✨│ai-help

[warped latch]

Hi guys, I've tried local RVC voice changer (I think it was W-Okada) a while ago using older AMD GPU with no tensor hardware (RX 5600 XT) and it worked but wasn't great.

Does it just use CUDA / OpenCL? Or does it also use ML hardware like tensor cores?

Now that I'm looking to upgrade my system, what kind of GPU is required to do it more smoothly?
I'd like to be able to run it while gaming too. Would having extra VRAM help with that?
I was thinking of getting an RTX 5060 Ti 16GB. Would it be good at that?

[simple ore]

you probably ran wokada with directml shim that is slow as hell

[warped latch]

Probably.. but I downloaded the one that is supposed to work with AMD GPUs

[simple ore]

5060ti is a waste of money, do not ever get 8GB

[warped latch]

Of course not the 8GB.. I'm only considering the 16GB.. and I don't see great options under 500 euros or even under 700 euros where I live

[simple ore]

extra vram can help.. if you play in 1080 ot 1440p there's usualy enough left to run a voice changer

but it is a competition between the game and ai, some tweaks are needed to run both smoothly

[knotty moth]

• it is actually onnx/directML for AMD cards under Windows, but the fork version has better optimization on it https://rentry.co/forkvoicechangerguide
• I'd suggest to get RTX 40-series such as 4060 Ti 16 GB. 50-series cards can't run the voice changer due to using the stable torch version. (too bad I lost the link to voice changer for 50-series )

[warped latch]

Yea I could set FPS limit to like 70 or 80% of max fps to leave GPU headroom but my question is if the 5060 Ti would need like 90% of the GPU to run it smoothly or if it can do it more easily while gaming

[knotty moth]

sorry but I can't decipher what you mean the percentage?
perhaps you mean the model fusion ratio?

[warped latch]

What is the stable torch version? Is it a matter of time until 50 series work well like 40 series? Because I'll be waiting a bit before I buy maybe 2-3 months

[knotty moth]

50-series can only work on nightly version of pytorch. the stable one is used in the voice changer as shown as well as the original one.

not only that 50-series are also notorious for several known issues plaguing them

[warped latch]

So the 50 series are in like a beta phase? and will get full support once the stable gets an update?

[warped latch]

That's also why I want to wait but the 40 series aren't produced for a long time and are overpriced..
How do the latest AMD card work for voice changer? Like 9070

[knotty moth]

I suppose it should work fine for directml, despite lack of native ROCm support (for linux)

[warped latch]

Is using CUDA better than directml? I'm using a Ryzen 3600 CPU and will be upgrading to 5700 X3D.. so not the biggest CPU headroom..
And does voice changer use tensor hardware on RTX cards? Or just CUDA?

[knotty moth]

cuda is generally more optimized

[warped latch]

Ok so if all else equal (which it isn't) I should get an RTX card

I'm not very hyped for 5060 Ti but it's like the least bad card under 600 euros where I live.. I'll see what AMD releases soon but I have doubts..

[knotty moth]

for most AI applications, 4060 Ti is a solid choice, its 16 GB vram is just enough

[warped latch]

At least it should be able run voice changer smoothly even while playing a lighter game like an esport game

[knotty moth]

to stretch your budget, you can consider 4070 Ti super

for the same 16 GB

[warped latch]

True but I'm also buying for gaming and such so if 50 series will get full stable support eventually, I should get 5060 Ti for the +15% for probably cheaper too

[knotty moth]

would be good if getting more affordable, but still a bit slower than 3080/4070 (or matches 3070 Ti)

[warped latch]

Where I live you can easily get the 5060 Ti 16 GB for 449 euros (basically the MSRP here) while 40 gen is usually same or more expensive

[knotty moth]

let's see how the market would respond to

[warped latch]

I think 4060 Ti 16GB is over 500 euros

[untold marten]

think of the percentage like volume: when u are setting to 0, u can't hear anything right? and at 100%, that's loud. Apply the same principle here:

processing slider:

(your voice) (model voice)
0%...........................................................100%

currently we can like either set to 0% and that would be the stop button
and start would be 100%. would there be a way to create a slider which allows to like change the way program process the voice input in increments of 5? like 5% voice model and rest 95% your original voice? and that slider is availably even when using it like pitch slider.

another example would be the DO-RE-MI-FA-SO-LA-TI-DO-RE exercise where u start with a low note (deep voice) and raise slowly to high notes then go low at the end again.

[analog obsidian]

you are trying to say that you want the model to not sound like you?

[simple ore]

@untold marten train your own voice model.. .then merge it with other voice model with some blend rates, 10%, 20%, 30%, etc

[knotty moth]

ye model merge is what I thought

[analog obsidian]

oh he meant he wants the model to sound more like him

[simple ore]

it wont sound like him, it will sound like a blend

a bit of this and a bit of that

[analog obsidian]

yea im aware of that, tried that once

i was trying to understand his message lol

[untold marten]

somewhat yes

[untold marten]

kind of that

[simple ore]

since it is just a merge of weights, the result wont be an actual blend between voices

it is a blending of data which may or may not sound good

[untold marten]

like a slider that when set to 0%, it's just your original voice. when raising, it's slowly/gradually changing the voice to the model set by you

[analog obsidian]

i remember there was a voice changer guide that also did that
supposedly it allowed the model to be able to inference your voice better

[untold marten]

lemme find a video to show better'

tbh idk if thats even possible

[knotty moth]

im sure it means to blend the voice

[analog obsidian]

he said he doesn't want the model to sound like him tho

[knotty moth]

and model merge can do it

[analog obsidian]

would be better to just speak like the model rather than merging

[untold marten]

ok, found a video to show what i mean. it's ok/am i allowed to post?

as i know so far, most links are prohibited and few like the rentry guide/aihub website are allowed

[rotund spindle]

what do i do with the models
i am not used to this new technology hehe

[untold marten]

it would be prob good if we can achieve such results by just merging the weights , since that means this feature could be done without starting from scratch

[rotund spindle]

what voice changer should i use?

[untold marten]

could you explain a bit better what you mean, what you have right now (files, what program etc) and what hardware u have? i'm not expert but can still try to guide you if u want.

on short, this is what u need to specify:

• what's your pc gpu
• what tutorial link did you use
• a screenshot of the program
• whats the issue

[rotund spindle]

its an intel 5500

how do i use these models

[restive kestrel]

Hi everyone, sorry to interrupt but is it ok to post a job announcement here?

[rotund spindle]

i did it nvm

[hallow thistle]

Intel HD Graphics 5500? That won't gonna work well.

[untold marten]

wanted to say the same thing but my message got being held as a hostage :))

[hallow thistle]

None in this server is allowed to post a promoting.

[restive kestrel]

Ok, thanks for responding.

[tough fiber]

--okada

what was the command

[tame mica]

-realtime

[patent trellis]

💻 Local Realtime RVC

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

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

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

[hallow thistle]

Also, even laptops with NVIDIA GeForce GT 7xx won't gonna work with W-Okada either. Both Intel HD Graphics and NVIDIA GeForce GT 7xx are old; only NVIDIA GeForce GTX 10/16 and up will work.

[untold marten]

would be fine to tag Lyery, Noobies and Makidanyee regarding what i mentioned before and allow me to post a link from a yt video showing what i mean?

[analog obsidian]

just share it

[untold marten]

hope it wont be deleted tho'

video: https://www.youtube.com/watch?v=s1HPMV2EMGo

short description: a feature that can achieve that transition from your original voice to the model you set and vice versa like this person can achieve using his own voice. (this person doesn't use rvc and solely using his own 7 years trained voiced and does voice trolling)

timestamps:

• 7:10
• 7:55+

oh, it works

[analog obsidian]

ai cant do that bro

[untold marten]

thats why i said before that i'm not sure if it's even possible

guess it's also taking more processing power than usually does

[analog obsidian]

w-okada has a bug that everytime you change a setting, it'll use more resources

idk why

[untold marten]

huh

a setting, not a model?

[analog obsidian]

to be honest, i believe doing anything will cause that

restarting the thing fixes it

closing it and opening it again

[untold marten]

but the fork from deiteris is not having that bug, right?

[analog obsidian]

yes the fork also has that bug

in the original w-okada is worse because every slot after 1 also has increased gpu resources

at least in the fork that doesnt happen

[untold marten]

weird. while i don't know how this entire thing works, but imagine that it would make sense if with each setting modified or a few key settings, a new instance of processing the model is created and the previous one isn't disposed of? or idk but hope the devs can manage to fix that (no offense towards anyone) 🙂

[knotty moth]

afaik there is Vocoflex by Dreamtonics, can do voice changing but not sure if could do smooth voice transition like you meant

https://tenor.com/view/gura-senzawa-get-real-gawr-gura-gif-12667446290887286758

[untold marten]

i see

[lyric fossil]

does anyone use deiteris on collab

ive used less than 5 minutes and colab disconnected me

it says due to executing code that is disallowed

[viscid moss]

isn't patched ig

[golden girder]

where to change epochs?

[valid spruce]

Hi guys

[golden girder]

hi tell me how to change epochs

[valid spruce]

How do I remove those low voice effects from vocals?

[golden girder]

there might be 2 reasons

[valid spruce]

You need to specify your problem

[golden girder]

because of how much

u put index

max

or

F0 DET..: u choosed is wrong

[lyric fossil]

which rvc cloud you guys using?

mainline colab is broken :/

[mossy ocean]

HELP!!!!

при открытии файла start_http.bat выскакивает ошибка

when opening a file start_http.bat I have files downloaded and then an error pops up and nothing opens what to do

[knotty moth]

if not sure, try the fork version that shouldnt have issue like that https://rentry.co/forkvoicechangerguide

[mossy ocean]

I can't access the site. You can send me a file in a private message?

[soft zealot]

for an llm with personality

[mossy ocean]

help

[wild bobcat]

why dont my pth and index files show up?

am i missing something

[oak hornet]

which ai trainer is better for google colab

[white ice]

I need help, there is no gpu option during the voice conversion I can use my cpu, how can I make the program use the GPU?? (in W Okada)

[wanton valley]

Yeah? what is this "better" method? hope is similar to this colab

-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

[wanton valley]

[craggy saffron]

is there any w-okada sport?

support?

wait i fixed it

[analog obsidian]

Latest mainline realtime inference runs better in nvidia than both w-okada and deiteris fork

Clock speed of mainline

Clock speed of deiteris

[proud hinge]

Im having troubles with the Real time voice changer, it keeps buffering and not play the full audio just parts of it how can I fix this. I have a AMD RX 6600 and I want to use the audio in discord or any other games.

[formal locust]

hi, when i upload my ai voice model, it doesnt seem to pick up my voice

compared to the presets

when i use passthru it picks up my voice but the vol right next to the picture doesnt fluctuate at all

[brittle wing]

im trying to download for 5000 series but when i try to extract it says the same volume can not be used as the same source and destination

@odd shale @rustic ridge

[viscid moss]

are u tring to install this one?

-rt

-realtime

[patent trellis]

💻 Local Realtime RVC

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

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

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

[viscid moss]

The first link one?

[brittle wing]

yes

for 5000 series

i have 5070ti

[viscid moss]

try using 7-zip to unzip the files u need

7-zip it's a program

[brittle wing]

so i downlaod?

from google?

[viscid moss]

ye

[brittle wing]

which one

first?

[viscid moss]

yep

[brittle wing]

omg im stupid it literally says to make sure u have 7zip in the guide

sorry

[viscid moss]

dw

[brittle wing]

i tried to extract again

[brittle wing]

i extracted with winrar

but when i open its just a black screen

[dusty rampart]

do not use watermark

[formal locust]

having some issues with using virtual audio cable atm

all my inputs are the vitual audio cable

and my virtual audio cable is set to default

[brittle wing]

do i need to change any of these

to help with voice lag

[oak hornet]

please say me free ai trainer

[kind drum]

I am using RVC and I couldn't see the usage of RTX in performance monitors

[oak hornet]

how to use mainline rvc in colab ?

[worthy saffron]

Hey, is anybody good at clipping/making AI voice overs. Can send references, dm me. Money ready.

[craggy bough]

hiring people doesnt belong in a support channel

[knotty moth]

are you sure you're using the version for RTX 50-series?

also you don't need to make a support tiket for your question

[sweet wren]

anyone know how to fix the pytorch installation on applio for 50 series?

been spending hours

[hallow thistle]

Don't ask anyone to direct message you like that. Also, not everyone gonna do it for money.

[sweet wren]

ik i need the nightly build, but cant figure out how to get my applio to run on the nightly build

[hallow thistle]

Set F0 to regular rmvpe.

[sweet wren]

cant get pytorch 2.8.0 nightly to install in the applio script

[sweet wren]

this really should be pinned. incredibly helpful thank you so much bro

[pastel oak]

F0 rmvpe

Increase chunk to 128

[pastel oak]

Change defaults in audio device settings to your normal stuff

https://rentry.co/ForkVoiceChangerGuide in step 3 virtual cable its explained

[analog obsidian]

#✨│ai-help message you might be interested in this

[pastel oak]

It doesnt surprise me all that much although i expected the updates to pytorch and all to make a difference by now but guess not
Mainlines functions like response threshold and how easy it is to mix audio drivers remains a rather annoying issue though

[analog obsidian]

fixed in latest update

[pastel oak]

There was an update?

[analog obsidian]

yuh

now it forces both input and output to use the same api

no longer mixing mme with wasapi

fcpe in mainline uses like 1gb of vram

and rmvpe 2gb

[pastel oak]

Nicee

[analog obsidian]

and you can do fp32 inference too

[analog obsidian]

got this with fp32 enabled

fp16 should use even less

fcpe

[pastel oak]

Did this come with a prepackage release? I remember the audio driver type was changed before, but the latest release which most people end up downloading is still mixed

[analog obsidian]

rmvpe

[analog obsidian]

sadly no

the new package is only available in pan baidu

for chinese users

but the changes are in both fumiama's repo and rvc-boss repo

the installation is easy, create a env, install requirements for the voice changer and after installing those, you can use it

they changed a lot of things since 2023

[analog obsidian]

new compiled version is from october 2024

boss also even added support for 50xx gpus

original rvc devs are aware of realtime being popular

for me it makes sense mainline realtime runs better now, its made by literally the people who created realtime inference in the first place

so uh yeah i wanted to mention that because i like how fast mainline is now, and the best part, it doesn't run in your browser

[pastel oak]

Sounds good, id still want to release an up to date prepackage for ease since people rlly dont know much about creating an env
I did that for mainline in april 2024 for the mainline voice changer guide

[hallow thistle]

Me who doesn't have a good GPU:

[low shard]

Buy 20 Rtx 4090s 🔥🔥🔥

[sour niche]

Does anyone want to learn AI with me?

[hallow thistle]

Too crazy. Training a model with 20 NVIDIA GeForce RTX 4090 GPUs in five different server PCs.

[brisk beacon]

Is the Voice Changer compatible for Discord?

[hallow thistle]

Yes, "W-Okada" is compatible with Discord, if everything is set correctly.

[brisk beacon]

Can you send a link for Windows?

[hallow thistle]

Don't skip step like that. What is your PC GPU?

[brisk beacon]

4070

[hallow thistle]

https://rentry.co/ForkVoiceChangerGuide#download-for-nvidia-gpu-on-windows

[brisk beacon]

Thanks its working for discord

[sly furnace]

what watermark ?

[outer isle]

I put in a datsset of Blackiana and I want to put it in crepe. How to do that on Applio in Google Colab?

[orchid flint]

I’m pretty sure it works for legit any application that uses a microphone. All it’s doing is just outputting the audio to a virtual audio device.

[latent yacht]

you guys know a way to get the funny autotune effect in realtime?

how do you do that?

I've seen people use that in-game

[orchid flint]

Voicemod most likely

[analog obsidian]

oki
i was using this https://github.com/fumiama/Retrieval-based-Voice-Conversion-WebUI/tree/dev

that is still technically "mainline" because the person who made this repo is part of the original rvc dev team

[errant parrot]

how do i use the model with huggingface

[low shard]

elaborate:

• what's your pc gpu
• what you want to do
• what are you talking about

[errant parrot]

Sorry I just figured out how to do it, but i want to know which pitch extraction is the best, because back then harvest was the best, I havent used this for a while so I want to know which one is the best now

[low shard]

be sure to never use video tutorials for rvc

the best since alot of time is rmvpe

[analog obsidian]

rmvpe

[median lance]

is it possible to merge two voice if they have different sampling rate? when i tried doin it, web-page wont allow me to select them

[crude bolt]

Define 'voice' ... Asuming you mean voice-models; you have voice models for rvc and for tts. with .pth extension ... So i know the answer in the context of rvc-voice models of same sample rates - in that case yes.

[outer isle]

I put in a datsset of Blackiana and I want to put it in crepe. How to do that on Applio in Google Colab?

[simple ore]

you can only blend models with the same SR

[median lance]

can I convert one to a higher one to merge?

[simple ore]

technically models can be converted, but they need some training after that

technically the mismatched parts can be taken from one of the model

would not recommend

but it is doable

[cedar sky]

Hi guys, does someone know great video-to-video model which is NOT changing background. Or its not possible without train custom LoRa?

[median lance]

how could i re-train a model i found online into the higher sample rate? I dont have the voice audio clip data they used?

[simple ore]

that's the problem with that approach

[median lance]

what would be a way?

[simple ore]

as I said, while merging copy the higher SR generator weights, blend other weights

[simple ore]

hm.. not quite

[nimble copper]

how can I connect rvc to discord?

any guide for it

[low shard]

Elaborate:

• your PC GPU
• what u want to do
• what tutorial link did u follow

[nimble copper]

I have rtx 4050

I want to connect the changed real tiem voice with discord instead of just the rvc client

if is there tutorial of the matter I am talking about

[low shard]

RVC = Retrieval-based-Voice-Conversion, the best Speech To Speech AI Models (on v2), Inferences (use models) pre-recorded audio (ai covers) and train (make) models

Wokada = uses RVC for realtime inference

Wokada has 2 main versions:

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

each version has it's own updates

You can uninstall all u got before since ud need Wokada

[nimble copper]

any offical download link for wokada?

by wok

[low shard]

Your GPU is good enough

[low shard]

It's better to use the deiteris fork

-realtime

[patent trellis]

💻 Local Realtime RVC

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

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

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

[low shard]

1st link

[north basin]

What’s with the black screen when I open the app?

[cedar sky]

Maybe someone knows?

[low shard]

elaborate:

• ur pc gpu
• what u want to do
• what program
• what tutorial link did u follow
• a screenshot of the error

[north basin]

The weights app. When I go to open the app, it shows the logo and then turns into a black screen

[olive cliff]

What's the best settings to use for Renegate?

[plain zinc]

Do you know if I can run the program with my laptop that has an RTX 2050? I'm not at home and therefore can't use the desktop, but I'll try when I get back.

[gilded trail]

Can anyone tell me how I can use voice models to modify my pre-recorded audio to have that voice? I already have the zip with the .pth and .index of the voice but i dont know how to apply it. My GPU is NVIDIA GTX 1050 Ti

[odd shale]

Welp, i'm not sure if these specs are good enough to use Applio locally.

But you can anyway try and install it

https://github.com/IAHispano/Applio

Just go to the "Releases" tab

In fact they released the 3.2.9 version today

Also there you have the docs.

https://docs.aihub.gg/

[low shard]

what program? there's thousands of ai programs

[plain zinc]

RVC w./ w-okada if I'm not mistaken

[low shard]

those are 2 different programs

RVC = Retrieval-based-Voice-Conversion, the best Speech To Speech AI Models (on v2), Inferences (use models) pre-recorded audio (ai covers) and train (make) models

Wokada = uses RVC for realtime inference

[plain zinc]

wokada

or either of the two

[low shard]

yes, you should be fine as long as it's not used for intensive games

[low shard]

well it depends on what you want to do

they serve 2 different purposes

[plain zinc]

I don't think I'll use it in heavy games, it would only be Roblox and VR chat.

[low shard]

as long as it's lowest settings for the games then yeah

[plain zinc]

Do you know what configuration is recommended in the program?

[low shard]

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

also cap games fps, and you might want to play with the extra (lower extra = less delay but also less bit of quality)

[plain zinc]

I guess what the GTX16 would be

[plain zinc]

I understand

[low shard]

let me know

[plain zinc]

Later I will try with my microphone that I play with but my brother is using it

still, thanks

[royal juniper]

guys I found some nice AI voices in weights. are all of them copies of someones voice? it says that it's an original voice, is it possible to make an original voice with ai voice that doesn't belong to anyone?

[inner trout]

all ai voices are trained off of real.. voices, or else there's no progress made.

[simple ore]

you can blend several voice model with some % to the point the voice is no longer recognizable as any specific person

but you cant built it out of nothing

[royal juniper]

I see. it needs real voice as data but if you blend it enough it'll be a whole new voice. is there any efficient ways of doing this? like blending voices while making them sound pleasent?

[glossy cipher]

Hey everyone, If I send someone a one minute dataset, can someone show me where Im failing? My AI model sounds nothing like I thought after I trained on Apollio

[simple ore]

forgot to slice audio in preprocess?

[outer isle]

I did put the Blackiana dataset in a folder and now I want to put in "crepe" on Applio in Colab. How do I do that?

A Ai model in "crepe" in good quality?

[glossy cipher]

I still need to even if I split the audio myself?

[simple ore]

how big are your splits?

[inner trout]

what??

i was answering someone's question are you good?

[glossy cipher]

9 seconds

[simple ore]

too large, let the app spli things

[glossy cipher]

I have Audio Cutting on, Training it now

Whats the reccomended?

Usually Ill do a minute set with 9 second splits

10 minute set*

[inner trout]

Looking for ways to make a voice sound more natural and realistic, so far I've got some background noise / taps in the back that play randomly. Does anyone perhaps know of an equalizer setup or badmic setup I could use?

[crude flame]

filter and bitcrusher would do good in making the mic sound bad

[inner trout]

What about things that would work live?

[crude flame]

yea a filter and bitcrush

lighthost and some vsts

for live

[inner trout]

Huh okay

Wow why didn't people tell me about lighthost before, this is pretty simple and light weight

Thanks

[knotty moth]

nvm

yea try lighthost

[knotty moth]

not sure but have you added it to the docs? (or ask Shad for it)

[crude flame]

I haven’t added it to the docs bec 99% of people don’t care and just want a e girl voice that is realistic without putting any effort into it

[knotty moth]

could be a nice addition so why not

[analog obsidian]

adding it on the fork guide would be nice

also we need a new mainline realtime guide because now its better than both w-okada and the fork

there's no compiled version for that tho 👀

[knotty moth]

let's summon @pastel oak

[crude flame]

idk where ill add it

any suggestions?

make a tab dedicated to making models sound "real"?

[knotty moth]

I think since it is voice changer related, it'd be in Shad's fork guide

[crude flame]

https://docs.aihub.gg/w-okada/local/w-okada/ there is a guide in the docs

it is just shad's guide

but its there

[hallow thistle]

[low shard]

Huh?

[analog obsidian]

#✨│ai-help message

[low shard]

Isn't mainline almost abandoned since they work on GPT-SOVITS

[analog obsidian]

is not abandoned

last update was in 2025

boss added support for 50xx

and the realtime inference update was in october 2024

[analog obsidian]

well to be fair, this is just a patch of the october 2024 version

last actual update was in dec* 2024

[low shard]

Wow, but pls next time use staff chats so more staffers see it 😭

[hallow thistle]

[analog obsidian]

sorry 😭

the compiled version for that is only in pan baidu

but the changes are in mainline second repo https://github.com/fumiama/Retrieval-based-Voice-Conversion-WebUI/tree/dev

[broken vortex]

what do u open to get the okada running by

i forgot

@analog obsidian Cn u belp.

[hallow thistle]

For fork W-Okada:

• A web browser like Google Chrome
• Virtual Audio Cable lite

[low shard]

@pastel oak , @knotty moth pls talk about that in staff chats, not here

[lyric fossil]

can someone help my #1364884749482590209 here. i didnt edit any code except adding ngrok token

[jovial fox]

hey is there anyone that can help me with the w-okada i have a high end pc but dont know good settings

[low shard]

elaborate

What’s your pc gpu? What tutorial link did you use?

[low shard]

replied

[jovial fox]

well i know how to use the mic inputs but idk what the best chuck settings are and the extra

[low shard]

Reply to: #✨│ai-help message

[hallow thistle]

That doesn't answer the last two questions.

[low shard]

there aren’t chunks and extra that go well with every single pc

[jovial fox]

i 9 13900K and a 4080

[low shard]

Nice, What tutorial link did you use?

If u don’t remember, you can send a screenshot of ur wokada

[hallow thistle]

There are two different versions of W-Okada; one is original and Detris fork.

[low shard]

!give-media-perms 1h @jovial fox

[knotty moth]

https://rentry.co/forkvoicechangerguide#known-working-settings-for-chunk-and-extra
which part of the guide don't you understand?

[jovial fox]

i dont remember

[jovial fox]

ill look

[hallow thistle]

If you've watched what "tutorial videos" on YouTube saying to you, those ones are outdated, they only tell you to install the original version of W-Okada.

[brittle wing]

Warning this AI server is bri*ish scary

[low shard]

send a screenshot of ur wokada

[hallow thistle]

I'm not a British.

[jovial fox]

i think mine is outdated then

[low shard]

Uninstall that

-realtime

[patent trellis]

💻 Local Realtime RVC

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

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

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

[low shard]

1st link

[hallow thistle]

As I expected. This is the original W-Okada.

[low shard]

** video tutorials are outdated**

that one is an old version of original wokada

get the wokada deiteris fork instead

[jovial fox]

okay ill uninstall then thank you

[hallow thistle]

You got any virtual line program installed?

[low shard]

if you installed vb audio, uninstall it too

[jovial fox]

yea vb audio i have okay i uninstall everything

[hallow thistle]

Download and use Virtual Audio Cable lite instead. https://software.muzychenko.net/freeware/vac470lite.zip

[jovial fox]

is it right that it opens a browser when i install ?

[low shard]

Yes read https://rentry.co/forkvoicechangerguide#why-does-it-run-in-a-browser-and-not-its-own-window

[glad scaffold]

How do I make a chat bot

[thorny cargo]

After two successful models I've made I wanted to step up the dataset and so I did. at first I ran into the issue surrounding the index not being able to be generated, but after finally fixing that I am now having this issue

I've had the same issue on Kohya (stable diffusion image model training software)

But I resolved it by quite literally removing the variable

I wanted to first ask here before anything

the line in question

I've seen a few others have this issue in the server but their questions were met with no help

[low shard]

Applio and that are 2 different software completely, applio is an rvc fork, it’s made for speech to speech

[low shard]

update to 3.2.9 and see if it fixes that

[thorny cargo]

I am well aware, im just surprised I got the same unbound local error

on two completely different platforms

🤦‍♂️

[thorny cargo]

Alright

[low shard]

luck 🔥

[low shard]

what’s ur pc gpu btw

[thorny cargo]

Could it be that a 40 minute dataset is problematic for my vram

actually nah I doubt it

this error seems unrelated

[low shard]

nah u should be fine

Be sure that applio, windows & your drivers are up to date tho

[thorny cargo]

Yup

[thorny cargo]

I forgot how long this takes

this maneuver is gonna cost us 51 years

ssd btw

[low shard]

have you got an hdd

[thorny cargo]

this is a gen 4 nvme ssd 😭

[low shard]

you use winrar/peazip/7zip to extract things right?

[thorny cargo]

oh...right....good point

I have 7zip

did not use it

[low shard]

usually 3rd party tools like that are better for extracting faster large files

[hallow thistle]

If file copying/moving takes forever, it could indicate an issue with software or the SSD failing.

[thorny cargo]

I used 7zip and it took 1 minute

but ill look into the ssd issues

I have samsung magician for that afaik

what the heck is this

I didn't see this in 3.2.8

What even is the call on this, like how do I gauge what I set it at

[low shard]

it’s so it doesn’t give weird noises when there’s silence

[thorny cargo]

ahhh

[low shard]

there’s a description for most features in applio btw

as you can read below the feature title

[thorny cargo]

Question is how do I gauge what I set it to

between 2 and 10

since I already know to put something (there isn't much silence within the dataset)

I'll just leave it at default for now

[thorny cargo]

It worked

much appreciated

It's gonna take a while to train but slow is better than nothing

[uncut raven]

am wanting just a rvc to put in audio buits and get the ai converted audio whats the best way to get that

[thorny cargo]

Wokada has an option to put in files

as in, put in an audio file as an input

and it will do the conversion

[uncut raven]

ahhh i have the old version

[thorny cargo]

I had the version from 2023

and that one does it

[hallow thistle]

-gui

[uncut raven]

welll....

[patent trellis]

RVC Easy GUI

This program is pretty abandoned and old, don't use it.

[uncut raven]

maybe im dumb

[thorny cargo]

you mean

clueless

[uncut raven]

that too

[thorny cargo]

[uncut raven]

i been using the realtime for years

no issues

not completely clueless

as such as out of the loop

[hallow thistle]

Applio is an RVC program that can both do AI cover and train voice model in one program.

[thorny cargo]

W okada my beloved nobody will ever stop me

[uncut raven]

i just installed wokoda

[hallow thistle]

W-Okada is realtime voice changer that uses RVC voice model for inference.

[uncut raven]

should i get this too

use this

[hallow thistle]

What is your PC GPU?

[uncut raven]

3060

[hallow thistle]

Did you download W-Okada from this guide? https://rentry.co/ForkVoiceChangerGuide#download-for-nvidia-gpu-on-windows

[uncut raven]

yeah

[hallow thistle]

Good.

[uncut raven]

figured it out on my own

im a big boy

so much easier than the old onme

[thorny cargo]

this guide....is quite nice

[uncut raven]

would u be a chad and te ach me real quick how to use it

[low shard]

yw

updating solves everything 🙏

[low shard]

everything is explained right in the guide tho

[uncut raven]

im dyslexic

[thorny cargo]

have you tried turning it off then on again?

[thorny cargo]

hi dyslexic im dad

[uncut raven]

ill kiss u if u do that again

[thorny cargo]

[uncut raven]

just how do i throw audio in and get ai back

[low shard]

oh im sorry, i can text but can’t really vc

[thorny cargo]

is there a dropdown for input

under the input section for devices etc

[uncut raven]

yah

lemme guess

file

[thorny cargo]

see if theres a "file" one

yes

...

[uncut raven]

big brain

[low shard]

wokada isn’t for that

[thorny cargo]

[uncut raven]

do you use crep

[low shard]

rvc is for pre-recorded audios

wokada is for realtime calls like discord vc

[uncut raven]

well how does one get this

[thorny cargo]

I mean, wokada CAN do prerecorded audio

I think?

i've done it before

huh'

[uncut raven]

just not as reefined

[thorny cargo]

[hallow thistle]

This is Detris' fork W-Okada.

[low shard]

iirc it could but it’s meant for speed

@uncut raven do you need only pre-recorded audios? Or also realtime usage in calls?

[uncut raven]

both

i used the old one

for gaming and stuff

but now i want it for pre recorded

aswell

[low shard]

I would suggest wokada deiteris fork for realtime, and then applio (rvc fork, modified version) for pre-recorded audios https://docs.aihub.gg/rvc/local/applio/

the issue is that video tutorials are old, since ai changes at sonic speed and you can’t easily update a video

it’s the reason why we have written guides instead

[thorny cargo]

You should see things on the img gen end, stuff from 2-3 months ago is out the window

lmao

[low shard]

gets outdated even more faster?

[uncut raven]

ohh

[thorny cargo]

I...think so, I have yet to really familiarize myself with rvc and how things are happening here

but It could be'

I should have batched 16 instead of 4

I was told batch of 4 means higher quality

but given the time it will take, is it worth the time increase

[low shard]

Not exactly, pls check https://docs.aihub.gg/rvc/resources/training/#batch-size

[low shard]

what do u use for imgs btw

i use comfyui

[thorny cargo]

oh

[thorny cargo]

forge UI

its similar to the old Automatic 1111

just better

[hallow thistle]

I use Automatic1111.

[thorny cargo]

comfy is nasty

I would never

[low shard]

i like comfyui for having day 1 support for shi

but yea it’s harder to use

[thorny cargo]

learning curve certainly isn't easy but it definitely has more refined control

[low shard]

yup

[uncut raven]

how much index do you suggest using or does it depend

[low shard]

index is the accent

[thorny cargo]

I was told to use 0 most of the time

since it has the least restriction

1 is like, closest to the original source

[upper furnace]

sorry one thing how do i create beatrice templates personlized for voice changer?

[low shard]

higher value means more accent (tho in realtime not suggested for causing autotune usually), less value means it more of the input accent

[low shard]

no one trains beatrice models

there’s very few, they aren’t preferred for being lower quality and mostly focused on being lightweight

[thorny cargo]

while I have you here, what do you think causes the slight robotic sound in rvc, when talking its subtle but you hear it

its subtle so not egregious but

you hear it

[low shard]

might depend on how good clean was the dataset and that you used the tensorbosrd

[thorny cargo]

Yeah I have to completely refamiliarize myself with tensorboard

focus on g loss total and the other loss total right

[low shard]

be sure to use rmvpe, since it’s more robust to noise compared to other pitch extraction algorithms like crepe

[thorny cargo]

Yes I use that

[thorny cargo]

the dataset is extremely clean, I was trying different epochs that I saved

ranging from 200 to 500

[low shard]

loss/g/total

[thorny cargo]

on a 14 minute dataset

[low shard]

there’s also a more advanced explaination in the docs about the tensorboard btw, explaining you each graphs

[thorny cargo]

got this

[thorny cargo]

oh let me see where that is

[low shard]

https://docs.aihub.gg/rvc/resources/training/#usage-guide tells how to set it up + easy way + advanced way

i would really suggest you to give the docs a good read since they got alot of good info

and also still thanks to @crude flame for maintaining the docs, you did a great job

[thorny cargo]

Yeah I am now refreshed on it

always stop once the graph starts to perk back up

atp its overtraining

I remember when this server had to be remade

been around a long time

[thorny cargo]

it got deleted or smtn

[low shard]

-# idon’t remember u tbh

[thorny cargo]

oct of 2023?

i think

[low shard]

2024

[thorny cargo]

hmm

[low shard]

actually

[thorny cargo]

odd

[low shard]

4 october 2023

iirc

[thorny cargo]

why did you reply with 2024

[low shard]

got confused lol

[thorny cargo]

oh lul

yeah

it was tragic cuz a lot of the models were gone or supposedly uploaded to weights

[low shard]

yup

[thorny cargo]

question, under #1175430844685484042 there are a lot of models people upload without including the index

why is that

don't we need the index for rvc

[low shard]

what models exactly?

[thorny cargo]

Uh some voice models whether for characters or just in general

been a while so idr them off the top of my head

why do you ask

[low shard]

index is optional since it’s just the accent, but model makers should upload both pth and index

[thorny cargo]

yeaaa, ok so next time i encounter that

do I let someone know or what

[low shard]

might be sometime ago a model maker accidentally forgot to upload it but yeah u should prob ask the model maker

[thorny cargo]

oki

[thorny cargo]

once I make a model I want to upload, I can get the model maker role and actually comment/engage with posts on the thread?

[low shard]

if u do the model maker verification role yes https://docs.aihub.gg/extra/model-maker-role/

[thorny cargo]

Got it

Nice

I assume that this is like this because I had cancelled the training and restarted it with a different batch size

[low shard]

dude what the fuck 🔥

just redo the training atp

[thorny cargo]

back to square one

LMAO

[low shard]

Bro was going to recreate #✦│chat message

[thorny cargo]

LMAOOOOOOOO

https://tenor.com/view/akiyama-yakuza4-yakuza5-yakuza-meme-gif-25779818

oh

MY EMBEDSSS

how selective

what a sham

[analog obsidian]

@low shard we need this to be a sticker

[knotty moth]

skill issue

but there are lot of customizable comfyui workflows you can search in civitai, etc

[outer isle]

Where’s the mangio-crepe in Applio in Colab? I don’t see it

[thorny cargo]

I know

Ive used it though

so I can at least attest to that

anyway WHAT THE HELL

the graph is meant to go DOWN

then UP

@low shard

nvm

this is the real graph

way overtrained

gonna stop it now.

[upper furnace]

help

[crude flame]

what happened here?

[thorny cargo]

it did a flip

[crude flame]

did you change the batch size?

[thorny cargo]

(i stopped training, changed a parameter and started it again)

yes

[crude flame]

oh

dont do that

[thorny cargo]

just start a whole other one?

or wha

[crude flame]

yea

[thorny cargo]

oki

[thorny cargo]

yipee its normal

[simple ore]

give me a bit, i'll fix it

[upper furnace]

ok

[simple ore]

all right, open the fresh noUI link and go from the top

let me know how it goes

[upper furnace]

👍

[simple ore]

keep going - extract features and training, all of them had to be fixed

[thorny cargo]

can't I use a hugging face link?

does it have to be Weights?

[thorny cargo]

nvm

[coarse pier]

ey i wanna clone my own voice, how can i start?

[low shard]

by telling your pc gpu

[coarse pier]

is there no collab for it?

yikes

mine is a gtx 1650

[low shard]

there is but we ask the pc gpu because we need to check if it's good enough to run it locally first, there's people with a good pc too, and it happened in the past that people with an rtx was using colab bc they didn't know any other way

[low shard]

it could actually train a model but eh limited by vram

I would suggest cloud

Train (make) RVC Models on cloud:

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

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

3. Be sure to know about the tensorboard

Google Colab = Easier but risk of getting disconnected
Kaggle = Harder but way more gpu time
If you are looking for the easiest way and for free, is using https://weights.com/ which ofc uses RVC

RVC Inference (use models) on pre-recorded audio on Cloud

You can use either:

• Weights.com: Easiest Possible Ever Automatic
• Ilaria RVC Zero: Fastest free on cloud
• Applio UI Colab: RVC Fork with some extra features like TTS
• RVC AI Cover Maker UI: Automatically Separates the Vocals and Instrumentals, converts the voice and mixes them back

Mainline is broken as right now though

[left dirge]

Which is the current best RVC version? For AMD, I kind of can't find where to download the updates.

-realtime

[patent trellis]

💻 Local Realtime RVC

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

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

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

[left dirge]

I don't think the version from 7th of december is the latest one

By the way, I'm currently using DirectML from a year ago, just want to know if there's a better one now for AMD (I have a good card, the RX 6800), or if there's an update on DirectML

[low shard]

it is, the latest wokada deiteris fork stable release is from december 2024, and ofcourse it's better than the over year old versions of original wokada

[left dirge]

Yes, but I was using DirectMl as I have AMD, I'm not sure which one is better

Or if there's some other fork I should use.

[low shard]

are you using the latest (version b2332, 7th december 2024) wokada deiteris fork directml version?

that's the best for amd rn

[left dirge]

No clue if its that version, can you send me the link? I'll download it and just use that one

[low shard]

-realtime

[patent trellis]

💻 Local Realtime RVC

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

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

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

[low shard]

1st link

it's inside the wokada deiteris fork written guide

i feel like you were using an old version of original wokada off youtube lol

[left dirge]

This one then?

[left dirge]

Not off youtube, I was givena link here

[low shard]

yep

[left dirge]

hMMM

Just a few versions behind

Let's see if my model won't sound like a fake robot now with this version, by the way, got any reccomendation for English sounding models?

I tried searching a few months ago but I didn't get anything decent

[analog obsidian]

realtime changes are only for perfomance
also rvc does not mean realtime voice changer just in case

[left dirge]

But- It says Realtime Voice Changr

[low shard]

RVC = Retrieval-based-Voice-Conversion, the best Speech To Speech AI Models (on v2), Inferences (use models) pre-recorded audio (ai covers) and train (make) models

Wokada = uses RVC for realtime inference

[analog obsidian]

blame wok for naming his realtime gui like that

[left dirge]

Ohh, fair enough

[analog obsidian]

so you are running a realtime inference gui for rvc models

[left dirge]

Makes sense, any good rvc models then? I didn't really have performance issues before. Mainly just model performance. Youtubers have these models that sound perfect but even after trying 30 different ones they all sound robotic

[low shard]

I mean, wok's program is a realtime voice changer

[low shard]

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

[elfin nebula]

Is there an AMD version of the Tortoise tts?

[low shard]

Tortoise tts is pretty old

https://docs.aihub.gg/tts/tts-tools/ but not sure if those have AMD support, Zluda could be your friend

[brisk steeple]

Random Question, is this the most recent version ?

MMVCServerSIO_win_onnxgpu-cuda_v.1.5.3.18a

[low shard]

Nope, that's an over year old version of original wokada

If you followed a video tutorial, it's better you just forget it since most of those are old

What's your PC GPU and what you want to do?

[brisk steeple]

i have a Ge Force RTX3060

i just wanna know if theres another version just to know

[fading lodge]

does anyone know the best mic for wokada?

i got told before by a voicemdoel creator it was a maono pd400x

but im just making sure if thats a good one

or if i can get something better

[low shard]

Wokada has 2 main versions

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

Currently we Would suggest the latest deiteris' fork

-realtime

[patent trellis]

💻 Local Realtime RVC

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

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

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

[low shard]

1st link

[brisk steeple]

Perfect thank you so so much

Do I need Vb cable ?

[fading lodge]

yes

you do

you always need vb audio cable for the voicechanger to work

[brisk steeple]

Perfect

[fading lodge]

do u need any help?

i have expirience in helping people here i can help u with whatever else

actually wait

add me rq

[brittle wing]

hii im trying to make a female voice a male one, does anyone know how much to pitch down by?

[low shard]

Vb audio cable gives random issues on windows

Uninstall it and get vac lite at the 3rd step of the guide I sent

[low shard]

Vac lite is more suggested

[low shard]

No one perfectly knows, depends on the model and your voice, it's better to play with it

[fading lodge]

oh, yeah

ill tell her that instead

[low shard]

Already did

[fading lodge]

also, @low shard can u awnser my question

[fading lodge]

.

[low shard]

What question

[fading lodge]

i replied to it

[low shard]

You'd just need a decent one, don't need an extremely pricey 1k mic because rvc conversion downsamples the voice input for pitch estimation anyways https://rentry.co/forkvoicechangerguide#do-i-need-an-extremely-expensive-mic-for-good-quality

[fading lodge]

but is there a reccomended mic?

the mic an old voicemodel person gave me was the MAONO PD400X

its 150$

its a podcast microphon

[low shard]

Ehh don't think there's an universal suggested one, just being decent and not a 1 cent microphone off Temù will work

[brittle wing]

okk tysm

[low shard]

@analog obsidian lmao do people really buy expensive mics for wokada

[fading lodge]

it depends on ur voice

but if ur looking for a voicemodel creator i have one for you

[brittle wing]

im doing lorde to harry styles

[fading lodge]

he charges though

but there insane voicemodels

i personally use him and hes great

[brittle wing]

no thats ok i have voicemodels

[fading lodge]

alright

hes in the model master shop if u want him tho

i think its @tulip cloak

yeah its him

[analog obsidian]

and also pc

[fading lodge]

i spent 1100$ on a 4080 super for the voicechanger to improve

im deadass planning to buy a double 5090 for it to have no delay

😭

[analog obsidian]

._.

okay just in case
for like the 100th time
rvc inference downsamples the source audio to 16khz

[low shard]

PC makes sense, but microphone not really

[analog obsidian]

so to anyone reading this
your 1900000khz mic will still get downsampled to 16khz

[low shard]

There's no such thing as 0 delay

[fading lodge]

i just wanna know a good mic i should use to pick up my voice better

[fading lodge]

there is a thing such as barley any to a point where you dont notice any delay

like a 4090 super has almost 0 delay

close to none

[analog obsidian]

technically possible with rmvpe_onnx and onnx model while using fp16

and crossfade 0.05

but rip stability and naturalness

[low shard]

There is little that is barely noticeable, but not 0

[analog obsidian]

not exactly 0 but like 10ms or so

[fading lodge]

i just need a good microphone to pick my voice up better for wokada to function properly

right now i have a blue yeti but the shitty usb-b cable is breaking

[analog obsidian]

well the thing that affects the most in realtime inference is noise and random sounds

[fading lodge]

thats the thing, the mic i got reccomended removes all backround noise and sounds really crisp

so i was wondering if it was a reccomended pick by otheres

like i said i got it reccomended by some vociemodel person in here

[crude flame]

dont live on the high way and just have a mic and you good 👍

[fading lodge]

alr

i might buy the mic anyway i need a new one

😭

[fresh cairn]

ahh

the linux verson of the discord voice changer is still busted

[knotty moth]

you still want to buy that broken Tesla car?
note that the fork version for 50-series isn't guaranteed to completely work like the normal one

[fading lodge]

then ill buy 2 4090's lol

[knotty moth]

a separate 2nd pc or 4060 laptop for voice changer & streaming could be better choice

[fading lodge]

i see

alr

thanks

[visual meteor]

i saw that RVC-GUI is made in 2023 is there any better ones now? "VC-Project/Retrieval-based-Voice-Conversion-WebUI"

[violet badger]

I have been here for a hot minute and I use AI voices once every few months and the way to make the voices changes so much. What do I use to make an AI voice with this Link? - https://huggingface.co/Freaky98/Tom_Robinson-rvc-v2-320e/resolve/main/Tom_Robinson-rvc-v2-320e.zip?download=true

I can prob figure it out myself if someone just shows me where to go

[knotty moth]

find ones in https://docs.aihub.gg/

[shy spruce]

5090 works 100% fine now. I have one, and trained plenty of models since. You just need to use the new pytorch 2.7 with CU 12.8 that was recently released, before the stable release things were a bit weird. all good now though

[burnt saffron]

hi was wondering if this is overtraining?

model is at 275 epochs

[brittle wing]

I was trying to download a model in Applio then this happened

It turned out like this

It is the wrong model

[thorny cargo]

Yeah youre supposed to stop around the lowest point

[thorny cargo]

Around there

You should smooth it to 0.987

[elfin nebula]

I mean if it's not AMD then it's useless to me

[simple ore]

not for avg_50 charts, they are smoothed already

[simple ore]

tortoise is garbage

can try https://github.com/patientx/fish-speech-zluda

[agile kelp]

Need help

i downloaded Ducks voice changer and it only works in Beatrice i want to use RVC

[fervent rover]

Hey y’all it is be an while since last time, but I just wanna to know, is RVC Mainline Colab Finally Worked and Fixed, I Just Wanna To Know

[pastel oak]

Delete that its outdated
Whats your gpu

[compact linden]

Hi everyone! Does anyone have the "checkpoint_v2" file for OpenVoice 2? I'd really appreciate it if you could share it or point me to a working download link. Thanks a lot in advance! 🙏

[simple ore]

https://huggingface.co/myshell-ai/OpenVoiceV2/tree/main/converter ??

actually https://myshell-public-repo-hosting.s3.amazonaws.com/openvoice/checkpoints_v2_0417.zip

[lyric fossil]

hello

why cant i share screenshot here? need some help

[feral bane]

really?

[lyric fossil]

yes i tried copy pasted img and doesnt work

so i'm currently using wokaga rvc on kaggle (https://www.kaggle.com/code/suneku/voice-changer-public). last night it worked perfectly, realtime quality to mic is convincing. today i wanna open up my kaggle again but i cant run the server

there is no run button on the left cell

is it because i need to "edit my copy" every session?

this error prompted on the last call [Errno 2] No such file or directory: '/kaggle/working/MMVCServerSIO' /kaggle/working Go back and run first and second cells.

[knotty moth]

try F5 the page

[lyric fossil]

i did refresh. the first cell had button while the last cell wasnt

i ended up making new notebook and all buttons are there

[compact linden]

I already downloaded it. But v2 models are removed. Cannot be accessed

Has anyone downloaded and used it before?

[humble dock]

What is the top ai for 25$ a month in my case? Should i move out to grok 3/ claude 3.7? Im using chat gpt plus and have alot of information it knows there. I can start again but it will take like some time.I chat alot with o4 and it does alot of the things i need maybe- it doesnt give me ideas about my project and doesnt understand me only follows my commands on what to do so i need to program it only by seeing the answers and he dont know how to solve some problems of mine and dont understand some of what i explain to him( i am explaining only by examples but it doesnt recognize the patterns in other things in the world. For example i need a menu for something and i have that in mind but it doesnt understand what im telling him and i am assuming he will understand the connection to things happening in movies and in real life but it doesnt)

[merry kernel]

is there any version of RVC that support intel arc card?

[simple ore]

rvc mainline?

[low shard]

What's your PC GPU? What you want to do?

[low shard]

RTX 9090 ti for a voice changer 🔥

[knotty moth]

"I'll buy everything!"

https://tenor.com/view/all-the-things-hyperbole-and-a-half-do-all-the-things-excited-gif-5207369

[low shard]

AMD isn't that much supported for ai, anyways you can use zluda (cuda emulator) for basically everything

[merry kernel]

i can use beatrice v2 but can't use any RVC model
i try many version but still not work
anyone know how to solve this?
im using laptop with intel arc gpu

[low shard]

Elaborate:

• what's your exact PC GPU name
• what you want to do
• what tutorial link did you use (be sure to NEVER use video tutorials, they are old)

If you don't remember the tutorial link, you can also send a screenshot of your program

!give-media-perms 1h @merry kernel

[merry kernel]

i want to use rvc model but it so buggy and lag

[pastel oak]

Cuda (the version you downloaded) is for nvidia gpus

https://rentry.co/ForkVoiceChangerGuide

Download amd/intel version

[merry kernel]

i also try this but still not work

[pastel oak]

Wdym "still not work" did you download it and run it

[merry kernel]

i follow this intructions https://discord.com/channels/1159260121998827560/1365096442061918239

i saw this and tried

still not work some how :v

[knotty moth]

not explaining your "not working", no help

[pastel oak]

What about it doesnt work

[low shard]

How

Did you read the guide

[knotty moth]

!howtoask

[patent trellis]

How To Troubleshoot

__**GIVE CONTEXT.**__ 📝

• Don't simply mention your issue, like "my rvc is not working".
• Describe the step you are on, what you're trying to do, the RVC 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.

[merry kernel]

yes

i did exactly what you guy said :v

[low shard]

What's the issue then?

What doesn't work?

[merry kernel]

wait

let me

re dowload

and send you picture

[low shard]

You were using original wokada, you shouldn't use that

Uninstall original wokada and vb audio cable

Install Wokada deiteris fork and vac lite from the guide

I guess you followed a video tutorial lol

[merry kernel]

i uninstall vb audio ✅
install vaclite ✅

install the old version ✅

[low shard]

Nuh uh

[merry kernel]

i have a question

[low shard]

Maybe u meant uninstall lol

[merry kernel]

i mean install this

[low shard]

That's the latest version of wokada deiteris fork yeah

[low shard]

What is it

[merry kernel]

can i keep the version im using rn and install this version

[low shard]

It's better to just uninstall the version you got rn

Install the version we told you

[merry kernel]

done

[low shard]

Alright, take a look for the guide and let us know for any issues or questions

[merry kernel]

downloading ...

ok

im stuck at this

what now?

[low shard]

It was just downloading internal files

[merry kernel]

so what should i do now

[low shard]

GPU: ur Intel arc
Input: microphone
Output: line 1
Chunk= perf number + 60, to find your own chunk

Then upload a model

[merry kernel]

ok wait

i need to unzip the model file right?

[knotty moth]

you should first consult the guide to get started and set up everything you need, then you can tell us if there is some problem

[merry kernel]

its said this

[low shard]

Yes https://rentry.co/forkvoicechangerguide#uploading-models

[low shard]

What did you do? What's your browser?

[merry kernel]

i upload the model

and boom

[low shard]

What model? Send the link

Also what browser?

[merry kernel]

chrome

https://www.weights.com/models/clm72qeuf0jlncctcr7zvp4rc

this model

[low shard]

Did you click edit, click a slot, upload, then upload the .pth for model and .index for index?

[merry kernel]

yes

what should i do now

[agile kelp]

Rtx

[merry kernel]

can i retry and screen shot it

[low shard]

Retry, if it doesn't work maybe try using firefox

[merry kernel]

how to retry btw :v

[low shard]

Restart the program and redo the process

[merry kernel]

oke

[pastel oak]

https://rentry.co/forkvoicechangerguide

Download nvidia rtx (if you have rtx 5000 use the 2nd nvidia download)

[merry kernel]

like this?

[low shard]

Yup upload

[merry kernel]

and what now

last time it crash when i tried to rename it

[pastel oak]

Guide says what to do for linux virtual cable in the virtual cable step, might have to google extra stuff if you dont know Linux well

[low shard]

You can't rename it on the ui

It's bugged

[merry kernel]

oh

that why

[pastel oak]

Rename before uploading

[low shard]

You need to rename the model files and reupload the slot

[low shard]

Hello shady

[pastel oak]

Yuuurr

[merry kernel]

bruh it so lag :v

[low shard]

What programs are you using it with

Close all useless programs in background too

[merry kernel]

only discord rn

[low shard]

Are you gonna use it in games too?

[merry kernel]

yeah maybe :v

[low shard]

Show a screenshot of ur wokada, be sure everything else is closed, you can even check the ^ at the bottom right of windows to check programs in background

Steam is an example of programs in background that you see in ^

[low shard]

Can't recognize all the programs just be sure to exit those that aren't needed

[low shard]

Be sure to not use super intensive games, and to lower graphics to the lowest ever possible in every game

[merry kernel]

i mean

i cant even hear what i said rn

it killin my cpu

not gpuy

[low shard]

You need to set monitor to your headphones if you want to hear yourself

[merry kernel]

i set it

[low shard]

Do you see a little triangle in the UI?

Did you set the GPU option to your Intel arc?

[merry kernel]

what ui

yes

[low shard]

User Interface, the program's web user interface

[merry kernel]

i just stop

[low shard]

Set extra to 0.3 and chunk to 800, don't expect low delay on a not so powerful gpu

Also don't expect it to run with things like marvel rivals lol

[merry kernel]

erm

it said this

[merry kernel]

lol hell nah

im playing valorant

[low shard]

Yes exactly, do it

It's bc non Nvidia GPUs aren't that good for ai

And for non Nvidia you gotta change back like that to make sure it utilizes the GPU in a good way for the program

[merry kernel]

oh

it

kinda work

[low shard]

Yeah, you gotta do that everytime you change settings