✨│ai-help

[gloomy mortar]

wait i cant upload images

it spams a message with "xhr poll error"

the full message is [SIO] rconnection failed Error: xhr poll error

any fixes?

[stoic oasis]

Thanks guys. I was thinking about merging two models too, I guess I'll just experiment with that for more unique voice.

[knotty moth]

model merge is the easiest thing ppl without capable gpu can do

[stone lynx]

okay i fixed it

the problem was my headset

it leaks my sound

i lowered the volume and its fine rn

[unreal linden]

Yeah and my headphone does not support 2 channels 48000 hz

[low shard]

maybe try an SR that ur ur headphones support

[abstract glade]

Y'all, I'm confused as hell. I'm just trying to get a quick and free voice changer for games but everything is either just a "Free trial"/has paywalled options or isn't what I'm looking for (can't change the voice from my actual input and then output it into the game in question)

Isn't there just a simple tool where I can select my audio devices, drop a voice model and have it work without needing to pay for additional functions/time?

[low shard]

use wokada deiteris fork, wokada is a program designed to run rvc (retrieval-based-voice-conversion, speech to speech models) in realtime, and the deiteris fork, modified version is the best one

it's FOSS: Free and Open Source Software

those paywalled programs like voice.ai use the same type of technology, RVC, which is why you can upload the same models

the difference is that those paywalled program are just slightly easier to use and install, but we don't suggest things like voice.ai since it also uses your pc power to support their services

[low shard]

what's your pc gpu?

[abstract glade]

Roger, sounds like exactly what I'm looking for.

[abstract glade]

RTX 2050.

[low shard]

you should be fine, what game are you planning to use it in?

[abstract glade]

For now either R.E.P.O., VRChat or Lethal Company. None of which are very graphics-intensive games.

[low shard]

Nice then

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Unavailable, the guide is outdated and the program is worse compared to the ones above, and much less updated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

wokada deiteris fork

all you gotta do is read the guide and follow the parts related to windows nvidia gpu

[abstract glade]

Yeah, looked it up in the channel history, already set everything up and it's working nicely. Thanks for the help.

[low shard]

u might lower game graphics settings to the minimum

[low shard]

do u may want me to check your wokada settings?

[abstract glade]

It's working nicely, so no need. If I need anything else I'll just ask.

[low shard]

alright if you're sure about that

let me know for any further issues

so, how's it going?

are u using https://rentry.co/forkvoicechangerguide now?

[sharp crescent]

im not home, i already have but not sure its latest, i will check once im home , ty again

[low shard]

The one you showed me was an ancient old version, it's better to uninstall it along with vb audio cable and follow the guide I sent

Let me know

[sharp crescent]

Im using Steel series virtual cable, microphone for it. I think it does the same thing, virtual audio cable software that is suggested.

[austere snow]

hi

anybody here

i have to automate notion for task management

through llm model

anybody here who can help

[knotty moth]

what is it? choppy sounds?

[sharp crescent]

Oh i found i a nice Ai voice, its very good, but i commissioned a custom one, and then you could tell immediately its Ai , so its either dependant on voice, or the guy did a bad job xD. i have no issue with steel series virtual microphone.

and just need to check if i have latest version, maybe it can help with the audio.

[low shard]

alright then

are u home now btw?

[knotty moth]

a model master commission?

[sharp crescent]

Yes, model master AI god, tags. Im guessing is just my voice is not compatible with the character that i commissioned.

[knotty moth]

AI god what?

you can adjust pitch and index rate depending on the accent

otherwise perhaps it is ur speaking style that may make it sound too far away for you

[sharp crescent]

https://prnt.sc/z9ndXerh4lfL

[knotty moth]

I've to use vpn to open this shit
you can ask me for the pic upload perms

[sharp crescent]

LOL

[low shard]

!give-media-perms 1h @sharp crescent

[sharp crescent]

Idk single words are good, but if you speak more it GG , you can easily tell its AI xD

[low shard]

honestly skill issue, use cloudflare's dns

[sharp crescent]

can send u the model if u want to try it ur self ?

[low shard]

for issues related to a model master that did a bad commissions, it's better you open a ticket by dming @vital hedge so all staff can check it out

[knotty moth]

jk but im too lazy to open links, mainly while in mobile

[low shard]

double skill issue tbf

[knotty moth]

more of willingness issue

[sharp crescent]

Nah clicking sus links is bad, i understand.

[restive birch]

Hello, I asked a question one day ago but I still got no answer

[low shard]

what's your issue?

[restive birch]

.

[low shard]

you have to put the game input to line 1, while the output as ur headphones

[restive birch]

Thank you ^^

[low shard]

in wokada, u have to put the input to your mircrophone and output to line 1

[low shard]

want me to also check your settings while we are at it?

[restive birch]

My settings are good, i checked up yesterday and everything is working perfectly

Thank you for the help 😁

[low shard]

alright, you're welcome

[lofty lichen]

does wokada support MRF hifigan trained models?

[low shard]

I'm not sure that wokada supports model trained in hifigan yet

[toxic cosmos]

Hiiii

[proper plinth]

Hello, I am making a minecraft map game show, I wanted to have a host that's full of energy, but all the voice models are just kinda flat. Right now I am using elevenlabs.io, let me know how I can improve this https://cdn.discordapp.com/attachments/1362540689862295633/1362540690524864603/welcometo100waystodie.mp3?ex=68036d27&is=68021ba7&hm=4985becc1f4416ccfef47cde96cfc72d950fee92ad23cb6126bc38ba7b57b67a&

[toxic cosmos]

does anyone know why when i open the shortcut on my desktop it says only main.exe cui --https false --no_cui True and on banana mans one it comes up with c : \users\gavin\desktop\vboicve changer/application and some other stuff

mine dont say that

[simple ore]

what bad person keeps pinging me

[brittle wing]

How do I resume on applio no UI

Cause I'm planning on training a model soon

When I switch accounts what cells do I run

Also

Like in depth

[simple ore]

restore backup, run training cell

[long oyster]

hmm can i have guide to download Virtual Audio Cable lite please

[simple ore]

https://vac.muzychenko.net/en/download.htm

scroll down to Lite link

[toxic cosmos]

@low shard do u know
does anyone know why when i open the shortcut on my desktop it says only main.exe cui --https false --no_cui True and on banana mans one it comes up with c : \users\gavin\desktop\vboicve changer/application and some other stuff
???

[analog obsidian]

game characters aren't the best in realtime, for natural results, you have to train natural speech, not random voice lines

[low shard]

are you using RVC for TTS?

[proper plinth]

I have no idea what that is

[low shard]

be sure u didn't use video tutorials

share a sceenrecording of ur issue, the tutorial link u used, and also ur pc gpu

!give-media-perms 1h @toxic cosmos

[low shard]

what are u using and doing exactly

[proper plinth]

I am using Liam on https://elevenlabs.io/app/speech-synthesis/text-to-speech

[low shard]

11labs is one of the best TTS you could get, maybe it's better you try playing around with the settings and voice

[proper plinth]

I quite like it usually, we have done some projects before with it and have been satisfied

but that was all where the characters were just talking, for a game show host you would want them to be so excited they are basically shouting

[low shard]

I honestly dont pay for 11labs, but I remember it being the best out there and there are options to change the emotions iirc

not sure if maybe you're expecting a bit too much out of AI, have you tried messing with the settings voice ?

[proper plinth]

I have just been using the free plan, they let you do like hundreds of lines a month which is plenty for me

[proper plinth]

yes, this is what i came to https://imgur.com/undefined

oops

https://imgur.com/VDeVEsJ

[low shard]

try to increase tone exxageration?

[proper plinth]

hmm its a little better

I also had a suggestion from a friend to increase the volume as it goes on, which helps somewhat

That is in the original clip as well

this is just for like, the grand introduction (I can send you a clip of that if you want some context) usually he would not be shouting as much

[toxic cosmos]

How his is coming up with stuff at the top but he did say it takes a bit of time but even after a bit of time it dont come up @low shard

[low shard]

you could also try playing with the text like adding more letters or punctation like ! or ?

[proper plinth]

this is my prompt atm

[low shard]

the updated wokada deiteris fork has a .exe file, i feel like you're using an outdated old version of original wokada

please reply to everything I said

[low shard]

reply to what tut link u used and what's your pc gpu

[toxic cosmos]

OH bruh how do i get the updated one

[hallow thistle]

What is your PC GPU?

[hallow thistle]

That's CPU, not GPU. Did you read that right?

GPU = graphics processing unit
CPU = central processing unit
A GPU is simply a graphic card found in your PC.

[low shard]

i need to check if your pc is good enough to run it first

i feel like u used a video tutorial lol

[toxic cosmos]

i did

[proper plinth]

i dont understand

[hallow thistle]

To check your PC GPU, open Task Manager.

[toxic cosmos]

Soo what do i do now to fix the problem

[hallow thistle]

!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.

[proper plinth]

but im not doing anything with my gpu

im using elevenlabs

i think maybe they responded to the wrong person lol

[low shard]

all videos are outdated

[low shard]

You can check your pc gpu via:
ctrl+shift+esc (task manager) -> Performance tab -> GPU

and tell me the gpu name

[low shard]

I replied to the wrong msg my bad

[proper plinth]

lol

[sharp crescent]

[rugged crane]

How do they do these "what if x was a 80s dark fantasy live action movie" videos?

[low shard]

To generate images for free (text2img), either:

• Use @brittle wing in https://discord.com/channels/1159260121998827560/1202754985255764060 (It's powered by DALLE3, from ChatGPT+), pretty easy
• Another easy and good ways with weighs.gg are:
  • Use /image with @earnest musk in https://discord.com/channels/1159260121998827560/1202754985255764060
  • Create an image on their site https://www.weights.gg/ (which you can also use LoRAs, Low-Rank Adaptations, basically a small trained additional model to adjust your generation)
• Use Open Source Models like stable diffusion & flux that could be a bit **harder **but good, what's ur pc gpu? As you could run them locally (on ur pc) or on cloud (remote good pc)

[earnest musk]

:wave: @low shard, How can I help?

Available Commands:
• @weights find <query> or /find <query> - Search for RVC Voice Models
• /create - Create an AI Cover
• /image - Generate an Image

[low shard]

all the best u can do is playing around with the prompt and tone

[rugged crane]

Thanks for the guide! I actually have weights premium, I was wondering how to do this specific kind of video: https://www.instagram.com/reel/DEzrUT0JeE4/?igsh=MTEzMnduZWVscW0wbg==

How do they produce these clips? Camera slowly moving in that certain way, characters look precisely like their copyrighted selves, and the setting looks truly like if it was a 80's live action movie

Any ideas how do they prompt these

[near axle]

Hey, How many minuites should I run for a 280 epoch drum model?

@broken prairie

[sharp crescent]

@low shard 2.0.73-beta cuda

[low shard]

yep you're using an old version of original wokada

uninstall it along with vb audio cable

-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]

read the 1st link

wokada deiteris fork is better

@sharp crescent also what's ur pc gpu?

[sharp crescent]

3060 ti

[low shard]

yep it's good

read the guide and lmk

[sharp crescent]

so basicaly i download this

https://huggingface.co/Shadicti/deiteris-Fork/blob/main/voice-changer-windows-nvidia-b2332.zip

and unzip

[low shard]

yes

be sure to not skip 3rd step

[olive plover]

Hi I use FORK in Windows 11 But I am facing some stuttering and hanging.
I was facing this in Windows 10. What is the solution?

RTX 4070
i7-14500

[broken prairie]

as long as you have at least 1 minute of training dataset...

[near axle]

any examples?

[olive plover]

??

[trim sparrow]

does anyone have a pretrained modal that for singing (r&b)

[analog obsidian]

https://discord.com/channels/1159260121998827560/1339155300720054316

[trim sparrow]

is there a male one?

[analog obsidian]

are u fully aware that the original pretrain also has a female voice

you can train any voice in any pretrain

[trim sparrow]

oh so it doesnt matter?

[analog obsidian]

yeah

your dataset will replace the pretrain's voice

[trim sparrow]

wait so im super new to this still so can you explain pretrain a little bit more im not sure still only thing i know is someone told me that its useful for what my goal is and that its like a guide

[broken prairie]

It's like training an RVC voice model but for drums

[near axle]

How come the klm version does not work on my programme? im using Mangio-RVC-v23.7.0 @broken prairie

[analog obsidian]

without a pretrain the model has no idea what sound is, it has to learn everything from 0 (this requires a dataset of around 50 hours to give good results)
with the pretrain, the model already knows what sound is and how to create new audio, with this you're able to train small datasets like 10 minutes just fine

[trim sparrow]

ahh okay but it wont take the voice from the pretain only the dataset?

[analog obsidian]

your small dataset is meant to change a small portion of the pretrain, without losing too much knowledge

[analog obsidian]

no, the model will replace the pretrain's voice with your dataset's voice

[trim sparrow]

ooo okay do you think i should probably stop my training then and use a pretrain im getting okay results but its not great as i was wanting

[analog obsidian]

always train with a pretrain lol

[trim sparrow]

thank you so much and this will work just fine for singing r&b style etc

[analog obsidian]

yes, it's a singing pretrain

[trim sparrow]

do i download the g and d?

[analog obsidian]

[trim sparrow]

Do you know what folder i put it inside of or do i create one ? @analog obsidian

[analog obsidian]

https://docs.applio.org/applio/getting-started/pretrained

[trim sparrow]

thank you!!!

[low shard]

@sharp crescent how's it going?

[low shard]

share a screenshot of ur wokada

!give-media-perms 1h @olive plover

[sharp crescent]

Good.

[low shard]

do u want me to check ur wokada settings?

how's it going with that model you said sounded weird?

[sharp crescent]

im using the one that worked good, and it is still good

[low shard]

I'm talking about the one that you made a ticket about

if it's all good you can send a message saying it's good now that you updated the program and close the ticket

[low shard]

Also I can give you perms for this

[low shard]

the settings seems fine, are you having lag only in games?

[olive plover]

When I play Fivem
Or sometimes when running the program, I encounter stuttering and freezing while speaking. This didn't happen to me in Windows 10, only in Windows 11.

[sharp crescent]

tried a bit but i think its just as "game characters aren't the best in realtime, for natural results, you have to train natural speech, not random voice lines "

[olive plover]

yes

[low shard]

lower the game graphics to the minimum possible

[olive plover]

I also experience lag sometimes without games.

[olive plover]

yes i do this

The comment has eased
But why am I facing such strong device settings?

[low shard]

also lower the extra to 2.0

then show a screenshot of ur wokada while playing fivem

[low shard]

you're using 2 very intensive programs at the same time

[low shard]

so what are we going to do about the ticket ?

[olive plover]

What is the solution then?

[olive plover]

why 2.0 not 2.7 ?

[olive plover]

Will the sound quality change?
extra to 2.0

[sharp crescent]

"unitedshoes:
Hi, thanks for reaching out, we'll have an RVC pro test that model out soon" i mean i would apreacete if they tested it and confirmed, else i dont mind closing

[low shard]

extra kinda controls model quality, lowering it can kinda lower the quality but at the same time it will lower the resources it needs

also be sure everything of the game settings is lowest possible

and to close all background programs

[olive plover]

But my device settings are strong, why should I do all this??

When I was on Windows 10 I was facing these problems only on Windows 11

[low shard]

AI and a game are taking expensive resources

so that's why, you're both running expensive programs at max settings

[low shard]

windows 11 is more bloated and uses more resources than windows 10

for example linux gives more performance since it's not as bloated and more minimal

[olive plover]

oh okay

[olive plover]

yes yess

Do you advise me to go back to Windows 10?

[low shard]

also games are more prioritized than wokada in windows since the voice changer isn't rendering anything on the screen

what I would suggest you is to first of all lower the game graphics to the minimium

you could even go back to windows 10, but this will mean that after october 2025 you will lose any future security updates from windows and your pc might be more vulnerable

[vast elm]

man everytime i come in here to check if i can assist anybody i see nick is abosloutely demolishing it 😭

[low shard]

https://www.youtube.com/watch?v=FczDo1YWvWY as you can see, yeah windows 11 is more bloated and that's true

[low shard]

lmfao

[vast elm]

leave some unanswered questions for the rest of us lol

[low shard]

wow bro literally summoned an unanswered question 😭

[olive plover]

@low shard okay thank you so much🩷

[vast elm]

as the error says it seems the your current applio installation is located inside onedrive, have you tried moving it somewhere else?

[crystal gull]

YIKES

Apologies that was NOT meant to be copy pasted

[low shard]

dude.

[crystal gull]

Hellooo, I use MMVC Server SIO nvidia b2332 for live voice changing. Does anyone have any suggestions for what I should use for a live tts using my same models?*

[crystal gull]

-# It was a part of a report on another server, for some reason discord refused to properly copy paste what I was asking from #🧬│ai-chat

-# I'm just glad it wasn't the video

[low shard]

nope, RVC is STS, not TTS

the max you could do is use another tts program, use the output as an input for rvc

then use the vac lite to re-route the output audio to the input of disocrd settings

[crystal gull]

cough I don't think you understood my question, apologies. I'm not looking to pass through one through another or whatever, just whatever might be similar to that one but in a TTS form

And perhaps have an api of some kind

I've seen ones where you type in text it'll convert to tts then pass it through a live all by itself, can't recall the names or if they are relevant anymore.

[low shard]

welp, at max https://github.com/w-okada/ttsclient/tree/master
or https://docs.aihub.gg/tts/realtime-tts/

[trim sparrow]

how do i fix this " self._target(*self._args, **self._kwargs)
File "/kaggle/working/program_ml/rvc/train/train.py", line 412, in run
net_g.module.load_state_dict(
File "/kaggle/tmp/.venv/lib/python3.10/site-packages/torch/nn/modules/module.py", line 2189, in load_state_dict
raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
RuntimeError: Error(s) in loading state_dict for Synthesizer:
size mismatch for dec.ups.0.parametrizations.weight.original1: copying a param with shape torch.Size([512, 256, 24]) from checkpoint, the shape in current model is torch.Size([512, 256, 16]).
size mismatch for dec.ups.1.parametrizations.weight.original1: copying a param with shape torch.Size([256, 128, 20]) from checkpoint, the shape in current model is torch.Size([256, 128, 16]).
The parameters of the pretrain model such as the sample rate or architecture do not match the selected model."

the sample rate is 48000 idk why its doing that

[low shard]

are you using a refinegan pretrain?

[trim sparrow]

its this one
no clue

ah it might be this woops i forgot to put at 48

nope still

untimeError: Error(s) in loading state_dict for Synthesizer:
size mismatch for dec.ups.0.parametrizations.weight.original1: copying a param with shape torch.Size([512, 256, 24]) from checkpoint, the shape in current model is torch.Size([512, 256, 16]).
size mismatch for dec.ups.1.parametrizations.weight.original1: copying a param with shape torch.Size([256, 128, 20]) from checkpoint, the shape in current model is torch.Size([256, 128, 16]).
The parameters of the pretrain model such as the sample rate or architecture do not match the selected model.

[lofty lichen]

are u using pretrain

[trim sparrow]

yeah

https://discord.com/channels/1159260121998827560/1339155300720054316

[lofty lichen]

its not updated

[trim sparrow]

ah okay

do you know which one i can use for singing

[lofty lichen]

try 4.4

[trim sparrow]

do i use the 32k

because theres no 41 option in applio or i dont see at least

https://canary.discord.com/channels/1159260121998827560/1240679287036579901

@analog obsidian do yk if this one work good or should i look for another one the klm 4 one doesnt work

[analog obsidian]

no

[trim sparrow]

alright sorry im not sure which one to use i was hoping to use the klm 4 onee but for some reason i just get a error do you have another pretrain singing model in mind perchance

[cursive vault]

hi there

are there any good image model to selfhost without using gpu?

[simple ore]

you need to use a pretrain that maches the sample rate you selected. The error means 'you have 48k model, but trying to load 32k pretrain'

[trim sparrow]

hmm let me retry i downloaded the 48k pretrain and my dataset was 48 and then retried with 40 and changed the dataset audios to 40

TypeError: stat: path should be string, bytes, os.PathLike or integer, not NoneType

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "/usr/lib/python3.10/multiprocessing/process.py", line 314, in _bootstrap
self.run()
File "/usr/lib/python3.10/multiprocessing/process.py", line 108, in run
self._target(*self._args, **self._kwargs)
File "/kaggle/working/program_ml/rvc/train/train.py", line 425, in run
torch.load(pretrainD, map_location="cpu")["model"]
File "/kaggle/tmp/.venv/lib/python3.10/site-packages/torch/serialization.py", line 1004, in load
with _open_zipfile_reader(opened_file) as opened_zipfile:
File "/kaggle/tmp/.venv/lib/python3.10/site-packages/torch/serialization.py", line 456, in init
super().init(torch._C.PyTorchFileReader(name_or_buffer))
RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory

yeah idek

[simple ore]

RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory -- .pth is damaged/not downloaded fully

if you cant open it with WinZip it is cooked

[trim sparrow]

it might of been something i did i redownloaded the 40k and its training with the pretrain and the dataset now thankfully

time to wait 5 hours ill be back yall 🙏

[lofty lichen]

hifigan and refinegan doesnt support wokada

[trim sparrow]

idek what that is im using applio😭

[lofty lichen]

the thing you talk on

[trim sparrow]

oh the voice changer?

[lofty lichen]

yea

[trim sparrow]

nah im not using that i might look into using it sooner or later it kinda looks cool just to mess around with ill keep that in mind ty

[lofty lichen]

r u using tts

[trim sparrow]

no im training a model and ima use the convert thingy

[analog obsidian]

hifigan works in wokada, thats literally the original vocoder of rvc

[lofty lichen]

mrf hifigan

[analog obsidian]

a dat one

guys just use hifigan

its great

🦈

[simple ore]

same thing

[maiden oar]

how to fix that?

[low shard]

What tutorial link did u use and what's your PC GPU

[broken prairie]

idk

[low shard]

mangio rvc is outdated asf don't use it

[simple ore]

move the whole thing into C:\Voice_Changer or D:\Voice_Changer

it can not write into a restricted folder "Program Files"

[viral mason]

help

after clicking continue it closes

[steel forge]

I suspect something is already open on that port somehow

Idk reboot PC should fix

If you have another copy of the voice changer open, only run one (or manually change a port)

[viral mason]

hopefully

probably glitched because I use it a lot

[trim sparrow]

does anyone know how id fix this?

@simple ore

is that due to poor dataset?

[crystal gull]

It's more likely to be a shot model than to be whatever software was suggested here

[torn bluff]

what's a good free to use real time voice changer where i can upload my own models? i've tried w-okada but it's just.. eh

[knotty moth]

https://rentry.co/forkvoicechangerguide

[torn bluff]

thanks

[old island]

anyone know where i can train my own model?

for voice

[glossy vortex]

don up date to windows 11 is bad

[olive plover]

Yess

What's the solution then?

[hallow thistle]

That's your opinion. But not everyone would use Windows 10 until 2025.

[hallow thistle]

Running another W-Okada at the same time on the same port by accident, or there's another program that uses port 18888, can cause this problem. You can try restart your system and run the program again.

[simple ore]

has anyone seen this recently ?

[lusty whale]

does anybody know how to reset gpu value to -1

[hallow thistle]

W-Okada? Try delete stored_setting.json file inside MMVCServerSIO folder.

[karmic egret]

i dont understand

what i should downlod from this Link for VoiceChanger
https://github.com/w-okada/voice-changer/blob/master/README_en.md

[knotty moth]

https://rentry.co/forkvoicechangerguide

[karmic egret]

who is the Downlod for the vc there i only see AudioCable

[knotty moth]

||I am|| ctrl+F

[karmic egret]

?

pls sent the Downlod link for the voice changer 😫

[hallow thistle]

You don't know how to find a download link? Lol. What is your PC GPU?

[karmic egret]

i have it

[karmic egret]

but a Nvidia

[hallow thistle]

To check the "full name" of your PC GPU, open Task Manager, go to Performance tab, spot where GPU 0 or 1 is in the left panel and click one of it.

[karmic egret]

ah i have an Nvidia Geforce 4090

[hallow thistle]

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

[karmic egret]

ty

[gentle robin]

hi guys does anyone know what video can help me to learn how to make a custom character tts

[golden walrus]

Guys. One question, is it okay to use WavLM as embedder to train a model in Vietnamese?

i asked GPT and it said it's better to use that than Contentvec

[trim sparrow]

does anyone know which tool i can use to remove reverb delay and other effects from vocals?

[golden walrus]

Uvr ui

it's pretty good

[trim sparrow]

i have it but i dont see any options for it to remove that stuff

[golden walrus]

So first tab. Try to find reverb, i forgot which one

[trim sparrow]

ah i have to update i got a pop up thing let me update real quick

[trim sparrow]

also do you have any clue on how to fix those cackles and stuff by the way perchance?

im not sure if its my dataset or if its something else

[golden walrus]

i'm a noob too

[trim sparrow]

ahh okay no worries thank you

im still not seeing the reverb thing?

[golden walrus]

No no. The UI

[trim sparrow]

?

[golden walrus]

Try in #📰│dev-updates

Not Eddy made it

The uvr UI

[trim sparrow]

found it

downloading rn thank youu

[hallow thistle]

It's still be UVR5. There are specific UVR5 models that remove reverb and echo.

[trim sparrow]

is this inside of the ui for uvr ?

[hallow thistle]

I don't know. I don't use UVR5 as a local program. If the models aren't available in UVR5 program by default, you can download them and put them in UVR5 folder.

[trim sparrow]

perchance would you mind sending me this?

[hallow thistle]

[trim sparrow]

ill take that as a no

[trim sparrow]

perchance do you know how to fix this @hallow thistle

is it a dataset issue?

[stray steeple]

tysm!!

[gentle robin]

guys how do i use the Ilaria or the zero rvc

[compact mantle]

Hello, sorry to bother you, but I have a problem. I want to make an AI model of an anime character, but I don't know what settings I should have in the UVR5 Locally, could you please help me, or do you know someone who could help me? @low shard

[silk pike]

Check this out https://docs.aihub.gg/rvc/cloud/ilaria-rvc-zero/

[silk pike]

Hey, no bother at all! Could you tell me a bit more about what you're trying to do with Ultimate Vocal Remover? Are you trying to isolate the character’s voice from background music or something else?

[compact mantle]

Well, what I wanted to do was extract the voice from the audio of an anime character since I want to create an AI model of that character.

[gentle robin]

OMG THX SO MUCH I FINALLY GOT IT

🩷

[silk pike]

I personally use the BS-Roformer-Viperx-1297 Model for that, so you can try that one.

[compact mantle]

I'll send you a private message so you can explain to me exactly what I need to do.

[silk pike]

No problem at all—happy to help! 😄

[crystal gull]

Currently bashing my head against my desk rn, I'm trying to get llama py to properly use my gpu with cuda to generate llm text. and no matter how I install the pip packages, no matter how i set my build flags. It keeps saying it's assigning layers to my cpu, any help?

I've tried pip install llama-cpp-python --upgrade --force-reinstall --no-cache-dir --compile --extra-index-url=https://pypi.nvidia.com, set CMAKE_ARGS=-DLLAMA_CUBLAS=on before building, CMAKE_ARGS="-DLLAMA_CUBLAS=on" pip install llama-cpp-python --force-reinstall --no-binary llama-cpp-python, $env:CMAKE_ARGS = "-DGGML_CUDA=ON". etc etc

ive just cleared my env so im working from fresh rn..

[crystal gull]

I mean I suppose it's still generating pretty quick but

[trim sparrow]

what toolkit ver r u on, py version & gpu model and im guessing ur on win?

[tranquil mortar]

Hi! I’d like to request a custom voice model based on the Italian voice of Lucia from Mermaid Melody (voiced by Denise Misseri). I’m not sure where to post this request — could someone point me to the right channel? Thank you so much!

[hallow thistle]

This is not where you request a voice model. To make a voice model request, go to #1159289738314919936.

[tranquil mortar]

Thank you ❤️

[low shard]

What's your PC GPU

[low shard]

I haven't seen anyone report that yet

[low shard]

Basically you were downloading an old version of original wokada, the helper gave you the Wokada deiteris fork

[low shard]

How's it going?

[low shard]

There are different Text To Speech (TTS) AIs:

GPT So Vits: RVC isn't as good as GPT So Vits for tts, but gpt so vits (few shot tts, which means needs just a lil training for models) can't use rvc models (and viceversa), and its only limited to: english, chinese, Cantonese, japanese & korean, if you wanna check gpt so vits instead, read https://docs.ai-hub.wtf/tts/gpt-sovits/

Freemium 11labs: Easy way to do TTS is https://elevenlabs.io/, you can't use RVC model on this but its a mostly premium easy way for good quality TTS

FishSpeech: FishSpeech is a 0 shot (no explicit training needed) TTS, if you got a good pc you can use it locally else use their site

You can check TTS in our tts index

With RVC Models:

RVC is natively for Speech To Speech, but forks such as ilaria rvc mainline & applio have built in tts (using Microsoft Edge TTS to make a generated tts audio, which i suggest you to choose a tts model that is the same gender and language of the rvc model you wanna use, and then convert it with rvc)

If you wanna do tts locally with RVC Voice Models (if you got a good pc):

• You can get Applio in our docs
• While Ilaria RVC Mainline here (no guide as of right now)

If you don't got a good pc you can do tts with RVC Voice Models on cloud:

• Ilaria RVC Zero (Running on A100 GPU, free fasted rvc on cloud) and the guide

• Use Applio UI Colab (with google colab T4 free daily limit gpu)

• You could try another tts from our tts index and use the output as an input in rvc

[low shard]

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

[signal hinge]

Will the Voice Changer work on Linux?

[nova relic]

I have a trading strategy, and i'd like to algorithmize it, for that I need to code in PineScript(TradingView) and NinjaScript (Ninjatrade, the broker),
I have been using the free ver of GPT for some time and the outputs for indicator and strategies so far have exceeded my expectiations, the problem is free gpt is too Limited in terms of attachements and reasoning, so im planning to run LLMs and coding ai on cloud gpus.
Another major reasor for me to use cloud based ai is-
Both Tradingview and Ninjatrader have complete documentation on their websites on pinescript and ninjascript, which i want to scrape (HTML, JSON, vectors? gpt confused me here) and feed to the ai, so that the outputs are more precise....the ai will know what it is doing.....

How can i actually do this stuff?

[hallow thistle]

Of course, W-Okada indeed works on Linux. https://rentry.co/ForkVoiceChangerGuide#linux

[low shard]

Yes, be sure to use wokada deiteris fork

Be sure to also check your PC GPU

[zealous hatch]

HOW I CAN LISTEN ME IN VOICECHANGER HELP PLEASE

[low shard]

Share a screenshot of your wokada

!give-media-perms 1h @zealous hatch

[low shard]

Holy shit that's extremely outdated

The settings are awful too

Lemme guess

You used a YouTube tutorial?

[zealous hatch]

yes

[low shard]

Harvest is literally ancient

[low shard]

All video tutorials are outdated asf

They don't make newer ones bc it's a dead trend

The version you got is an over year old version of original wokada

[zealous hatch]

but how can i listen to myself

[low shard]

The performance gonna be shitty asf

[low shard]

You shouldn't use that version at all

The performance and quality is so bad that you shouldn't even bother fixing that version

[zealous hatch]

can you send me which version to download

[low shard]

Uninstall that along with vb audio cable which gives random issues on windows

-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]

Read the 1st link

Wokada deiteris fork

Only written guides are updated

[low shard]

I just did

[zealous hatch]

the site is not working

[low shard]

Where are you from

[zealous hatch]

russia

[low shard]

Yup that's why, Russia blocks a lot of sites

Rentry.co is a markdown sharing site, its made for like sharing .txt files and that's blocked in Russia too

[low shard]

Try https://docs.aihub.gg/w-okada/local/w-okada/

If it doesn't work either use a VPN

[zealous hatch]

it works

[low shard]

Alright, read it and let me know for any issues

[zealous hatch]

I read

[low shard]

Alright, let me know for any issues

[zealous hatch]

I have this program now, how can I listen to my voice?

[low shard]

Share a screenshot so I can tell you that and also the best settings

[zealous hatch]

I mean about the old program

[low shard]

You shouldn't use the old program

It's more bugged and worse performance

It's better you uninstall it and read the guide I sent

[zealous hatch]

just answer

[low shard]

The program you got is extremely outdated tho, you really sure you want extremely worse performance especially on your not so good GPU?

It affects more bugs, worse quality and worse performance

[hallow thistle]

!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.

[low shard]

The version is over a year old

It's useless to still use that version

[hallow thistle]

Or I can write 69 pages of why using the "original" version of W-Okada in 2025 is bad, if you want to.

The W-Okada that opens as a separate program is the original one. The one that opens your browser is the fork one, and it works better than the former. Unless you wanna use the original version so bad, I can leave you to use it, but at the same time I won't be helping you if you do so.

[zealous hatch]

я нихуя не понимаю что в этом гайде

[hallow thistle]

Please speak English.

[zealous hatch]

I don't understand anything in this guide. I just want to know how to listen to my voice

[craggy bough]

uninstall it, install the new version

[hallow thistle]

Please listen to me. What Nick meant is to let him know if you have any issue after using the better program, not still using the original version. Did I say it right?

[zealous hatch]

just how can i listen to my voice in the old program. i only need this

[hallow thistle]

Please no.

Please.

[craggy bough]

whats the point of asking for advice if you arent gonna listen to the advice

[hallow thistle]

It always be the language barrier. If you're here just for the original W-Okada, just get out I guess.

[wet wing]

my sliders aren’t working and i keep a “type error:failed to fetch” pop up

[hallow thistle]

What program are you trying to use?

!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.

[zealous hatch]

какой совет

я блять спросил как слушать себя и все

нахуй мне чет дополнительное

[wet wing]

vcclient_win_std_2.0.76-beta

[hallow thistle]

Again, please speak English. Not everyone here would talk Russian to you.

[craggy bough]

#✨│ai-help message "can you send me which version to download"

[zealous hatch]

ало блять

just how can i listen to my voice in the old program. i only need this

[hallow thistle]

What is your PC GPU? There's a better W-Okada available.

[craggy bough]

you literally asked what version you should download, the advice was to download a newer version

[zealous hatch]

where i can listen myself

[craggy bough]

uninstall the old version, install the new version

[hallow thistle]

[zealous hatch]

how to listen to yourself in the old version

[hallow thistle]

This is the better W-Okada.

What you're doing is using the original version. Sorry, but I can't help you with that. I can only focus on fork W-Okada.

No, thanks.

I don't direct message a stranger I don't know.

[wet wing]

AMD Radeon (TM)

[hallow thistle]

Any other GPU than this?

[zealous hatch]

what's the problem with telling me how to listen to myself

[craggy bough]

thats like asking a modern day person how to fix a 100 year old machine

no ones gonna know how to use antiques

[zealous hatch]

вы динозавры ебучие

че никьл

[wet wing]

no i dont think so

[hallow thistle]

The problem is that you didn't read to what Nick said to you, and you act like you know things better. That's what I thnk about you.

[zealous hatch]

никто этой версией не пользовался

[zealous hatch]

да идитынахуй

[craggy bough]

thats right, no one uses it because its old.

[hallow thistle]

If there's no GPU other than this, that sounds bad. But there's another option.

[craggy bough]

🇲 🇺 🇹 🇪

[zealous hatch]

ролику год по которому качал

[zealous hatch]

че мут

[wet wing]

what’s the other option?

[zealous hatch]

ишаки позовите того кто знает

хули вы так ведете себя

[hallow thistle]

Don't be a dick here. Just speak English. It ain't that hard.

[zealous hatch]

где одна блядская кнопка которая отвечает за прослушивание себя

да вы заебали меня уже

[hallow thistle]

You can try Kaggle instead. But you'll have to register with your phone number on this one, and also register an ngrok account for a token. https://www.kaggle.com/code/suneku/voice-changer-public

[craggy bough]

me when i follow the ancient instructions of an old video instead of listening to people recommending a far superior update with actual support docs:

[zealous hatch]

на русском гайды делай

скачаю

[craggy bough]

use google translate

[zealous hatch]

я ебал в рот написали и нихуя не понятно

[hallow thistle]

When people still use RVC-GUI in 2025:

[zealous hatch]

теперь сам юзай гугл транслатор

[hallow thistle]

No one's gonna use the old version of W-Okada, lil bro. Get over it.

The problem is on you now, not us.

[zealous hatch]

ты проблема ходячая

[craggy bough]

if you dont want to listen to the advice you were given, why are you here?

people here can only give you advice, they cant understand it for you

[hallow thistle]

Not everyone gonna understand anything you say, bro. You should at least follow what people telling you, not being a selfish person and act like you know better than everyone.

[zealous hatch]

хуесосы тупые

надо было на одну кнопку монитор нажать

ебанные дауны бестолкоевын

[hallow thistle]

Again, if you're still here just for the original version, and still saying things in Russian like everyone could read what you said, just better give up already and then go do something in your life, or you can come here again if you feel guilty for what you did.

Don't try to curse in Russian, I already know it. If it ain't reaching the agreement, I may call a mod to take an action.

[knotty moth]

Sorry, only English is allowed. You can use translator to do so.

Извините, разрешено говорить только на английском языке. Вы можете использовать переводчика.

[hallow thistle]

[low shard]

@zealous hatch only english is allowed here, we can't moderate other languages

Uninstall the old version you got along with the vb audio cable

Video tutorials are outdated, the one you have has worse performance

It's better you read https://rentry.co/forkvoicechangerguide, the wokada deiteris fork is 100 times better than the one you got

If you really really really don't want to follow this advice and want to use a way worse voice changer, you can set the monitor option to your headphones, but this is extremely NOT SUGGESTED since the version has WAY WORSE QUALITY AND PERFORMANCE

[potent bone]

how do i make my own voice model

my gpu is 3060 ti

[knotty moth]

https://docs.aihub.gg/essentials/how-to-make-voice-models/

[low shard]

promos ain't allowed

[stone lynx]

guys im trying to download pytorch

but program gives error

[low shard]

are you using a youtube video to download RVC or Wokada?

[knotty moth]

global installing torch is NOT recommended

dont follow video tutorials that may instruct such that

[stone lynx]

oh okay

[simple ore]

use regular cmd.exe, not powershell

also global install wont help with a venv project

[craggy bough]

not what this server is for, go look for jobs on seek or indeed

[sharp crescent]

@low shard Tested new version for some reason its worse than the "old" that i have, makes the audio sound like im in a metal Cylinder lol

or speaking through a 1

dafuq was that xD

[low shard]

show a screenshot of ur wokada

[hallow thistle]

What new version? Is it W-Okada?

[low shard]

@keen pasture promos aren't allowed here, please don't ask where to find jobs here

[lofty lichen]

promo? wrong word

[hallow thistle]

So you think you know better?

[sharp crescent]

Is better for me than B2332 the "latest"

[low shard]

the version you showed is an old version of the original wokada (latest original wokada is beta 2.0.76)

could you share a screenshot of the wokada deiteris fork settings?

[sharp crescent]

the wha

the 2.0.73 ? or the b2332

[analog obsidian]

ive compared the original w-okada with the fork like 100 times and every time the fork was better

like, the only reason you ever want the original w-okada is if you're an advanced user and your model is using a custom embedder

[sharp crescent]

well b2332 i tried adjusting the setttings and i always sound througha metal cylinder lol

[low shard]

the one you showed a screenshot for is an old version of original wokada

the b2332 is an update of the wokada deiteris fork

original wokada is made by Wok
deiteris fork wokada is made by deiteris

each one has it's different updates

share a screenshot of the wokada deiteris fork b2332 settings

[low shard]

uncheck sup1, it's worse noise suppression, use sup2 instead if you're having noise issues

[hallow thistle]

This is the correct one.

[low shard]

you can also set force fp32 mode to on for better quality and model stability, so it will use 32bit floating point values for inference rather than 16fp

thought this will slightly impact performance

[analog obsidian]

dont use formant shift

[sharp crescent]

yeah i tried with and without

[analog obsidian]

its meant to change the timbre if the model

[low shard]

try now with the settings we suggested you

[hallow thistle]

Formant shift can make the audio to sound different from the original.

[analog obsidian]

again, its a timbre change

[hallow thistle]

You got no Virtual Audio Cable? The chunk number is too large, which is what makes the audio to delay further. Try reduce down chunk number at that point where perf number is green for less delay.

[sharp crescent]

i have its the same

i mean now it sounds ok i guess

must have been the sup1

[low shard]

sup1 is not suggested as it's a worse noise suppression shortly

technically highering up the extra could also help with quality but could cause cutoff issues

[analog obsidian]

sup1 making things worse is a fairly occurrence here

[hallow thistle]

[sharp crescent]

Im using Fifine T683 so the mic is good i think

[low shard]

so, is it all solved now?

[sharp crescent]

yeah

[low shard]

alright, have a nice day

[waxen pike]

invite me friend pls

help

[low shard]

wdym?

elaborate your issue

[waxen pike]

I'd like to lay it out in a text message.

[low shard]

elaborate it more please

[low shard]

• what's your pc gpu
• what you want to do
• what issue are you having
• what tutorial link are you using

[waxen pike]

4060
I have such a problem that when I communicate with a friend he hears a quality voice of a girl or any other model without effects. But when I go into a group chat or voice channels on different servers, I get an echo effect and repeat the same phrase
RVC GUI

[low shard]

RVC GUI is extremely outdated

you're talking about the go-realtime.bat in rvc mainline right?

you should never use video tutorials for RVC / Wokada, those are old

[waxen pike]

yea

[low shard]

uninstall rvc gui and vb audio cable

[waxen pike]

How do I make a quality female voice then, more specifically Leia Organa and Darth Maul? So that when I go to my friends I don't have echo in group chat. Please give me instructions

[low shard]

Wokada is the program to use RVC (Retrieval-based-Voice-Conversion, Speech To Speech) Models in realtime for calls

There's 2 versions:

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

Both of those are better than the one you're using right now, but we suggest Wokada deiteris fork for best performance and quality

wokada deiteris fork guide: https://rentry.co/forkvoicechangerguide

@waxen pike read the guide and let me know how it goes

[waxen pike]

how are they different?

[low shard]

rvc gui is outdated

wokada is more focused on rvc for realtime inference

while the deiteris fork gives better performance and quality for that

video tutorials are really outdated

since the trend died, there are more than a year old videos, and unfortunately people watch those

[waxen pike]

can we use the same models? And how do I know what settings are needed?

[low shard]

can we use the same models?
Yes, all of 3 use RVC models

And how do I know what settings are needed?
It's all explained in the guide, and I can help too

[waxen pike]

I don't get it. I'm used to watching video guides. What do I need to download to open the application?

I have nvidia 10 windows.

[low shard]

Welp, the issue is that all video tutorials are very old

The only updated guides are the written ones

[low shard]

windows 10 is the operative system

but which nvidia gpu?

[lofty valve]

how to make model

[waxen pike]

5600 amd

[low shard]

what's your pc gpu? and what type?

[waxen pike]

4060 ventus

[low shard]

huh you just said nvidia

You can check your pc gpu via:
ctrl+shift+esc (task manager) -> Performance tab -> GPU

It's better you send me a screenshot of your task manager

or do you got 2 gpus ?

!give-media-perms 1h @waxen pike

[low shard]

alright

[waxen pike]

this processor

sorry

my bad

[low shard]

alr

firstly read
https://rentry.co/forkvoicechangerguide#virtual-audio-cable
getting vac lite

then read
https://rentry.co/forkvoicechangerguide#download-nvidia-on-windows

[low shard]

it's fine

[low shard]

@lofty valve btw if u don't know your pc gpu

You can check your pc gpu via:
ctrl+shift+esc (task manager) -> Performance tab -> GPU

[crystal gull]

Hey! Sorry I went to bed for the night

Python 3.12.6
RTX 4070
Toolkit 12.8

[hallow thistle]

This is the NVIDIA GPU. AMD Ryzen is a processor (CPU), while Radeon and Radeon RX are AMD GPUs.

[crystal gull]

Please ping me if you have a response, as I might not see it otherwise

[hallow thistle]

What do you mean by "model"? Is it RVC, Stable Diffusion or some physical clay figure? And what is your PC GPU?

[low shard]

@waxen pike @lofty valve how's it going for y'all btw?

Any issues or questions?

[vale osprey]

-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

[crystal garnet]

guys, help me in cs2 and discord and obs, the voice mod does not work.

[low shard]

elaborate:

• your pc gpu
• a screenshot of the program and its settings
• what tutorial link did you use
• what exactly doesn't work

!give-media-perms 1h @crystal garnet

@crystal garnet also, if you don't know what's your pc gpu,
You can check your pc gpu via:
ctrl+shift+esc (task manager) -> Performance tab -> GPU

and if you don't know how to share a screenshot, you can use the stamp/print button or do windows+shift+s

[hallow thistle]

Explain why the "voice mod" thing isn't working.

Wow, that must be taking so long to respond huh. Otherwise, I won't be able to identify the cause for you.

[low shard]

@ancient karma don't use #1159289738314919936 for help related issues

ask there or in #1192011222023950368

also you need to tell your pc gpu first to check if it's good enough

[crystal garnet]

The Voice mod is working, but it's not in the game.

[hallow thistle]

Still.

[low shard]

elaborate

!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.

[hallow thistle]

At least just sent the screenshot of it to here, otherwise you're gonna waste your embed permission time til it run out.

Make sure to focus at least one thing at the moment instead of games.

[austere snow]

anybody here

i need help

😫

[low shard]

please elaborate your help request

[austere snow]

i n e e d h e l p 😫

[low shard]

elaborate what you need help with

[austere snow]

ok i understood

i need help with ai engineering, anybody here who is doing ai engineering and wanted to do

[trim sparrow]

hey does anyone know how to fix those cackles? it sounds like a dying cat ive been trying and messing around with things but cant seem to understand why or how to fix it

[austere snow]

which sound is this

[trim sparrow]

?

[low shard]

what exactly do you need ?

[austere snow]

i need guidance

[trim sparrow]

☠️

[austere snow]

can you give 😩

i will be too much helpful

[low shard]

give you what?

[austere snow]

guidance

[craggy bough]

The more context, the better.

[low shard]

that's general

[austere snow]

i need roadmap

[trim sparrow]

😭😭

[austere snow]

do you have road map

[low shard]

AI has many different fields

[trim sparrow]

gotta be trolling

[austere snow]

i have to do ai engineering

[low shard]

exactly, there's many different fields

[austere snow]

ok give me information

i dont know

i will be happy to learn

[low shard]

@austere snow maybe you could start by checking huggingface's docs https://huggingface.co/docs

[austere snow]

i have to make ai,

[low shard]

AI is too vast, start by checking online and huggingface docs along with pytorch ones for example

[austere snow]

but programming, ml is common so

how can i start

[low shard]

are you training a model or using wokada?

[low shard]

#✨│ai-help message

[trim sparrow]

training a model with dataset & pretrain

[waxen pike]

Why would I upgrade amd and nvidia if everything was working before?

[austere snow]

yeh i have to make model

[austere snow]

wow

that sounds cool

but i didnt understand

[low shard]

things change, things upgrade

you can't be using always the same outdated program for the rest of your life

I'm telling you to use the more updated version of the program, it's for the best, just use that

I'm not telling you to change your gpus, your rtx 4060 is fine

you can't use windows xp for 50 years

technology is about making revolution and upgrading, and AI progresses at sonic speed

[austere snow]

yes

[low shard]

the version i'm telling you to use is way better than the one you're using right now

[austere snow]

it is multiplying

[low shard]

are you looking to make an rvc model?

[austere snow]

and tha google come back is awasome

[austere snow]

what is rvc

[low shard]

did you clean welll the datset?

might also be that the dataset didn't contain a various pitch range

[austere snow]

no

[low shard]

Retrieval-based-Voice-Conversion

Speech to Speech AI models commonly used for realtime inference and ai covers

[austere snow]

i have to make ai avatar

ai character

[trim sparrow]

i normalized it cleaned the truancey thing cant rember what its called its mainly clear the dataset is around 22 minutes long do you think i should add another 5-10 minutes of me going through various pitch i thought i did but i guess its not the pitch needed

[low shard]

a virtual 3d character? sorry but I don't personally know of any ai that does this, maybe other helpers will know

[austere snow]

but something will help to that

[viral mason]

forgot how to fix this issue

[austere snow]

the programming, ml, dl is same in all

all generative ai

[low shard]

• what's your pc gpu
• what do you want to do
• what kaggle/tutorial link are you using

[trim sparrow]

lower batch size for gpu

[viral mason]

it's batch 4 it should work fine

[austere snow]

ok im giving you more context, i have to make generative ai

what

?

[low shard]

that still has many subfields, please start reading the pytorch and huggingface docs, and also start by checking the basics

I can't explain the entirety of ml with all subfields in 5 minutes 😭

[austere snow]

ok i have to make voice agent

voice ai

[low shard]

so, a speech to speech voice model?

[austere snow]

like gemini and alexa

[austere snow]

yes

[low shard]

that's RVC

read https://docs.aihub.gg/essentials/whats-rvc and https://docs.aihub.gg/essentials/how-to-make-voice-models/

[low shard]

could you reply to my questions?

[austere snow]

no my avatar will speak

[low shard]

I don't know how to help you making a 3d avatar with AI, i can help you with RVC and image gen for example

[austere snow]

i have to create one i dont want to use

i used too many ai models

[austere snow]

this link telling how to use model

that open source model

[low shard]

no that's for explaining you what's rvc and how to train them

please

read it

just read the docs, they got everything on how to train models

[austere snow]

how do you get such resources please

tell me

[low shard]

https://docs.aihub.gg

[austere snow]

so that i can find by my self

[low shard]

everything is in there

the ai hub docs

[austere snow]

i hve to make npc

for game

the ai character

hey you are understanding right, i cant express in words i dont have too much termanoligies

i just need help

[low shard]

@austere snow you're changing topic and request every 5 seconds, firstly focus on what you want

[austere snow]

thats the same i have to create ai avatar in game

the character

who can speak

to our response and act as other person

[austere snow]

i didnt tell that i need for game, i thought it would be odd to ask then i saw your profile

[craggy bough]

(so basically create what multibillion dollar companies struggle to do)

[austere snow]

there is written ai gaming

[viral mason]

2. I want to train my model
3. I'm not using a tutorial I know how to use kaggle applio (https://www.kaggle.com/code/sillyslugcat/applio/edit with the applio fix made by a user in this server (I forgot who already) )

[low shard]

not in a related way

[austere snow]

yes

[low shard]

be serious here

[austere snow]

but im really serious 😔

[viral mason]

https://cdn.discordapp.com/attachments/1131011308792324159/1362874159914418606/speechmeme.com_18_04_2025_20_34_27.gif

[austere snow]

chat gpt, perplexity wasnt mulitbillion dollor company before

they created

so why cant i

[low shard]

oh damn you got a gtx 1660 super, that could theoretically train but would be slow asf and limited by vram here so it's better to use cloud

it's better you use the official applio kaggle https://github.com/IAHispano/Applio/blob/main/assets/Applio_Kaggle.ipynb

[austere snow]

?

[waxen pike]

I need to download these two files for Windows.

[untold marten]

Heya, just checked the link for how to use w-okada vc and seems like some stuff may be a bit different? I mean the way to add models and some stuff remain the same but i don't think i got at all how i can use the F0 Est option which i thinki it's the algorithm for pitch? idk whats the best one to use or most 2-3 recommended. also what can i do with the button "export to onnix"?

[viral mason]

aren't I using the original? or is the link I sent different

[austere snow]

please reply

[low shard]

nope you are in the wrong part, https://rentry.co/forkvoicechangerguide#download-nvidia-on-windows redirects to https://huggingface.co/Shadicti/deiteris-Fork/blob/main/voice-changer-windows-nvidia-b2332.zip

don't follow the nvidia 5000 serie part, you don't got an rtx 5000 serie

[low shard]

I'm not joking, be serious here, don't continously change topic and questions

[austere snow]

.

[austere snow]

but my topic is same

i just need road map

[low shard]

the link you sent is your own private kaggle edit, i feel like you used the outdated link https://www.kaggle.com/code/deiant/applio and then modified the code, but that link is outdated sine 2 months

use the link i sent before

[craggy bough]

1. get a loan
2. hire people who actually know what theyre doing

[analog obsidian]

export to onnx converts your model to onnx format, it's faster and gives less delay at the cost of slightly worse quality (smol but audible quality loss) and also i feel onnx is a bit unstable compared to .pth, but im not sure about that last one

[viral mason]

ah alright, is it the same when training a model?

[low shard]

the updated link should fix your issue, try it out

[austere snow]

no i will first join any other gaming company to see how npc works and racts

[analog obsidian]

and you have to use the f0 of the model, if your model was trained in rmvpe, you have to use rmvpe

[austere snow]

just give me roadmap till to make npc

then i will gain experience by myself

[viral mason]

the link u sent is what I use already, I import it into the kaggle space after opening a fresh one when training a model

[craggy bough]

1. get a loan
2. hire game devs

[austere snow]

pliz

[analog obsidian]

if you convert your model to onnx, you also have to use an onnx f0 so you get even less delay

[austere snow]

but i have to make it alone

[viral mason]

you smell..

[low shard]

you mean the github link? but you shouldn't need any user made fixes, it should be working all fine

[untold marten]

I see. Can i ask if the "slot index" means the position in the list of models where i can choose/? or ? i don' wanna overwrite something as it says 0-5 used but others doesn't so 6+ should be safe ?

[viral mason]

wdym?

[craggy bough]

1. get a loan
2. hire game devs
3. pay them extra to keep their names out of it

[low shard]

@viral mason please be respectful to everyone

@austere snow seriously you can't expect to be explained the whole ML in 5 minutes and make an entire new AI without any knowledge and partners

[viral mason]

sorry

[low shard]

you used the latest https://github.com/IAHispano/Applio/blob/main/assets/Applio_Kaggle.ipynb ? it has been updated last week

[low shard]

it's fine dw, I'm taking care of the situation

[austere snow]

but game devs cant make ai npcs, they know only programming and some engines

[analog obsidian]

i have not idea what are you talking about, can u send a screenshot or something

[craggy bough]

1. get a loan
2. hire game devs
3. hire ai devs
4. pay them extra to keep their names out of it

[austere snow]

please explain

[viral mason]

I don't understand how to use this as it doesn't open the kaggle space when I click it

[untold marten]

i could but no perms here at least. do i send one in dm or where can i?

[analog obsidian]

just read this https://rentry.co/forkvoicechangerguide

takes 5 mins of reading

[low shard]

In Kaggle you can create a notebook, then click import from GitHub and insert the link

[analog obsidian]

explains all about the voice changer

[viral mason]

oh, yea I already have this then

think I was confusing you on how I explained it

[craggy bough]

1. get a loan (cuz you clearly have no money)
2. hire game devs (to develop the npc parts)
3. hire ai devs (to develop the ai integrations)
4. pay them extra to keep their names out of it (cuz your ego wants you to do it "by yourself")

[viral mason]

-# damn

[untold marten]

i've checked that out and thats why i asked here. i looked several times on https://docs.aihub.gg/essentials/whats-rvc/

[austere snow]

ok i have to become ai developer

how can i be

[craggy bough]

university

[austere snow]

without university by courses

[viral mason]

https://cdn.discordapp.com/attachments/1159290139609137264/1363207818013966597/Walterspeechbubble.gif

[austere snow]

like elon musk

[craggy bough]

yeah well elon musk has money

you dont

[analog obsidian]

don't be paranoid, just add your model, the thing is not going to break

lol

[low shard]

Did you do it during this week? Because it got updated around last week

[austere snow]

self learning dosent require money

[viral mason]

what are you trying to do, if it's with the okoda voice changer I can help u, I use it every day

[viral mason]

ye

[untold marten]

ehhmmm..?

[analog obsidian]

are you using rvc or w-okada?

[low shard]

Are you using T4x2 GPU?

[analog obsidian]

because u literally asked for a setting thats only in w-okada

[viral mason]

I am not sure, the gpu I use is in that video I sent a second ago

[low shard]

They also could be using a weird rvc off youtube

[analog obsidian]

slot index... i think he's using the very old w-okada

like the one used in v1 days

[low shard]

That's your local GPU, you're using Kaggle, cloud (remote good PC), so kaggle's GPUs, you have to set the GPU, verify your phone number and turn in internet in the session options

[low shard]

God fuck those outdated video tutorials

[viral mason]

I've done that, but what do you mean set the gpu?

this yes?

[untold marten]

basically i wanted to ask 2 things:

• recommended F0 Est choices. i understood thats like pitch algorithm ? but just said "*don'tr choose crepe just because some ytber said that... *" and nothing else
• when tring to export as onnix a model, it asks the index slot to save but i'm wondering if that is about the list of models saved and asks me where i want to save it. i could have posted an image but i have no perms to upload one.

[analog obsidian]

1. depends in the model
2. index slot is the model index file

[low shard]

Yep you set it all right, could you show me also your applio training settings?

[analog obsidian]

add your .index file there

[untold marten]

sadly the model i got had initially no index file

[austere snow]

no one giving me answer for how to become ai developer

😔

[analog obsidian]

but converting to onnx doesn't ask for the index

[untold marten]

for me it does tho

[analog obsidian]

nick

can u

give him image perms

[craggy bough]

university

[analog obsidian]

@low shard

[low shard]

!give-media-perms 1h @untold marten

[analog obsidian]

ok @untold marten send a screenshot of what are you seeing rn

[viral mason]

ok so uh use this one
https://huggingface.co/Shadicti/deiteris-Fork/resolve/main/voice-changer-windows-nvidia-b2332.zip?download=true

[untold marten]

okay a sec

[austere snow]

but there might be some courses,

[austere snow]

online

[craggy bough]

why do you think companies hire people with university degrees?

[viral mason]

@low shard is this the most updated version of this okoda?

[craggy bough]

cuz theyre educated

[austere snow]

they are hiring now according to skills

[low shard]

Yup, latest version of wokada deiteris form

[viral mason]

sure!

[craggy bough]

(which you have none of)

[austere snow]

elon hires people according to skills

no degrees

see

[untold marten]

one img is the client and the other what appears after pressing the export button

[craggy bough]

explains why twitter/x and tesla are shit then

[analog obsidian]

just like i guessed it

you're using the original w-okada

[untold marten]

got it some time ago

[austere snow]

what abut space x

[analog obsidian]

and yes, index slot is where you want to place your model, it doesn't matter

[austere snow]

nuralink

[craggy bough]

are you seriously saying that engineers dont have degrees?

[analog obsidian]

you can keep using it if you like that, but just remember that the fork runs way better

[austere snow]

he hires according to skills

[analog obsidian]

also don't use onnx in the original version, it's bugged

[craggy bough]

ok blocked.

[analog obsidian]

increases cpu usage like crazy

[craggy bough]

trolls arent worth my time

[austere snow]

he himself chief engineer of space x, without having degree

[viral mason]

agreed

[analog obsidian]

onnx works fine in the fork

[austere snow]

so you are out fo the reality

[craggy bough]

especially cave trolls like this one

[viral mason]

he's like.. got a fish in his head and not a brain/j

[austere snow]

people in todays date, for asking help ego comes first

[untold marten]

i see. it's good or bad if i use the original one? i remember the index file being something additional to the pth file thats fine even if it's missing

[austere snow]

then why it is support channel

[untold marten]

aaa ok

[austere snow]

if not have to support

[analog obsidian]

same results, worse perfomance

[craggy bough]

remove the /j and itll be true

[viral mason]

real

[analog obsidian]

index increases cpu usage to like 70% for me in the original version

[untold marten]

then i'll download the fork one from deiteris right?

[untold marten]

YIKES

[analog obsidian]

yup

[viral mason]

I sent the link here just a second ago

[untold marten]

thankyu ❤️

[viral mason]

https://huggingface.co/Shadicti/deiteris-Fork/resolve/main/voice-changer-windows-nvidia-b2332.zip?download=true

[analog obsidian]

quality wise they're the same so dont worry

[viral mason]

ur welcome ^^

[analog obsidian]

settings are the same

[untold marten]

sorry if i type slower but i have right hand in cast and can't type 100% and takes time to correct'

[untold marten]

also on this one, it's same guide as the link from github of deiteris, right? just search deiteris on github too a few sec ago.
https://github.com/deiteris/voice-changer?tab=readme-ov-file#for-cpu-only-voice-conversion

[analog obsidian]

it's the same yeah, the link in the guide merged the two nvidia zips to make the installation easier for people

[untold marten]

ah okay then doesn't matter what i follow as it's the same. thanks for the help^^

sorry if i asked obvious stuff but i didn't dabble with them since like idk..., 2022-2023 and wanted to get back and heard some stuff changed and so on and seen more options + the new groundbreaking change: beatrice that uses 100% cpu and real time conversion etc and wanted to give it a try (to both as i got a new beefy pc with good gpu for AI)

ehmmm... wanted to send a sticker image with shylily heart but it's blocked here... oh well i just wanted to express gratitude ^^

[viral mason]

@low shard sorry for pinging u so much but it is working now idk what fixed it

think it was just tweaking

[austere snow]

i will fight, hasta la morte

bye

you are missing something

[analog obsidian]

beatrice models are like a downgraded version of rvc, they're unrealistic and sound a bit weird, but they run very fast

[viral mason]

https://media.discordapp.net/attachments/1171122607165276202/1171122927916294274/attachment.gif

[fickle acorn]

unmute me guys i will be polite

i swear

[untold marten]

I see. guess they are just in pre-alpha /alpha so it will take a while until they can hold up to the normal level of rvc with gpu and stuff

[analog obsidian]

they released their v2 version not so long ago

[gusty cloak]

when I run codename rvc for I getting this error:
ERROR: Exception in ASGI application
(Big traceback, cant send through Discord)
TypeError: argument of type 'bool' is not iterable
An error occurred launching Gradio: When localhost is not accessible, a shareable link must be created. Please set share=True or check your proxy settings to allow access to localhost.

[analog obsidian]

it's def an improvement compared to what it was years ago

[untold marten]

thats cool

[analog obsidian]

https://x.com/prj_beatrice

Beatrice作ってる人@声質変換VST (@prj_beatrice) on X

Project Beatrice / 完全無料のストリーム声質変換VST「Beatrice」を作っています / Beatrice 2 ベータ版公開中 / VC Clientでも使えるようになりました / もっといろんな声の表現ができるようになることを目指しています

X

everything about beatrice is there

(yes, we don't have beatrice english resources, everything is in japanese)

[untold marten]

took a file of beatrice v2 but idk whats with all the names (just more choices with diff pitch algorithms? or what?

[untold marten]

thats prob last thing to worry :))

[analog obsidian]

things work differently there yeah

they have multispeaker models

its possible to do that in rvc atm but the results arent the best (at least for finetuning)

beatrice runs light fast in your cpu

no gpu used at all

about f0 estimation used there... i have 0 clue

i know that beatrice models have an inbuilt noise supressor

[untold marten]

it removes that

guess it's just for gpu models the estimation and choices for chunks and extra so no need for those options to appear on beatrice

[analog obsidian]

makes sense, this is not rvc

is a whole new thing

[viral mason]

wait beatrice isn't rvc?

[analog obsidian]

no

[viral mason]

oh

[analog obsidian]

they do have a f0 estimator but its weird

https://huggingface.co/fierce-cats/beatrice-trainer

this is their trainer

looks like this requires a big dataset in order to give good results, last time i tried a 5 min one and the results were horrible

[viral mason]

wait what even is this, I'm slow

[analog obsidian]

wok is involved in beatrice's development i think

he's here btw (the author of w-okada)

[viral mason]

peak

[analog obsidian]

this is their trainer... to train.... models

[viral mason]

oh

I use these settings in the okoda

idk if these are the best or not

[analog obsidian]

i was about to say enable fp32, but you've got a 16xx gpu so you're forced to use fp32 anyways lmao

[viral mason]

lol

[untold marten]

so compared to gpu models, it's worse at training if your dataset isn't big, right?

[analog obsidian]

thats what i've got last time i tried it

i only trained once so idk

[viral mason]

any other setting I should change to improve my delay or anything?

[untold marten]

i've used to train before (1y ago+) a model with about 21/30min dataset and got good results after like 300-450+ epochs with gpu model ofc after a friends voice

[analog obsidian]

convert your model to onnx, use rmvpe_onnx, use crossfade 1s

[untold marten]

wasn't great dataset but good nonetheless

[analog obsidian]

rvc it's better bro

[viral mason]

I thought onnx made the quality worse tho..

[untold marten]

and used google collab as my pc was bad :))

[analog obsidian]

yeah but most probably you wont notice a difference

lol

[viral mason]

o

👍

[analog obsidian]

its so minimal

for lowest delay possible you can set crossfade to 0.05

the model will sound more unnatural

but hey, technically its the lowest delay possible

[viral mason]

ah

[analog obsidian]

[viral mason]

I still want quality tho ;-;

[analog obsidian]

ok so crossfade 1s

[viral mason]

0.1?

[analog obsidian]

1s

but also one more thing, onnx increases cpu usage

🦈

[viral mason]

bruh

[analog obsidian]

i think its like 5-10% more