✨│ai-help

[gusty cedar]

I don't dual boot.

Plus, I all of my stuff is on windows

[low shard]

huh, you said you were on Linux Mint

[gusty cedar]

Fugg

My brain wasn't braining

I just woke up.

Sorry, I meant to say all my stuff is on Linux mint

@low shard imma check the guide u gave later

[low shard]

ohh

unfortunately none of the wokada forks contain a precompiled build for Linux with Intel Arc Support, you can either:

• use it on CPU (easiest local for your case, but probably high latency)
• use the dml windows build via a VM (with GPU Passthrough) or dual boot
• use cloud methods
• if you're a coder (i can't test this but based on research): install the Intel Arc Compute Drivers, overwrite the default pytorch installations to use XPU, modify the source code to detect XPU

I think you're the first user on Linux with Intel Arc GPU

[gusty cedar]

Damn.

Are there any other voice changers out there that AI is optional?

[winged adder]

Rx580
Kobold.cpp Linux no cudA
Vulkan driver
Model qwen3 tts

Hey everyone, I’m running into a frustrating issue with my local TTS setup and could use some insight from those more familiar with Vulkan/AMD offloading.
The logs show that Vulkan is detected, but my GPU (RX 580) is sitting at idle while my CPU is pegged at 100%.
The Problem
Even though the log says:
ggml_vulkan: Found 1 Vulkan devices: AMD Radeon RX 580
The actual inference backends are refusing to move over:

• TTSTransformer backend: CPU
• AudioTokenizerDecoder backend: CPU
  As a result, I’m getting about 0.07x – 0.08x realtime performance. It’s painfully slow.
  My Specs & Config
• GPU: AMD Radeon RX 580 (Polaris)
• Software: KoboldCpp / Qwen3-TTS
• Settings: gpulayers=-1 and usevulkan=[0]
  What I’ve Noticed
  The log also mentions fp16: 0 | bf16: 0. I suspect my RX 580 might be too old to support the specific math required for these models, or perhaps the Vulkan implementation for this specific TTS model just isn't there yet.
  My questions for the experts:
• Is the RX 580 simply a "dead end" for this type of inference because it lacks FP16/tensor cores? But It work on llama.cpp
• Is the TTSTransformer backend in KoboldCpp currently CPU-only for Vulkan users?
• I dont want switching for ROCm actually help an older Polaris card, and i Will not get new RTX card for CUDA!
  If anyone has managed to get GPU working on older AMD hardware for TTS, I’d love to know how you did it!

https://sharetext.io/t8mghj64

[low shard]

about the making your own release option, maybe you could also check out https://github.com/tg-develop/voice-changer/tree/master-custom/.github/workflows and see if you could edit the workflows to add linux mint on intel arc gpu support

Are there any other voice changers out there that AI is optional?
Vonovox uses the same RVC technology but is currently closed sourced freemium and on Windows on modern Nvidia GPUs only

So you can either:

• try https://github.com/wwmm/easyeffects (which seems like FOSS Voice Mod alternative with basic audio effects, like Pitch shift)
• try one of the methods I told you before about running the wokada forks (like tg-develop)

[gusty cedar]

I can't make my own release version 😭

Computer science has lost its value to AI, meaning AI can just write all the code and push the release...

Then I chose electrical engineering, I already have my own fair share of workload.

I will have to rely on EasyEffects..

Thank you so much :<

[fringe swan]

I need help

I am looking for a image to video

[low shard]

i mean an easier way to run it was using a Windows VM with GPU passthrough or using cloud

Computer science has lost its value to AI, meaning AI can just write all the code and push the release...
About this, i wouldn't say it lost its value, vibe coding can be harmful in some sectors since it's just token prediction at the moment, tho its true that vibe coding can be helpful in some cases, I saw it help with things like building Emulators, but anyways it's not what the convo is about

EasyEffects will work fine if you don't care about sounding like a specific character (like Homer Simpsons)

You're welcome, unfortunately your setup is kind of niche so not many programs support it

[gusty cedar]

Well. Sort of Don't care about sounding like a specific character, but be able yo shift my voice to sound either different tones of male and female, or just make goofy voices

[low shard]

ohh then it's all good for you then, enjoy it

you didn't need an AI Voice Changer to begin with for that so don't worry

[low shard]

are you looking for an open source model you could run locally or cloud? what's your gpu?

[fringe swan]

I am trying to make a music video

I am a music artist

[gusty cedar]

Oh yea I remember.

For context, my home PC with the INTEL ARC, I tend to run AIs like stable diffusion or Ollama models

However, when I bought a new laptop last year, I was unaware and discovered just yesterday that the CPU of my laptop (RYZEN 5 8640HS) has an NPU and I am wondering what AI models could I be able to run on the NPU

[fringe swan]

Can you make a music video for me

[gusty cedar]

Bro. I don't know how.

I can only do images

[fringe swan]

I already have a character

[gusty cedar]

You do realize you could try asking Gemini or GROK

[fringe swan]

I don't have subscription

[fast scarab]

Hey, quick question: for AI covers, is AI Cover Maker good enough, or is it better to use a full RVC workflow with Applio + UVR for higher quality?

[tame oracle]

People that dont know about free teirs be like:

[low shard]

LLMs surely are more popular and have support for your setup, it's just that not all technology does

I haven't head of many projects having NPU support, I remember right now: https://github.com/rupeshs/fastsdcpu (CPU + Intel Arc + NPU support, nice)

[fringe swan]

Do you know any

[tame oracle]

Know any what

there's many things i know

[gusty cedar]

Bro I literally said Gemini or GROK 🤦‍♂️

[fringe swan]

I am trying to make a music video for my new song

[low shard]

AICoverMaker is the easiest automated way, the manual setup with Applio and UVR gives you more control to be sure you like your output

[fringe swan]

They ask me for subscription

[gusty cedar]

Cool!

I will check it out

[low shard]

there's no company that's going to give you free unlimited chatbots, they would all just go bankrupt

[tame oracle]

I don't know about AI videos much

[low shard]

so there's neither any company that will give you free unlimited image to video services, they would just burn money

you either use the free trials, pay, or buy your own beefy pc to run open source alternatives

[fringe swan]

Do you know how to make fight scene video or sad

[tame oracle]

I just said i don't know about AI videos much

[fringe swan]

It's not a.i

Just edit

[tame oracle]

I still don't know how

i dont do editing

[low shard]

the other users aren't helpers, and this isn't an editing server

[low shard]

you could also try llama cpp on your NPU https://github.com/ggml-org/llama.cpp/issues/1499, not sure about that tho

[spare dove]

Please, how do I extract the files pth and index

[viral mason]

Are you trying to use the files in real-time or for applio?

[primal river]

Hello everyone!

is deepseek, z.ai, qwen models are still good to use as api with n8n? anyone using it?
I found a website which is freely providing api of all these platforms for free and unlimited for now. I don't know how much time they will provide it. But not sure if these models are still worth it to use or not.

[wraith coral]

Laptop gpu rtx 4050
Windows 11
AI covers and realtime voice change

[low shard]

did you download an rvc voice model from #1175430844685484042 ? just unzip it

[low shard]

-realtime

[patent trellis]

🔊 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

• Local 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.

• Local Guide
• Cloud Kaggle Guide
• Cloud Google Colab Guide

• Applio Realtime

A Realtime Voice Changer with similar performance to Vonovox & Wokada Tg-Develop Fork, with extra features.

• Local Guide
• Cloud Guide(s)

• Wokada Deiteris Fork

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

• Local Guide
• Cloud Kaggle Guide
• Cloud Lightning.AI Guide
• Cloud Google Colab Guide

⛔ 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

[low shard]

you might want to check out Applio, it works for both of your needs

[torpid jackal]

would you guys say that kaggle is good?

[viral mason]

I stand by it being the best for cloud since it's easy for new users and also has a decent free user limit of 30 hours per week

Google colab is just horrid with it giving a random time between like 2-4 hours for free users and lightning is kinda just confusing as hell

[fringe swan]

I will figure it out and fuck off

[oak wadi]

I need help can someone help me create AN RVC model for lakyus alvein dale aindra from the anime and light lovel seriees overlord please

[torpid jackal]

thank you, man! i'll defo use this advice to my advantage i love your models btw, i use a lot of them

[viral mason]

Ty so much! Glad I could help btw ^^

[oak wadi]

i want to request an rvc model of lakyus alvein dale aindra from overlord please

[wintry scaffold]

hey i was coding in one vibe coding tool, and i want to shift it to another vibe coding tool. how do i do that?

[steady urchin]

does anyone have the old model that's trained on dial up?

[steady urchin]

i don't remember the name

[viral mason]

just go here bro https://discord.com/channels/1159260121998827560/1175430844685484042

[safe pine]

Hey im using vonovox

and i was wondering how i can adjust the chunk rate

[viral mason]

it's called block size in vonovox, just adjust it with your mouse

[wraith coral]

link?

[reef isle]

Guys what's the best voice changer ???
I tried
With a good settings and everything
And the result was so bad
Any one can help
?

[fossil dune]

where can i get the latest version of w-okada

[winged adder]

Im looking for It too
Which GPU do you suggest tò run It, how many vram?

[normal zinc]

anyone able to help me re-install RVC i'm not sure why its not working. have i installed it wrong?

[spare dove]

Hey friends, listen, I've done everything, but when I try to extract the pth and index files and fill in the last fields, no black text appears when I click on them.

Is there a solution to this?

I wasn't given permission to post on Vioes-Models. I want it for my personal use... I really hope to train my voice.

[steel mortar]

HEY

[spare dove]

Welcome

[steel mortar]

THANKS

[spare dove]

You're welcome. Can you help me?

[steel mortar]

IN WHAT ?

[spare dove]

The following is above your first message.

I'll go eat and then come back

[steel mortar]

SORRY BRO I AM NOT REALY SURE THAT LAST I CREATED AN INDEX FILE

SO I AM PRETTY SURE USELESS FOR THIS

WHAT ABOUT U CAN U HELP ME ?

I WAS USING GROK AND NOW IT'S SAYS THAT THE AUTHENTICATION FAILED SO ALL THE THINGS I WAS DOING ARE PROBABLY GONE WITHOUT MY GROK ACCOUT , I WANT TO KNOW IF U KNOW ANOTHER FREE AI GOOD AS GROK TO COMPLETE MY WORK

[normal zinc]

@spare dove may i dm message you?

[spare dove]

I am so, so, so sorry for what is happening to you

[spare dove]

Of course I'm here

[low shard]

Be polite to everyone here or you will get higher moderation actions.

[low shard]

#1159289738314919936

[normal zinc]

i put a thing in help forums.

[white shadow]

why do you speak in yelling? ( @viral mason for the joke. )

[low shard]

there are different programs, elaborate:

• your pc gpu
• your pc os
• the tutorial link you're using

[spare dove]

Please, I'm on the phone. Please help me. I'm dreaming of that.

[viral mason]

can't use it on phone

[spare dove]

truly?

Can you train my friend's voice? Please, she's disabled and dreams of it.

What about Abellio?

Please answer

okay...

[viral mason]

I'm sorry I don't take commissions but you can learn how to train models yourself on applio

nah, just look through https://discord.com/channels/1159260121998827560/1175430844685484042 or https://voice-models.com

I don't train regular people voices just game or show characters

[spare dove]

Yes, please

[wide perch]

Using tg-develop's voice changer and it's not working. Using my default in and out to test and even that isn't working

Passthrough doesn't work

So idk what to do

https://i.imgur.com/T9iyB30.png

someone tell me what is wrong pls

[gleaming wasp]

Does anyone know how to reduce the delay on Vonovox?

[viral mason]

gotta turn down then block size

[undone aspen]

Can someone recommend some good llm subscriptions? I was on chatgpt free -> go -> pro (20 bucks). I mostly use it for learning math in the conversation mode (uploading pictures of the problem, then engaging the mode). Even after upgrading to pro, i seem to run out of advanced voice. Is there any llm that's more generous in voice and is good with math (converstational as well). Thanks!

[safe pine]

hey if i am not wrong the index file is for training the model right?
so like if there was an index file trained to speak a specific language which i speak? will it be better then?

and if yes how hard is it to train the model

[stiff cloud]

whats the best boy voice

[hallow thistle]

An index file stores accent of a voice model, typically achieved before or after model training. The actual RVC voice model file is in "pth" format. To train one, you would go for a dedicated RVC software (like Applio RVC).

[hallow thistle]

Don't wanna know.

[stiff cloud]

why

[low shard]

You can't use realtime on the phones

[low shard]

You need the best boy voice to troll/catfish right?

[hallow thistle]

You have got help about the voice changer multiple times earlier. Some settings look wrong there because you set chunk number too high, extra value should be "2.7" not 0.5, pitch extraction model should be "RMVPE" not rmvpe_onnx. You wouldn't set both output and monitor as the same speaker either.

[deep wasp]

i just use okada

anyone having problems with the weights website? i cant seem to download anything there atm, even though i could download normally a few hours ago

[low shard]

are you trying to e girl troll / catfish too like the other user before lol?

[deep wasp]

ya there was a dr strange and ultron voice model that i wanted to use during ranked lol

[olive monolith]

Hey, can someone send me the link to the correct Colab notebook to use to train an RVC model?

[olive monolith]

I just need the colab link, that's all

[pseudo sail]

guys

i cant install the vb virtual audio cable

always an errpor

[low shard]

elaborate:

• your pc gpu
• your pc os
• what you're trying to do, like AI Covers, LLMs, E Girl Trolling / Catfishing or Roleplaying as goku
• the tutorial link you're using

[low shard]

it would be better you first tell your pc gpu and os to see if you could do train RVC models locally

if you don't care about that, check this guide https://docs.aihub.gg/rvc/cloud/applio-cloud/

[low shard]

oh so just roleplaying as characters right?
what tutorial link are you using?
btw weights is shutting down on march 31st 2026

[summer ingot]

guys i got the ai voice changer now but does my mic have to be good in order to work flawless

[kindred kernel]

Which roles or Permission do i need to be able to text in the voice-models Channels? and how do i get them?

[viral mason]

No you can't train an existing model

[viral mason]

You have to first train a model and then submit it here https://discord.com/channels/1159260121998827560/1452279591317278840, once it's approved which can take a while you can then post models there and I believe that also lets you text there since you'll have the model maker role

Tho I'm not 100% sure you have to do that just to text there specifically

[viral mason]

Where'd you get the voice changer and what gpu do u have (Nvidia or AMD) btw your mic doesn't have to be perfect just make sure there's not a lot of background noise or people talking being picked up by your mic

[hardy yew]

I think so, I couldn't comment on models without the role either.

Maybe it's a limitation of discord

Or maybe bad permissions? 🤔

Wonder if it's possible to enable commenting but disable posting by default

[viral mason]

That's probably possible but I'm not sure, I've dealt with role management before in some servers but it's a little confusing

Maybe it was disabled to keep people from just going in there and being like "boo this stinks" and making someone quit model making

[hardy yew]

Dunno if it's like that on purpose but I think commenting should be enabled

Model feedback, discussion etc

[viral mason]

True yea

I've used my brain too much typing those big blocks of text I think it's resetting

I just woke up too lol

[hardy yew]

Meanwhile 5pm here

[viral mason]

Good afternoon Capy

Or is it evening

[hardy yew]

Doggo walk, gotta eat something, and then I can rest and enjoy the weekend xd

[viral mason]

We love Fridays

[hardy yew]

Not dark yet! Late afternoon, I'd say

[viral mason]

Nice!

[gritty rune]

Hi guys and gals, can anybody explain to how I can runs Stable Diffusion with muli-GPU?
I have two 3060, I read that comfy supports multi-GPU, but I'm mainly interested in automatic/forge, because of inpainting.

[kindred kernel]

thxthx

[winged adder]

Im looking for ready tò use voice assistant with theese feature: speech to text ,LLM text_gen ,text to speech pipeline with (STT → LLM → TTS)
I setup the following tool

whisper.cpp

llama.cpp (other doesnt work )

qwentts.cpp It use kobold.cpp only (other doesnt work )

distributed multi machine linux vulkan amd

It must be Separate PCs / distributed nodes

AMD + Vulkan only (VERY IMPORTANT for me)

4. Ready-to-use stacks

example

PC1 (AMD GPU Vulkan)

→ whisper.cpp server

PC2 (AMD GPU Vulkan)

→ llama.cpp server

PC3 ( AMD gpu vulkan )

→ kobold.cpp / Qwen-TTS server

[safe pine]

i see

the thing is im facing issues when using the rvc model and speaking in my local language

[safe pine]

hmm

[low shard]

elaborate:

• your pc gpu
• your pc os
• what you're trying to do, like ai covers, e girl trolling / catfishing or roleplay
• the tutorial link you're using

[low shard]

Automatic1111 stable diffusion web ui latest commit was 2 years ago

https://www.reddit.com/r/StableDiffusion/comments/1pswsw6/is_automatic1111_still_used_nowadays/ most people that don't want to use comfyui bc it's harder to use switched to forge

it doesn't seem currently possible to do multi gpu with forge based on https://github.com/lllyasviel/stable-diffusion-webui-forge/issues/48

I think your only way might be ComfyUI https://github.com/pollockjj/ComfyUI-MultiGPU

@simple ore iirc you used forge? not sure if you may know anything about this topic

[safe pine]

hey nick

u free?

can u come on vc i wanted some help tweaking my vonovox

[low shard]

i can't help in vc, if you want you can elaborate via text

[simple ore]

long time ago with AMD

[charred kiln]

hi guys, do yall know an opernsource ai that generates lip syncs videos from images?

[low shard]

iirc you can do that with wav2lip or facefusion

[winged adder]

How tò use vulkan driver tò run this
https://github.com/ShayneP/local-voice-ai/tree/gpu_enabled

[winged adder]

what size llm should big enough 2b 4b 8b 14b for the following task<capabilities>
The system acts as a specialized linguistic reconstruction engine. It possesses the ability to parse disjointed keywords, infer logical context, and synthesize them into a singular, cohesive, and grammatically standard sentence.
</capabilities>
<behavior>

• Tone: Maintain a strictly flat, neutral, and expressionless persona.
• Style: Avoid all unnecessary chatter, warnings, disclaimers, preambles, or conclusions.
• Constraint: You must generate exactly one sentence per input. Do not provide multiple variations or additional explanations.
• Logic: Interpret the relationship between keywords to create a realistic or contextually appropriate scenario.
  </behavior>
  <output_format>
  All responses must be wrapped in structured XML tags. No text should exist outside of these tags.
  Format: <result> [Reconstructed Sentence] </result>
  </output_format>
  Examples:
  Input: saw bear webt camping Majestic
  Output: <result> I saw a bear last time I went camping, and it was majestic. </result>
  Input: Snake terrariun naturecenter
  Output: <result> There is a snake inside a terrarium located at the nature center. </result>
  Input: car road fast mountain
  Output: <result> A car traveled quickly along the winding road through the mountain pass. </result>
  </result>

[regal dew]

How yall make voice models

[low shard]

it depends, what's your pc gpu and os?

[cosmic epoch]

How do i get a permission to post in #1175430844685484042 ?

[viral mason]

You have to submit your trained here https://discord.com/channels/1159260121998827560/1452279591317278840, once it's approved which can take a while you can then post models there and I believe that also lets you text there since you'll have the model maker role

[cosmic epoch]

umm, that link you gave me, i have no acess to it

[viral mason]

odd, maybe there is a role you're missing

[cosmic epoch]

what do you mean by "role i'm missing"?

[viral mason]

these

[cosmic epoch]

is this where the link you tried to send me leads to?

[viral mason]

no that screenshot is from the channels and roles section

[low shard]

@viral mason you linked the wrong channel

the submission channel can be view only by users who are already model makers as they have to rate them

@cosmic epoch you need to check #📤│model-maker-role

[viral mason]

oops

[cosmic epoch]

wait... i don't get it, how do i share models there without a file?https://discordapp.com/channels/1159260121998827560/1453581938538315847

[viral mason]

you can use the voice models using a realtime voice changer, depending on what gpu u have (Nvidia or AMD) I can get u what u need

what are you planning on using it for?

[viral mason]

Anything off yt is outdated

But uh wdym troll..

Why are people like this

<@&1159293140440723499> kill him

Catfishing is not good so uh yea good luck getting what u need

Ahem

Pretending to be a woman to get stuff from people

Yeah..

💔

[hallow thistle]

[gray dagger]

how many epochs should i aim for when using a 17 minute dataset

[plucky sleet]

so if i wanna get a model from weights that isnt on #1175430844685484042 what do i do?

i wanna get this dj drama model but its not letting me

https://www.weights.com/models/cm7ziq6nkc50ljy14429qfcuz

bc i dont have an acc and they disabled sign ups

[viral mason]

if you don't have an account and aren't signed in you're screwed

[plucky sleet]

i see, could you help me access it please

[gray dagger]

i got you, imma dm you the model

[plucky sleet]

cheers tysm!!

[gray dagger]

yw!!

[viral mason]

W ai hub member

[hallow thistle]

If you ever feel like voice models on Weights.com were outdated and wanted newer ones as replacements, you can request in #1159289738314919936.

[wooden void]

Hello! I was here before weights created. And soon it's shutting down sadly...

1. Can anyone suggest any similar websites for creating character voice covers?
2. If there are no websites, what programs are usually used to create AI covers?
3. How powerful does a PC need to be to create such covers? My laptop like, 15 years old... And have windows 7... What are the chances that I can make something with these technical specifications...
   Oh, or maybe not only PC! I also have android phone. It's more powerful than my PC i guess xd

I asked accidentally it in general chat, oopsie. Maybe someone else could talk to me. I would be very grateful for help in answering these questions!

[smoky plaza]

i need help with vcclient

i have everything setup with the VB cable as well and stuff and the right settings but NONE of the rvcs r detecting my input my voice yet the beatrice does detect my voice why isnt the RVC working

[viral mason]

What you're using is really outdated

[smoky plaza]

?

[viral mason]

What is your pc gpu? (Nvidia or AMD)

[smoky plaza]

nvidia and im on the MMVCServerSIO_win_onnxgpu-cuda_v.1.5.3.18a

[viral mason]

Yea that is the original w-Okada super old

[smoky plaza]

oh

what am i supposed to use

[viral mason]

Most likely you got it from a yt tutorial or just the official link

[smoky plaza]

yes yt tut

so what do i use then

[viral mason]

I'd use Vonovox, it's the current best for real-time on Nvidia

[smoky plaza]

free>

[viral mason]

I'll get u the downloads

Yea it's free

[smoky plaza]

ok

[viral mason]

I use it as well

[smoky plaza]

arent the voice models in this server meant for okada only?

[viral mason]

The models are just files that go to pretty much any software that does real-time ai with rvc

[smoky plaza]

ohh okay

[viral mason]

Even can be used on a crappy site like voice.ai

https://huggingface.co/dr87/vonovox/resolve/c8034f5f6d50648a8109bb4f847182362e2b779b/Vonovox_beta_17_11.zip

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

[smoky plaza]

do i use VB virtual cable for this too

[viral mason]

Anyways here's the two downloads you need, first one is Vonovox which is the voice changer and second one is a virtual audio cable called vac lite which is recommended over VB cable since it's less likely to mess up on windows

Yea you'll need a virtual cable for this to work on games or discord

[smoky plaza]

ok

thanks

if i have any issues ill plop them here i guess

[hallow thistle]

This is where I used to run "Applio RVC". While your old laptop couldn't run Applio RVC locally, this option might work.

[wooden void]

Thank you! I'll look into it, thanks for the tip.

I'm just looking for options with what I have. I know I don't have much, but at least I have something

[viral mason]

Btw if I do disappear it's due to it being late and me falling asleep

[smoky plaza]

i have one question what is index and do i need to bother with it

[hallow thistle]

An index file stores accent of a voice model.

There are different contexts on whether "index rate" should be enabled; index rate of non-zero value in any realtime voice changer program (W-Okada, Vonovox) will use an accent of that voice model, but then it would cause CPU to spike up more which is not very ideal for latency-sensitive scenarios. In non-realtime RVC (like Applio RVC), setting index rate to any value would also use CPU as well, but the performance is usually unnoticable.

[smoky plaza]

ok thanks alot

[summer ingot]

No its my laptop mic and it picks up a lot of noises is it because that?

[signal mauve]

hello, i need voice changer okada

[signal mauve]

to download it

[formal kindle]

rvc v2 still working?

[hardy yew]

why would it stop working? 🤔

[hardy yew]

what's your OS and GPU?
Do you want realtime conversion or inference on audio files?

[signal mauve]

rtx 5060

realtime

[hardy yew]

vonovox or tg develop

-rt

[patent trellis]

🔊 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

• Local 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.

• Local Guide
• Cloud Kaggle Guide
• Cloud Google Colab Guide

• Applio Realtime

A Realtime Voice Changer with similar performance to Vonovox & Wokada Tg-Develop Fork, with extra features.

• Local Guide
• Cloud Guide(s)

• Wokada Deiteris Fork

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

• Local Guide
• Cloud Kaggle Guide
• Cloud Lightning.AI Guide
• Cloud Google Colab Guide

⛔ 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]

What do you use the voice changer for?

[signal mauve]

for girl,zoro.Luffy

i need

w okada

[hallow thistle]

[signal mauve]

any help?\

sry my eng very

bad

[hallow thistle]

If you mean W-Okada, download Tg Develop's W-Okada fork, 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.

[signal mauve]

when i extract its show me like this

[hallow thistle]

Make sure to download both .zip.001 and .zip.002 into the same folder.

[green quail]

Hello, i asked gemini where i can find a alternate for Weights.gg since its shutting down, and it said something about w8ights.cc or w8.ai and it told me to come to this server to find something like weights-migration or smth.

[hallow thistle]

Not everything said by Google Gemini would always be accurate. This Discord server you're here is AI Hub, and yes it's mostly about "RVC".

[sudden laurel]

I am using w okada but when I talk it says something like 'trial' how can I fix that?

[hardy yew]

you download VAC trial version

download lite instead

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

uninstall the VAC you have and install this instaed

[hallow thistle]

You're using trial version of Virtual Audio Cable. Use the free Virtual Audio Cable "lite" instead.

[signal mauve]

do i use cable

?

[hallow thistle]

There's the guide. https://docs.aihub.gg/realtime-voice-changer/local/tg-develops-w-okada-fork/

For input/output settings, I usually do this. https://cdn.discordapp.com/attachments/1159290139609137264/1456249974525526185/image.png?ex=69c90ef9&is=69c7bd79&hm=62c75ab409606ec88e4f5370b09c829ef117b759b3ef4b0678cedc5c76cf62d5&

[signal mauve]

ok

[signal mauve]

how to fix perf

[spring socket]

cuda error

[hallow thistle]

Chunk: 120 ms
Extra time: 2.7 s
Pitch extraction: rmvpe (not rmvpe_onnx)

Sample rate: 48000 Hz

Where'd you get this W-Okada voice changer from? Tg Develop's W-Okada fork and Vonovox are only voice changers that can work with GeForce RTX 50 series.

[spring socket]

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

[spring socket]

can I get working link?

Before I use 4060, change to 5070 tonight maybe the prob

[hallow thistle]

Tg Develop's W-Okada voice changer is a continuation of Deiteris' W-Okada but made by different author and better, 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.

[spring socket]

both? why the file named amd, i use nvidia

[hardy yew]

CUDA -> Nvidia GPU
amd64 -> CPU architecture (it's the same as intel CPUs, it's just a name)

all's good

[spring socket]

I see, thank you capybar and namari!

[signal mauve]

still very high

[spring socket]

[hallow thistle]

AMD64 is basically another name for x86-64 CPU architecture, which is used by both Intel and AMD in their CPUs respectively. If you see "CUDA", it usually refers to NVIDIA GPU.

[hallow thistle]

Send your full screenshot.

[hallow thistle]

Nah, you still have processing unit set to CPU, which is why the voice changer lagging a lot.

[signal mauve]

ty

[regal dew]

w11 and 1650

[hallow thistle]

A Linux distro or Microsoft Windows 11?

[regal dew]

windows 11

[formal kindle]

RVC v2 Disconnected this one right?

[hallow thistle]

Old news, but there's why RVC Disconnected has long ended. #📰│dev-updates message Applio RVC is a better bet.

[formal kindle]

no ui?

[hallow thistle]

There are main Applio RVC (one with UI) and one with no-UI command line.

[low shard]

wdym? maybe check https://docs.aihub.gg/extra/model-maker-role/

[low shard]

there's no perfect value

[low shard]

if you didn't already have an account, you're kinda cooked

[low shard]

it depends on your pc gpu, maybe if you could tell that it could be helpful, but considering you're on Windows 7 you're probably cooked for local AI

If you just want to make ai covers easily, check https://docs.aihub.gg/rvc/cloud/aicovermaker-cloud/

[low shard]

RVC is Retrieval-based-Voice-Conversion, not realtime voice changer

elaborate:

• your pc gpu
• your pc os
• what you're trying to do, like AI Covers, E Girl Trolling / Catfishing, Roleplaying as goku
• the tutorial link you're using

[low shard]

never heard of those

can you elaborate your pc gpu and os and if you're trying to train RVC models or just make AI Covers?

[low shard]

check https://docs.aihub.gg/rvc/local/applio/, you can train rvc voice models here

[low shard]

RVC v2 disconnected is gone, elaborate your pc gpu and os

[hallow thistle]

What is your NVIDIA GPU? And what do you use Vonovox for?

Why film?

[hardy yew]

vonovox is the wrong tool for that

[hallow thistle]

You could have record your audios and convert them in non-realtime RVC like Applio RVC instead. "Realtime" voice changer is only if you mean to live stream on TikTok.

You would need OBS Studio for this one. But how would you make such scenes the same time you configure the voice changer?

[hardy yew]

I mean, I said it before you mentioned the livestreams

for streams you definitely want realtime (and vonovox is the right tool for that)

I just meant that for videos you'll be much better off with applio or any other offline inference tool. You can record the audio, then convert it later and also tweak it in postprocess which is not gonna be possible with vonovox

But regarding your "laggy and bad" sound in vonovox

Which version of vonovox are you using? Is vonovox the only GPU-heavy app running? Additional load from other apps might affect the performance

What vonovox settings are you using?

no way to send pictures here without explicit permission (which only admins can do). You can either describe in text or perhaps send a screenshot via DM to me

Can you scroll lower to the voice settings section?

In general what you might want to try is tinkering with the block size slider and see if it improves the issue of sound cutting off

The slider I was talking about is actually between the first and second screenshot

Anyway, try to play around with the block size. Maybe a value of 0.2 etc

See if it affects the choppy sound

In terms of quality strictly - did you ensure its not just a poor quality model?

Try some other models as well

[low shard]

are you perhaps trying to do e girl trolling / catfishing?

[hardy yew]

Also might disable Smart SINE for test. Depending on the microphone, it might sometimes cut off your speech abruptly. If it doesn't happen, you can keep it on

[low shard]

anymore?
with a girl account?

[lapis osprey]

hi people. I am someone who is working on a university project, and one of the things I need is object detection. I have 0 experience in AI. (I am a telecom and computer engineer and have lots of experience in electronics and programming). I have no clue where to start. what do I need to learn?

[hardy yew]

Did you extract the zip before trying to install the program?

[low shard]

Maybe you could check the Yolo Models, the latest is Yolo26

[lapis osprey]

1660ti
windows 11 (I can change this if needed obviously)
I need to do simple object detection like doors, stairs, etc
I dont have any tutorials that Im following, that's the whole reason Im here

[low shard]

maybe tutorials like https://www.youtube.com/watch?v=r0RspiLG260 and https://docs.ultralytics.com/tasks/detect/#export could help you, the yolo models are the best for this task

Nvidia GPUs are generally better for Local AI tasks because of CUDA support
which specific AI you're trying to run?

[lapis osprey]

nice. also, how efficient would it be to train my own AI model. I want it to be good enough to handle 24fps processing. are these yolo models customizable?

like, can I change what they can "see"?

[low shard]

you can finetune it to detect specific objects, like video game characters or cars

[lapis osprey]

does finetuning it change its speed?

tryna add more bots in ssl lobbies? 😭 😭

[hallow thistle]

Why cheat?

[hardy yew]

meanwhile "educational purposes" 3 messages earlier

You're not even trying

[lapis osprey]

thanks nick, I'll check the yolo models out

[low shard]

😭

[hallow thistle]

AMD Radeon GPUs are good as the budget alternative to NVIDIA GPUs, but then certain AI setups gonna be tricky because most AIs are made for NVIDIA GPUs which natively support CUDA better, while AMD GPUs need some specific workarounds to work. This also happens with something like W-Okada voice changer as well.

[low shard]

you're welcome, I haven't played around with YOLO models honestly but they might be great for your object detection task

your GTX 1660 Ti isn't that new so not sure if it might be able or fast enough to do YOLO finetuning locally, but you can try

[lapis osprey]

I saw that there are different sizes with a tradeoff between speed and accuracy so I might be able to work something out

all I need is basically a model that can detect cars, humans, signs, doors, and stairs. that's it. Ive ran 7B models on my pc before so these models SHOULD be fine, fingers crossed. Im a little lost because I have no clue how AI works except that it uses probability math, which is very interesting

[low shard]

maybe look out for YOLO26n (nano) or YOLO26s (small)

[hardy yew]

@low shard

free money glitch

[tame oracle]

Please remember that not all glitches work as intended

[viral mason]

Fuckass hacked account

Die

[simple escarp]

hi

[robust locust]

Now, Seedance is too secure a model and has so many filters that it won't let me make a video where a person transforms, flies, and attacks an aquatic monster. What do I do?

[winged adder]

, I’ve been looking into this
stack

Claude Code / MCP / Browser-use / MiniMax / NoVNC.

Before I dive in, I’m trying to figure out the hardware overhead.
: Does this specific workflow rely on CUDA
for the local components, or is it compatible with
Vulkan ?

: How much VRAM is typically "eaten up" by the browser automation and NoVNC layer on top of the LLM? What model via MCP?
what’s the "safe" VRAM ceiling I should aim for?
!

https://www.reddit.com/r/LocalLLaMA/comments/1qlfete/claude_code_mcp_browser_use_minimax_llm_novnc/

[vital igloo]

I need help with that everything work but when im testing myself on discord, its coming from my monitor and not in my ears. When i join talk, i hear people in ears , but from monitor my girly voice

i dont understand it, can someone help me please ?

[cyan panther]

Hello
A quick question
I bought the xm8500
And have been looking for a cheap way to connect it to my PC
Is the Behringer U-Control UCA222 good for that? it's pretty affordable from my findings.

[viral mason]

I've never even heard of that

[cyan panther]

Not sure if I can post links but here's the full name
(Behringer U-Control UCA222 Ultra-Low Latency 2 In/2 Out USB Audio Interface with Digital Output )

Else
What's the cheapest way to connect it to the PC?

[viral mason]

not sure, I'm not super tech savvy with stuff like that, is there anything on youtube or maybe reddit that you could find?

[hardy yew]

why UCA222? It doesn't even have a proper input (it uses RCA instead of jack/XLR).
I mean, sure, you can make it work with that, but might as well purchase an interface that is actually designed for the purpose

the main problem is that I assume you want to limit the budget as tightly as possible. Which also makes sense as the microphone itself is a cheap one

honestly, just try to find some USB interface with XLR input and you'll be fine

(or balanced 6.3 jack, but usually it'll be part of the XLR anyway)

behringer's UM2 is one of such, it'll do the job and it's cheap

(obvious side note - both the mic and DAC are fairly cheap, so it'll be reasonable to not expect supreme quality, but it should sure be better than e.g. a headset mic)

[cyan panther]

Same. It's the reason for my big confusion on what exactly to get. I got the mic as a gift so it'd be a waste not to use it, but I can't afford what reddit suggests unfortunately.

[cyan panther]

Oh. Thank you so much! I was about to ask for a specific suggestion actually.
However, the Behringer U-Phoria UM2 is still a bit expensive, it's double the mic's cost (as I checked how much it was on amazon.eg) ... is there no other cheaper alterantive for my use case?

[hardy yew]

Oh, I guess the prices vary a lot between countries. In my area the UM2 is priced very similarly to UCA222, which is also the same cost as the microphone

That was mostly why I recommended that one, because i assumed it's within price range

But apparently the prices are quite different in your case

Honestly I don't know, don't have any recommendations in this price range. Mostly because honestly it's kinda not a thing that can be bought for that money

[cyan panther]

Yeah unfortunately the mic is sold for around 2,5k EGP, while the UM2 is 4,5k EGP

[hardy yew]

But as I said, you might try to find some as-cheap-as-possible XLR USB interface and perhaps it'll be sufficient. Obviously those won't compare to the more expensive ones but they should do the job nevertheless

dunno if that's acceptable for you but i'd search among used stuff instead of new

much more likely to find something nice

[cyan panther]

I don't mind used stuff at all. My PC was built from some used stuff and it's been working perfectly fine. However, in this case, I don't have experience on where to get reliable used audio hardware, unlike my experience with PC hardware, hence why I prefer finding something new from a renowned place like Amazon.

I found this
What do you think?
M-Audio M-Track Solo 48-KHz 2-channel USB Audio Interface with 1 Crystal Preamp, Phantom Power and Instrument Input

[hardy yew]

Oh, sure, this one will do

[cyan panther]

Very well, I'll save up for it.
So now all I need is this audio interface, an XLR cable to connect the mic to it (my xm8500 came without cables at all) and that's it. I don't need anything else, correct?

[hardy yew]

Yeah, that will allow you to use it in PC

[cyan panther]

Many thanks

[hardy yew]

No problem. Good luck!

[finite light]

Quick question... is there any alternative website to use AI voice free... without credits and some?

[hardy yew]

"free without limits" - probably not.
But places like Kaggle grant some free GPU time. Strictly in case of Kaggle - 30h per week

[finite light]

Hmm... yeah, probably... it's really sucks that 2 days left and i really can't use Weights.gg when it shutdown... i really had a lot plan for it.

[low shard]

elaborate:

• your pc gpu
• your pc os
• what you're trying to do, such as AI Covers, E Girl Trolling / Catfishing or Roleplaying
• the tutorial link

[low shard]

there's no company that would give you free unlimited AI cloud usage, they would just go backrupt

the only free unlimited way is running local AI on your beefy PC
what's your pc gpu and os? you just want to do AI Covers or Train RVC Voice Models?

[midnight pebble]

my model says TypeError: Trying to convert Float8_e4m3fn to the MPS backend but it does not have support for that dtype.

m3 mac

24gb of ram

i use flux schnell

yep,everyone in this server is offline

this server has been dead since early 2023

no one has this issue

anyone

please

i want to run the model on my mac

someone listen

hello?

can anyone reccomend me a similar model

dead chat

anyone

[low shard]

@midnight pebble

please stop begging and demanding

this server is run by volunteers, no one is forced to actually help you

start by elaborating your issue

[midnight pebble]

my model says TypeError: Trying to convert Float8_e4m3fn to the MPS backend but it does not have support for that dtype.

i want to run a funny model on my m3 mac

i bought a m3 mac for nothing

[low shard]

elaborate:

• what you're trying to do, like AI Covers, E Girl Trolling / Catfishing, Training RVC Models
• The tutorial link

[midnight pebble]

image models (generic ones)

tutorial = none

[low shard]

m chips aren't as much widely supported as Nvidia GPUs on Windows / Linux in the AI field

[midnight pebble]

what model do i use

[low shard]

what program and models are you using then?

[midnight pebble]

flux schnell

don´t make me use sdxl turbo because it puts people with 200 eyes and floating head

look at what sdxl makes

[low shard]

what program are you using? comfyui?

[midnight pebble]

comfyui

going to fine tune it on tons of images of people

[low shard]

check https://github.com/filipstrand/mflux or https://www.reddit.com/r/StableDiffusion/comments/1ls2nqj/macos_comfy_error_trying_to_convert_float8_e4m3fn/

[midnight pebble]

mflux is too large

[low shard]

mflux would probably the best way for you

else use the other fix i sent

[hollow dome]

i'm sure this has been asked a billion times so sorry for perceived spam lol, i tried searching myself but didn't see anything that immediately stuck out to me. Is there a generally agreed upon replacement for weights.gg now that it is closing down? or is everyone just moving to local?

[hallow thistle]

Not true.

[stable current]

How do I post images in ai-images?

[white shadow]

for tg forks version of w-okada i keep getting this error over and over whenever I start the conversion

[sharp jungle]

?colab

[tame oracle]

-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

[sharp jungle]

Ty

[tame oracle]

You're welcome

[sharp jungle]

Can u tell me which is the best, there r so many...i wanted to train an model

[tame oracle]

For training a model you should use Applio

it'd be better to use it on Kaggle tho

[sharp jungle]

-kaggle

[patent trellis]

📘 Kaggle Notebooks

Kaggle is a Cloud (Remote Good PC) Service that offers 30 hours of GPU weekly, but needs a phone number verification

• **Applio Notebook**

by IAHispano
Kaggle

• **Hina Mod Original Wokada**

by Hina
Kaggle

• **Wokada Deiteris Fork**

by Hina & Deiteris
Kaggle

• **UVR5 UI**

by Eddy, ArisDev & Nick088
Kaggle

• **UVR5 NO UI**

by Eddy
Kaggle

• **RVC AI Cover Maker UI**

by Shirou & ArisDev
Kaggle

• **Music Source Separation**

by Shirou
Kaggle

[sharp jungle]

Alright

[tame oracle]

@sharp jungle Here are the Applio docs https://docs.aihub.gg/rvc/cloud/applio-cloud/

[sharp jungle]

Super useful 🙏

[tame oracle]

Yes it is

[hardy yew]

Have you checked the logs in terminal?

[white shadow]

yep

[hardy yew]

What did it say? There should be some info that should help debug it

[finite light]

Well... yeah, I'm doing AI Covers... and even try make my own RVC Voice Models too... and i look for something useful as Weights.gg... by any chances... i hope you understand?

[hardy yew]

Nick's reply remains valid though. There's no unlimited, free service like this. Either use limited cloud services, your own PC or pay for extended cloud service

[finite light]

Oh, that's shame... i bet one day, someone creating other website AI Cover and some as Weights.gg... Unlimited and free... for we can do whatever we want... Besides, Paying money for premium and credits is no nessacery to use it for what... to bill up to be fill with money of ours, For love of god... Why can't people just enjoy making AI Cover for free... without Premium and credits, It's not like end of the world or something... right?

[tame oracle]

How else are you supposed to pay for the power to do them

[finite light]

Hmm... if i had my own website AI Cover... i would just let people do let them do RVC making and do AI Covers for free... without premiums, credits and paying money... We really don't want to being a greedy and selfish for some reason, do we now?

[hardy yew]

Lol

[finite light]

Lol your kind ¬__¬

[tame oracle]

Its not being greedy or selfish

i just told you

you need money to pay for the power

to run the computers

[hardy yew]

"i'd give away a ton of free money for anyone that wants it, don't wanna be greedy or selfish"

[tame oracle]

to makee the AI covers

[tame oracle]

exactly lmao

[hardy yew]

And then people would just take that free money and turn it into egirls and fraud 💀

[tame oracle]

LMFAO

[finite light]

Sheesh...

[tame oracle]

So true

[finite light]

Especially in roblox.

[tame oracle]

https://tenor.com/view/dj-khaled-inspiration-inspirational-quotes-inspirational-quotes-for2023-inspirational-quotes-about-life-gif-27540397

[finite light]

Okay, that got me... bit laugh about, good one.

[white shadow]

2026-03-29 19:03:48,345 ERROR [VoiceChangerManager] 'Pipeline is not initialized.'
Traceback (most recent call last):
File "voice_changer\VoiceChangerManager.py", line 300, in change_voice
audio, vol, perf = self.vc.on_request(receivedData)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "torch\utils_contextlib.py", line 116, in decorate_context
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "voice_changer\VoiceChangerV2.py", line 190, in on_request
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.'

[hardy yew]

damn, indeed not very descriptive

I think it often happened in cases where audio device couldn't be initialized. Have you tried with a setup without the VAC? Just mic + speakers for a test

Other than that dunno, maybe someone else knows what else might cause it

[white shadow]

yeah tried that

[safe pine]

i just want help to tweak my vonovox

like i been trying different stuff but nothing seems to be working that well

[spring socket]

What happend with the zip

2x tried to redownload, same corrupt

[hardy yew]

did you add the .zip part at the end of the file names..?

you shouldn't

it needs to end with .001 and .002

[spring socket]

Ohh.. I'm not adding it, just realized it has 2 zip lol

automaticly

[hardy yew]

maybe the browser added it while saving, for some reason

dunno

remove the .zip endings and try again with unpacking the 001

Vonovox is unrelated to weights

No reason for it to become unavailable

In general - yes, vonovox has the most improvements in terms of sound quality and conversion robustness.
But lots of robotic sound cases are just poor models, which vonovox won't help with

(BTW: vonovox is for windows + nvidia GPUs only, just in case)

[low shard]

Like e girl trolling / Catfishing?

[low shard]

What's your PC GPU and OS?

[low shard]

No one will, so many sites already lost money, the AI Cover hype died too, would you throw money out of the window just to offer people free unlimited things for no reason? Companies are made to make money, when you see services like google and YouTube being free, they make money with your data and ads

[low shard]

elaborate:

• your pc gpu
• your pc os
• what you're trying to do, such as AI Covers, E Girl Trolling / Catfishing or Roleplaying
• the tutorial link you're using
• a screenshot of your settings

[slim prism]

Hello, it's been a long time since I've been here and done AI cover work. Where can I find AI Cover locally on Google Colab like in the old days?

[low shard]

it's best to check if you can do it locally first since there's no free unlimited cloud way,
what's your pc gpu and os?

[fiery raptor]

what is the alternative now that weights died? Where can i do sum similar

(ik it aint dead yet but it will in 2 days)

boi

[low shard]

what's your pc gpu and os first?

[fiery raptor]

windows 11 and 3070

[finite light]

To be honest... i've making AI Covers on working laptop.

[fiery raptor]

how dude

[finite light]

Well... i've start 2 years ago... and it really keep me good in process... and i use the Wondershare filmora too.

[pearl scarab]

i keep hearing my thing being distort

how do I fix it

[finite light]

Really, what you use...?

[pearl scarab]

wdym what i use?

[finite light]

uh... i thought you mention of making video, sorry.

[pearl scarab]

why you type so formal?

[finite light]

I really dunno, i don't usually type that formal in discord...

[fiery raptor]

you make ai covers w filmora?

[finite light]

Yep, and i have over 13K subs in youtube channel.

[fiery raptor]

i didnt know filmora has ai covers

[finite light]

Me either... but i mostly use Weights.gg AI Cover vocals... to put in Wondershare Filmora to make video with it... to be fit perfect. with along instrumental of any song we ever heard of it....

It takes quite long to process... but it worth to shot.

[fiery raptor]

ohhhhhhhhhhhhhhhhhhhh dude so you just use weights

well weights is dying in 2 days. Thats why i need a replacement.

[hardy yew]

https://docs.aihub.gg/rvc/resources/dataset-isolation/

https://docs.aihub.gg/rvc/resources/training/#introduction

take a look those two links. First for general dataset information + preparation, second for actual model training

[finite light]

True, i really need replacement too...

[fiery raptor]

Pfft and I thought filmora actually has ai covers😭😭

You just edit w it

[finite light]

I know... i bet filmora would add ai voices in it... even thought Capcut does have some.

[fiery raptor]

Yeah but not like weights

[finite light]

Yeah, nothing can't beat Weights... i wish it could come back... one day.

[fiery raptor]

CapCut has a bank of tts voices and a view covers that u can use to maybe hide your voice or make funny tt videos

[finite light]

Oh yeah, i even thought about making animation video too...

Yolted did that once...

[low shard]

you can do local AI using your PC GPU, without needing to relay on a site, it won't be as easy as using weights, but do you want to do that? It will be free and unlimited, check https://docs.aihub.gg/rvc/local/applio/

[low shard]

this is a general ai discord server, elaborate:

• your pc gpu
• your pc os
• what you're trying to do, like AI Covers, E Girl Trolling / Catfishing, Roleplaying, Chatbots

[hardy yew]

https://tenor.com/view/no-noo-nope-never-sarcastic-no-gif-4776564746551978937

[viral mason]

Ew

Guess we have to remove all of Poland/j

[hardy yew]

https://tenor.com/view/angry-cat-gif-25783770

I don't take responsibility for that fella

let me be

[viral mason]

Wait are you from Poland?

[hardy yew]

actually, yeah xD

[viral mason]

Oop

[uneven parcel]

Hey everyone! Does anyone happen to have a spare ChatGPT Plus referral link (guest pass)? I've been hunting for one for hours but haven't had any luck so far. Thanks in advance!

[winged adder]

How tò image gen model using vulkan

[hazy zenith]

yo, im trynna use the light yagami voice model, i can hear my real voice, but no the ai + real voice, INPUT is set to my mic, output is VB cable, and my monitor is set to headphones, passthough is on, how can i fix/use the ai voice model?

[crimson arrow]

Hello, I need some help. About 11 years ago, I was looking for a way to make video game graphics smoother, specifically to remove jagged edges (aliasing) from walls and textures, and also to fix mouse movement. After a lot of searching and struggling — it took about a year and a half — I think I managed to do it using things like CUDA, cuDNN, CNTK, Python, and similar tools. Could anyone help me figure out where to start learning and how to do this again?"

[low shard]

elaborate:

• your pc gpu
• your pc os
• the tutorial link you're using

[hazy zenith]

1. GeForce RTX 3050 2. windows

3. https://discordapp.com/channels/1159260121998827560/1481774803609194647

[viral mason]

I don't think that's a tutorial

[low shard]

I'm guessing you're on Windows 11

can you also send the tutorial link you have used? there are different programs

[plain berry]

ai for studying?

[trim yew]

i did all the steps but it isnt working and i also cant hear anything

[rare flint]

so what would you recommend with RTX 5060? locally for train and voice changer and cover?

(win 10)

[viral mason]

It's dependent on what Gpu you have, for AMD there isn't anything besides Wokada tg fork that's up to date and it hasn't been updated for a long time

And for Nvidia Vonovox is the best to use as it's got a ton of cool features that make it sound more realistic and much better

Oh nice

U should use Vonovox then ^^

To keep yourself up to date on that specific one there's the official discord server you could join

Btw I'll get the download to the most recent beta version since the beta release as of now is better than currently public release of it

I believe it's in the guide but I could also just send it to you in dms as I can't send it here

https://huggingface.co/dr87/vonovox/resolve/c8034f5f6d50648a8109bb4f847182362e2b779b/Vonovox_beta_17_11.zip

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

Here are the two downloads you'll need, also same for u @rare flint

This is just for real-time and for training download applio, I don't have the local version link or know how to set that up but here's all I can get u @rare flint

-Rvc

[patent trellis]

📚 RVC Documentations

AI HUB Docs

https://docs.aihub.gg

🍏 Applio Docs

https://docs.applio.org

[viral mason]

Is it alright if I send you the server invite to Vonovox?

Btw if you do need help feel free to ask here

[low shard]

are you planning to do e girl trolling / catfishing or just roleplay?

do you need any other help?

[tame oracle]

😭

[hallow thistle]

No, no. Applio RVC and Vonovox are two different programs. Vonovox is supposed to be in its folder, while Applio RVC in its folder as well.

Are you trying to do non-realtime (Applio RVC) or actual realtime (Vonovox)? Let's be real.

For "realtime", because your earlier query stated to be "realtime", what do you use Vonovox for?

Already answered in #🧬│ai-chat, so the answer about paying someone is still "no". If you mean a program to use RVC voice models, there's Applio RVC.

What is your PC GPU? And what scenario will you use W-Okada voice changer for?

Which NVIDIA GeForce RTX GPU?

Do you use the female voice model or something? Just wanna make sure.

I'm asking this to avoid one catfishing as other mods here suggested, so if you could directly answer so.

[plucky crescent]

hello@hallow thistle

[hallow thistle]

Do you need help about something?

Cool. Vonovox gives better audio quality than typical W-Okada voice changer versions, so download Vonovox. https://huggingface.co/dr87/vonovox/resolve/main/Vonovox_beta_17_11.zip?download=true

[plucky crescent]

I just joined

[hallow thistle]

The setting "block size" is the one that controls audio delay in Vonovox. This is the guide for Vonovox. https://docs.aihub.gg/realtime-voice-changer/local/vonovox

[hallow thistle]

For a regular conversation, #✦│chat would be better.

[plucky crescent]

ok thinks

[rare flint]

What's the difference?

[rare flint]

Anyway I mostly gonna use it for Ai Covers, Roleplay and honestly i just for shenanigans with friends. how come Wokada stuff got irrelevant?

[low shard]

please answer to be sure you’re using the right program, there are things for making rvc models, for ai covers, for e girl trolling / catfishing and for roleplay

[rare flint]

^

[low shard]

wokada isn:t for ai covers

it’s only for roleplay

would you rather using applio that can do both?

[rare flint]

I said i want both AI cover and roleplay

[rare flint]

i read that it's real time is kinda bad?

though idk how much it affects when it's 5060ti

[low shard]

yeah I’m saying that program won’t be good for both

it’s best you first try it here: https://docs.aihub.gg/rvc/local/applio/

[low shard]

It’s just newer, i would give it a try first and see

[rare flint]

out of curiosity, is it worse than Wokada for Real time?

does it have to be C drive? i don't want it to Hog it all

[low shard]

its suggested to

[willow lotus]

Sorry for the stupid question, I want to know how to make an AI cover...?

[low shard]

What’s your pc gpu and os?

[willow lotus]

RTX 3060 12 gb windows 11

[low shard]

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

[odd sedge]

whats the best local RVC for MacOs (macbook pro "M3 pro") ?

[mellow hinge]

Hi sorry to bother you all, i need some help on my applio 3.6.2. I would like to use the pretrained KLM6.2 voice model to create my own voice model, but it keep logging out

Loaded pretrained (G) 'rvc\models\pretraineds\custom\G_KLM6_Exp3_L6_32k.pth'
The parameters of the pretrain model such as the sample rate or architecture do not match the selected model.
Weights only load failed. In PyTorch 2.6, we changed the default value of the `weights_only` argument in `torch.load` from `False` to `True`. Re-running `torch.load` with `weights_only` set to `False` will likely succeed, but it can result in arbitrary code execution. Do it only if you got the file from a trusted source.
Please file an issue with the following so that we can make `weights_only=True` compatible with your use case: WeightsUnpickler error: Unsupported operand 60

other than setting
Sampling Rate: 32000
customer embedder: spin 7-12
is there anything i missed? thank you so much

[odd sedge]

all tutorials on windows

there is no macOs solution

[mellow hinge]

pytorch seems not really supporting macos.. so i think its a bit tough to make it work on macos. im new tho, correct me if im wrong

[mellow hinge]

sorry just found original RVC support macos

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

you may give it a shot

[low shard]

RVC means Retrieval-based-Voice-Conversion not realtime voice changer, can you elaborate if you're trying to do AI Covers, TTS, E Girl Trolling / Catfishing or Train RVC Models?

[mellow hinge]

tried set weights_only to false and it works now.

[glass rivet]

does appolio's real time thingy not work ingame, anyone?

Is there any way to do that?

[low shard]

elaborate what's wrong

[glass rivet]

I think appolio can't create a virtual audio device all by itself, that's the reason why the game picks up the normal mic input

I'm following that guide wherein we're required to install vac-lite

[low shard]

it can't do that, in the ai hub docs guide it says you have to setup the virtual audio cable yourself

what's your pc gpu and os btw?

[glass rivet]

rtx 5080, win10

ryzen 5700x3d

[low shard]

did you use setup64.exe?

[glass rivet]

yep. I'm following that guide now, will see if it works

downloaded the MMVCServer files

will it be able to use an rvc model trained using appolio's refine-gan?

[hollow tusk]

I need a bit of help, I am trying to reinstall MMVCServer on my Linux PC

KDE Plasma Version: 6.6.3
KDE Frameworks Version: 6.24.0
Qt Version: 6.11.0
Kernel Version: 6.19.10-1-cachyos (64-bit)
Graphics Platform: Wayland
Processors: 12 × AMD Ryzen 5 5600X 6-Core Processor
Memory: 32 GiB of RAM (31.3 GiB usable)
Graphics Processor: AMD Radeon RX 9060 XT
Product Name: B550M Pro4```
I am following this that was given to me last time I asked for help 
https://docs.aihub.gg/realtime-voice-changer/local/tg-develops-w-okada-fork/#download-for-amd-gpus-on-linux
I am following the AMD one since that is what I am running
But when I try it
```❯ ./MMVCServerSIO
2026-03-30 11:26:40,460 INFO     [main] Python: 3.10.19 (main, Oct 10 2025, 01:02:36) [GCC 11.4.0]
2026-03-30 11:26:40,460 INFO     [main] Voice changer version: b2397 AMD-ROCm
2026-03-30 11:26:40,460 INFO     [main] Server settings: model_dir='model_dir' sound_dir='sound_dir' content_vec_500_onnx_on=True host='127.0.0.1' port=18888 ssl_enabled=False ssl_certfile=None ssl_keyfile=None allowed_origins=[] edition='AMD-ROCm'
2026-03-30 11:26:40,460 INFO     [main] Checking for mandatory models...
2026-03-30 11:26:40,503 INFO     [ModelManager] Found 3 missing mandatory models. Starting download...
2026-03-30 11:26:40,503 INFO     [ModelManager] Downloading mandatory model: RMVPE
2026-03-30 11:26:45,783 INFO     [ModelManager] Successfully downloaded RMVPE                                                                         
2026-03-30 11:26:45,783 INFO     [ModelManager] Downloading mandatory model: RMVPE (Onnx)
2026-03-30 11:26:56,565 INFO     [ModelManager] Successfully downloaded RMVPE (Onnx)                                                                  
2026-03-30 11:26:56,565 INFO     [ModelManager] Downloading mandatory model: ContentVec / Hubert
2026-03-30 11:27:07,958 INFO     [ModelManager] Successfully downloaded ContentVec / Hubert                                                           
Traceback (most recent call last):
  File "client.py", line 22, in <module>
  File "asyncio/runners.py", line 44, in run
  File "asyncio/base_events.py", line 649, in run_until_complete
  File "main.py", line 93, in main
  File "webserver/server.py", line 132, in start
  File "uvicorn/server.py", line 70, in serve
  File "uvicorn/server.py", line 77, in _serve
  File "uvicorn/config.py", line 435, in load
  File "uvicorn/importer.py", line 19, in import_from_string
  File "importlib/__init__.py", line 126, in import_module
  File "<frozen importlib._bootstrap>", line 1050, in _gcd_import
  File "<frozen importlib._bootstrap>", line 1027, in _find_and_load
  File "<frozen importlib._bootstrap>", line 1006, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 688, in _load_unlocked
  File "pyimod02_importers.py", line 457, in exec_module
  File "app.py", line 17, in <module>
  File "pyimod02_importers.py", line 457, in exec_module
  File "voice_changer/VoiceChangerManager.py", line 12, in <module>
  File "pyimod02_importers.py", line 457, in exec_module
  File "voice_changer/VoiceChangerV2.py", line 17, in <module>
  File "pyimod02_importers.py", line 457, in exec_module
  File "voice_changer/common/deviceManager/DeviceManager.py", line 2, in <module>
  File "pyimod02_importers.py", line 457, in exec_module
  File "onnxruntime/__init__.py", line 58, in <module>
  File "onnxruntime/__init__.py", line 23, in <module>
  File "pyimod02_importers.py", line 457, in exec_module
  File "onnxruntime/capi/_pybind_state.py", line 32, in <module>
ImportError: /home/mouse/MMVCServerSIO/_internal/onnxruntime/capi/onnxruntime_pybind11_state.so: cannot enable executable stack as shared object requires: Invalid argument

Press Enter to continue...

2026-03-30 11:27:45,343 ERROR    [base_events] Unclosed client session
client_session: <aiohttp.client.ClientSession object at 0x7fa3e7c804f0>
2026-03-30 11:27:45,344 ERROR    [base_events] Unclosed connector
connections: ['deque([(<aiohttp.client_proto.ResponseHandler object at 0x7fa3e7cb2620>, 5640.111927896)])', 'deque([(<aiohttp.client_proto.ResponseHandler object at 0x7fa3e7af8f40>, 5651.28619272)])']
connector: <aiohttp.connector.TCPConnector object at 0x7fa3e8bf5540>

~/MMVCServerSIO 1m
❯ 

it should have been a simple install, but I am getting that odd error

[hollow tusk]

Ok I managed to fix it but it only runs on the CPU

2026-03-30 12:29:22,333 INFO     [main] Python: 3.10.19 (main, Oct 10 2025, 01:02:36) [GCC 11.4.0]
2026-03-30 12:29:22,333 INFO     [main] Voice changer version: b2397 AMD-ROCm
2026-03-30 12:29:22,333 INFO     [main] Server settings: model_dir='model_dir' sound_dir='sound_dir' content_vec_500_onnx_on=True host='127.0.0.1' port=18888 ssl_enabled=False ssl_certfile=None ssl_keyfile=None allowed_origins=[] edition='AMD-ROCm'
2026-03-30 12:29:22,334 INFO     [main] Checking for mandatory models...
2026-03-30 12:29:22,391 INFO     [ModelManager] All mandatory models are already downloaded.
2026-03-30 12:29:25,770 INFO     [loader] Loading faiss with AVX2 support.
2026-03-30 12:29:25,800 INFO     [loader] Successfully loaded faiss with AVX2 support.
2026-03-30 12:29:25,803 INFO     [__init__] Failed to load GPU Faiss: name 'GpuIndexIVFFlat' is not defined. Will not load constructor refs for GPU indexes. This is only an error if you're trying to use GPU Faiss.
2026-03-30 12:29:25,981 INFO     [VoiceChangerManager] Initializing...
2026-03-30 12:29:25,986 INFO     [DeviceManager] Initialized DeviceManager. Backend statuses:
2026-03-30 12:29:25,986 INFO     [DeviceManager] * DirectML: False, device count: 0
2026-03-30 12:29:25,986 INFO     [DeviceManager] * CUDA: False, device count: 0
2026-03-30 12:29:25,986 INFO     [DeviceManager] * MPS: False
2026-03-30 12:29:25,986 INFO     [DeviceManager] Switched to CPU (cpu). FP16 support: False
2026-03-30 12:29:25,987 INFO     [IORecorder] -------------------------- - - - /tmp/tmpnwld7ooe/tmp_dir/in.wav, /tmp/tmpnwld7ooe/tmp_dir/out.wav

it should be fine, my CPU can handle it, just dont like that it is

[north pilot]

Hello
Guys can you please help
I dont have a good realistic quality using vonovox

I tried a lot of models

[median monolith]

Since weights.gg is shutting down
I wanted to upload lots of voice models from weights users somewhere to archive them.
what would be the best choice?
Internet Archive? Hugging Face? something else?
(yes, im going to try my best to credit the original uploaders)

[urban forum]

Do we have any documentation for Seed VC?

[urban forum]

Hugging would be your best bet.

[viral mason]

what's that?

[urban forum]

https://github.com/Plachtaa/seed-vc

[viral mason]

ah zero shot

[urban forum]

Apparently this works for AMD?

[urban forum]

What GPU do you have worm? 😮

[viral mason]

I have a rtx 5070ti :3

[urban forum]

I need to get my 2nd PC up and running.

[viral mason]

what about you Loamy?

[urban forum]

So does SeedVC not work?

[viral mason]

idk never heard of it

[urban forum]

I'm on an 9070 XT

😭 However my W-Okada refuses to work.

It just randomly broke one day.

[viral mason]

are you using Tg fork?

[urban forum]

Yeah.

[viral mason]

hmmmm

[urban forum]

It just shuts down.

Lmao.

[viral mason]

that's weird

[urban forum]

Yeah idk.

[viral mason]

I use vonovox for realtime but the nvidia version of Tg fork seems to work fine still

[median monolith]

Alright. im going to " + New Model " on my profile, to create the "stach" of voice models, would that be right?

[viral mason]

basically just

do that

[median monolith]

Aight, thank you so much! 🙏

also, its gonna be easier on this case, since weights already downloads the model in zip format

[viral mason]

https://discord.com/channels/1159260121998827560/1175430844685484042

last time I helped someone see it they just had to click on it again

[urban forum]

I'm gonna open a thread.

[urban forum]

I went ahead and put a bounty on my issue to give folks some incentive.

If you guys can figure it out, I'll pay you.

#1488310350603616476

Surely someone is a brainiac. ❤️

[viral mason]

should ping the helper role in case

[urban forum]

Oh sure.

[midnight fern]

How do i use a ai model?

[drowsy forge]

Since Weights is shutting down tomorrow, is there any Weights alternatives?

[outer rover]

hugging face

[drowsy forge]

Hugging Face?

[viral mason]

ew

[somber moon]

Why are people like this?

I mean, I get it if someone’s transgender (like FTM, or uncomfortable with their voice, for example). It can be hard to sound like the opposite gender, so using RVC makes sense in that situation.

But e-girl trolling? That’s just malice and ill intent for the sake of your entertainment. Like. grow up dude.

[viral mason]

I agree, only good reason is for trans people that dislike their voice or wanna change it

[hoary zinc]

started up my notebook as usual (which is the deiteris fork for w-okada), and i keep getting the error "HTTP Error 500: Internal Server Error", made a post about it in #1192011222023950368 just in case, but from what i looked up it's an issue on kaggle's end

anyone familiar with that error?

[somber moon]

People who troll as e-girls end up making the trans people reputation worse. I kinda feel bad for them, honestly. It’s not just about their voice. A lot of the time they get crucified the moment they use an AI voice changer. Like, a trans person could literally just say “hi,” and someone instantly jumps in with “crazy voice changer” or something. The comments are pretty unnecessary. Most of the time it only happens because the other person assumes the trans person is trolling. 😕

[grizzled arch]

"PyngrokNgrokInstallError: An error occurred while downloading ngrok from https://bin.ngrok.com/c/bNyj1mQVY4c/ngrok-v3-stable-linux-amd64.zip: HTTP Error 500: Internal Server Error" what is this error on kaggle?

[hoary zinc]

i'm getting the same one, it's probably something on kaggle's end considering multiple people are getting it

[grizzled arch]

oh fair enough

[drowsy forge]

Any Weights alternatives besides HuggingFace?

[median lance]

Is RVC Google Colab down for anyone else, or just me?

[hallow thistle]

Horrifying.

[hallow thistle]

Google Colab website loaded fine. What do you mean the website is down?

[median lance]

oh, Ngrok is currently doen

[median lance]

https://status.ngrok.com/

[low shard]

huh weren’t you using applio

[low shard]

did it use to work before? It doesn’t even load the UI?

[low shard]

Elaborate:

• your PC GPU
• your PC os
• if you're trying to do e girl trolling / Catfishing
• the settings

[low shard]

HuggingFace

[low shard]

It's archived and 0 shot, it's easier but lower quality than RVC

@median lance @grizzled arch @hoary zinc Ngrok Download Link is down

[grizzled arch]

Any eta on how long it will be down?

[low shard]

nobody knows

[low shard]

it’s just better you use other tunnels if available, what guide link are you using?

[grizzled arch]

ehm i was using the kaggle guide one. And ive been using it for a few months now. But it seem to be gone now (?)

[hollow tusk]

Yes, I was using it under a Nvidia driver and then upgraded to a new GPU so I had to uninstall all Nvidia drivers to put in the AMD drivers

So knowing that, I also deleted the whole program and reinstalled it under the instructions of AMD

I did managed to get the UI to open but it is only recognizing my CPU now

[low shard]

This is a general ai discord server, we use multiple programs, elaborate:

• your PC GPU
• your PC os
• what you're trying to do
• the tutorial or notebook link

[low shard]

AMD isn't as much supported and good for AI compared to Nvidia btw

So the UI always opens, but it doesn't run if you use the GPU right?

If so, have you tried using client audio processing or server?

[hollow tusk]

Server? Hmm I have not tried that, I'll do research on that another day

Not sure what client audio processing is

[low shard]

its the option above the input and output

[topaz smelt]

guys what do i put on models: when i upload a voice in w okada

[sullen pecan]

arxiv

[short hazel]

HI

I NEED HELP

My laptop is very low end and I want to use some voice changer for edit and stuff

Should I use okada I mean

Idk where to download okada too

[karmic trellis]

Hello, i have been using a voice model for while, but i am curious if i can make an index file for it myself? (the voice was not made by me, i downloaded it)

also. in advanced settings, what protocol should i use? sio or rest?

[viral mason]

you can't make an index file for an existing model no, you can use another model's index file on it tho in realtime or even applio I believe

[lethal fractal]

Hi, what is the best ai model for seo writing
cheap & not cheap

[low shard]

Elaborate:

• your pc gpu
• your pc os
• what you're trying to do, like AI Covers, E Girl Trolling / Catfishing, Roleplaying, Chatbots
• if you're using any tutorial link

[low shard]

this is a general ai discord server, elaborate:

• your pc gpu
• your pc os
• what you're trying to do, like AI Covers, E Girl Trolling / Catfishing, Roleplaying, Chatbots
• if you're using any tutorial link

[winged adder]

Looking for AI Vision suggestions for Desktop Automation (Excel → Flutter UI

[pure quiver]

Hey question in general is realtime voice changer good when you laugh or does it sound awkward and robotic?

[broken urchin]

yo

can someone help me make my AI Voice realtime sound as realistic as possible on ApplioV3.6.2

i am trying to make the voice as close to a normal person speaking as possible

[broken urchin]

sounds bad usually

[hardy yew]

Did you select it as input in the target app (game, discord etc?)

[hardy yew]

Applio's not a great choice for realtime. What is your GPU?

[broken urchin]

RTX 3060 Ti 8GB

[hardy yew]

#✨│ai-help message
see here for recent Vonovox version, it should be much better

[broken urchin]

oh hold on

how did you get the beta?

what's the difference with the new version?

i have Vonovox 1.6.8

[hardy yew]

It's publicly available

[pure quiver]

What helps it sound better?

[hardy yew]

Oh, a lot has changed since then. There's been multiple improvements. Better pitch extraction, audio stitching etc.

[broken urchin]

i dont know, i never was able to pull of laughing personally

but disabling noice cancelation makes the app pick it up

[pure quiver]

On discord?

Or realtime?

[broken urchin]

yes

[pure quiver]

How do I do that?

[broken urchin]

it will sound bad though, expect it to be pretty bad

[hardy yew]

how RVC converts your laughter is basically an outcome of two things:

• the model itself
• the way you laugh

In general RVC is rather bad at it usually, but some models handle it much better than others.

[pure quiver]

Is there a way I can find the models that work well with it?

[hardy yew]

and about the second point - obviously, being more in control of your voice and paying attention to what the model handles poorly can help with avoiding sound issues

[hardy yew]

I'd say just try various models that you can find available

[pure quiver]

Sounds good!

[hardy yew]

Other than that... training a model on your own, i guess

[broken urchin]

@hardy yew i made my voice model on Applio though for realtime purposes, is that bad

[hardy yew]

there's also places where models can be bought and usually they should be of better quality than the public ones, but no guarantee about that, also I don't have experience in that area

[pure quiver]

Thank you!

[hardy yew]

Applio is the right tool for training models, it's just that running realtime conversion is better done by dedicated apps (vonovox/tg develop's w-okada fork, etc.)

[broken urchin]

okay

whats the difference exactly

ive noticed vonovox sounds the most real

[hardy yew]

Lots of differences really, most of which I'm probably not even aware of.
But in general, volume handling, stitching algorithm, pitch extraction and smoothing

Applio's realtime is more of an "additional" feature, it wasn't really dedicated for it

[broken urchin]

where did you get the beta from

i cant find it in vonovox discord or website

[hardy yew]

All betas are publicly available and are announced in the Vonovox discord when they're released

[hardy yew]

it is on the discord, should be pinned in the beta channel

(the announcements channel is outdated, that's true)

[broken urchin]

ohhh

i see it now

[hardy yew]

the betas are mostly dropped in-between messages, so it's easy to skip those unless you watch the chat in that discord daily

[broken urchin]

whats the differences between this vonovox version and the older ones

how better is it

i got v1.6.8

downloaded the beat

[hardy yew]

this, mostly

among other things and perhaps bugfixes

[broken urchin]

does it sound more real

[hardy yew]

it should be more robust and less prone to glitches

The core sound is determined by the model

So no app can make it sound "more realistic" if the model isn't good enough in the first place

[broken urchin]

will AI realtime ever be able to express emotions fully

[broken urchin]

my model is good

[hardy yew]

look at how it all progresses over years, to me it's rather obvious that it sure will be good some day.
Though at the moment it doesn't seem like it's gonna happen any time soon

RVC has been stuck in its current state for some time now

some work is being done but no breakthroughs for now

[broken urchin]

yeah i wish it was indistinguashable from human voices

its okay but not the most realistic

[hardy yew]

with a right model, "compatible" user voice and appropriate handling (knowing the model's boundaries and speaking accordingly) it can be quite real

but that's a lot of conditions to fulfill

[broken urchin]

right

how can i make a voice like that myself

my current voice model is 320 epochs, 90 minutes of a person speaking

[hardy yew]

https://docs.aihub.gg/rvc/resources/dataset-isolation/
https://docs.aihub.gg/rvc/resources/training/

The AI Hub docs are a good starting point

[broken urchin]

but it doesnt sound the most realistic

[hardy yew]

There's a ton of things that affects how the model will turn out.
Data quality, diversity, the voice itself

There's also the factor of "how well the pretrain clicks with the particular voice", some are just naturally handled better

[broken urchin]

what

my data quality was really good

the voice was normal

the accent and voice is Serbian

is that bad?

[hardy yew]

Nothing good/bad about accents, the model should just learn it appropriately

[broken urchin]

so then, what could be holding me back?

can you give me Vonovox settings for the beta

best settings for my goals??

i have the premium

[patent trellis]

<:error:865860636084142150> Message error

The message template:sticky-ai-help could not be displayed. More details can be found in the error log.
Please report this to a server admin.

[hardy yew]

the defaults, really. I mean, all you need to adjust is the model settings (pitch, index rate, maybe the formant slightly if needed)

the effects are more for tweaking the voice sound, it won't affect the "realism" per se.
Maybe the low-quality-mic filter can help make it sound a bit more genuine

[broken urchin]

nothing else though???

i tried the low mic filter before

yeah you end up sounding like you got a bad mic

which helps

but not the thing im focusing on

my goal is to make the sound more real like a human speaking

more emotion and more natural

[lyric kindle]

what is the best wokada version for amd graphics card, i have call with my boyfriend in a hour and i need to get girl voice

somebody help please

[thick dagger]

bro how to create my own model?

[torn trench]

How to overlay voice on my file

[low shard]

what's your PC GPU and OS? are you trying to do E Girl Trolling / Catfishing?

[low shard]

what's your PC GPU and OS?

[low shard]

elaborate more pls

[broken urchin]

can someone help me do my vonovox settings

[dull shard]

Using kaggle to fork notebook in order to use gpu quota as my pc is weak
Ive encountered a problem

[low shard]

This is a General AI Discord server, we don't do only one program, Elaborate:

• Your pc gpu
• Your pc os
• What you're trying to do, like TTS, E Girl Trolling / Catfishing, Training RVC models, Roleplaying, AI Covers
• The tutorial link used

[dull shard]

12th Gen Intel(R) Core(TM) i5-1235U (1.30 GHz)
RAM: 16,0 GB
Window 11
I want to use voice to roleplay
The tutorial I used before is gone, but this is the link: https://www.kaggle.com/code/suneku/voice-changer-public

I am from a poor country and that's why my pc sucks, also don't know coding at all, consider me dumb

[low shard]

that's wokada deiteris fork kaggle jupyter notebook, wokada deiteris fork didn't have updates in a long time, and it doesn't work as it uses Ngrok and Ngrok downloads are down https://status.ngrok.com

You can either wait, or use the updated wokada tg-develop jupyter notebook lightning ai guide: https://docs.aihub.gg/realtime-voice-changer/cloud/tg-develops-w-okada-fork-cloud/#lightningai

be sure also to not miss https://docs.aihub.gg/realtime-voice-changer/cloud/tg-develops-w-okada-fork-cloud/#virtual-audio-cable

[dull shard]

Thank you very much, seems like a temporary issue indeed
Another alternative is definitely helpful, I will check it out

[low shard]

yw, btw it's just better to switch overall, wokada deiteris fork seems currently abandoned and isn't suggested

[dusk sphinx]

anybody else suffering the ngrok problem ?

[low shard]

This is a General AI Discord server, we don't do only one program, Elaborate:

• Your pc gpu
• Your pc os
• What you're trying to do, like TTS, E Girl Trolling / Catfishing, Training RVC models, Roleplaying, AI Covers
• The tutorial link used

[graceful scaffold]

SAVE WEIGHTS PLEASE

[low shard]

it's shutting down in like 2 hours

[hardy yew]

yeah, i've seen people mention this yesterday here

also just had it when i started another training ~1h ago

I don't really need it so I just removed the connection

[dusk sphinx]

yesterday I mean like 20 hrs ago everything was fine I was changing my voice during roleplay I thinks thats a temporary problem

thanks for the update @hardy yew

[hardy yew]

see Nick's message in #📰│dev-updates

[low shard]

shared more info about it in #📰│dev-updates message, if you elaborate what I asked I could find you alternatives like I did for other users tho

[dusk sphinx]

Thanks Nick appreciated

[supple rain]

Now that Weights has officially shut down, what are we gonna do?

[low shard]

What's your PC GPU and OS?

[untold escarp]

hey how do i set the voice changer to my mic\

becuasse im not seeing it

i hear it in the client but like in discord

what could you be typing bro

[rich harbor]

you need a virtual audio cable,

you make your microphone ur normal mic in the voice changer software, the output virtual audio cable, and the micrphone should be virtual audio cable in discord,

at least im assuming thats how it works, i use it for making vids but that should work

[untold escarp]

oh ok

ily

[rich harbor]

ya, i use this one

https://vb-audio.com/Cable/

[dusk cosmos]

Hi everyone! Can anyone help me set up a realtime voice changer client?

[rich harbor]

@dusk cosmos whats up

[dusk cosmos]

the voice is very robotic

[rich harbor]

a lot of rvcs are kinda bad

it probably ust that

or

[dusk cosmos]

I don't know, I just downloaded this app for the first time.

[rich harbor]

you gotta change the chunks and extra to be higher

i can send u a decent one in dms, it shouldnt sound robotic

[dusk cosmos]

I'm using chunk 192 and max extra (I have a 4060ti if that matters)

[dusk cosmos]

yeah come on

[rich harbor]

oh yea i usually put chunk to like 900

[untold escarp]

so i got the software

[rich harbor]

or higher

[untold escarp]

and

idfk

[dusk cosmos]

It's just a really big delay.

If anything, I apologize for my clumsy English, I use a translator.

[rich harbor]

its not the best software if you want to communicate in real time

i think its better at making audio

[dusk cosmos]

Which one can you recommend?

[rich harbor]

maybe if ur pc is amazing u could use it

tbh idk i dont use it for that

i use that software for making vids

not for like discord calls or trolling

[dusk cosmos]

My friends and I downloaded it just for a laugh, but my friends didn't want to tell me how to set it up.

[rich harbor]

ya, the delay is normal tho, not much u can do about it other than lower chunks, but the quality gets worse

[untold escarp]

vcan u make this more simple

[rich harbor]

theres probably decent balance you can do after testing it yourself for your computer but

[rich harbor]

audio cable makes output audio into input audio, so basically speakers into your microphone

[untold escarp]

mmm

yeah

[rich harbor]

so you just use virtual audio cable in those two places and it should work

can be used for a lot of stuff

[untold escarp]

should i change it in windows too

[rich harbor]

nah

only software needs it

[untold escarp]

so u want me to chnage my headphoens to the vb \

thats it

[rich harbor]

headphones dotn change

[untold escarp]

listen

im special

ok

[rich harbor]

LOL

[untold escarp]

u need ot go oogva booga style on me

?

[rich harbor]

voice changer microphone: normal microphone
voice changer output: audio cable

ya

[untold escarp]

okok

but

how ould i hear it

then

[rich harbor]

make the cable go to where your microphone would be in the software u want it to go

microphone should be audio cable

speaker doesnt change

[untold escarp]

so swap

for discord

?

[rich harbor]

ya just make the microphone audio cable

doesnt matter where u use it

[edgy dew]

Darn bot tryin' to mess with me, when I already understood the first time. X3 But hey, guys.

[odd shale]

Yes.

[subtle cedar]

Since that Weights is already shut down so which is/are modern Google Colabs or Gradios that create AI Covers?

-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

[subtle cedar]

Which is those Colabs or Gradios that we can create AI Covers?

Can someone pls answer me thanks

Sorry but because that guidelines doesn't help a lot and some probably are paid or only works on PC or laptops ✌️

[subtle cedar]

Oh only that RVC-AI-COVER-MAKER-WebUI one is for AI Covers but I think the rest have also a features that can create AI Covers too

[supple trail]

the only time i return to use wokada somehow ngrok is offline 🥀

[hallow thistle]

Help topics about W-Okada are getting too boring, so it's a good time period to take a break from voice changer all of a sudden.

[hallow thistle]

You have mentioned about using Applio RVC years ago. Do you remember that? While the core features in Applio RVC are the same, there are some more features added in latest versions.

[subtle cedar]

Oh okay thanks

[supple trail]

hey i have a question is hrzn.run usuable in kaggle considering kaggle projects usually use ngrok? but i recently saw a project in colab use hrzn an alternative to ngrok

[hallow thistle]

Tg Develop's W-Okada Colab notebook has HRZN option available.

[supple trail]

yea but colab

requires me to havea paid plan no

?

[hallow thistle]

What? Applio RVC Kaggle notebook now has HRZN option available, but for W-Okada voice changer Kaggle notebook, I don't see anyone ever done that yet.

[supple trail]

yeah im gonna code it

just asking if its possible

[plain berry]

guys

are there any ai that can auto highlight important information or the like

i dont want a summary, i want the ai to highlight things like definitions straight on the text if u get what i mean

[supple trail]

done 😄

finally after a while i finished and made it work on hrzn instead

[dusk sphinx]

yes

-kaggle

[patent trellis]

📘 Kaggle Notebooks

Kaggle is a Cloud (Remote Good PC) Service that offers 30 hours of GPU weekly, but needs a phone number verification

• **Applio Notebook**

by IAHispano
Kaggle

• **Hina Mod Original Wokada**

by Hina
Kaggle

• **Wokada Deiteris Fork**

by Hina & Deiteris
Kaggle

• **UVR5 UI**

by Eddy, ArisDev & Nick088
Kaggle

• **UVR5 NO UI**

by Eddy
Kaggle

• **RVC AI Cover Maker UI**

by Shirou & ArisDev
Kaggle

• **Music Source Separation**

by Shirou
Kaggle

[low shard]

This is a General AI Discord server, we don't do only one program, Elaborate:

• Your pc gpu
• Your pc os
• What you're trying to do, like TTS, E Girl Trolling / Catfishing, Training RVC models, Roleplaying, AI Covers
• The tutorial link used

[low shard]

we don't suggest vb audio cable for windows, did you came form a youtube tutorial?

[low shard]

This is a General AI Discord server, we don't do only one program, Elaborate:

• Your pc gpu
• Your pc os
• What you're trying to do, like TTS, E Girl Trolling / Catfishing, Training RVC models, Roleplaying, AI Covers
• The tutorial link used

[rich harbor]

oh what yall use, is there a better thing for better quality audio?

and nah i just knew about it

only thing id care about is better quality

[hardy yew]

the quality is the same, it's just that Muzychenko's VAC lite is recommended over VB cable because it usually causes less issues

but if you got VB and it works fine for you, no reason to change

[low shard]

Can you elaborate what tutorial link do you use?

[supple trail]

gosh i was so wrong for saying that only now did i finish 🥀

had to redo the whole uploading system because of hzrn limitation

[low shard]

there are way better and easier alternatives, please elaborate:

• your pc gpu
• your pc os
• what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
• the tutorial or kaggle link used

[supple trail]

1. im on a 1050 ti i discussed this before alot of times so i prefer staying on a cloud hosted version till i get a new pc.
2. windows 10.
3. roleplay, realtime tts.
4. tg-develops det fork of w-okada but i just coded it so that it runs on hzrn instead and uses a whole new uploading method

[low shard]

1. you could actually try to run it locally, but your choice

2. w-okada only does realtime RVC, not tts, for tts check https://docs.aihub.gg/tts/realtime-tts/

3. that's great, you could also use wokada deiteris fork lightning ai which is safer and has already multiple tunnels https://docs.aihub.gg/realtime-voice-changer/cloud/tg-develops-w-okada-fork-cloud/#lightningai

[supple trail]

1. tried it before quite really had my pc struggling when used on heavy games + causes bad lag
2. my bad i meant real time rvc forgot
3. no money + already had an account on kaggle just more used to being on kaggle instead

[low shard]

about locally: you should run always games with the lowest quality settings when using the ai realtime voice changer locally btw

about lightning.ai, it's free btw, would be nice if you make an account for it since it actually allows Web UIs in the free tier and doesn't have a ban risk unlike Kaggle, and I mean having 2 accounts (one for kaggle, one for lightning.ai) might be better

if you want you could PR your solution https://github.com/tg-develop/voice-changer to help everyone, or I might just PR all my tunnels code I use in the lightning.ai port in it instead