, because what I wanna do is: I run the model on the ram, and then I offload some layers to the gpu, this works fine in text models, but vision models... idk, they just dont wanna work. Any recommendations?

[serene swallow]

-vonovox

-help

!vonovox

Bro. Can someone give me this disc's vonovox repository

[hallow thistle]

The hell. Make sure to read help guidelines before start asking. What is your PC GPU? And did you follow any tutorial before?

[hallow thistle]

The "repository" in question: https://github.com/dr87/Vonovox

Assume you have NVIDIA GPU since you never stated anything about it just demanding for the voice changer.

[serene swallow]

Why do you sound aggressive about it

It was just a question bro

You dont have to reply if you dont want to

[hallow thistle]

Because literally everyone thinks I sound angry when I never meant it.

[hallow thistle]

For "free" Colab users, Google Colab can disconnect your current runtime (especially the one with Web UI) anytime, which is an expected behavior. With Colab Pro, you could run Applio RVC UI or any Web UI notebook on Colab without a problem. This is not a bug or defect that persists with Applio RVC UI Colab notebook, it also happens with certain notebooks that use Gradio, ngrok or other Web UIs like "W-Okada voice changer" as well.

The same issue** doesn't** happen with Applio RVC no-UI because this specific notebook doesn't use Gradio or any web UI codebase, certain features/commands are separated into code cells like this.

[loud warren]

This file was blocked because files like this from the internet arent safe

how do i fix

@hallow thistle

[hallow thistle]

What is this file supposed to be?

!howtoask

[patent trellis]

❓ How to Ask for Help

✅ Before You Ask!

1. Check Docs & Guides: Your answer may already be in the AI Hub Docs or the https://discord.com/channels/1159260121998827560/1159513888199540817 channel.
2. Search the https://discord.com/channels/1159260121998827560/1192011222023950368 : Look for existing posts that solve your issue. Do not invade someone else's post.

📝 How to ask?

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.

🚫 Prohibited Topics (We Will NOT Help With These)

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.

<:matsuripray:1159685390156967936> Community Expectations

• Be Polite & Patient: Our helpers are volunteers. You may ping the Helpers role once.
• English Only: Please keep all conversations in English.
• Don't Ask To Ask.

[loud warren]

check my forum it has the full image

[weak sorrel]

Full GPU Name: AMD Radeon RX 7900 XT
Operating System: Windows 11
Detailed Description: I use the basic W Okada client but wanted to see if i can maybe use something that can work better for AMD but ended up making it worse lol just wanna know if i messed up or if my system is cooked and keep with what ive been using (when i say worse i mean my system lags and my ms is INSAINE i cant get it to lower at all even at chunk at 800ms)
Tutorial Used: https://github.com/deiteris/voice-changer?tab=readme-ov-file

[hallow thistle]

On your b2332 W-Okada fork, try these settings:
Chunk: 128 ms
Extra: 2.7 s
GPU: AMD Radeon RX 7900 XT

[weak sorrel]

okay

its laggy still it keeps being "close" to the perf but when i raise it it will only keep going up

[viral mason]

that looks old

[weak sorrel]

so if i match 208 i would need 250 and etc

oh?

i clicked the download link it said as recommended

[viral mason]

are you using wokada deiteris possibly?

[weak sorrel]

yea

[viral mason]

oh nvm ur good

[hallow thistle]

This is Deiteris W-Okada fork. https://cdn.discordapp.com/attachments/1425025269949010070/1426904432058171503/image.png?ex=696d760f&is=696c248f&hm=3bff9bb8fb32cc30f374bf21a423c692904f67d6987d607ad19f22ee53b8d6cf&

[weak sorrel]

oh did i fuck up

[hallow thistle]

What do you mean you're running the older version like v.1.5.3.18a?

[weak sorrel]

i wont lie im not sure where to check to see what verison i am on

[hallow thistle]

[weak sorrel]

[hallow thistle]

On top right of voice changer interface, it's obvious.

[hallow thistle]

You set GPU to AMD Radeon(TM) which is an integrated GPU, not the mentioned AMD Radeon RX 7900 XT which is a dedicated GPU.

[weak sorrel]

OOPs

yes

i kept switching to see what would work

like better

[hallow thistle]

Simply, a dedicated GPU is better than an integrated one.

[weak sorrel]

i saw that too yee

im just trying everything mbmb

this is the ms im at with the right gpu^^

[hallow thistle]

Set F0 det to "rmvpe_onnx".

[weak sorrel]

we at 260ms now

any advance settings i should try?

[hallow thistle]

Try set crossfade overlap to 0.15 and enable "force fp32". These settings are not needed but useful if you want audio quality to get higher.

[weak sorrel]

thank u so much!!

https://cdn.discordapp.com/attachments/1324428256103104593/1426953953483952128/IMG_7332.gif

[hardy hare]

went through the normal process of setup with colab applio, and when i open the public url the colab "completes" the task, making gradio unusable... is this a bug?

[wind yacht]

ask ai

both chatgpt/gemini can guide you through that

[hardy hare]

well. trying to work with chatgpt on this but i'm not exactly getting very far. i don't really code things, i just like using colab because it's usually simple..

[hallow thistle]

ask AI

[wind yacht]

ai's your best helper in 2026

[wind yacht]

whats ur issue mainly

ill get an answer for u

[hallow thistle]

I don't believe that.

[wind yacht]

yes it is

youd be surprised what ai can do

[hallow thistle]

Ok, AI bro.

[hardy hare]

according to gpt at least, the colab is cancelling the gradio public link upon opening it.

in as simple terms as i can explain, i click start server, i wait for the public url, i click the url, gradio says a session isn't open, and coming back to colab it displays as "complete" with a check mark.

atp gpt gave me a bunch of code to add so i probably already need to refresh anyway. idk why i even tried to see if it could help at all.

again, i don't code, and whenever colabs have bugs like this i'm usually at the mercy of other people coming out with new ones or something.

i've been using the same one (afaik) for months without issue.

it's uh.. linked to the applio colab guide.

[hallow thistle]

Try delete your current Applio RVC notebook, and then re-import the link https://github.com/IAHispano/Applio/blob/main/assets/Applio_Kaggle.ipynb to your account again.

[hardy hare]

alright, tried that, didn't seem to make any difference, but i also don't know if i deleted it properly. it ran as if it was still in.

i could try a different google account maybe?

i have spares, i'm a free user and i hate time limits

i swapped accounts and it worked, so... i have no idea why it happened to that account in particular

either way i guess i'm all good now.

[outer valley]

Is there a good tool for voice/speaker seperation ? Like if i want to remove TTS or 2-3 other speakers and only want one specific voice? I used WhisperX with a little script but i was wondering if there is a more easy way or maybe even better working then what i use.

[worn lodge]

Anybody?

[flat trail]

why is it so delayed or sometimies i aint even hearing it

[patent latch]

The tensorboard extension is already loaded. To reload it, use:
%reload_ext tensorboard
Reusing TensorBoard on port 6006 (pid 3062), started 0:02:05 ago. (Use '!kill 3062' to kill it.)
Ngrok URL: https://hiltless-marbly-brianne.ngrok-free.dev
WARNING:ngrok.tunnel_ext:error connecting to upstream error=Connection refused (os error 111)
An error occurred connecting to Discord: Could not find Discord installed and running on this machine.

• Running on local URL: http://127.0.0.1:6969

i'm getting an error like this, what's the reason?

[magic lynx]

/content/Applio
An error occurred extracting the index: need at least one array to concatenate
If you are running this code in a virtual environment, make sure you have enough GPU available to generate the Index file.

huh?

same here

/content/Applio

No wav file found.
/usr/local/lib/python3.11/dist-packages/torch/utils/data/dataloader.py:626: UserWarning: This DataLoader will create 4 worker processes in total. Our suggested max number of worker in current system is 2, which is smaller than what this DataLoader is going to create. Please be aware that excessive worker creation might get DataLoader running slow or even freeze, lower the worker number to avoid potential slowness/freeze if necessary.
  warnings.warn(
Not enough data present in the training set. Perhaps you forgot to slice the audio files in preprocess?
An error occurred extracting the index: need at least one array to concatenate
If you are running this code in a virtual environment, make sure you have enough GPU available to generate the Index file.

[random forge]

how to addapt this ai voice?

[carmine glade]

https://www.youtube.com/watch?v=trF7lTckNXc

[hallow thistle]

What AI voice?

[random forge]

Like how to use the ai voice in the game

[hallow thistle]

No, but like is it about Applio RVC (non-realtime) or W-Okada voice changer?

!howtoask

[patent trellis]

❓ How to Ask for Help

✅ Before You Ask!

1. Check Docs & Guides: Your answer may already be in the AI Hub Docs or the https://discord.com/channels/1159260121998827560/1159513888199540817 channel.
2. Search the https://discord.com/channels/1159260121998827560/1192011222023950368 : Look for existing posts that solve your issue. Do not invade someone else's post.

📝 How to ask?

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.

🚫 Prohibited Topics (We Will NOT Help With These)

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.

<:matsuripray:1159685390156967936> Community Expectations

• Be Polite & Patient: Our helpers are volunteers. You may ping the Helpers role once.
• English Only: Please keep all conversations in English.
• Don't Ask To Ask.

[cosmic epoch]

@hallow thistle is there a version of replay that works?

[hallow thistle]

No. Applio RVC and UVR5 are only your hopes.

[cosmic epoch]

what's UVR5?

[cosmic epoch]

two questions: 1, can this app be used on PC? and 2, is it possible to train rvc models with it?

[hallow thistle]

How do you not know how UVR5 supposed to work?

UVR5 doesn't use RVC voice model. It's a program that separates audio into stems.

[cosmic epoch]

can you make AI covers in UVR5?

[hallow thistle]

[cosmic epoch]

i don't understand this answer. just say yes or no.

[hallow thistle]

UVR5 is not your typical AI cover maker program, the what.

[jaunty shale]

in an online method of Tg Develop's W-Okada (Kaggle), the option to use server is greyed out, is there a reason why?

[viral mason]

No, use applio for that

[outer valley]

Anyone?

[viscid scarab]

I need some help

[real depot]

Hello everyone.
I’m testing inference in Applio with models trained by me using SPYN V2 as the embedder.
I’ve noticed that if I leave everything at the default, like the index at 0.75, the voice sounds very close to the original, but the pronunciation of the phonemes becomes a bit strange.
I also notice this with models trained with cvec: in Applio I need to tweak things to try to make it sound better, while in Replay, for example, it already sounds perfect, with correct pronunciation.
I’ve heard comments that Applio doesn’t seem to be very good for inference, but it’s the only one I know with support for SPYN V2.
Has anyone experienced this and has any suggestions?
If I lower the index to a value below 0.75, for example 0.3, the pronunciation gets better, but it loses some of the voice characteristics.

[analog obsidian]

nothing can be done, it's a side effect of the index files

[real depot]

Yes, but when I do inference in Replay with a model trained with Cvec I don’t notice this, and the index practically stays at the same value…

Although there is a difference, with SPYN V2 and the index at 0.75 in Applio the voice sounds very close to the natural one, but both with SPYN V2 and with cvec I need to make adjustments, whereas in Replay I just run inference.
Well, at least that’s been my experience with inference using both.
Replay still doesn’t have support for SPYN V2 to compare.

[analog obsidian]

maybe they're not really using the index file

bc index files will always have some negative impact in the pronunciation, it's not normal to have perfect flawless pronunciation with them

[real depot]

It could be… It’s strange, because when I import the model it shows that it has an index. But it makes sense, because when training in it the index is not generated.

[real depot]

Do you think it’s better not to use it to mitigate this problem?

[analog obsidian]

yep, index files use a quite old technology (circa 2020?), they're good at increasing the similarity between the model and the dataset but they cause weird/bad pronunciation sometimes

[swift yarrow]

@analog obsidian can u help me

i think its a simple problem but i cant seem to find a solution too

please ping me when ur here so i let u know what i need help w

[craggy bough]

https://dontasktoask.com/

[swift yarrow]

okay, i have a problem where im on intel gpu so i downloded the vc client win std beta, i extracted open file opened dist and i opend the start_http in the dist folder and it didnt load the voice changer all it said in the terminal was
C:\Users####\OneDrive\Desktop\ai voice changer\dist>main.exe cui --https false --no_cui True so i need help loading the voice changer

@craggy bough

[craggy bough]

did you get it from here https://github.com/tg-develop/voice-changer/releases/tag/b2397

https://docs.aihub.gg/realtime-voice-changer/local/tg-develops-w-okada-fork/

[swift yarrow]

i got it from here

bruh

not that

https://huggingface.co/wok000/vcclient000/tree/main

i got it from there

[craggy bough]

yeah thats pretty old

[craggy bough]

this ones recommended

[swift yarrow]

ok i need help with smth else now

and

Intel(R) Core(TM) i3-1005G1 CPU

is that reccomended to download vc client

[craggy bough]

gpu matters more

[swift yarrow]

okay my gpu it doesnt say, it says intel uhd graphcis

graphics

[craggy bough]

ok nvm

https://docs.aihub.gg/

use one of the cloud options

[viral mason]

they're so cooked

[swift yarrow]

where that at

[craggy bough]

you havent seen my hd4000 yet

[swift yarrow]

idk which one to pick, i downloaded vc client for girl voice changer

i js followed a tut

turns out its outdated

[viral mason]

ew don't help them anymore

[swift yarrow]

what

[craggy bough]

dont follow tutorials cuz theyre all outdated

and its for people with proper gpus

[viral mason]

what contraption is that lol

[swift yarrow]

okay thank you for being helpful

[craggy bough]

https://docs.aihub.gg/

[craggy bough]

hamburger menu, realtime voice changer, cloud

[austere stream]

i swear this happens every time i use kaggle and i forget how i got around it all the other times

[sudden violet]

where do i get the zip files for these?

[gusty ferry]

why is it not letting me upload any voices

[hallow thistle]

Are you sure you wanna use this specific W-Okada version that made to use Beatrice voice models especially? Because Beatrice model is rare in "AI Hub by Weights" here. "RVC (retrieval-based voice conversion)" models are more common since they give significant better audio quality, especially RVC v2.

[hallow thistle]

W-Okada voice changer or Applio RVC? What is your PC GPU? And did you follow any tutorial or guide before?

[hallow thistle]

For help, use "@ helper" instead of a specific user.

[hardy yew]

HI! I have a question regarding training RVC models, more precisely the choice of sample rate. So far I've been mostly sticking to legacy core 1.5 48kHz pretrain so that was my dataset's sample rate, but I wonder if there's a chance the models would train better on lower frequency samples (with appropriate pretrain, e.g. 32/40kHz legacy core).
Is sticking to 48kHz alright, or should I rather choose a minimum sample rate that fits the frequency spectrum of my dataset input? I haven't checked it so far so perhaps my input samples don't actually utilize the entire 24kHz frequency range

[viral mason]

That is ancient technology bro

I can see the dust on it

What gpu do u have

[analog obsidian]

models train 32k better since hifigan is not great for 40k and 48k, i know this personally because i train pretrains and i know what rvc is capable of

and its just bad for high sample rate, too many problems, like worse breaths and esses

[hardy yew]

That's some great info, thanks!

[sudden violet]

a 5090

but i dont know what version to download

[viral mason]

you should use wokada tg fork, I'll get you the links

[sudden violet]

thank you brooooooooooo

[viral mason]

https://software.muzychenko.net/freeware/vac470lite.zip

https://github.com/tg-develop/voice-changer/releases/download/b2397/voice-changer-windows-amd64-cuda.zip.001

https://github.com/tg-develop/voice-changer/releases/download/b2397/voice-changer-windows-amd64-cuda.zip.002

u will need the first one to connect your voice changer to games or discord

[sudden violet]

ok

[viral mason]

and the other two are for wokada tg fork

[sudden violet]

ok they're all downloading

whyd u send 2?

[viral mason]

for vac lite run setup 64 after extraction and for wokada tg fork after u download both extract 001 then place 002 into the folder of 001 since it cannot be extracted but needs it anyways

[viral mason]

needs both to work but idk why

[sudden violet]

so its making me reinstall vb cable?

[viral mason]

no? it's a different one that works the same but doesn't cause issues on windows

vb cable isn't as recommended since it causes weird issues on windows sometimes

[sudden violet]

ok its done on vac4lite

oh ok

then i open which one first?

[viral mason]

?

[sudden violet]

0.001?

or the 0.002?

[viral mason]

extract 001 and then place 002zip file into the folder of 001

after u do that just run mmvcserversio in the folder of 001

[sudden violet]

wait theyre the same thing no?

[viral mason]

they're two different files

[sudden violet]

oh ok

where did u find this new version from because i still look here for the main things

https://huggingface.co/wok000/vcclient000/tree/main

[viral mason]

oh that's really old

if you do this

-rt

[patent trellis]

💻 Local Realtime RVC

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

• Vonovox

A Realtime Voice Changer with similar performance to Wokada Tg-Develop Fork, with extra features, but it supports only Nvidia GPUs on Windows 10/11 unlike other options that have wider support. and without cloud options GUIDE

• Wokada Tg-Develop Fork

A personal fork (modified version) of Wokada Deiteris Fork, it just adds some Quality of Life improvements to it like supporting Spin Embedder and Audio Effects. Don't expect too much about it since the creator made it originally as a personal project. GUIDE

• Wokada Deiteris Fork

Deiteris' fork (modified version) of wokada that doesn't get updates anymore. GUIDE

⚔️ Wokada Tg-Develop Fork vs Vonovox

For Windows Nvidia, Both Wokada Tg-Develop fork and Vonovox have similar performance & quality. Users should read the pros and cons for both and choose based on their differences, such as UI and Vonovox's paid effects.
Read Wokada Tg-Develop Fork Pros&Cons & Vonovox Pros&Cons

⛔ Outdated/Discouraged

These options are not recommended for use.

• Original Wokada

Not suggested, older versions in youtube tuts are even way worse. GUIDE

• RVC GUI Mainline Realtime

The program is worse compared to the ones above, and much less updated. GUIDE

[viral mason]

then u can find the three main voice changers used now

I would recommend you vonovox since your gpu can definitely handle it but it's currently in a beta stage adding new things and I'd rather give u the newest version of it than the current version that has less features than the beta

[sudden violet]

how do i pick that?

[viral mason]

what?

[sudden violet]

the vonovox

[viral mason]

if you'd rather use that I can give you the download

you'd need to delete the one I just gave u tho to save on space

[sudden violet]

whats the difference

[viral mason]

there's too much stuff but it's basically the best one of the three

[sudden violet]

oh ok

[viral mason]

if you'd like to use it here's the download
https://github.com/dr87/Vonovox/archive/refs/tags/v1.6.9.zip

just run setup first then run start

they're .bat files

[sudden violet]

do i extract it

[viral mason]

yup

[sudden violet]

ok so how is it the best?

[viral mason]

it produces more natural speech, has a lot of different features to block out bg noise messing with the model so weird sounds don't come out or give random voice cracks

there's an optional paid effects like adding reverb or a 8bit effect although you can easily bypass that by using a DAW like fl studio or just using voicemod

it's mainly to support the creator which is always nice

[sudden violet]

so its the best version of the three?

whats the third one?

[viral mason]

the third one is a little outdated compared to the othe two, it's wokada deiteris fork

basically the original wokada but made more up to date code wise and has a few quality of life features, tho the other two have them as well

[sudden violet]

oh okkkkkkkkkk

i like the web version that they made for the rvc one

easier

[viral mason]

so wokada tg fork?

that one runs on a web browser

vonovox doesn't tho

[sudden violet]

yeah

so waht do i do when its done

[viral mason]

depends, which one are you going to use

[sudden violet]

vonoxo

i was waiting for the setup to finish

[viral mason]

alrighty, after the setup is done run the file called start

[sudden violet]

ok it worked thanks so much gangyyy

[viral mason]

you're welcome!

if u need more help ask me here

btw to import a model just download one from here, extract and press one of the empty slots and insert the .pth file

and underneath is a box that says index I believe

import the index file for the same model in that

[sudden violet]

on vonovox

?

[viral mason]

yep

[magic lynx]

im finishing revamping applio notebook ||which wont be even accepted but idgaf, they dont want autotune infer fixed||
and i wonder which badge set is the best
i think 2, 3* or 4
* - uses website colors

[fringe swan]

How do I train a voice for free

@magic lynx can you help me

[shadow sparrow]

anyone know if https://colab.research.google.com/github/iahispano/applio/blob/master/assets/Applio.ipynb?authuser=1&pli=1#scrollTo=IlM6ll0WDuOG
is having issues? i keep recieving "No interface is running right now" message and can't use it

[viscid scarab]

Guys I need some help

Help me

[hallow thistle]

Say something.

[hallow thistle]

Applio RVC can train a voice model. I don't know what to say about this.

!howtoask

[patent trellis]

❓ How to Ask for Help

✅ Before You Ask!

1. Check Docs & Guides: Your answer may already be in the AI Hub Docs or the https://discord.com/channels/1159260121998827560/1159513888199540817 channel.
2. Search the https://discord.com/channels/1159260121998827560/1192011222023950368 : Look for existing posts that solve your issue. Do not invade someone else's post.

📝 How to ask?

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.

🚫 Prohibited Topics (We Will NOT Help With These)

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.

<:matsuripray:1159685390156967936> Community Expectations

• Be Polite & Patient: Our helpers are volunteers. You may ping the Helpers role once.
• English Only: Please keep all conversations in English.
• Don't Ask To Ask.

[viral mason]

I think they're an NPC stuck repeating the same thing

[hasty solstice]

im here

[tame oracle]

okay

what voice changersing

*u using

[hasty solstice]

Voice Changer Client Demo

[tame oracle]

Dats old as balls

-rt

[patent trellis]

💻 Local Realtime RVC

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

• Vonovox

A Realtime Voice Changer with similar performance to Wokada Tg-Develop Fork, with extra features, but it supports only Nvidia GPUs on Windows 10/11 unlike other options that have wider support. and without cloud options GUIDE

• Wokada Tg-Develop Fork

A personal fork (modified version) of Wokada Deiteris Fork, it just adds some Quality of Life improvements to it like supporting Spin Embedder and Audio Effects. Don't expect too much about it since the creator made it originally as a personal project. GUIDE

• Wokada Deiteris Fork

Deiteris' fork (modified version) of wokada that doesn't get updates anymore. GUIDE

⚔️ Wokada Tg-Develop Fork vs Vonovox

For Windows Nvidia, Both Wokada Tg-Develop fork and Vonovox have similar performance & quality. Users should read the pros and cons for both and choose based on their differences, such as UI and Vonovox's paid effects.
Read Wokada Tg-Develop Fork Pros&Cons & Vonovox Pros&Cons

⛔ Outdated/Discouraged

These options are not recommended for use.

• Original Wokada

Not suggested, older versions in youtube tuts are even way worse. GUIDE

• RVC GUI Mainline Realtime

The program is worse compared to the ones above, and much less updated. GUIDE

[tame oracle]

first guide is vonovox

[hasty solstice]

I figured it could be the thing itself

[tame oracle]

is the best of those three

[hasty solstice]

awsh man i cant send media

but does it say

For Windows
Download this: VAC Lite (Virtual-Audio-Cable by Muzychenko)??

[tame oracle]

yes

that is the recomended one

[hasty solstice]

is there a step by step on how to set this up completely?

[tame oracle]

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

dis guide

:D

[hasty solstice]

thank uu!

i see it now

[tame oracle]

yess ofc

[hasty solstice]

ill see about what i could do and hopefully itll stop making me sound like a robot

[tame oracle]

im sure it will, its a lot better code wise

[hasty solstice]

@tame oracle ive gotten far to the point where the files is already exported and i installed the "setup64", what now?

[tame oracle]

for the VAC lite?

[hasty solstice]

Do i open that?

i think so

[tame oracle]

yes

[hallow thistle]

[hasty solstice]

its opened now

its just

wait

is this it?

[hallow thistle]

Words alone won't always imagine as image. For better understanding, there's screenshot.

[hasty solstice]

i just now got image perms

hold on

[hallow thistle]

That's not the actual program, it's "help" doc for Virtual Audio Cable.

[hasty solstice]

where do i find that?

[hallow thistle]

This is the actual control panel for Virtual Audio Cable.

[hasty solstice]

OHHHH

[hallow thistle]

No, you don't need to set anything in this control panel.

[hasty solstice]

okay so im basically set?

[hallow thistle]

This is what you gotta do. https://cdn.discordapp.com/attachments/1159290139609137264/1459041990804115497/image.png?ex=697056be&is=696f053e&hm=f9f501b5af541dd9aa4c577d76111b767982c91367065446636f86a7a9b3546a& https://cdn.discordapp.com/attachments/1159290139609137264/1459041991290650760/image.png?ex=697056be&is=696f053e&hm=f9cad34a51bd2908bbcab47a043851feaab807b00b82f640845f114a2b9e6797&

[hasty solstice]

All of them are now checked

is there an option where like

okay im going to sound so slow but like possibly add the models?? 😭

[hallow thistle]

This is Tg Develop's W-Okada fork.

This is where you upload a voice model to the voice changer.

[hasty solstice]

so i have to go to that website?

[hallow thistle]

No way, you're looking for Vonovox, not the W-Okada fork.

By the way, if you wish to stay for Vonovox, I can help you a bit about Vonovox. But if you'd like to try that one W-Okada fork, the easier interface, I can send you links.

[cedar sky]

Hi guys, Ive spend all dat yesterday to try to find cheap api to kling models video provider. Does anyone know the workflow or any api or apy way to get kling video generation price less than 0.5$ per 5 sec? 🙏

[spark citrus]

-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

[viscid lily]

anyone know how to fix this issue of fuzzy output? I'm new so I don't really know what I'm doing yet
RTX 4060 laptop GPU
I'm using airpods but nothing is using airpods input, so I know it's not bluetooth distortion
This is the voice im using:
https://discord.com/channels/1159260121998827560/1403422896680079390

[chrome geode]

So I've got the "generating buffers" thing goingon and I can't find a single way to fix this on discord or online

Anyone have any actual assistance with this?

[sturdy oracle]

Hey guys
I'm new here
I joined because I don't have a good idea of how gen ai models work, but find their use of copyrighted materials problematic
I want to create software that efficiently and quickly poisons Gen AI through encoding music, audiobooks, images etc
Does anyone here have an idea of how I could go about doing that?

[shy torrent]

Hey guys any quick help I installed via 'W-Okada Fork Guide' but whenever I try a model my voice never changes anyone has a fix ? GPU 5070TI

[viral mason]

which guide did you use?

[shy torrent]

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

this one

downloaded the 5000 series one

[viral mason]

hmm I'd switch to either vonovox or wokada tg fork tbh

both are better than deiteris

-rt

[patent trellis]

💻 Local Realtime RVC

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

• Vonovox

A Realtime Voice Changer with similar performance to Wokada Tg-Develop Fork, with extra features, but it supports only Nvidia GPUs on Windows 10/11 unlike other options that have wider support. and without cloud options GUIDE

• Wokada Tg-Develop Fork

A personal fork (modified version) of Wokada Deiteris Fork, it just adds some Quality of Life improvements to it like supporting Spin Embedder and Audio Effects. Don't expect too much about it since the creator made it originally as a personal project. GUIDE

• Wokada Deiteris Fork

Deiteris' fork (modified version) of wokada that doesn't get updates anymore. GUIDE

⚔️ Wokada Tg-Develop Fork vs Vonovox

For Windows Nvidia, Both Wokada Tg-Develop fork and Vonovox have similar performance & quality. Users should read the pros and cons for both and choose based on their differences, such as UI and Vonovox's paid effects.
Read Wokada Tg-Develop Fork Pros&Cons & Vonovox Pros&Cons

⛔ Outdated/Discouraged

These options are not recommended for use.

• Original Wokada

Not suggested, older versions in youtube tuts are even way worse. GUIDE

• RVC GUI Mainline Realtime

The program is worse compared to the ones above, and much less updated. GUIDE

[shy torrent]

ill try rn

[viral mason]

vonovox is currently the best of the three but I personally use tg fork for the fast model swapping time

if u do use tg fork u will need the 001 zip and 002 zip

[shy torrent]

i'm trying vono first

[viral mason]

alrighty, all u need is just to download the current version 1.6.9 and run setup first, the run start

[shy torrent]

ty bro

[viral mason]

np!

[shy torrent]

also i've seen people talking abt malware inside of w-okada soft were there triping or what ?

[viral mason]

if it isn't from here it most likely is a scam

a lot of people are scummy and try and trick people with versions of old wokada that have viruses

[viral mason]

these are the only trusted sources

[shy torrent]

yeah i figured

alr ty

[frigid stone]

Is there a standardized regular speaking audio sample for both male and female voices that can be used for testing inferences?
I'm looking for something that covers a good amount of speaking sounds just to know if a model works well

[brittle wing]

yo anyone ?

i wanna learn as much as i can about aiii

[viral mason]

what version of wokada are u using?

also u cannot upload it if it's a json file sadly

[viral mason]

not really but I do have these

that's ollllllld

[frigid stone]

Thanks for these, it's good to know I'm not the only one just ripping audio clips off whatever shows up on YouTube

[viral mason]

no problem! if you'd like more talking audio or just singing samples I got plenty

what gpu do u have btw, since ur using a really old version of the voice changer u should upgrade

[viral mason]

oki doki

[abstract comet]

What’s this new model you speak of?
I’d be really interested to learn about it

[analog obsidian]

failed experiment

[abstract comet]

What’s the latest architecture you recommend

Specifically for singing models

[abstract comet]

Also what’s the other rvc server you speak of