✨│ai-help

[undone orbit]

@viscid moss @simple ore HUGE thanks for your help guys 🙏 and once again applogies for the spam but this was driving me crazy

[viscid moss]

dw ur welcome bud

[simple ore]

just remember that the console window is where you actually should look for errors

gradio Ui that everyone uses is generally not helpful

[undone orbit]

I recently read that using two GPUs will not accumulate the VRAM amount for the purposes of AI. Is that true? As I was also thinking about buying another 5060ti to get 32gb of vram in total.

[simple ore]

you can run a big LLM if you have two cards of the same type

you can't do shit if you have one nvidia card and one amd card

you're using like 2-year old program

-rt

[patent trellis]

💻 Local Realtime RVC

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

• Wokada Deiteris Fork

Most suggested WebUI with the best general support for many platforms. GUIDE

• Vonovox

A Realtime Voice Changer with similar performance to Wokada Deiteris Fork, with extra features, but supported only for Nvidia GPUs on Windows. and without cloud options GUIDE

• Original Wokada

Not suggested nor maintained, older versions in youtube tuts are even way worse. GUIDE

• RVC GUI Mainline Realtime

Unavailable, the guide is outdated and the program is worse compared to the ones above, and much less updated

[simple ore]

1st link

[low shard]

you're using an over year old version of original wokada

and prob using vb audio cable, which causes issues on windows, don't trust yt tuts

delete the zip, folder and uninstall from windows app settings

[low shard]

@fossil cosmos read wokada deiteris fork

[simple ore]

you're installed what in AI terms is considered and ancient piece of mammoth shit app

no wonder it is ass

read the guide above and uninstall your garbage as well

there was some effort made to improve the apps performance

[brittle wing]

@simple ore i just rented the gpu because now im free to train the model, do you have any idea on how i can train on it?

[low shard]

yes, better performance

working doesn't mean it works good lmao

[simple ore]

you should have an option to SSH into your instance

pytty or some other ssh client

[low shard]

yes you can play certain games on an old laptop, it works, but would it work good on high graphics? mostly not

[brittle wing]

what's SSH bro 😭

[simple ore]

secure shell

[brittle wing]

what does that do, and how can i do it?

[simple ore]

does it say anything about how to connect via ssh?

[brittle wing]

Based on how you plan to connect to your Vast instances, you should have your account and device set up accordingly. You can learn more about our launch modes here. If you are interested in SSH, you will need to create an SSH key and upload to your account, or if you want to use Jupyter you may need to install the Jupyter certificate
for Vast. To make sure these modes of connecting to your instances are set up properly, we recommend you follow our set up documentation step-by-step, as we find some users have trouble with this portion of set up.

this is what it says

[simple ore]

Jupyter Terminal perhaps?

look, i've never used it, so I can't really help here

[brittle wing]

all good im watching a tutorial rn hopefully it works bruh 😭

this shit too complicated

[simple ore]

I'm sure there are a million guides how to run a python cuda project on vast.ai

[brittle wing]

bro i think i got it it's on jupyter notebook

i'm in it

[tribal fog]

Hi! I have a question. I'm new to this and I'm using the voice changer, but when I speak, it takes a while to change my voice, and sometimes it can't even get what I'm saying right. I suspect it has to do with the section that says "res" in the RVC box because it's the only thing I see with different values in the YouTube tutorial video compared to mine. What does that "res" value mean?
Does it have anything to do with the processor?

[low shard]

YouTube tutorial video
All video tuturials are outdated, you shouldn't trust them

Please Elaborate:

• your PC GPU (Such as Nvidia RTX 3060, AMD RX 9070 XT..)
• your operating system (Like Windows 10, Windows 11, Linux, MacOS 26..)
• what you want to do? There isn't a program that does everything, there's a program for each thing:
  • AI Covers
  • Train RVC Models
  • TTS
  • Roleplay in VC
  • Roleplay in Games
  • etc
• what tutorial link are you using
• a screenshot of the program

[tribal fog]

Intel(R) Iris (R) Xe Graphics
Windows 11 pro
I want to use it for changing my voice to give new voice to a Vtuber Avatar to stream on twitch with that avatar. So neither my face and voice would be the real ones XD
Tutorial: https://www.youtube.com/watch?v=q4i68Uxcv8E&list=LL&index=1&t=297s&ab_channel=EnzoRE%3AView
Screenshots:

I can't send pictures here...

can I DM them to you?

[low shard]

Tutorial: https://www.youtube.com/watch?v=q4i68Uxcv8E&list=LL&index=1&t=297s&ab_channel=EnzoRE%3AView
it uses an over year old version of original wokada and vb audio cable
simply, forget it ever existed and its settings, it was a waste of time, delete the zip folder and uninstall vb audio cable from windows app settings

[low shard]

Intel(R) Iris (R) Xe Graphics
that's integrated graphics, weak asf, have you checked if you have any other GPU 0 or 1?

[tribal fog]

It's the only one the laptop has

[low shard]

Your pc is too Weak for Wokada locally, You got 3 options:

• Buy a better pc
• Run it locally (on ur pc) using the CPU mode of the wokada fork which has better performance https://docs.aihub.gg/realtime-voice-changer/local/deiteris-w-okada-fork/ (but this isn't suggested as it could be unstable and unreasonably high ping)
• Use **cloud **(remote good pc):

About Cloud, there are different services:

• Google Colabs (4 hours daily of free T4 gpu, easy to use, require only a google account) :
  • W-Okada's Deiteris' Fork Voice Changer Google Colab (currently works only on google colab PAID tier)
• Kaggles (30 hours weekly of better GPUs, T4x2 & P100, harder to use, requires an account and a phone number):
  • W-Okada's Deiteris' Fork Voice Changer Kaggle (the best and only working one currently for free)

[tribal fog]

I'm using that one because i'm new to this and I don't know if i'll understand how to use the tools without a tutorial XD

[low shard]

I understand, just never trust video tutorials for realtime ai voice changers, they are all outdated

based on your hardware situation and needs, I gave you the tools that fit you

AI is intensive, it needs good hardware to run it locally
you can't expect to run it on an old weak laptop

[tribal fog]

What does high ping means?

[low shard]

it will lag, it won't be good in most cases
what's your PC CPU btw?

[tribal fog]

Also the laptop is not that old, it's not even 2 years old

[low shard]

that doesn't mean that it's good, you can buy new laptops that are weak,
the date of when you bought it or was made is not the only factor that defines how good the performance is, that depends also on the actual components inside it, and your laptop doesn't have any dedicated GPU

[tribal fog]

13th gen intel (R) core TM i5 -1335U

[low shard]

ehh, you can give it a try, don't expect it to work in games though since you're going to stream

you can test, but i feel like it would be just a better option to use Cloud, aren't 30 hours weekly for free good enough?

[tribal fog]

yeah! the only thing that takes me back is that I don't know how to use the tool but I can give it a try 😅

basically, I need the laptop for the voice and the Avatar, the games are on the PS5

[low shard]

You can start by reading https://docs.aihub.gg/realtime-voice-changer/local/deiteris-w-okada-fork/ and asking me for any issues

[tribal fog]

Thank you!

[low shard]

Yw, lmk!

[tribal fog]

The tutorials for Colab and Kaggle are down (404)

[low shard]

they seem to work fine?

[low shard]

you sure you used those links

ohh you're talking about https://docs.aihub.gg/realtime-voice-changer/local/deiteris-w-okada-fork/#online-alternatives-colabkaggle, gonna fix it right quick
btw this is only for cloud, not for CPU local

[tribal fog]

the part for online alternatives

yeah

yeah, I'll use a cloud option since you said it would be better considering my laptop. I just want to use it to make streams with my avatar so 30 hours per week is more than enough

since problably i'll stream just on weekends and in case i stream on a week day it wont be more that 2-3 hours that day

[low shard]

should be fixed now! retry pls

[tribal fog]

Yeah! they work now! thank you very much!

[low shard]

Yw!

[tribal fog]

Sorry, but which is the good version of the voice changer program I have to download?

[random dune]

im a teenager and im looking for a realistic ai boy voice model just to make me sound different but still like a teenager

[simple ore]

git clone applio repo

run python, try import torch print(torch.__version__)

if you get cuda version then next would be cd Applio (clone applio folder)
and uv pip install -r requirements.txt

[tawdry shell]

Is this normal?

[tribal fog]

on models search for Adit or Gotcha force. Those could help you sound a bit diferent but still young

[hallow thistle]

If you see this, it means that Weights bot was down for a brief moment or it took too long to respond so it timed out.

[tawdry shell]

Every time I use the /find it does that hahaha

[hallow thistle]

Although the bot still works for me.

[tribal fog]

I was running it on Kaggle, it opened the page in ngrok, and when I tried to change the name of the model the page colapsed, i pressed initialize and it didnt work, I tried to rerun the last cell and it says:

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 140, in main
File "main.py", line 65, in runServer
File "main.py", line 49, in check_port
OSError: [Errno 98] Address already in use

Press Enter to continue...

but i'm pressing enter on my keyboard and nothing happens

I re run everything again

and it worked

But when i try to edit the model's name it crashes again

[low shard]

you shouldn't rename the model on the Web User Interface, you can just rename the file and then reuplload the slot, it's a common bug

[tribal fog]

ah okey

[hallow thistle]

Additionally, you can rename the model name directly in MMVCServerSIO folder with Notepad.https://cdn.discordapp.com/attachments/1159290139609137264/1374779239437701220/image.png?ex=688f87f0&is=688e3670&hm=494e14a9e2505be92468baeacf6434ce9fef2023abe863613264d7e8a26d8d02&

[tribal fog]

Thank you!!!

also, if the voice sounds a bit robotically which parametre should i modify on the dashboard to make it sound a little bit more natural?

(It's just some times that the sentences end up sounding with a bit of robotical sound on the end, for the rest the tone and everything is perfect)

[cold cave]

Why do I keep getting this message?

"Not enough data present in the training set. Perhaps you forgot to slice the audio files in preprocess?"

[viscid moss]

@spark mesa

show me the GPU on ur task manager

[spark mesa]

omg, this is the same experience my friend is experiencing

i was supposed to message that in here too

[spark mesa]

wait i forgot how to check it

what's the command again

[viscid moss]

Ctrl + Shift + Esc

[spark mesa]

someone in here already told me about it

oh yeah that

where do i go after?

[viscid moss]

Performance or something like that

then GPU

[spark mesa]

i can't screenshot it..

wait

there, i opened the snipping tool manually lmao

[viscid moss]

u have a GTX

:3

isn't the best thing but it's better than just a CPU

I think UVR5 is using it, but can confirm at all

also Roformers are so computer intensive

[spark mesa]

ohhh that explains why

guess the only thing i can hope for is to find something like the uvr on huggingface to use gabox models

i personally like the v4

thank you, eddy!

[golden walrus]

Sire, i want to ask about how can i get another model. Like on the guide it's said Dereverb v2 by Anvuew but i only see v1

[viscid moss]

ur weldome. U can use Colab or Kaggle it's pretty ez

[viscid moss]

V2 is normal one iirc and V1 is aggressive one

[golden walrus]

Oooohhhhh.

Thank you sire

[viscid moss]

ur welcome

[hasty kite]

Hi. I'm trying to run Stable Diffusion with zluda, but this error pops-up, what could be the issue?
hiplnfo.exe - System Error

The code execution cannot proceed because amdhip64.dll
was not found. Reinstalling the program may fix this problem.

[turbid rover]

https://www.instagram.com/reel/DKZJxAWt_ms/?igsh=bzFpM3JoaTlwazhj
hello plz can someone tell me where do i find this voice over?

[latent kettle]

External links are not allowed.

[echo crown]

could somebody help me figure out how to train an ai im not experienced and i wanna get into ai voice training if not thats totally fine

[vital mesa]

Hii, is there a way i can fix the ai voice in games such as OW or marvel rivals? is keeps buffering

[velvet solar]

Hi, so, I had a small question, and im not sure where else to put it. What voice changer app would I use with the models here? I'm not sure which one it'd be. Im assuming W-Okada, but when i do the httpstart.bat file, Windows defender pops up and im not sure whether to allow it access or not.

[brittle wing]

Hello, i once used Pinokio to quickly set up Forge webui for SD, and i generated some pictures here and there, ii had 2 problems so far with it, it takes quite alot of time to generate 1 image (25 min or so for 1 image especially at 512x512 or less resolution), i have 4gb vram so that might explain it, second problem is that i can't find a way to make prompting conveniently easy, i struggle with english alot so i can't describe things in a prompt that i want from imagination and therefore rely on other peoples generated images to copy their prompt and change it to my needs, but this is very time consuming and tedious especially with the time it takes to generate an image, i was wondering if there is a way to make prompting much easier and comfortable, and also a way to improve performance for faster generation time (if it's possible to host it on google colab i wanna do that instead).

This is my PC Specs:

     Memory: 16384MB RAM
            Page File: 21011MB used, 16871MB available

Operating System: Windows 10 Enterprise 64-bit

            Processor: AMD Ryzen 5 5600H with Radeon Graphics          (12 CPUs), ~3.3GHz


       Card name: NVIDIA GeForce RTX 3050 Ti Laptop GPU

[minor isle]

are models still bad with laughs?

[fast osprey]

if you are talking about RVC well it depends on the quality of the voice model you're using, actually theres quite alot of model that can laugh realistcly

[minor isle]

all the ones I tried cant laugh at all

which ones should I try?

[fast osprey]

are you looking for a male or females voice?

[minor isle]

any works, but rn looking the female ones

[fast osprey]

hm

wait

[fast osprey]

lemme check what female voices i have rn

[minor isle]

[fast osprey]

well i do got a few, altho you may need to increase the chunk for them to actually sound good i had a few questions for ya tho, whats you gpu, and are you looking to do roleplay or e girl trolling/catfishing or like what're u tryna do, btw i'd assume you have used rvc models before if not theres alot of guides on that

[minor isle]

i use rtx 2060

how much chunk

[fast osprey]

ahh f mb i can't find where i downloaded the models from, and theyre too large to send over here. sorry
but could just search for a little more i believe you can find a high quality one pretty quickly

[fast osprey]

about 700 is really good, but i'd say atleast 400-500

[minor isle]

how do i know its high quality, most of these sound robotic asf

[minor isle]

isnt that too much

[fast osprey]

you just gotta try the models, just hear the preview sound and see which one sounds less robotics and download those ig, keep trying until you find a high quality one

[undone orbit]

Hi, I'm watching a yt tutorial on preparing datasets for training RVC models and the guy is using whisperx. However, I don't understand whats the purpose of it (my plan is to use Applio).

[fast osprey]

not really

[vast elm]

Most yt tutorials are outdated

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

check out this guide on preparing the dataset, its the most recently updated

you gotta keep in mind, preparing the datset is the most important part of making a model, so a higher quality dataset = a higher quality model

if you had any questions needed help following the guide just tell me and ill try my best to help

[undone orbit]

@vast elm Thank you Sir! 🙏

[vast elm]

ofcourse, glad to help

[hallow thistle]

The correct name for audio track separator is UVR5. This "Dataset & Isolation" doesn't tell you about a specific RVC program. However, this section somehow located in **rvc**/resources/dataset-isolation/ which made you think it was about a specific RVC program named dataset-isolation.

[knotty moth]

try some other models that may sound better

also I'd recommend wokada deiteris fork esp if you're using non-Nvidia gpu and directml version

[undone orbit]

Does anyone know if iZotope RX11 standard is enough for the RVC purposes or do I need an advanced one ? 🙏

[hallow thistle]

I use UVR5 and demucs to remove background noise audio, but I used some models supposedly for extracting vocals and instrumental on vocal-only audio files.

[undone orbit]

Yeah was also wondering of it's worth the extra 'hassle' to use iZotope

[hallow thistle]

I searched it on Google, and I see this pricing option. I ain't gonna "pay" on this one, regardless of its tiers available for purchase.

[undone orbit]

standard one is available to rent at $20 per month so I could probably bite the bullet but no such option for an advanced one 🙁 hence my question

[hallow thistle]

You could've go for UVR5, which is a free and open source program; the UVR5 itself won't come as a VST plugin for DAW software, and may require you to install some extraction models to use the program.

[simple ore]

🏴‍☠️

[hallow thistle]

You can read more from this docs. https://docs.aihub.gg/rvc/resources/dataset-isolation/

[simple ore]

advanced on high seas

[undone orbit]

I'm already using UVR5. I thought RX11 would be the next step to clean the audio even further?

[simple ore]

only if you really want to go into advanced audio editing

it would take probably a few months to get familiar with it, watch tutorials and stuff

[dusty scarab]

Guys I need some TTS. I can run locally or use a site but I don't have nvda but amd radeon

[simple ore]

what gpu?

[minor isle]

I'm using 2060 and using the normal fork version afaik

[paper widget]

1. RTX 4070 8GB
2. win11
3. Like Urdu Hindi, only xtts was able to do it, but it does not clone well.

[low shard]

I'm not aware of any other tts tools that does hindi, maybe @simple ore knows

[simple ore]

kokoro has hindi, I think

[spark mesa]

eddy here

[viscid moss]

yep

[spark mesa]

what alternatives can i use again?

for uvr hf

[viscid moss]

Colab/Kaggle/Lightning ai/Local

[spark mesa]

where can i find those?

[viscid moss]

I think Colab it's the easiest way

[spark mesa]

cuz one of my friends who uses uvr hf a lot, also wants to find an alternative to use the other models

[viscid moss]

-colab

[patent trellis]

📒 Google Colab Notebooks

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

• **Applio**

by IA Hispano
Google Colab

• **RVC Mainline**

by Hina
Google Colab

• **UVR5 NO UI**

by Eddy
Google Colab

• **UVR5 UI**

by Eddy
Google Colab

• **Wokada Deiteris Fork**

by Deiteris & Hina
Google Colab

• **Hina's Modified Original Wokada**

Google Colab

• **RVC-AI-Cover-Maker-WebUI**

by Shiro & Eddy
Google Colab

• **FaceFusion UI**

by Nick088
Google Colab

• **FaceFusion NO UI**

by Nick088
Google Colab

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

by Jarredou & Makidanye
Google Colab

[viscid moss]

just type "-colab", "-kaggle", "-lightning" and bot will show u the links

[spark mesa]

ahhh alright

oh

[viscid moss]

yep

[spark mesa]

i use that colab one for making models

i didn't know there was a uvr there

[viscid moss]

https://github.com/Eddycrack864/UVR5-UI
Local is here

[viscid moss]

lol

[spark mesa]

how do i check the uvr in there tho?

do i just run the applio?

[spark mesa]

AHHHH

LMAOOO

😭

[viscid moss]

😭

[spark mesa]

not me clicking the first one

you saved me

wait let me try it first

[spring socket]

Today I want to open my Okada (start_http.bat), Then this happend. Can someone explain what to do?

[viscid moss]

sure, go ahead

[simple ore]

look at this guy using start_http.bat

pathetic

[spark mesa]

oh

[spring socket]

😭 ALWAYS USE THAT SINCE LAST YEAR

[simple ore]

I also used a horse and buggy 40 years ago

[simple ore]

you need to click the popup and enable its use of google drive

[quasi gyro]

What does the ring form architecture do for code name?

[viscid moss]

u should give it GDrive perms or disable that mark if u don't want to

[spark mesa]

alright let me do it

[spark mesa]

but is it required?

[low shard]

start_http.bat is apart of the old original wokada

[viscid moss]

nah, but it's useful for some features like.. batch separation

[low shard]

exacty, it's outdated asf

you deffo have vb audio cable too, which can create sometimes issues on windows

delete the zip, folder and uninstall vb audio cable from windows app settings

AI moves at sonic speed

what's your pc gpu, operating system and what you want to do?

[spring socket]

Yes, I use Okada & VBCable for my daily content, also start_http.bat
My GPU is 4060, Win 11

I want to be able to use the app again. Thankyou for helping

So, reinstall the VBcable first?

[low shard]

for my daily content
Oh you're a vtuber or smt?

[spring socket]

Yes

[low shard]

nope, don't reinstall anything, everything you got is outdated, i guess you also used crepe which is outdated lol

[spring socket]

RMVPE

hahahaha

[low shard]

remember to never use yt video tuts, since 99.99% of them have outdated info related to realtime ai voice changers

[low shard]

that was the only updated thing you were using lol

anyways

[spring socket]

I see

[low shard]

delete everything as i said

then

-realtime

[patent trellis]

💻 Local Realtime RVC

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

• Wokada Deiteris Fork

Most suggested WebUI with the best general support for many platforms. GUIDE

• Vonovox

A Realtime Voice Changer with similar performance to Wokada Deiteris Fork, with extra features, but supported only for Nvidia GPUs on Windows. and without cloud options GUIDE

• Original Wokada

Not suggested nor maintained, older versions in youtube tuts are even way worse. GUIDE

• RVC GUI Mainline Realtime

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

[spring socket]

So, any suggestion which or where do I get the updated version so I can start using it again?

[low shard]

you got 2 options, either wokada deiteris fork, or vonovox

[spring socket]

Hoooo wokada

[low shard]

the major difference is the User Interface
vonovox got a newer user interface, some bugs fixed but it's not mature yet, and some paid voice effects like low quality microphone

[spring socket]

Wokada deiteris fork is the newest / updated version from what I use then?

[low shard]

don't use the original wokada, i'm talking about the first 2 links lol
you can also check both of their guides and read each pros&cons if you want

[spring socket]

Yes sir!! I think I'm gonna go with the Wokada Deiteris Fork, anyways, thankyou so much for replying

[low shard]

yes, vonovox is another realtime ai voice changer, you could check also that if you want
they both do the same thing, and have some differences though (vonovox is newer)

[low shard]

alright, let me know for any issues!

[spring socket]

For now, I'm gonna try the first one. Thankyou Nick and others!

[low shard]

great, just keep me updated if you need anything

[spring socket]

It works! Thankyou for the guide
So from now on, the Wokada will open on the browser instead from seperated application?

Sorry for another dumb question, thanks!!

[low shard]

do you want me to check your settings while you're at it?

I can give you temporary image perms

[low shard]

So from now on, the Wokada will open on the browser instead from seperated application?
Yep, https://docs.aihub.gg/realtime-voice-changer/local/deiteris-w-okada-fork/#why-does-it-run-in-a-browser-and-not-its-own-window

[spring socket]

I don't understand what u mean I'll try my usuall setting like the old outdated okada first!!

[spring socket]

Thanks!!

[low shard]

I'll try my usuall setting like the old outdated okada first!!
some settings are different, you shouldn't use the same

!give-media-perms 1h @spring socket

please share a screenshot of your program settings

[velvet solar]

Still wondering about this if anyone can help

[low shard]

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

• Advanced Settings -> Force FP32 mode: on (THIS IS OFF BY DEFAULT! Turning this on improves stability. Increases VRAM usage by 200 MB)
• Advanced Settings -> Disable JIT compilation: off for faster loading speed of the program, on for slightly better performance (10-15 ms) for Nvidia only)
• Advanced Settings -> Crossfade Lenght: Controls how smoothly the AI stitches different processed parts "chunks" of your voice back together. 0.1 for fastest voice, 0.15 for improved quality but increases delay by ~50 ms
• Reduce the delay on Windows via the Wasapi / Asio Guide

[low shard]

Hello, remember to not use yt video tuts,
Please Elaborate:

• your PC GPU (Such as Nvidia RTX 3060, AMD RX 9070 XT..)
• your operating system (Like Windows 10, Windows 11, Linux, MacOS 26..)
• what you want to do? There isn't a program that does everything, there's a program for each thing:
  • AI Covers
  • Train RVC Models
  • TTS
  • Roleplay in VC
  • Roleplay in Games
  • etc
• what tutorial link are you using (if any)
• a screenshot of the program (if any)

[low shard]

Please Elaborate:

• your PC GPU (Such as Nvidia RTX 3060, AMD RX 9070 XT..)
• your operating system (Like Windows 10, Windows 11, Linux, MacOS 26..)
• what you want to do? There isn't a program that does everything, there's a program for each thing:
  • AI Covers
  • Train RVC Models
  • TTS
  • Roleplay in VC
  • Roleplay in Games
  • etc
• what tutorial link are you using
• a screenshot of the program

[low shard]

I'm guessing you mean RVC (STS) AI Voice Models Training
Please Elaborate:

• your PC GPU (Such as Nvidia RTX 3060, AMD RX 9070 XT..)
• your operating system (Like Windows 10, Windows 11, Linux, MacOS 26..)
• what you want to do? There isn't a program that does everything, there's a program for each thing:
  • AI Covers
  • Train RVC Models
  • TTS
  • Roleplay in VC
  • Roleplay in Games
  • etc
• what tutorial link are you using
• a screenshot of the program

[velvet solar]

AMD Radeon RX 580, Windows 10, and im wanting to do game roleplays. Use voices to sound like others, like Master Chief, or Serial Designation N. And tutorial link I had used was https://youtu.be/xwdbwbtO-FQ?si=I6nTl-m14_OfbQYK
I cant get a screenshot of the program, but i know on the github or huggingface website, it said the download link was at least over a year old for the W-Okada I downloaded

Is the W-okada i downloaded the outdated one or something?

[spring socket]

Thankyou so much! I already tweak the advance settings when I read the guide!! Thanks for the advice too!! 👍

[low shard]

that video uses an over year old version of original wokada, delete the zip and folder
also vb audio cable could possibly create issues on windows, delete zip folder and uninstall from windows app settings

AMD Radeon RX 580, Windows 10, and im wanting to do game roleplays. Use voices to sound like others, like Master Chief, or Serial Designation
Your GPU is the bare minimum, wokada deiteris fork will help in performance, but you need to not play any intensive games, and only 1080p 60fps cap, which games do you want to play?

[low shard]

You're welcome
So is everything solved now?

[spring socket]

RuntimeError: Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same

[low shard]

could you please tell me what did you do exactly?

[spring socket]

I got this little error and can't find the Torch.cuda the Wokada mention, can you help me?

[velvet solar]

I was mainly planning on using it for vrchat stuff. I had a master chief avatar and itd be fun to sound like him. And the VB audio cable Im not really sure about deleting since i use it for a lot of other stuff too, but I can try it with and without

[low shard]

sure, I just need to understand what did you do exactly to cause this error, even a screenrecording can help or just telling what you did via text

[velvet solar]

And ill look for that deiteris fork version of W-Okada

[simple ore]

if you changed precision to fp32, you may need to restart the application

[spring socket]

I only press "START", and the error pops up

[spring socket]

I SEE, I'll try it thankyou

[low shard]

And the VB audio cable Im not really sure about deleting since i use it for a lot of other stuff too
the VAC Lite does the same thing, and doesn't cause issues

vrchat
be sure to play on lowest settings

-realtime

[patent trellis]

💻 Local Realtime RVC

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

• Wokada Deiteris Fork

Most suggested WebUI with the best general support for many platforms. GUIDE

• Vonovox

A Realtime Voice Changer with similar performance to Wokada Deiteris Fork, with extra features, but supported only for Nvidia GPUs on Windows. and without cloud options GUIDE

⚔️ Wokada vs Vonovox

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

⛔ Outdated/Discouraged

These options are not recommended for use.

• Original Wokada

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

• RVC GUI Mainline Realtime

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

[low shard]

1st link in your case, wokada deiteris fork

[spring socket]

@low shard @simple ore Restarting just like Noobies said solve the error! Thankyou so much for helping me. The Conversion is now going smoothly

Have a nice evening~

[low shard]

Huh I didn't need to restart when changing the precision, is it a rare bug?

[low shard]

You're welcome

[simple ore]

if it has a model loaded, I guess it has f0 loaded with one precision and voice model with another

[low shard]

I just tried to load a model and start in fp16
then stop
use fp32
use the same model and start
and didn't need to restart

might it happen only sometimes?

[cyan gull]

yo, did they change anything in the applio colab? it doesn't seem to save all weights to my drive even though i enabled the checkmark. Also i don't see an option to load a backup

anyone help me.. i got 30 mins left on colab and i noticed that all the weights arent saved in my drive..

[paper widget]

How to use that

[simple ore]

https://github.com/hexgrad/kokoro

demo https://hf.co/spaces/hexgrad/Kokoro-TTS

[viral mason]

is it possible to train using batch 32 on the kaggle applio space? I get an error every time I try, I'm using a 1 hour 25 minute long dataset, I edited the video to cut out the waiting where it was just sitting there loading

[simple ore]

if you enable checkpointing maybe

and dont cache in gpu

[viral mason]

how would I do both of those things

[simple ore]

hm.. no checkpointing in noui colab

select view code, then change this to True

[dense flower]

why my voice stay with delay

i say something and comeback 3 sec later

[low shard]

This is a General AI Server, AI has many fields, so we can't know your issue with little info

Please Elaborate:

• your PC GPU (Such as Nvidia RTX 3060, AMD RX 9070 XT..)
• your operating system (Like Windows 10, Windows 11, Linux, MacOS 26..)
• what you want to do? There isn't a program that does everything, there's a program for each thing:
  • AI Covers
  • Train RVC Models
  • TTS
  • Roleplay in VC
  • Roleplay in Games
  • etc
• what tutorial link are you using
• a screenshot of the program

!give-media-perms 1h @dense flower

[viral mason]

where is this?

[dense flower]

gtx 1660 super
w10
roleplay

[low shard]

you're using an over year old version of original wokada

with also bad settings

dont trust yt tuts

delete the folder and zip

-realtime

[patent trellis]

💻 Local Realtime RVC

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

• Wokada Deiteris Fork

Most suggested WebUI with the best general support for many platforms. GUIDE

• Vonovox

A Realtime Voice Changer with similar performance to Wokada Deiteris Fork, with extra features, but supported only for Nvidia GPUs on Windows. and without cloud options GUIDE

⚔️ Wokada vs Vonovox

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

⛔ Outdated/Discouraged

These options are not recommended for use.

• Original Wokada

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

• RVC GUI Mainline Realtime

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

[low shard]

either use wokada deiteris fork or vonovox

[dense flower]

oh thanks

make sense

[low shard]

yw, lmk

[scenic vine]

boys i tried wokada but it doesnt work on discord

any help please ?

[honest sparrow]

use virtual audio cable

[scenic vine]

i did
still

and when i test on the app the voice is toooooo much delayed

i dont know why

can someone show me best settings ?

[honest sparrow]

hmmm

[scenic vine]

any solution ?

[honest sparrow]

id recommend trying deiteris fork

[scenic vine]

where to install it ?

[honest sparrow]

uhhh

docs.aihub.gg

there should be a download link somewhere

it’s js a better version of w okada btw

[tender zephyr]

I don't get it, I can hear myself in the program but when I switch input in discord it doesn't display the sound

Like it's only in the app

Anyone know something

Nvm im apparently a dumbass

[severe meteor]

Is there a way to have the deiteris fork RVC into an app on your computer instead of a tab on your web browser? Or is a tab on the web browser the only way to use that fork?

Thank you.

[simple ore]

you can try vonovox instead

w-okada uses browser for UI and it is trash, but it works everywhere

[vital mesa]

GeForceRTX 4060, Windows 11, and i use the an AI program that changes my voice in real time into characters i find in this discords AI models, I play games and call while i use the AI prgram. https://youtu.be/SxdnGxicJOg?si=EhG07BO540t-mto5, this is a link to the video i watched, and the program i use is also in it.

[low shard]

that tutorial uses an over year old version of original wokada
delete the zip and folder
it also sues vb audio cable which isn't suggedted
delete the zip, folder and uninstall from windows app settings

I see the tutorial is about "ai girl voice", are you trying to do e girl trolling/catfishing?

[low shard]

Please Elaborate:

• your PC GPU (Such as Nvidia RTX 3060, AMD RX 9070 XT..)
• your operating system (Like Windows 10, Windows 11, Linux, MacOS 26..)
• what you want to do? There isn't a program that does everything, there's a program for each thing:
  • AI Covers
  • Train RVC Models
  • TTS
  • Roleplay in VC
  • Roleplay in Games
  • etc
• what tutorial link are you using
• a screenshot of the program

[pulsar yarrow]

Alright I wanna know why was the realism section deleted in the guide? I was trying to figure out how to turn off playback with lighthost (even though I dont even think the guide told you how)

Then I dont even know why it was specifically named "Realism" when it was just an extra feature to add plugins via VST to the voice if you wanted

[viral mason]

I know how to set it up, I did it through trail and error

idk what the realism part meant tho

[pulsar yarrow]

I get that it might've been unstable at times maybe but you gotta let the user decide that instead of just outright removing it me thinks

[pulsar yarrow]

I have the whole thing set up with voicemeeter and all that, thank you still though. I just wanted to turn off the playback or "monitor" on LightHost

Unless u already figured that out

[viral mason]

I'm not sure what Lighthost is so I can't help with that sorry

[pulsar yarrow]

Oh

[low shard]

The great majority of users of the users reported issues with it, it's not suggested and won't be maintained nor support

also, you don't really need it as long as you have a good model tbh

[pulsar yarrow]

Wait so what other way were you talking about

Through trial and error

[pulsar yarrow]

Ehh Idk

Having the extra autonomy is nice

[low shard]

there's also vonovox, which is a newer ai realtime voice changer that has built in (so less delay) voice effects, some free some paid

but we won't help for that old lighthost guide, it was broken

[viral mason]

free fx through a daw like fl studio is 100x better than buying fx

[low shard]

again, those are built in, and offer less delay, and easier to use

and actually work, rather that old realism guide

[viral mason]

if u wanted I could show you how I have set up mine, although I believe it causes a slight bit of extra delay

[pulsar yarrow]

Rn Im using Dieteris and I assume that its better than Vovonox, no? I thank u for the suggestion and I have tried vovonox but Im confused on why it doesnt have an index slider

[viral mason]

not too noticable tho

[pulsar yarrow]

Even though I know if you have a good model, index isnt really necessary. But again still nice to have

[low shard]

Rn Im using Dieteris and I assume that its better than Vovonox, no?
Nope
They both are based on RVC, meaning they have similar quality and performance
Vonovox is still getting updated, and fixed some bugs (even though it isn't super mature yet)
Wokada deiteris fork update was over 8 months ago (december 2024)

[pulsar yarrow]

Yuhh my friend

As long as I dont have the playback Im good

[low shard]

Im confused on why it doesnt have an index slider
I think that is being worked on, this is basically an updated work in progress

[viral mason]

Vonovox, in the most recent beta has index support and it's really good tbh, although I still use Dieteris since it has more model slots

[pulsar yarrow]

Right okay. Well I might try again soon, I still like that it can use spin embedder so it has a slight edge over deiteris in this case

[low shard]

Vonvox has actual potential, deiteris isn't being active since over half a year unfortunately
maybe he will comeback one day but who knows

also the creator of vonovox (dr87) said he might work on an open source version

[pulsar yarrow]

Ohhh alright then yeah Ill try it again thanks for letting me know

[echo crown]

can smbdy help me with training ai nobody seems to help i need somebody to elxplain how to do it in bite sized pieces or a video tutorial please 😭

[viral mason]

I've tested actually and using the same model you can get more nuance from vonovox than Deiterus

[low shard]

all video tutorials are outdated

Please Elaborate:

• your PC GPU (Such as Nvidia RTX 3060, AMD RX 9070 XT..)
• your operating system (Like Windows 10, Windows 11, Linux, MacOS 26..)
• what you want to do? There isn't a program that does everything, there's a program for each thing:
  • AI Covers
  • Train RVC Models
  • TTS
  • Roleplay in VC
  • Roleplay in Games
  • etc
• what tutorial link are you using
• a screenshot of the program

[pulsar yarrow]

Ok cool. I see him around in #🔊│ai-development and I learn from him sometimes. Im a little upset that Codename and SSS left though

[scenic vine]

guys what is the best ai girl ???

[low shard]

I'm not so sure about the SSS situation
but codename was related to an useless drama

[low shard]

are you trying to do e girl trolling/catfishing?

[scenic vine]

no just pranking my boys haha

[low shard]

so e girl trolling people?

[scenic vine]

i wont dare to do such thing

[scenic vine]

if u say its for catphising and scam thats an another talk

[pulsar yarrow]

Yeah I like having a lot of model slots to mix and match indexes and models, also having the picture is nice as well. So what do u use for the DAW and VSTs and all that

[viral mason]

I use FL studio for the DAW, and honestly I use a lot of different plugins and vsts, most of the time tho I just have at most a denoise plugin for voices that don't need it although for something like GLaDOS I use a paid Autotune vst or the battle droids I use Ultrapitch which is also paid, only two I have ever bought in my life tho

[pulsar yarrow]

Yeah I saw that whole thing take place, honestly I do think that channel should be a little more obfuscated due to that "redirecting" thing happening pretty commonly

I normally just watch what they talk about in the background, not knowledgeable enough on this to pitch in

[pulsar yarrow]

Ok Ill try that, I got a couple DAWs but I didnt know FL studio worked for real time. Thank you

[echo crown]

alright well i wanna do ai voice covers and roplay in vc and i am running windows 11 with a AMD Radeon RX 550

[viral mason]

I can send u the peak setup in dms, I'm recording how to train in applio kaggle rn for someone

[low shard]

AMD Radeon RX 550
that is lower than bare minimum

[echo crown]

dam

[low shard]

Your AMD GPU is might be able to do inference (use models) locally (on ur pc), ofcourse not train

You can:

• Locally (runs on your pc so the speed depends on that, you will have to set it up with the guides):
  • Applio (AMD Windows) : A fork of RVC with some extra features like Applio TTS, kinda faster and simpler but same quality tho
  • Mainline (AMD Linux/Windows) : The original RVC
• Cloud (remote good pc, easier and faster than ur PC but it's limited):
  • Weights.com: Partnered with AI Hub, lets u do them easily but u may be in a queue
  • Applio Colab: max 4 hours daily, not granted, of GPU
  • RVC-AI-Cover-Maker-U Colab & Kaggle: Automatically separates the vocals and instrumentals, converts the voice and mix all together back

Easiest possible (automatically separates vocals & instrumentals) : weights.com & rvc-ai-cover-maker-ui colab/kaggle
easiest cloud: Ilaria rvc zero
easiest local: Applio

[low shard]

Your pc is too Weak for Wokada locally, You got 3 options:

• Buy a better pc
• Run it locally (on ur pc) using the CPU mode of the wokada fork which has better performance https://docs.aihub.gg/realtime-voice-changer/local/deiteris-w-okada-fork/ (but this isn't suggested as it could be unstable)
• Use **cloud **(remote good pc):

About Cloud, there are different services:

• Google Colabs (4 hours daily of free T4 gpu, easy to use, require only a google account) :
  • W-Okada's Deiteris' Fork Voice Changer Google Colab (currently works only on google colab PAID tier)
• Kaggles (30 hours weekly of better GPUs, T4x2 & P100, harder to use, requires an account and a phone number):
  • W-Okada's Deiteris' Fork Voice Changer Kaggle (the best and only working one currently for free)

there's 2 different programs for realtime and ai covers basically

[echo crown]

okk

[pulsar yarrow]

Alright cool, no rush. DM anytime!

[echo crown]

Yeah cause i use UVR5 perfectly fine is it just that it takes like alot of processing power

[low shard]

uvr5 is only for separating vocals and instrumentals

[echo crown]

i know but im talking about the second part of the process of throwing the ai onto the vocals

[severe meteor]

What are some common reason why Deiteris fork isn't picking up my Virtual Audio Cable? It works perfectly on the browser but it won't pick it up anywhere else however when I use VAC with other stuff that works it's just Deiteris.

[potent folio]

한국어로 할수있나요

[grizzled sapphire]

why does my rvc sound muffled when i use it 😭

[fallow jasper]

Yo i havent made ai covers in a long time, what's the best non-local model trainer?

[hardy cradle]

how do i know whhat the best setting sr for my mic and my voice so they work good

[tulip lintel]

how can i make my voice less laggy

it lags so much and idk why

[viral mason]

here you go

btw if any mods or ppls who saw anything I did wrong wanna point anything out I don't mind

[brittle wing]

Hello, can someone please help me with this?

[simple ore]

Pinokio is a tool for total noobs that is not maintained quite right to keep requirements up to date

but your problem is that your mobile 3050 only has 4GB vram

it is barely enough to run SD1.5 models

[simple ore]

for prompts you can read here

https://vladmandic.github.io/sdnext-docs/Prompting/#prompting-tips

[brittle wing]

So there's no way to use SD to generate, right? I don't wanna use online services because of censorship, so i was wondering if Google Colab would work for this case

[simple ore]

if you want to have a good time with image gens you really need at least 12GB card

[fluid lion]

is rvc outdated? if so whats the newer version

[hardy crow]

is there any difference between rmvpe_onnx and normal rmvpe other then one is made for amd. is one faster then the other or sound better?

[safe musk]

Can someone help me setup a voice changer(im trying to egirl troll people)

[viral mason]

Onnx smells really bad and icky 🐴

Trust

[hardy crow]

unfortunately i have to use it cause im on amd 😔

[viral mason]

I'm sorry for your loss 🙏

At least it's quicker

[hardy crow]

😢

im guessing it makes models sound worse then

oh well, i'll just deal with it

[viral mason]

It's probably not too noticable tho, it'll be ok

[analog obsidian]

.onnx is a model format made for compatibility

afaik amd deiteris wokada can use .pth files too lol

about speeds idk i dont have an amd gpu

they should be the same thing, no accuracy loss

ive read if done wrong there can be some precision lost, not sure how its done in wokada tho

[hardy crow]

thanks for the info. i dont have any issues with speed on amd so thats not an issue

[analog obsidian]

nice, onnx is actually meant to be faster than .pth

[viral mason]

Wdym, is onnx a file type like .pth?

I'm confused

[analog obsidian]

they're the same thing
models

[viral mason]

Ohh

[analog obsidian]

onnx is just faster for inference

[paper widget]

Does this support voice cloning? If so, how?

[wind obsidian]

how do i remove delay when italk

[mossy crater]

what's ur device?

[grim raptor]

does anyone know why the voice cuts out

[flint vapor]

Where? Voice changer?

[grim raptor]

yes

[flint vapor]

Which version you have downloaded and which is your gpu?

[grim raptor]

MMVCServerSIO_win_onnxdirectML-cuda_v.1.5.3.18a

with rx 5500xt

[flint vapor]

I will give you deiteris fork, follow the guide and delete this version is outdated

-wokada

[patent trellis]

💻 Local Realtime RVC

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

• Wokada Deiteris Fork

Most suggested WebUI with the best general support for many platforms. GUIDE

• Vonovox

A Realtime Voice Changer with similar performance to Wokada Deiteris Fork, with extra features, but supported only for Nvidia GPUs on Windows. and without cloud options GUIDE

⚔️ Wokada Deiteris Fork vs Vonovox

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

⛔ Outdated/Discouraged

These options are not recommended for use.

• Original Wokada

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

• RVC GUI Mainline Realtime

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

[flint vapor]

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

[grim raptor]

is the ui the same

[flint vapor]

Yes but is on browser

[grim raptor]

do i click deiteris fork embed

nvm

[flint vapor]

You have to follow the guide and download amd version

[grim raptor]

okay

do u also know why stuff that i listen to in my headphones i can hear in the playback

[proven hill]

catfish

[quartz moon]

yall what voice changer is there to use for rx 570 16gb of ram and i3-8100(ik very weak)
im trying to use v1 voice lmao

[flint vapor]

You have to remove your speaker in monitor option on the gui

[flint vapor]

In what sense v1 voice?

[quartz moon]

uh v1 from ultrakill yk

like use it in game idk

[flint vapor]

Download here wait

-wokada

[patent trellis]

💻 Local Realtime RVC

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

• Wokada Deiteris Fork

Most suggested WebUI with the best general support for many platforms. GUIDE

• Vonovox

A Realtime Voice Changer with similar performance to Wokada Deiteris Fork, with extra features, but supported only for Nvidia GPUs on Windows. and without cloud options GUIDE

⚔️ Wokada Deiteris Fork vs Vonovox

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

⛔ Outdated/Discouraged

These options are not recommended for use.

• Original Wokada

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

• RVC GUI Mainline Realtime

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

[flint vapor]

Follow the first guide for wokada

And download amd

[quartz moon]

oki
also idk i think i tried ts but it didnt work

maybe it will work ts time

[quartz moon]

uh i downloaded it

is there like specific settings for every model?

[simple ore]

no, there are very small number of tts that clone voices

[paper widget]

what can i do then? i want to tts with one voice how can i do it locally?

[simple ore]

run tts using kokoro, if you want to change the voice using rvc

the only option for hindi with voice cloning is xtts using a hindi finetune

https://huggingface.co/Abhinay45/XTTS-Hindi-finetuned

i dont know whether it is good at all

[quartz moon]

https://i.imgur.com/vnNzcJy.png
does ts look right

[mossy crater]

why dont u use kaggle

cloud voice changer

[quartz moon]

uh is it good?

[mossy crater]

ik it has some delay, maybe like 0.3s (?) but its worth for low end laltop ngl

yeah w okada kaggle

[quartz moon]

i never heard of it

does it accept custom models or is it like voicemod?

[mossy crater]

i mean it is basically a wokada in kaggle

jt does accept custom models

[quartz moon]

is there a guide for it?

[mossy crater]

wait

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

@quartz moon however you only get 30hrs of GPU runtime PER WEEKS, but it is more enough (for me anyway)

[quartz moon]

ye 30h is plenty

ty

[mossy crater]

also i forgot to tell ya, dont forget to stop session after you finished with the voice changer otherwise it'll continue to drain the quota, and eachtime you want to use it you need to repeat the installation step including uploading the models again.. it will only takes about 10 mins tho usually

[quartz moon]

oki thanks

[mossy crater]

yeah np

what?

what voice changer r u using, the w okada one? and what is ur device specification

yeah what voice changer r u using? @tranquil osprey

[tulip lintel]

can anyone help me make my voice lag less?

[mossy crater]

can i see the screenshot?

whoops yeah its the W okada

hmm

[cedar valve]

https://www.youtube.com/watch?v=g6wbK3f8Y_g

[hallow thistle]

What is your PC GPU? The version of W-Okada you're using is the outdated one. There's a better one to use.

[tulip lintel]

which one is it?

[hallow thistle]

Which one is what?

[tulip lintel]

what is the new version of the W-okada

[hallow thistle]

The better W-Okada version is Deiteris W-Okada fork.

[tulip lintel]

i also have the outdated one

[tulip lintel]

can i also find that one on git hub?

[hallow thistle]

While you can search Deiteris W-Okada on Google, it will show GitHub links of both fork and the original one. There won't be any guide website for them. Also, make sure to answer my "PC GPU" question.

[tulip lintel]

i have the GTX 1060

[formal crater]

;O

[hallow thistle]

Download and use the better W-Okada from this link instead, especially when you have AMD Radeon RX GPU. https://docs.aihub.gg/realtime-voice-changer/local/deiteris-w-okada-fork/#download-amd-intel-and-cpu-on-windows

[hallow thistle]

What's up?

[tulip lintel]

yeah i gotta upgrade💔

[hallow thistle]

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

[formal crater]

yeah but atleast you could use the realtime voice changer right

[formal crater]

im using kaggle lolololol

[tulip lintel]

yeah

[hallow thistle]

You can still use W-Okada if you got GeForce GTX 10/16 GPU or greater, anything else below this one is not usually usable.

[formal crater]

still good tho, but i need alternative bcs i almost used all my GPU quota 💔

[hallow thistle]

!howtoask

[patent trellis]

How To Troubleshoot :AIHC_WaitWhat:

__**GIVE CONTEXT.**__ 📝

• Don't simply mention your issue like "my rvc is not working".
• Describe your PC GPU Name, Operating System, the Guide and Step you are on, what you're trying to do, the Program you're using, a screenshot, etc.
• The more context, the better.

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

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

__**BE PRODUCTIVE.**__ 🤝

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

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

[formal crater]

naahh, my gpu is R5 M330 and or intel HD 520 from my i3 6006U, in other words i only got igpu 💔

[hallow thistle]

I know. An integrated GPU is not suitable for AI. Even so, the AMD Radeon R5 M330 is unlikely to be used for AI as this is the old and low-end mobile one.

[low shard]

Please Elaborate:

• your PC GPU (Such as Nvidia RTX 3060, AMD RX 9070 XT..)
• your operating system (Like Windows 10, Windows 11, Linux, MacOS 26..)
• what you want to do? There isn't a program that does everything, there's a program for each thing:
  • AI Covers
  • Train RVC Models
  • TTS
  • Roleplay in VC
  • Roleplay in Games
  • etc
• what tutorial link are you using
• a screenshot of the program

[low shard]

@potent folio Hello, sorry but this server is english only, please don't speak korean

[low shard]

i replied in https://discord.com/channels/1159260121998827560/1401701535179735061

[hallow thistle]

Your Windows Explorer looks like this? Yes, MMVCServerSIO is the program itself. Double click to run (don't run as admin).

[low shard]

Please Elaborate:

• your PC GPU (Such as Nvidia RTX 3060, AMD RX 9070 XT..)
• your operating system (Like Windows 10, Windows 11, Linux, MacOS 26..)
• what you want to do? There isn't a program that does everything, there's a program for each thing:
  • AI Covers
  • Train RVC Models
  • TTS
  • Roleplay in VC
  • Roleplay in Games
  • etc
• what tutorial link are you using
• a screenshot of the program

[low shard]

Please Elaborate:

• your PC GPU (Such as Nvidia RTX 3060, AMD RX 9070 XT..)
• your operating system (Like Windows 10, Windows 11, Linux, MacOS 26..)
• what you want to do? There isn't a program that does everything, there's a program for each thing:
  • AI Covers
  • Train RVC Models
  • TTS
  • Roleplay in VC
  • Roleplay in Games
  • etc
• what tutorial link are you using
• a screenshot of the program

[low shard]

for rvc ai voice models? you sure you want to do it on cloud? it would be better you try local first if u can, what's ur pc gpu?

[low shard]

Please Elaborate:

• your PC GPU (Such as Nvidia RTX 3060, AMD RX 9070 XT..)
• your operating system (Like Windows 10, Windows 11, Linux, MacOS 26..)
• what you want to do? There isn't a program that does everything, there's a program for each thing:
  • AI Covers
  • Train RVC Models
  • TTS
  • Roleplay in VC
  • Roleplay in Games
  • etc
• what tutorial link are you using
• a screenshot of the program

[formal crater]

yep that's why im using kaggle one.

also, i'm curious, can i use sagemaker studio labs for cloud realtime voice changer? as it is similar to kaggle, had a "gpu" runtime if im not wrong

[low shard]

your gpu is lower than bare minimum
Your pc is too Weak for Wokada locally, You got 3 options:

• Buy a better pc
• Run it locally (on ur pc) using the CPU mode of the wokada fork which has better performance https://docs.aihub.gg/realtime-voice-changer/local/deiteris-w-okada-fork/ (but this isn't suggested as it could be unstable)
• Use **cloud **(remote good pc):

About Cloud, there are different services:

• Google Colabs (4 hours daily of free T4 gpu, easy to use, require only a google account) :
  • W-Okada's Deiteris' Fork Voice Changer Google Colab (currently works only on google colab PAID tier)
• Kaggles (30 hours weekly of better GPUs, T4x2 & P100, harder to use, requires an account and a phone number):
  • W-Okada's Deiteris' Fork Voice Changer Kaggle (the best and only working one currently for free)

[tulip lintel]

i downloaded the newer version, should i delete the old one ?

[low shard]

nope, you got an over year old original wokada and vb audio cable
dont use yt tuts

[quartz moon]

so i use cloud?

[hallow thistle]

Yes, sure. Also, if you have installed VB-Cable, download Virtual Audio Cable lite instead. https://software.muzychenko.net/freeware/vac470lite.zip

[low shard]

you could try the local one, but your specs aren't that good so it's not suggested
id suggest cloud

[quartz moon]

ill only use it in mc so ill try
also is the cloud free?

[tulip lintel]

also delete the other one then right?

[low shard]

ill only use it in mc so ill try
AI is more intensive than games lol, i can't guarantee you it will be good

also is the cloud free?
only the kaggle version, 30 hours weekly, and you need to verify your phone number (it's a google service)

[hallow thistle]

A virtual audio program that works similar to VB-Cable. If you don't know, you can search for more on Google or read the guide. https://docs.aihub.gg/realtime-voice-changer/local/deiteris-w-okada-fork/#virtual-audio-cable

[chilly furnace]

nope; if it opens a page called localhost or something, your desktop is infected.

[hallow thistle]

If you see this specific W-Okada version opens your browser, it's completely normal.

[quartz moon]

30h weekly is plenty i think
its not like im gonna use it every time i play an online game
maybe in mc and roblox only

[hallow thistle]

1. Bait used to be believable. 💔
2. Also joking in a serious conversation. 💔

[quartz moon]

also do i need vb cable for the cloud one?

[low shard]

alright, V1 ultrakill, tbh i'd suggest you cloud for the least delay in your case, so you won't have trouble 1v1ing Gabriel in mc pve

[chilly furnace]

im not yapping im for real

trust me

[low shard]

you do need a VAC (Virtual Audio Cable) for re-routing the output
But you'd need VAC Lite mentioned in the kaggle guide, and uninstall vb audio cable as many users reported trouble with it

[hallow thistle]

This is not a downloading, you have just extract the files from zip after it finished downloading. Double click on "setup64.exe" to install.

Ignore him.

[chilly furnace]

them

[hallow thistle]

Still.

[tulip lintel]

which one of the 2?

[quartz moon]

is there an online guide?
someone send me one but i lost it lol

[tulip lintel]

one with the a or without

[hallow thistle]

setup64.exe

[quartz moon]

alot of catfishers on ts server lol

[tulip lintel]

i dont have the .exe

[quartz moon]

uh
idk

[low shard]

yeah, if you click the kaggle hyperlink i sent you, it's already a guide

[quartz moon]

oh ok im blind sorry lol

[hallow thistle]

Yes, it's basically "setup64" since you didn't enable file extension on your Explorer.

[tulip lintel]

okay thank you

[low shard]

catfishers aren't allowed here lol, we take actions against them

[quartz moon]

good thing im not a catfisher lol

[low shard]

https://tenor.com/view/ultrakill-v1-gabriel-gif-26595155

man when is layer 8 out

[hallow thistle]

"What should I do after I download this thing" sounds like you tryna ask for every step-by-step instruction, which is annoying for sure.

[low shard]

it's fine lol
let me know for any issues

[quartz moon]

fr

also uh btw is it like 30h per week for every gpu or does every gpu has like uh its own 30 hours

idk if u understand

[low shard]

30h per week for every gpu
they give 2 gpus (P100 or T4x2, even tho the last one is 2 gpus in 1 but you can't use 2 gpus at the same time in wokada), if you use one of them, it takes account for the total gpu time of the week which is 30 hours

[quartz moon]

but uh there is a 3rd gpu with 20h per week so i thought every gpu is seperate

[low shard]

are you talking about smt named like "TPU"? that isn't a GPU, that's a TPU (Tensor Processing Unit), another component speciliazed in a very specific AI complex framework, Tensorflow, but most AI projects use PyTorch like wokada deiteris fork, which have poor support

TL;DR: that's another component, not a gpu, you can't use it in wokada deiteris fork specifically

[fallow jasper]

Yeah rvc ai voice models. Back when i used to make them (which was like a year ago) i used collabs cuz i had problems setting up the local shit and it just gave me a headache.
my gpu is NVIDIA GeForce RTX 3080

• if its easier now then i can try doing it locally ig

[lethal iron]

bruh my comfyui wont work on my rx 6750xt

[quartz moon]

oh ok thanks

[low shard]

it would be way better you do it locally in your case, cloud can be unstable and has limited free gpu time,
there's an easier RVC fork (modified version) named Applio, you could try that: https://docs.aihub.gg/rvc/local/applio/

[low shard]

yw! keep me up to date

[hallow thistle]

When I'm looking for "directml-amd-cards-on-windows" on ComfyUI GitHub, I just see this.

[quartz moon]

uhm i may be dumb but i did all of the uh guide but uh idk what to do after the vac download cuz the guide stopped

[astral perch]

I put my audio into spek and got this. Can anybody tell which sample rate is ideal for this dataset?

[hallow thistle]

Did you follow this?

[quartz moon]

ye

like how do i add models and uh open the thing to start it and change it i meant

[hallow thistle]

Can you say that again? I'm not sure what "the thing" mean.

[quartz moon]

the thing

from fantastic 4

jk

uh wakoda or whatever its called

[hallow thistle]

You want a voice model of Ben Grimm to be used in W-Okada? That's a cool idea.

[quartz moon]

actually of v1
i meant how to open wakoda
thats the thing

its named wakoda right im not dumb?

oh w-okada u wrote it

same thing

[hallow thistle]

Realtime Voice Changer Client and W-Okada refer the same program.

[quartz moon]

uh
im dumb js tell me how to open it using kaggle or whatever

i did the guide

but idk what to do after adding vac

[hallow thistle]

To add a voice model on W-Okada, click on edit button.

[quartz moon]

IK HOW TO ADD IM NOT SURE HOW TO OPEN THAT MENU THO

LIKE THE WHOLE PAGE

im stil on the kaggle.com page

still**

i tried running it

[hallow thistle]

Huh? You use Kaggle? Did you even click on run button?

[quartz moon]

yes i did what the guide said

i ran the first cell then the 3rd then put my token and ran the last

then i ran everything after adding vac cuz i didnt know what to do

[hallow thistle]

Did you copy this link and paste it in your browser?

[quartz moon]

uh where is that even located

ik in the output

when u run all the cells it gets stuck at session is starting

i**

[sharp lantern]

I'm trying run "start" code AICoverMaker Kaggle, and it gives me this prompt:

/bin/bash: line 1: fuser: command not found
/bin/bash: line 1: fuser: command not found
Traceback (most recent call last):
  File "/kaggle/tmp/main_program/main.py", line 3, in <module>
    from tabs.full_inference import full_inference_tab
  File "/kaggle/tmp/main_program/tabs/full_inference.py", line 1, in <module>
    from core import full_inference_program
  File "/kaggle/tmp/main_program/core.py", line 9, in <module>
    from audio_separator.separator import Separator
ModuleNotFoundError: No module named 'audio_separator'```
All Tunnel get same result

[quartz moon]

wait nvm im dumb im pressing on the button that looks like a power button instead of the run all button

tyty

[hallow thistle]

[quartz moon]

uh another question

do i copy and paste the ip looking thing or the one in ur picture

nvm

i thought both r broken

btw what settings do i put here and edit
https://i.imgur.com/uXY4Pn3.png

ik how to add the models idk about the other settings tho

[low shard]

after this https://docs.aihub.gg/realtime-voice-changer/cloud/deiteris-w-okada-fork-kaggle/#discord--games ? i forgot to mention, but after doing all the kaggle stuff and you ppen the user interface url, the program is the same as the local one, so you can continue via https://docs.aihub.gg/realtime-voice-changer/local/deiteris-w-okada-fork/#settings-explained

[hallow thistle]

If you see http://127.0.0.1:18888/ from its Kaggle output, it is their server localhost, not what in your PC. So this localhost won't be accessible in your PC, until you run an actual W-Okada program in your PC which also uses 127.0.0.1:18888.

[quartz moon]

ye i figured it out
idk about the settings tho

[low shard]

the settings are explained in the guide, you can also share a screenshot of ur settings and ill help u out if u want

!give-media-perms 30m @quartz moon

[quartz moon]

ye ye i tried them both

[quartz moon]

so uh here is my settings

[quartz moon]

so i js use these?

for input i put my mic for output i put vac input and for monitor i put my headset if i wanna hear myself right?

[hallow thistle]

Boom. "F0: rmvpe" and extra 2.7 s always preferred, while chunk can be around 56-64 ms.

[low shard]

input: microphone
output: line 1
monitor: headphones to optionally hear urself
gpu: t4 or p100
chunk: 64
f0: rmvpe
extra: 2.7

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

• Advanced Settings -> Force FP32 mode: on (THIS IS OFF BY DEFAULT! Turning this on improves stability. Increases VRAM usage by 200 MB)
• Advanced Settings -> Crossfade Lenght: Controls how smoothly the AI stitches different processed parts "chunks" of your voice back together. 0.1 for fastest voice, 0.15 for improved quality but increases delay by ~50 ms
• Reduce the delay on Windows via the Wasapi / Asio Guide

[quartz moon]

why does it take so long for models and stuff to load

[quartz moon]

ok ty yall

[low shard]

yw

[low shard]

it just takes some seconds to load models loll

[quartz moon]

hopefully it works
i never made a wokada work lmao

[quartz moon]

btw isnt chunk supposed to be high for better quality and stuff?

[low shard]

chunk controls the delay
extra controls the quality (higher than 2.7 can cause cutoffi issue on some models in wokada deiteris fork, and add delay ofc)

[quartz moon]

also uh what do yall recommend i put for the pitch format shift and index

[quartz moon]

oh ok

[hallow thistle]

More like "extra" number for indicating the audio quality and how much data it should process at a time.

[low shard]

you just gotta play around with the pitch, there's no perfect settings

[quartz moon]

uh id rather not since id prolly break smthn lmao

uhm im not hearing anything

oh i forgor to change the cpu

i think

no im still not hearing myself

[hallow thistle]

Change "GPU" to Tesla T4, not CPU.

[quartz moon]

ye i changed it

still

nvm im hearing myself but it kinda sucks

it took a moment to load or smthn i think

[hallow thistle]

If perf number at top left is red, change the chunk number that would make perf number to stable. https://docs.aihub.gg/realtime-voice-changer/local/deiteris-w-okada-fork/#finding-my-own-settings-for-chunk

[quartz moon]

no it is green

maybe the model sucks

nvm i kinda understand it now

[low shard]

you can't break anything

lower pitch = more masculine
higher pitch = more feminine

it depends on your voice and the voice model you know, so it can be different for everyone, you only gotta play with that 1 specific settings, leave the rest as we said

[quartz moon]

uh

is v1 masculine

idk

[viral mason]

It's a robot with a tts voice

[low shard]

v1 is a male robot i think?

[quartz moon]

nope its genderless

[low shard]

well

be sure to play with different models, not every is perfect, and to play with the pitch

we dunno which value is perfect based on your voice, or which out of the thousands is good

also, be aware that rvc has limitations, it can't do super realistically always do non natural speech sounds

[quartz moon]

ye ofc im js trying it for mc scpsl and roblox lmao

btw does it save my settings

cuz i heard someone says it doesnt and i need to reload my uh models everytime

also does it still uh take time even when im not using the voice(like for example uploading a model or smthn)

[trail fractal]

hello, i want opinions from you guys as your experts about which AI softawares will be the best for starting up content creation.

[low shard]

yeah, it's on kaggle so you restart the thing everytime, since it's not local

[low shard]

starting up content creation
what type are you talking about?

[quartz moon]

it takes 5 minutes to make everything anyways
uh ty for the help

[low shard]

is everything solved now?

[quartz moon]

yes thanks

[low shard]

yw!

[astral perch]

In the rvc guide it says that MelBand Roformer is the best vocal remover tool but then it also states that ZFTurbo's finetuned BS Roformer is currently the best vocal remover on MVSEP

[viral mason]

I looked through MVSEP and idk if I'm blind or what but I didn't see ZFTurbo's model

[astral perch]

I used melband reformer for my dataset but now im confused

[viral mason]

I mean I can't find it in the selection menus

[analog obsidian]

where is that message? bs roformer is very old

sdr score is also very innacurate

[simple ore]

no it wont

[viral mason]

it'd probably run better

[simple ore]

with a chance to burn your house due to a shitty 12v connector

[viral mason]

hmm nevermind don't run it

[light igloo]

i cant get w okeda to work well at all

it only has static that somewhat sounds like the character after a 40 second delay

[low shard]

This is a General AI Server, AI has many fields, so we can't know your issue with little info

Please Elaborate:

• your PC GPU (Such as Nvidia RTX 3060, AMD RX 9070 XT..)
• your operating system (Like Windows 10, Windows 11, Linux, MacOS 26..)
• what you want to do? There isn't a program that does everything, there's a program for each thing:
  • AI Covers
  • Train RVC Models
  • TTS
  • Roleplay in VC
  • Roleplay in Games
  • etc
• what tutorial link are you using
• a screenshot of the program

[stray wyvern]

anyone know why I can't upload my own models? (in w okada)
I hit upload and it goes through but the slot still shows as blank

[low shard]

I hope you aren't using some youtube video tut lol

Please Elaborate:

• your PC GPU (Such as Nvidia RTX 3060, AMD RX 9070 XT..)
• your operating system (Like Windows 10, Windows 11, Linux, MacOS 26..)
• what you want to do? There isn't a program that does everything, there's a program for each thing:
  • AI Covers
  • Train RVC Models
  • TTS
  • Roleplay in VC
  • Roleplay in Games
  • etc
• what tutorial link are you using
• a screenshot of the program and issue

[astral perch]

Its in the guide where they explain RX

[stray wyvern]

I only found out through youtube idk about any other guides, I tried reading the manual but there's so much and I assumed downloading a model and uploading was going to be simple

[low shard]

i feel like you're using an outdated one, please elaborate the infos i asked you

[stray wyvern]

I'm reading the guide on Deiteris's fork now, indeed I was using the original

I'll update if the same issue happens with this fork

[low shard]

from docs.aihub.gg?

[stray wyvern]

just from the #1159513888199540817 (so yes)

[low shard]

great, lmk

[cyan raptor]

what does wait web server . . .mean in command promt whenever i launch it

[viral mason]

https://cdn.discordapp.com/attachments/1194442192111013888/1316496021039349880/attachment.gif

[stray bison]

How can I use it on vr and hear what I say!!

[tepid cargo]

You need to set your output audio to your virtual audio cable, which then needs to be set as the input audio to your VR. To hear yourself, you will need to set the monitor on the RVC client to your headset (or whatever you are using for your audio output)

[stray bison]

Oh ok tysm!!!

[neon trout]

Hey! I'm looking to subscribe to an AI tool, but I'm not sure which one is best for professional video generation. Could you please help me choose?

[simple ore]

google veo3, good way to set money on fire attempting to roll a dice for a good generation

[neon trout]

What about Runway

[mossy crater]

i wouldn't use Runway (personally), because there's a lot of other models that do the job better than Runway (just my personal opinion)

[neon trout]

Suggest me one!

[mossy crater]

well it depends, VEO 3 did a good job on generating "professional video" while kling is good for image to video. Pika also doing a great job when it comes to anime style video generation

but, wait.. i've just saw the newest Runway models and it actually did a great job @neon trout, you should do some research and watch some review from youtube to get a better ideas about each models (ofcourse watch the review or comparison of the newest models)

[storm shard]

Can I make a question about ComfyUI and Wan.video here?

I followed the 2.2 guide and I get stuck at 45%. The server seems to crash and it stops at 45% every single time
https://docs.comfy.org/tutorials/video/wan/wan2_2

[simple ore]

probably running out of memory

[neon trout]

Thanks for the suggestion

[storm shard]

I have 12GB of VRAM tho...
I got the 5B model cuz apparently it only requires like 6GB of vram.
Is there a way to troubleshot or verify?
I don't really get an error message. it just dies...

Thank you for the answer btw

[simple ore]

start comfy from a manually opened window

or if you're staring it with a bat file there should be pause at the end

so you can see the error when it crashes

does it show anything?

5B model is still 20GB

you need GGUF model that is quantized

https://huggingface.co/QuantStack/Wan2.2-TI2V-5B-GGUF/tree/main

Q8 here

[storm shard]

it doesn't really say anything after it crashes. lemme try again...

I will try that one next and see what happens

hmm yah it just dies. Lemme try that one you sent

[simple ore]

it may be logged in event log

check event viewer, check task manager/performance tab to see whether you're running out of vram / ram

make sure paging file is enabled

[livid orbit]

how to fix lagging with the mic

[storm shard]

I can't seem to be able to add GGUF

i tried moving it to a folder named gguf but no change

okay,... maybe i don't have enough vram...

i dunno why i can't access all of it...

I might not be able to run anything here

[viral mason]

btw I never figured out how to do either of these because I'm very slow, could you send a video showing how to do each

[storm shard]

I think i'm running out of... everyhing tbh

[viral mason]

@pulsar yarrow I sent u how to do the vst daw thing btw for the voice changer

[storm shard]

yeah i think i just run out of resources

[knotty moth]

@storm shard you haven't considered checking event viewer at the crashing time

[storm shard]

uhh i could but I don't know what to look for in it

[knotty moth]

win+R then enter eventvwr.msc

[storm shard]

what info do you need from here?

[knotty moth]

then search something like "Application Crash"

[storm shard]

i have... error.

Faulting module name: torch_cpu.dll, version: 0.0.0.0, time stamp: 0x6837cdb4
Exception code: 0xc0000005
Fault offset: 0x0000000006013f94
Faulting process id: 0x1554
Faulting application start time: 0x01dc05c0a3cb4117
Faulting application path: D:\AIgen\ComfyUI_windows_portable\python_embeded\python.exe
Faulting module path: D:\AIgen\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\lib\torch_cpu.dll
Report Id: 531d5d58-f0ba-4400-96fc-663a4b45cf41
Faulting package full name: 
Faulting package-relative application ID: ```

Event Name: APPCRASH
Response: Not available
Cab Id: 0

Problem signature:
P1: python.exe
P2: 3.12.10150.1013
P3: 67f515a7
P4: torch_cpu.dll
P5: 0.0.0.0
P6: 6837cdb4
P7: c0000005
P8: 0000000006013f94
P9: 
P10: 

Attached files:

These files may be available here:
\\?\C:\ProgramData\Microsoft\Windows\WER\ReportQueue\AppCrash_python.exe_4884a4c6ded95a2c89cb525556929e9ee4d5d1e2_636aa578_eb373fd1-1699-4898-9758-1ee2fe816f2d

Analysis symbol: 
Rechecking for solution: 0
Report Id: 41faa861-03dc-452d-9cb0-479180129048
Report Status: 100
Hashed bucket: 
Cab Guid: 0```
The other errors are like this

[knotty moth]

try googling or ask chatgpt/claude about that crash log with the exception code 0xc0000005

[storm shard]

hmm i can try updating python i guess

but that would break other RVC that I have that need that specific version pf python to run

hmmmm

[knotty moth]

you can install multiple python versions and you should install packages for different applications within venv/conda

particularly the comfyui portable should have its own environment and prob using python 12 or 13 for the latest one

and I'd recommend using comfyui manager to install custom nodes required by some workflows

[storm shard]

I got the portable one cuz I thought it would be easier as it seemed i just download the things and put them on folders and it should work but it won't be that easy it seems

I really have a hard time with ths things

I will try it another day.
I've been at this for a while

Thank you for all your help btw. I had no idea where to start troubleshooting this

[craggy bough]

how do you have 14gb of ram

[merry eagle]

<@&1159293204038955078> is there is any light weight rvc?

[storm shard]

i dunno. should be 16

[craggy bough]

thw poor 2gb dedicated to integrated graphics

[merry eagle]

-collab rvc

-rvc

[patent trellis]

📚 RVC Documentations

AI HUB Docs

https://docs.aihub.gg

🍏 Applio Docs

https://docs.applio.org

[sharp lantern]

@viscid moss can you help me check this thing, plz

[celest pier]

@desert mica

ask here

[simple ore]

14GB RAM? how?? lol

I guess igpu stole 2

[storm shard]

i have a 2060 that should have 12gb but only 6 are available

i don't know what's going on

[simple ore]

anyway, you are not in a position to run shit

[storm shard]

yeah

[simple ore]

barely good enough for image gens

[storm shard]

i hope that now that a lot of models are dystiled i could maybe run something

i mostly use image gens for quick concepting and shit

[patent gazelle]

Hello guys

Can anyone tell me how to use voice conversion ? In ai hub

[paper widget]

How should I use the link you provided?

Also, can I not train my own model? I have an I9 and an RTX 4070 8GB.B

[north barn]

unrelated AI question but does anyone experience their own mic echoing when streaming in discord?

[knotty moth]

there should be bios setting to change the igpu ram allocation, depending on your laptop OEM

[low shard]

be sure also you're using wokada deiteris fork b2332, and not off youtube tuts

[low shard]

Please Elaborate:

• your PC GPU (Such as Nvidia RTX 3060, AMD RX 9070 XT..)
• your operating system (Like Windows 10, Windows 11, Linux, MacOS 26..)
• what you want to do? There isn't a program that does everything, there's a program for each thing:
  • AI Covers
  • Train RVC Models
  • TTS
  • Roleplay in VC
  • Roleplay in Games
  • etc
• what tutorial link are you using
• a screenshot of the program

[low shard]

This is a General AI Server, AI has many fields, so we can't know your issue with little info

Please Elaborate:

• your PC GPU (Such as Nvidia RTX 3060, AMD RX 9070 XT..)
• your operating system (Like Windows 10, Windows 11, Linux, MacOS 26..)
• what you want to do? There isn't a program that does everything, there's a program for each thing:
  • AI Covers
  • Train RVC Models
  • TTS
  • Roleplay in VC
  • Roleplay in Games
  • etc
• what tutorial link are you using
• a screenshot of the program

[low shard]

replied in https://discord.com/channels/1159260121998827560/1402179820946853968

[low shard]

is this related to any ai realtime voice changers or just discord ?

[north barn]

i guess its related to the virtual audio cable, i remember having this issue before with voicemeeter and when i uninstalled it, it was fixed

[short oak]

i did it right from one of the videos but it still doesnt make any sound

[low shard]

could you share a screenshot of ur discord and ai realtime voice changer settings?

[graceful isle]

trying to turn Francis from L4D's voice into Duke Nukem's with this
https://discord.com/channels/1159260121998827560/1205699564330557525
how do i make him sound better as he sounds a little weird?

[low shard]

dont use video youtube tuts

Please Elaborate:

• your PC GPU (Such as Nvidia RTX 3060, AMD RX 9070 XT..)
• your operating system (Like Windows 10, Windows 11, Linux, MacOS 26..)
• what you want to do? There isn't a program that does everything, there's a program for each thing:
  • AI Covers
  • Train RVC Models
  • TTS
  • Roleplay in VC
  • Roleplay in Games
  • etc
• what tutorial link are you using
• a screenshot of the program

[low shard]

try playing with the pitch and other models, also remember that there's no such thing as perfect, rvc can't do always super realistically perfectly non speech sounds such as screams/laughing

[graceful isle]

i did put the right pitch and play with the slider but the results sound kinda different

[graceful isle]

yeah, i see
it

it's weird as i've seen a soldier tf2 ai mod on the workshop that sounded indeed finer

[low shard]

im guessing you're trying to make ai covers? you're using applio and not some mangio rvc off youtube right?

[graceful isle]

nope, i plan on turning his ingame lines with duke's voice

[low shard]

might also depend on the model, which is why i also suggested playing with other models

[graceful isle]

i have his crystal clear lines from the game files yeah

[low shard]

ohh, i mean you'd still need applio for that, just be sure to not use video tuts for rvc as they are mostly outdated

[graceful isle]

i have the webui if by applio u mean that

[low shard]

huh? by Applio I mean https://docs.aihub.gg/rvc/local/applio/

it is an rvc fork (modified version) that helps in ease of use, and performance

unfortunately original rvc devs left the project to rot in 2023, so there aren't any much quality improvements

[graceful isle]

oh i see

alr let me try

@low shardcause you see, i was using some webui that was half chinese in text

[low shard]

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

[graceful isle]

yeah that one

[low shard]

that's original/mainline rvc
it doesn't get much updated anymore
again, there wouldn't be quality improvements using applio, but it would be easier to use, has extra features (like TTS) and performance improvements, which is why i'm suggesting it

[graceful isle]

sorry for breaking your balls lol

[low shard]

I'm here to help lol 😭

[graceful isle]

ye but i kinda ignored the word applio earlier lololol

thx for the help ye

[low shard]

yw, lmk for any issues

[graceful isle]

👍

[low shard]

you can also find models in https://weights.com/models btw, in case you don't find something in #1175430844685484042

[graceful isle]

oh yeyeye thanks

[low shard]

yw 🔥

[graceful isle]

nevermind

@low shard now, where do i put the index and pth?

[low shard]

did u upload the model? https://docs.aihub.gg/rvc/local/applio/#1-upload-voice-model

[graceful isle]

oh it was written there

[low shard]

on applio you can easily upload it via the Web User Interface

[graceful isle]

argh sorry

[low shard]

it's fine lol

[graceful isle]

yeah this applio is better

nicer and easier interfaces indeed

[low shard]

yeah, unlike the original/mainline rvc uploading method which is worse tbh

[graceful isle]

now... why am i getting an error? i gave it the audio to convert trough the path

[low shard]

could you please share a screenshot of the error?

[graceful isle]

it was just an error, now that i uploaded directly it's fine

his voice is still quite weird, i gotta play with those sliders

feels like it's because of Francis' strong accent. @low shard what should i do in this case?

"eleKtrical" lol

[low shard]

you can also try playing with this setting, having an higher value uses more of the trained model accent, which is Duke Nukem's accent

[graceful isle]

oh hey thanks

then i gotta try that

way better already

but those crackling sound... are those maybe because the game has lower quality voices than modern games?

[low shard]

mm, might be related to the original audio quality but i'm not sure, be sure to get clean vocals, try also this feature maybe, and try better models

[graceful isle]

yeah they're files directly from the game so they're clear... might switch to other models maybe

[low shard]

yeah, i have a feeling this crackling might be related to the model itself

[graceful isle]

the model was the one with most epoches or whatever they're called tho

that means better no?

[low shard]

epochs are a unit of measuring the training cycles of the AI model

basically the amount of times the model went over its dataset and learned from it

they don't mean how good is the model, it's just an info provided on how they trained the model by the model maker
More ≠ better
Less ≠ better

There's no way to determinate how good the RVC model is until you try it out or listen to the audio samples if there are

TL;DR: No, epochs don't matter at all

[graceful isle]

yeah i see

@low shard i only gotta pick rvc models yeah?

i need a modern duke voice that sounds good in performance

gpt one i presume is not compatible

[low shard]

yup, if you see any other tag like "GPT-So-VITS" (which is a TTS program made by mostly the same devs), they aren't compatible

[gusty rock]

ello

uh im trying to find a teto voicechanger but the ones i see are SO outdated and i kind of want a real time one i have a decent gpu and cpu so

[graceful isle]

this is how he sounds now after switching models and trying more settings such as the audio split one... what would you reccomend me to make him sound even better?

[low shard]

Please Elaborate:

• your PC GPU (Such as Nvidia RTX 3060, AMD RX 9070 XT..)
• your operating system (Like Windows 10, Windows 11, Linux, MacOS 26..)
• what you want to do? There isn't a program that does everything, there's a program for each thing:
  • AI Covers
  • Train RVC Models
  • TTS
  • Roleplay in VC
  • Roleplay in Games
  • etc
• what tutorial link are you using
• a screenshot of the program

[low shard]

audio split one
ngl I'm not sure if that one would help on short audios

what would you reccomend me to make him sound even better?
I don't think you can do much more, you could also try post processing features like bitcrushing to make it hide some harsh sounds that RVC produce, but can't do much more

[graceful isle]

alr. thanks

[low shard]

is it solved now?

[brittle wing]

@simple ore bro what commands should i run on jupyter notebook on a rented gpu

cause i copied the kaggle code and it didnt work, it had some ngrok errors which is the main reason why im not doing it on kaggle

[simple ore]

since it has no ui, you need to use noUI notebook

not all of it

[brittle wing]

like copy the noUI code?

[viscid moss]

hmmm again that type of issues. Did u run the installation cell, right?

[simple ore]

not exactly

[brittle wing]

then what do i gotta do

im pretty sure i can open applio as long as the command is working

like from the notebook

[simple ore]

you can use python core.py and run every single function using command line

[merry eagle]

r5 5600 cpu and gtx 1660ti gpu
windows 10
roleplay in games
i tried w okadaa but i think its little bit heavy task for my pc
fk i don't really remember which i used but i think that was w-okada

[brittle wing]

i didnt get this ngl

what functions exactly

[low shard]

could you share also a screenshot, tutorial link, and tell which games at which graphics settings (like low or high) ?

@merry eagle i saw you edited the message, if you don't remember which tutorial link you used, do you atleast still have the program?

I have a feeling you used that outdated original wokada off youtube tuts

[merry eagle]

yes

[simple ore]

preprocess, extract features, train

[low shard]

delete the folder, zip
and uninstall vb aduio cable off windows app settings

-realtime

[patent trellis]

💻 Local Realtime RVC

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

• Wokada Deiteris Fork

Most suggested WebUI with the best general support for many platforms. GUIDE

• Vonovox

A Realtime Voice Changer with similar performance to Wokada Deiteris Fork, with extra features, but supported only for Nvidia GPUs on Windows. and without cloud options GUIDE

⚔️ Wokada Deiteris Fork vs Vonovox

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

⛔ Outdated/Discouraged

These options are not recommended for use.

• Original Wokada

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

• RVC GUI Mainline Realtime

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

[simple ore]

they got the right vc 😛

[low shard]

1st link, wokada deiteris fork improves performance

[low shard]

oh wait

you already have wokada deiteris fork

are you saying you previously used youtube tuts had issues and now are using wokada deiteris fork, or are you saying you have issues with wokada deiteris fork?

[brittle wing]

bro how am i gonna link jupyter to google drive tho

i just don't get that

[low shard]

because you have both wokada deiteris fork and both replied to using yt tuts so

[merry eagle]

like the wokadaa deiteres fork is little bit heavy for my pc
isn't there is any light weight real time voice changer like dubbing ai

[simple ore]

I dont know why I have to explain again that I've never used vast.ai and I have no idea what is available there

you're asking me to explain how to walk a tightrope and I only done hopscotch

[low shard]

dubbing ai
dubbing.ai uses cloud, remote good pc, it basically connects your pc to a remote better one, that's why it's "lightweight" lol

share a screenshot of your wokada deiteris fork settings, be sure to always play on lowest graphics too

[brittle wing]

there's a files folder and a notebook available

im not sure but i think that should be enough to run the training in there without ngrok and shit

[simple ore]

did you clone applio?

[brittle wing]

i copied the kaggle code and everything worked fine except opening the links, and that was because of ngrok

[simple ore]

@viscid moss do you have any idea about this shite? ^

you said at some point you used vast.ai?

[viscid moss]

I think ports, @brittle wing did u specify the ports u want to use?

[brittle wing]

nah bro, as i said i only copied the kaggle code, another helper told me to do that

and obv i put my ngrok token thats like the only thing i changed

[viscid moss]

i think that's the issue

[brittle wing]

how should i specify the ports

like should i use the same kaggle code but specify the ports?

[viscid moss]

ye u should use the same ports kaggle says

or change the code to set other port

how many port u need? just 1?

6969

can u show me ur template settings?

[brittle wing]

alright

i'm not too familiar with what ports are ngl but ima send u a screenshot of the jupyter notebook

does that work?

[viscid moss]

should be something like this

lemme show u

[brittle wing]

aight

[brittle wing]

oh shit and instead of '9999' it should say 6969

@viscid moss i dont think i can edit the settings now, ima need to rent a new one but this is how the current one looks like

[viscid moss]

yeah

[viscid moss]

yep u can't

only before launching ur instance

[brittle wing]

okay so i'll rent a new one with the new settings

[viscid moss]

I hope that works

[dapper shuttle]

my rvc voice sounds like bunch of mumbling, anyone know a fix ?

[brittle wing]

what should these look like

im sorry bro i just wanna make sure this works

[viscid moss]

I think that's fine, not sure why those ports are already there but u should the 6969 too

then that port should be on docker options

not sure if u need to add it manually tho

[brittle wing]

aint those the 'docker repository authentication' settings

"localhost:1111:11111:/:Instance Portal|localhost:8080:18080:/:Jupyter|localhost:8080:8080:/terminals/1:Jupyter Terminal|localhost:8384:18384:/:Syncthing|localhost:6006:16006:/:Tensorboard"

this is also how the local configs looks like in case thats useful

[viscid moss]

makes sense, that's why u have all those ports

are u trying to run applio?

[brittle wing]

yeah im trying to train a model in applio

[viscid moss]

u should add the tensorboard port too

[brittle wing]

and it should be the same? 6969?

[viscid moss]

can't see any other port, so prob ya just that one

[brittle wing]

aight

@viscid moss "localhost:6969:16969:/:Instance Portal|localhost:6969:6969:/:Jupyter|localhost:6969:26969:/terminals/1:Jupyter Terminal|localhost:6969:16969:/:Syncthing|localhost:6969:16969:/:Tensorboard"

this should work right

[viscid moss]

yep

[brittle wing]

now its not even opening bro