✨│ai-help

[wraith heart]

can someone give me tips on how to make my voicechanger sound smoother? it always sounds unnatural n weird (male -> female voice)

[dim jewel]

Will my audio sample model lose some of model's voice if I set index to 0?
it's really hard for me to wrap my mind how can accent(pronunciation of sounds) be separate from the actuall voice. I mean I understand that a person can have one voice, yet use different accents, like American or British. But how come pronunciation of sounds is not behind the person voice. Is it then pitch? Tember?
If accent is shaped by our mouth, then voice is shaped by our throat? And AI can distinguish it?

[tame mica]

if you use index to 0, it will change the model's accent exactly like the inferred sample

on higher indexes, it will try to change the accent exactly to the model

idk bout the other question tho that's outta my league

[low shard]

Wrong channel, use #🔍│help-w-okada

Don't use that software nor yt tuts

What's ur PC GPU and what u looking for

That's old, and use #🔍│help-w-okada

Hell yes

[tame mica]

#🔍│help-w-okada

[low shard]

@river smelt #🧬│ai-chat message this is the right channel for inference on prerecorded audios

tell me ur pc gpu

[river smelt]

I have laptop, but not with gpu. It's Dell Latitude 5285.

[low shard]

• Cloud (remote good pc, easier and faster than ur PC but it's limited):
  • Ilaria RVC Zero: fastest and simplest that you can get for free
  • Weights.gg: Partnered with AI Hub, lets u do them easily but u may be in a queue
  • Applio Colab: max 4 hours, not granted, of GPU

Easiest possible (automatically separates vocals & instrumentals) : weights.gg
easiest cloud: Ilaria rvc zero

Your PC is not that good for it locally, it could run on CPU, but will be very slow, I would suggest to use cloud since you got integrated gpu and only 8gb of ram ddr3

[hallow thistle]

You can run an RVC fork program locally on your laptop with only CPU, but it won't be that fast. I tried it with my Dell Inspiron laptop.

[river smelt]

What do you mean very slow? It does make the result of audio become bad?

[river smelt]

Thanks for this. I'll try it now

[hallow thistle]

Uh.

[knotty moth]

8 GB DDR3 in 2025

[hallow thistle]

Whether if it's slow or fast, it has nothing to do with the quality of the output audio, unless you did some settings on RVC.

[river smelt]

I planned to make cover voice ai from beginning, not the RVC

[hallow thistle]

RVC isn't like W-Okada the realtime voice changer, where the speed of CPU or GPU matters to the quality of the output audio.

[night bone]

Wha-

[night bone]

Thats already alright...

[river smelt]

Misunderstand

[low shard]

In the conversion

[low shard]

So, use models on pre-recorded audios or for calls realtime?

[hallow thistle]

Why'd you ask me like that? You'll need a "Line In" or microphone recording program or use the built-in monitor recorder on W-Okada to record its output audio.

[night bone]

Is their a voice model for "suisei" in blue archive, im still a noob anyways on this shtuff

[hallow thistle]

You mean Sensei?

I found some on Weights.

[river smelt]

No, i didn't mean that. I mean ai can cover audio from music.

[night bone]

Oh wait suisei is not in blue archive mb

[hallow thistle]

https://cdn.discordapp.com/emojis/1016021422809817160.webp?size=48

So, that's mean I've never heard of this girl name.

[night bone]

Its a vocaloid i think idk a fellow filipino challanged me to ai cover an opm using her voice:|

[hallow thistle]

To download a voice model, go to #1175430844685484042. To find a voice model, use Weights bot in #🔍│find-models.

[knotty moth]

https://tenor.com/view/hoshimachi-suisei-thinking-pondering-thonk-rolling-eyes-gif-24617936

[hallow thistle]

https://tenor.com/view/blue-archive-tendou-alice-thinking-loading-gif-1346706331989991068

[night bone]

Since kits.ai f me over i just sawr the video and said hell yeah

[night bone]

Yip

[proven hill]

nice!

[night bone]

Brvuh

[proven hill]

?

i like the artwork

[night bone]

Wait... since theirs no rules stating age, am i allowed?
Since in some servers i get banned and sheesh without an explanation...

and this is an alt acc

[proven hill]

how old are you?

[night bone]

Disi-singko...

[hallow thistle]

https://cdn.discordapp.com/emojis/1016021422809817160.webp?size=48

[proven hill]

@low shard underage user (minor of 13)

[night bone]

Im cooked then

[hallow thistle]

If you're under 13, this can get you banned everywhere. Huh.

[night bone]

Eh im 15 but some servers i get banned without a single explanation even though there are no rules stating a specific age allowed in the server im just gonna ask now if under 18 is allowed here.

[hallow thistle]

Using a Discord account at such underage is against Discord's Terms of Service.

[proven hill]

why would you get banned if youre 15

[night bone]

What age?
13?

[hallow thistle]

This number of age should be fine.

[night bone]

Idk, i made this alt for that server and i just got banned again, though the server is just a community server for a vtuber(yes im a dweeb)

[hallow thistle]

The underage means any who's under 13, not minor (less than 18).

[night bone]

Hmm i see, lets end this convo before it has the potential to become some political sheesh

[proven hill]

what

[night bone]

Cuz in some servers im in, the trivial matter were talking about just boils down into a political argument.
Though that will not happen here ig

[low shard]

Wait what

15 or under 13 then

[proven hill]

nothing hes 15

idk he was presenting himself as under 13

[hallow thistle]

[night bone]

Misunderstanding or all what i said was vague?

[hallow thistle]

Something like this can make me and anyone here worried.

[night bone]

Alright

[storm shard]

why not?

I'm on a ryzen 5 with a 2060

[low shard]

bc it's very very old, and youtube tuts aren't updated

you're looking for inference (use models) on pre-recorded audios right

[storm shard]

I'm looking for something like that RVC converter. Something simple. This pne is pretty much drag and drop. I like it cuz it's easy to use and convert.

[storm shard]

yes

[low shard]

You can:

• Locally (runs on your pc so the speed depends on that, you will have to set it up with the guides):
  • Applio: A fork of RVC with some extra features like Applio TTS, kinda faster and simpler but same quality tho
  • Mainline: The original RVC
• Cloud (remote good pc, easier and faster than ur PC but it's limited):
  • Ilaria RVC Zero: fastest and simplest that you can get for free
  • Weights.gg: Partnered with AI Hub, lets u do them easily but u may be in a queue
  • Applio Colab: max 4 hours, not granted, of GPU

Easiest possible (automatically separates vocals & instrumentals) : weights.gg
easiest cloud: Ilaria rvc zero
easiest local: Applio

[storm shard]

I'll go locally. Thank you. I will try the applio and see how it goes. I kinda wanted to update the other RVC anyway
Is there a way to keep the libraries and whatever it needs within the pkg folder?
I don't want to lose it again.

[low shard]

U can get the precompiled one

Microsoft: UPDATE RIGHT NOW

[valid spruce]

Hi guys

What is it batch size?

Could someone explain this to me?

[analog obsidian]

how many samples at the time the model will learn

[valid spruce]

Thanks for helping me

[analog obsidian]

a safe value is 8

[valid spruce]

I always leave it at 8 or 9

[knotty moth]

not 9

[analog obsidian]

perfect you fine

[valid spruce]

But usually it's always on 8

I have another question

[analog obsidian]

any batch size works, but i agree that 9 is a bit unsafe

[valid spruce]

How do I get model singers to sing falsetto perfectly without that screaming sound?

[indigo harness]

I've got an issue with my rvc, For some reason its not working on discord i hadnt messed around with the settings much but one day it just started not catching the voice anymore, anybody else got this issue before?

[twin magnet]

Can someone help me rq? I'm using the voice changer on discord but its just speeds my message up and makes it crispy

[latent kettle]

How do I make applio link public?

[viscid moss]

Adding --share but not sure where

[carmine siren]

-colab

[azure marsh]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• FaceFusion UI, by Nick088 Google Colab
• FaceFusion NO UI, by Nick088 Google Colab
• EasyGUI, by Rejekts Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

More commands: -audio, -hf, -local, -kaggle, -realtime, -rvc, /help

[latent kettle]

you Can use #🤖│bots for commands

[carmine siren]

well this also working fine as of now

[latent kettle]

i mean use correct channels

[carmine siren]

#🤖│bots

[olive night]

Hello everyone! i'm currently trying to set up google collab w okada, but i have no option to connect it to google drive, i need to connect it to google drive for everyday use. please help, thank you.

is someone online?

uh

[brisk orchid]

how do you make the output and input so you can use the voice changer in roblox?

[olive night]

also ngrock is not working for me but the other option is dissapeared too

[radiant thorn]

I was wondering if someone could vc with me and help me

[olive night]

[Errno 2] No such file or directory: '/content/voice-changer/server'
/content

ModuleNotFoundError Traceback (most recent call last)
<ipython-input-1-e49dd8a37163> in <cell line: 24>()
22 get_ipython().run_line_magic('cd', '/content/voice-changer/server')
23
---> 24 from pyngrok import conf, ngrok
25 MyConfig = conf.PyngrokConfig()
26 MyConfig.auth_token = Token

ModuleNotFoundError: No module named 'pyngrok'

---

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

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

[indigo harness]

gonn ask again since nobody really answered but I've got an issue with my rvc, For some reason its not working on discord i hadnt messed around with the settings much but one day it just started not catching the voice anymore, anybody else got this issue before?

[latent kettle]

better if you ask in #🔍│help-w-okada if you are talking about Realtime voice tool (okada)

[indigo harness]

Oh i thought i was using the correct channel since this is rvc

Alright

[pastel oak]

nope its just unfortunate thats the same abbreviation, rvc is retrieval based voice conversion

[empty ocean]

I'm trying to convert a lot of small files but I seem to be getting rate limited, is there anything else other than https://huggingface.co/spaces/TheStinger/Ilaria_RVC I can use? or is there another one I can use at the same time to double my conversions?

or I guess the other question is...can batch conversion be done with the same level of quality as that link

[tiny totem]

Is it generally safe to make the Learning rate lower for small datasets?

[raven burrow]

Hey guys quick question here. What's the fastest way to train in cloud? Paid Colab? Wich gpu? do we have some info about that?

[analog obsidian]

free tier kaggle and lightning ai are the fastest
for paid options no idea which one is the fastest, paid colab has an a100 which has decent speeds

[analog obsidian]

maybe but remember that a lower lr requires training more epochs than usual

depending in how much u decreased it

[crude flame]

runpod has h100s

paid tho

[raven burrow]

ty @analog obsidian @crude flame gonna check those later :3

[tiny totem]

The model I did train sounds pretty decent as-is but the total loss for the generator (I think) consistently stayed at 41 and the graph was really inconsistent so I'm just wondering if I either need to just train it more with the current LR or lower it more

I noticed that when training there was something about Multi speaker datasets but idk how that works or if I can even get the model to output the right voice if I do try that

[silent yacht]

how to get this rvc v2 thing

[analog obsidian]

thats just because the dataset is small, a lower lr is not gonna make it better

[silent yacht]

/??

[tiny totem]

#🤖│bots message

[silent yacht]

i dont get it

[silent yacht]

i just want the best local software that i can run locally which works with msot of the models in #1175430844685484042

[analog obsidian]

applio https://github.com/IAHispano/Applio

[tiny totem]

I just use colab.

[analog obsidian]

for simplicity use the compiled version https://huggingface.co/IAHispano/Applio/blob/main/Compiled/Windows/ApplioV3.2.8-bugfix.zip

[silent yacht]

thank you 🙏 does it use my gpu and cpu? i want it to use both

[analog obsidian]

yes, uses your gpu for converting audios

[silent yacht]

is there any special way i have to set it up or is there a file i just run which will do it for me?

[tiny totem]

Isn't it called run.*

[analog obsidian]

nope, just unzip the file in a folder and use run-applio.bat

place your models in the logs folder

[jaunty mason]

What can I use to make AI covers of songs live? does anyone know

[hybrid rune]

does this go here, or there?

[silent yacht]

uh does this support real time?

[analog obsidian]

no

[silent yacht]

ok welp what does support realtime

[hybrid rune]

who knows how to help me? I've been trying to fix this sense 11:00 am

[proven hill]

ita the wrong pth

[hybrid rune]

ok where would the right one be?

[proven hill]

show folder of the weights

lets go to #🔍│help-w-okada

[bold goblet]

For overclocking my gpu is core clock or memory clock more important for rvc

[proven hill]

no

[bold goblet]

What

It wasn’t a yes or no question

[proven hill]

its useless to overclock for rvc

[lyric hinge]

hi

how can run rvc with gpu amd

rx590 8g

[proven hill]

voicechanger or normal rvc?

[lyric hinge]

normal

[knotty moth]

you'll get 1% performance increase and within error margin

[lyric hinge]

I have a 30 minute audio that actually takes a whole hour and that's a long time so I want to use a graphics card

[proven hill]

then no, check online options

[hallow thistle]

That doesn't answer the question.

[knotty moth]

split it into 5 minutes each or use split audio feature in applio

[hallow thistle]

You might either mean the RVC if you wanna train or inference a 30-minute audio.

[proven hill]

yes but he have an amd carf

[lyric hinge]

inference

Yes, inference. I did not understand correctly at first. I sorry.

[knotty moth]

the gpu doesn't matter but that's recommended to do
note that applio supports amd gpus

[simple ore]

for inference you can use Applio

[proven hill]

really?

[simple ore]

https://docs.applio.org/applio/getting-started/installation#amd-gpu-support-windows

[lyric hinge]

Thank you Do you think this will take less time right?

I have a processor Rayzen 5 3600

and a graphics card RX590 8g

[hallow thistle]

It is possible to inference a single 30-minute audio. However, it would be taking too long to inference like that.

[knotty moth]

it could run OOM without splitting

[lyric hinge]

I can make a Python script that splits a 30 minute audio into 5 minute parts.

[hallow thistle]

https://cdn.discordapp.com/emojis/1016021422809817160.webp?size=48

Really?

[knotty moth]

better do it manually without cutting voice sentences

[lyric hinge]

Sorry what is oom I'm new to this

[knotty moth]

OutOfMemoryError

[lyric hinge]

yeah

[lyric hinge]

If it was run on colab, would it be better? I want the shortest possible time.

[proven hill]

colab is not reliable for this

[hallow thistle]

Inferencing a longer audio duration, it would be better to done on a super fast GPU with more VRAM.

[hallow thistle]

For free version of Google Colab, no.

[lyric hinge]

Well thank you
you helped me

[azure marsh]

Hey, ! JOE! Please use the command !howtoask to increase your chance of getting help by structuring your question in a way others can understand better. Also make sure you're asking in the right help channel:

• General RVC help: #✨│ai-help
• W-Okada / Realtime RVC: #🔍│help-w-okada
• AI image related: #🔍│help-ai-art

[hallow thistle]

However, the paid version like Google Colab Pro, you will get more RAM and GPU VRAM.

[knotty moth]

actually a voice sentence couldn't be more than 1 minute and that doesn't take much vram to infer

and you just need to denoise and noise gate before doing so

[lyric hinge]

In fact, I am converting the audio of a lecture video to another audio model. The lecture is on average 3 hours long. I know that it is crazy, but I am trying.

[viscid moss]

@humble cobalt After trying several times, I found out that the amplification threshold is the volume of the primary stem, and the normalization threshold is the volume of the secondary stem. I changed the default values ​​to these, compared the SDRs and these values ​​seem to be fine, since as I told you very high values ​​introduce noise.

[knotty moth]

again just do denoise and noise gate, then split it into 1-3 minutes each esp if you have 4 GB vram

[worldly marlin]

this is what am stuck on

[proven hill]

try restarting the training, its stuck it seems

you should loose much data

just re-input everything like before

[worldly marlin]

RESTART

[hallow thistle]

Not a real Windows XP.

[worldly marlin]

WAT U MEAN IT TOOK ME 10 HOUR

[hallow thistle]

Calm down.

[worldly marlin]

and i have to do it all again

[proven hill]

it will restart from the latest checkpoint dw 😭

[worldly marlin]

oef thank god

how do i restart

[hallow thistle]

Or you can wait for it to finish again, anything you want.

[proven hill]

stop the training

then start again

[worldly marlin]

this one

?

[proven hill]

yes

[worldly marlin]

doesnt even do nothing

just keep continuing

i just wanted to have a rvc deep voice for trolling on roblox

its for content

[hallow thistle]

Wait for it a while, it's kinda slow to suspend the training process.

[worldly marlin]

but i litterly dont understand how this works at all

@proven hill by any chance do you have a deep voice rvc model to troll with

[worldly marlin]

BRO STOP

i dont wanna be a girl

am a boy with a baby voice

[stable palm]

whats the best girl voice

fast

[worldly marlin]

ayo

[proven hill]

actually yes but its paid

[hallow thistle]

Nuh uh.

[worldly marlin]

paid my bum

[proven hill]

mines

[worldly marlin]

WHAT

PAID

oefnonon

2 mini files cost money?

[hallow thistle]

It would be a shame to go for e-girl, like fucking shame ass. However, if you go for a voice model of a real man voice, I'd leave it go anyways.

[proven hill]

yes

[worldly marlin]

oooh no am crying

[worldly marlin]

do u have a free one

[proven hill]

i mean its a market

[worldly marlin]

how much did u pay

[proven hill]

free mhh

[worldly marlin]

only got 2 euro

please

[proven hill]

theres something in my shop free

[hallow thistle]

I'm sorry, but I don't make a voice model.

[worldly marlin]

il pay 2 euro al my money

for somthing good

[proven hill]

i dont pay, i make them

[worldly marlin]

oef

you will dos me a deal

[worldly marlin]

well vat is it

[hallow thistle]

You can look for a voice model from #1175430844685484042. To #🔍│find-models, use Weights bot in #🔍│find-models.

https://cdn.discordapp.com/emojis/872078487639453707.webp?size=48

[worldly marlin]

dude i was camping there 24 7 only see playboi carti

[proven hill]

check link in bio

[worldly marlin]

the only deep voice am using rn is corpse husband

wich is bum cheeck

[hallow thistle]

[worldly marlin]

Hi, my name is Ilaria and im a 25yo girl from Italy 🇮🇹

i dont see no link

nvm i do

srry

[proven hill]

😭

[worldly marlin]

oh noo

i dons want girl

i want man

mama mia

ts too over 2 euro

😢

[proven hill]

mhhhh

why dont you search on #1175430844685484042 ?

[hallow thistle]

Boo. 👻

[worldly marlin]

all of em bad

most of em playboi carti

am very sad

my training completly failed

[proven hill]

did you checked weights.gg too?

[worldly marlin]

every website u name it

[latent kettle]

train on Weights.gg

[proven hill]

theres not much market as “male voices” tbh

[worldly marlin]

u litterly have it

[hallow thistle]

If you're looking for a generic voice model used in a commercial-type ahh website, then no. AI Hub by Weights doesn't have a voice model like that.

[worldly marlin]

but it over my budget

[proven hill]

its impossible theres nothing for free

[worldly marlin]

ow man

[hallow thistle]

https://cdn.discordapp.com/emojis/843447432174895164.webp?size=48

[worldly marlin]

hey not fanny

dont luaghs

https://www.weights.gg/models/clm72mqu008w8cctcaar21t8l

this is the one am using

[proven hill]

furry

[worldly marlin]

its corpse

[hallow thistle]

Skibidi Toilet.

[worldly marlin]

BUT ME CRYING

@proven hill is there any exeptions you make for me?

[proven hill]

regarding?

[worldly marlin]

for high five?

[hallow thistle]

I like anime women.

[proven hill]

trust me you can find it for free

[worldly marlin]

nos i cant pease can u find me a good or give me own of yours

[proven hill]

try finding an eboy voice

[worldly marlin]

they are alll badd

i wanna hear the one of yours

wait matter fact

this is what i want https://www.youtube.com/watch?v=SGFxMxxUboE&ab_channel=ItsKashB

[proven hill]

theres the demo

[worldly marlin]

oh ok but can u see like what i mean in that video

https://www.youtube.com/watch?v=SGFxMxxUboE&ab_channel=ItsKashB

this one like am tryna find that but it wont say nothing

[proven hill]

unlucky ig idk what to say anymore

[carmine siren]

#🤖│bots -kaggle

[hallow thistle]

https://cdn.discordapp.com/emojis/1016021422809817160.webp?size=48

[polar ridge]

Hi how many epoch is actually good? I set my epoch to 1000 and it has been 1 day already i am at 385 epoch lol

and my voice source is only less than 2 min like 1 min

Would it be bad to stop my training at 500 when i set it to 1000?

[proven hill]

epochs ≠ quality

[polar ridge]

so is not quality?

[proven hill]

no, the quality is made by the dataset mostly

the epoch is how many times the file is being trained on

[polar ridge]

Alright is there a way to enhance the trained model

[proven hill]

you cannot improve an already trained model

[polar ridge]

ok thank you

If the model sound

is quite low

when speaking is it possible to make it louder

like the pitch

do i need to use another software then? or ai tools

[proven hill]

interesting question, i think yes that you need something else

afaik theres no option to be louder in okada

[polar ridge]

👍

Do you have experience with Kits.ai?

[proven hill]

no, kits is bad

[wet agate]

hi

i need help

[polar ridge]

what do you prefer?

[wet agate]

how do i get rid of popping bubble sound or static

[proven hill]

i mean, im ilaria, the creator of ilaria rvc, so id say ilaria rvc 😭

[silent basalt]

yo guys can someone help me pls? Running with the system Python.
python.exe: can't open file 'C:\Windows\system32\rvcgui.py': [Errno 2] No such file or directory
Press any key to continue . . .
how can i fix that

[azure marsh]

Hey, levangabriel! Please use the command !howtoask to increase your chance of getting help by structuring your question in a way others can understand better. Also make sure you're asking in the right help channel:

• General RVC help: #✨│ai-help
• W-Okada / Realtime RVC: #🔍│help-w-okada
• AI image related: #🔍│help-ai-art

[proven hill]

could be a model problem

[proven hill]

youre using rvc gui?

[silent basalt]

yep tryna install

[polar ridge]

Llaria when done training do i need to press stop training or go to the settings menu and restart applio? How do i extract my pth files and other files?

[proven hill]

use something else like applio

[proven hill]

wait the train finishes

[polar ridge]

but is at 1000 epochs

lol

i think at 500 is enough

[proven hill]

then stop training

[knotty moth]

👎

[hallow thistle]

White noise was the only voice model I've ever trained with this shitass site.

[silent basalt]

import gradio as gr
ModuleNotFoundError: No module named 'gradio' ☠️

[proven hill]

did you set up everything correctly?

[silent basalt]

yep

im gonna try pip install maybe it's work

[proven hill]

LMFAO

[silent basalt]

oh theres some warnings :d

[hallow thistle]

Never use python -m pip install into a Python program if you don't know what you're doing.

[silent basalt]

im studying ce bro no problem i fixed it <3

[knotty moth]

-gui

[azure marsh]

https://cdn.discordapp.com/attachments/1122285248844144733/1203460490475343953/caption.gif?ex=65d12cec&is=65beb7ec&hm=bd2fb8d010006dd7c6e3c1c67d3ae846fd1478e1a3124c544c31b43086fe54aa&

[knotty moth]

better try download applio precompiled, so u dont need to worry about pip install shit

[polar ridge]

If you don’t finish all the epochs sessions

Will there be negative effects afterwards?

I set it to 1000

If I finish it at 500 is there a downside?

[proven hill]

no

[polar ridge]

Thank you

[urban cape]

Guys, does anyone know how to download a female voice changer and how to set it up?

[proven hill]

-rt

[azure marsh]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[proven hill]

first link

[glad mountain]

is rvc or w-okada bettter?

[proven hill]

its two different things

[glad mountain]

what are they both for

whats better

[proven hill]

what are you searching

[glad mountain]

wdym

[proven hill]

depends on what you need

[glad mountain]

girl voice changer

[hallow thistle]

RVC is the audio conversion program. W-Okada is the realtime voice changer program. These two programs have different usages.

[glad mountain]

oh ok

[hallow thistle]

A program that works better than RVC-GUI is Applio. A program that works better than W-Okada is Deitris fork W-Okada. Both programs can both run locally and on a cloud service.

[latent kettle]

Which audio is primary and which is secondary and what does SDRs means

[viscid moss]

U can check model's stems on leaderboard tab by filtering by vocals, instrumental, etc. Primary stem is showed on Stem 1 on gradio interface and secondary stem on Stem 2. Maybe a few models don't have their stems listed cause I'm still working on that cause there are too many models. About SDR..

[valid spruce]

Hello again guys

I heard that it is necessary to add a silent space in the dataset, is this true?

[latent kettle]

No

It don't make any sense

[polar ridge]

cant find my index file

in my model folder is this expected to happend?

[proven hill]

its in the logs folder

[polar ridge]

nothing

still

[proven hill]

did you train it?

[polar ridge]

yeah

[proven hill]

uh

[polar ridge]

is it important?

[proven hill]

train it again then, it will take no time

depends your use

id say nah

[proven hill]

what did you use to train?

[polar ridge]

applio

[proven hill]

idk if it was somewhere else then

[polar ridge]

they say is in'

model pth

[proven hill]

model pth is in the weights folder

[polar ridge]

there is no folder named weights

[proven hill]

oh

[polar ridge]

there is alot of pth.file

should i choose the closest one

[proven hill]

assests -> weights

[polar ridge]

there is no weights in assests

[hallow thistle]

You mean assets?

[polar ridge]

you know how to find index?

file

[polar ridge]

there is no index file....'

What is the fix?

Just train it agian?

[simple ore]

you can run 'create index' any time

not the full training

[proven hill]

https://tenor.com/view/interspecies-reviewers-crim-crimvael-annoyed-gif-10657062117549581565

[polar ridge]

there is nothing in logs folder i swear

[proven hill]

then it didnt traint the index

[polar ridge]

with the equipped model in the interface?

yah looks like

do i just use the model?

that i trained

to make the index then?

[hallow thistle]

Do you mean like some custom pretrained models?

[simple ore]

go to training tab, select your trained model, then click create index

[proven hill]

just do everytjing like if youre about to train the model but just press “train index”

[simple ore]

if you deleted the sliced audios and extracted features, you'll need to re-do those steps

[polar ridge]

what if i closed the program

[simple ore]

just re-open and do as I said

[polar ridge]

would the index have affect since i did not created the index from the start?

[simple ore]

no

[polar ridge]

Alright thank you

But what are the index used for?'

[azure marsh]

Faiss Integration (.index file): The Faiss library enables efficient approximate nearest neighbor search in RVC during inference, retrieving and combining training audio segments with closest embeddings. For your final RVC model, include the one which the file name starts with added.

Example: added_IVF157_Flat_nprobe_myModel.index

[polar ridge]

ok

still dont understqand'

lol

[hallow thistle]

The train index button is "Generate Index"?

[polar ridge]

i did'

whats the differences lol i know that the last one is stopped before being processed

the first one is fully finished

[polar ridge]

whats the best pre trained models?

is the original good?

[pastel oak]

original

[low shard]

there isn't a best one

it depends on ur dataset language and lenght

usually original (which has been trained on english) is the most used one

[polar ridge]

Thank you do you know where or best place to test your model/.pth file

Any website?

Or any ai tools that don’t require so much effort to download

[latent kettle]

Weights.gg

[polar ridge]

Thank you agian Dr Jr

[latent kettle]

Your welcome again @polar ridge

[lunar holly]

I got an ASUS Vivobook S 14 with Ryzen AI 9 HX 370 on board.
Still no such software, that uses NPU for processing AI voice changing?

Not something I bought my laptop for, I'm just curious, if there's anyone, who made it to work it on CPU's neural processing unit all of the sudden for the last 6-7 months instead of iGPU, when I've been considering getting one for myself.

[lavish lintel]

Starting pitch extraction with 4 cores on cuda:0 using rmvpe...
  0%|                                                     | 0/1 [00:00<?, ?it/s]An error occurred extracting file /kaggle/working/program_ml/logs/riffy_v3/sliced_audios_16k/0_0_0.wav on cuda:0: CUDA out of memory. Tried to allocate 1.93 GiB. GPU 
100%|█████████████████████████████████████████████| 1/1 [00:04<00:00,  4.54s/it]
Pitch extraction completed in 10.34 seconds.
Starting embedding extraction with 4 cores on cuda:0...
100%|█████████████████████████████████████████████| 1/1 [00:00<00:00,  2.18it/s]
Embedding extraction completed in 5.08 seconds.
An error occurred extracting the index: need at least one array to concatenate
If you are running this code in a virtual environment, make sure you have enough GPU available to generate the Index file.

Not sure what to do

[simple ore]

depends on NPU

[simple ore]

slice the audio

[lunar holly]

It's a Ryzen AI with 50 TOPS.
I also happen to have a mini PC with their older R7 7840HS, though it has 5 times less the NPU performance (only tops out at 10 TOPS in theoretical tests), and I doubt it will suffice for such a task.

[simple ore]

if it is AMD, expect pytorch on linux supporting it sometime in 2035

[lunar holly]

so that means never gets support on Windows then

[simple ore]

as I see there's support for ONNX

and directML

[lunar holly]

yeah, but is there any RVC software right now, that can take advantage of CPU's NPU?

[simple ore]

voice changer with onnx maybe?

[wispy lodge]

I don't think any software is capable because no one knows whether it's possible to choose NPU device

There're examples for DirectML in CPP, but none in Python. And no one reported whether python directml reports NPU device

And I think ryzen ai series use Vitis AI and separate execution provider, so it won't work with directml

[warm monolith]

Hello ! I have install Applio on my pc on local and I would like to share for my friends but i don't know where a create a public link with. "set share=True in launch()."

[low shard]

go in the applio folder, at the search bar of file explorer, write cmd, then in cmd write env\python.exe app.py --open --share

-# guessing that you're on windows and used the precompiled downloaded from HuggingFace

@simple ore btw adding another bat or another way to easily do this would be cool, just a suggestion

[simple ore]

can just edit the existing bat

why do you need another?

[low shard]

Was just saying for newbies to do this easier

[simple ore]

dont think it makes it available on public internet, does it?

[low shard]

It's the gradio share link so it works

It's the same way used for the Colab

But if you don't think it's useful,ofc u can disagree, I didn't mean this suggestion for myself nor going to force anyone

[brittle wing]

Hey I wanna make my own model soon I’m currently working on making the vocal batch. I wanted to ask how much all vocals should amount to, as in referring to time. I have a 40min studio session I want to use and I think I can get about 15 minutes of full vocals from it all ranging in different octaves

[low shard]

Nope, NPUs are basically never supported anywhere,
The only project I have ever seen support it was FastSDCPU iirc

-# from what I have seen personally

[plush compass]

Where can I download applio that uses refinegan? It's not in version 3.2.8...

[simple ore]

there was no official release yet

all you can do is clone the latest repository

[plush compass]

thank you i found it

[tight plank]

how is RVC still not AMD supported, I come here every 6 months I feel like

and I am once again disappointed

(on windows)

[median elbow]

-COLAB

[azure marsh]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• FaceFusion UI, by Nick088 Google Colab
• FaceFusion NO UI, by Nick088 Google Colab
• EasyGUI, by Rejekts Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

More commands: -audio, -hf, -local, -kaggle, -realtime, -rvc, /help

[simple ore]

You can run Applio on AMD GPU, ideally 5700+

[tight plank]

6600xt

[simple ore]

yeah, should be fine

https://docs.applio.org/applio/getting-started/installation#amd-gpu-support-windows

[analog verge]

-colab

[azure marsh]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• FaceFusion UI, by Nick088 Google Colab
• FaceFusion NO UI, by Nick088 Google Colab
• EasyGUI, by Rejekts Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

More commands: -audio, -hf, -local, -kaggle, -realtime, -rvc, /help

[fickle dust]

@low shard what do accent mean on Illyria zero like that setting? Is it like how accurate you want the voice to sound like them or?

[simple ore]

index value?

it is a slider between 0 (like audio) to 1 (like the model)

[velvet stream]

when i upload my audio on applio, the runtime is 0:00 to 0:00. would anyone know how to fix it or know why it's like that?

[quaint siren]

what does it mean?

feature retrieval rate\

[proven hill]

its when

[quaint siren]

sorry?

[proven hill]

dont say sorry

[quaint siren]

can you continue pls

[proven hill]

Feature retrieval rate basically indicates to the AI model how close or far off from the prompt/image it should work in terms of inference which dictates how close to the subject in input you want the output to look like.

[glacial pollen]

If you need more sophisticated answer, it controls the usage of feature index.
You see, when one trains a model, model learns features extracted from the voice ( it's own ) but, for simplicity I'll refer to features as " accent " or " way the voice articulates, stresses " certain phonetics etc.

Basically, if it's 0, model doesn't use the features from own index, heavily relies on what it can retrieve from the audio you use for cover ( inference ) and similarly but reversed, if it's 1, it'll use all it can from own index.

yet another take on it is:
see it as " strength of searching for own voice's features that'd be closely matched ( if possible ) with the audio "

Note: Poorly trained models or those trained on limited data will struggle with high index ratio; can manifest as artifacts or jammed / jittery pronunciation or phonetic handling

[lavish lintel]

if something like that happens (overtraining) where should i stop/export model

[lone nest]

How can I use a zip for training dataset instead of a single audio file on Applio Colab?

[proven hill]

you cant

[lone nest]

Aight lemme boot up Foobar again

[knotty moth]

unzip it first to a new folder

[lone nest]

Doesn't it take a single file input?

[proven hill]

you can have multiple files

[low shard]

The index value, which higher the value more accurate is the accent

[naive crow]

upload failed to import no module name utils

can you help with this?

[azure marsh]

Hey, Sophie! Please use the command !howtoask to increase your chance of getting help by structuring your question in a way others can understand better. Also make sure you're asking in the right help channel:

• General RVC help: #✨│ai-help
• W-Okada / Realtime RVC: #🔍│help-w-okada
• AI image related: #🔍│help-ai-art

[low shard]

Wrong channel, Wokada is different than RVC, tell the link tutorial you followed in #🔍│help-w-okada

[hallow thistle]

Use an entire zip to train? No, Applio doesn't read zip a file to train like that. You must extract its files to a dataset folder.

[regal epoch]

help me install rvc

[tame mica]

-docs

[azure marsh]

📚 Documentation

AI HUB Docs

https://docs.ai-hub.wtf

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

• How to use RVC Mainline Colab by Cauthess
• AICoverGen Colab Guide by Eddy (Spanish Helper)
• Create a model with RVC disconnected (colab) by Angetyde

[latent kettle]

Which gpu do you have

[regal epoch]

rtx 4070 laptop gpu

@latent kettle

[latent kettle]

And vram ?

[regal epoch]

ram?

[latent kettle]

No video RAM of your GPU

[regal epoch]

idk

one second

[latent kettle]

Go to task manager and open performance

[proven hill]

its useless to know the vram its good enough anyway

[regal epoch]

16 gb

[latent kettle]

Okay.

[latent kettle]

Go it

[regal epoch]

im install virtual cable

[latent kettle]

You can try applio

[latent kettle]

RVC is not Realtime voice conversation

[proven hill]

what are you looking for? what you wanna do?

[latent kettle]

If you want realtime conversation go to #🔍│help-w-okada

[hallow thistle]

RVC is the audio conversion program. W-Okada is the realtime voice conversion program.

[regal epoch]

Do you need a virtual cable?

[hallow thistle]

Bruh.

[proven hill]

^^^

[regal epoch]

AAAAAA

translation

sorry

[hallow thistle]

You will need Virtual Audio Cable if you wanna use W-Okada with Discord.

https://cdn.discordapp.com/emojis/1016021422809817160.webp?size=48

[regal epoch]

ok

[hallow thistle]

-realtime

[azure marsh]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[regal epoch]

im install realtime voice changer in disk c

[hallow thistle]

So many times people keep mistaking W-Okada as RVC. Click on the first link for Fork W-Okada, this fork W-Okada runs better than the second link.

[regal epoch]

windows 11 work?

[hallow thistle]

Any Windows version works, but Windows 10 is the minimum.

[regal epoch]

ok

[hallow thistle]

https://cdn.discordapp.com/emojis/1157877047805366343.webp?size=48

[regal epoch]

start start_htpp.bat?

[hallow thistle]

https://cdn.discordapp.com/emojis/843447432174895164.webp?size=48

Damn. I said click on the first link to download fork W-Okada, not the second link. Did I say it wrong?

[regal epoch]

blat

[hallow thistle]

Let's be real. For more information about using W-Okada, go to #🔍│help-w-okada. This #✨│ai-help is about RVC programs.

[regal epoch]

voice changer windows invidia

this?

[hallow thistle]

Anyways, screenshot your process at #🔍│help-w-okada.

[plush compass]

I ran it as explained in this link and succeeded in recognizing AMD.
RuntimeError: CUDA error: operation not supported
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions. An error appeared. Does anyone know about this error?

[simple ore]

card?

[hallow thistle]

Did you use the NVIDIA version instead of DirectML?

[plush compass]

Let me check in a moment.

[knotty moth]

are you using an AMD card? try follow the instructions carefully, including HIP SDK stuff

[simple ore]

I had no issues wit 6700xt

[plush compass]

AMD card? What is this?

[simple ore]

you also ened to post a bigger piece of log, not just cuda error.. I need to know where it happened

[plush compass]

I am also using 6700XT

[plush compass]

How can we find out? Would it be better to start from the beginning?

[simple ore]

hip sdk 5.7, these libraries https://github.com/brknsoul/ROCmLibs/blob/main/Optimised_ROCmLibs_gfx1031.7z

[plush compass]

Should I download that?

[simple ore]

that's what the guide says

does it not?

follow the guide, my dude

[plush compass]

Oh, I downloaded that too, but it still doesn't run. How can I fix the TT CUDA error... I think I'll have to try it from the beginning.

[simple ore]

show a bigger error message

[plush compass]

I will check and let you know. Thank you...

After checking, there are no other errors other than that. The AMD graphics card is also recognized well.

Thank you all for your replies. Let me start again from the beginning. I may have missed something.

[simple ore]

are you running applio with run-applio-amd.bat?

@plush compass

[plush compass]

YES

DEPRECATION: omegaconf 2.0.6 has a non-standard dependency specifier PyYAML>=5.1.*. pip 24.0 will enforce this behaviour change. A possible replacement is to upgrade to a newer version of omegaconf or contact the author to suggest that they release a version with a conforming dependency specifiers. Discussion can be found at https://github.com/pypa/pip/issues/12063
Installing collected packages: torch, torchvision, torchaudio
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
torch-directml 0.2.5.dev240914 requires torch==2.4.1, but you have torch 2.3.1+cu118 which is incompatible.
torch-directml 0.2.5.dev240914 requires torchvision==0.19.1, but you have torchvision 0.18.1+cu118 which is incompatible

Is this error a problem? There were no problems running it at all.

[simple ore]

you should be using a compiled version of applio

https://huggingface.co/IAHispano/Applio/blob/main/Compiled/Windows/ApplioV3.2.8-bugfix.zip

[plush compass]

I think you are using this. Is there a separate amd version of the 3.2.8 bugfix version?

[simple ore]

no

[plush compass]

Even though I use that, a cuda error appears TT

[simple ore]

use the compiled version, there's nothing to install (nothing that could trigger "DEPRECATION: omegaconf 2.0.6 has a non-standard dependency specifier PyYAML>=5.1.*. pip 24.0 will enforce this behaviour change. A possible replacement is to upgrade to a newer version of omegaconf or contact the author to suggest that they release a version with a conforming dependency specifiers. Discussion can be found at ")

make sure you replace torch as the guide explains

make sure you patch in the correct zluda version

make sure the path to HIP sdk is present in the environment variables

[plush compass]

I will try it all at once. Thank you.

[split grail]

why does rvc use the card so much

[proven hill]

because yes

[split grail]

On 4090 40%-60% while using it

[proven hill]

training?

or voice changer?

[split grail]

Voice changer

[proven hill]

wrong channel then

are you using the fork?

[split grail]

wdym by that?

[proven hill]

are you using the old version?

because the new version is less resource instensive

[split grail]

Rn im using smth with 72 or 73 i think

[proven hill]

send screenshot

[split grail]

2.0.0.73

Thats the version

In beta

[proven hill]

i think its the old version

-rt

[azure marsh]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[proven hill]

download the new one from the first link

[split grail]

Im using this

[proven hill]

yea its old

[split grail]

With this setup can i get good ovr performance?

[proven hill]

ovr?

[split grail]

Overall

Latency and quality

[proven hill]

yes you can ofc

[knotty moth]

pls discuss your topic in #🔍│help-w-okada
that spec is overkill already for the usual valorant etc. use cases tho

though 4090 should be more suitable for those who train models and flux loras

[split grail]

Idk what does it even mean

[proven hill]

woman stuff

[split grail]

Ah

[latent kettle]

@viscid moss Separation by Link is not working I'm UVR UI (Local) How to fix it ?

[viscid moss]

Can u show me the error on CMD?

If the video has any age, region or similar restrictions you will not be able to download it.

[latent kettle]

i try with multiple videos but it not worked

[viscid moss]

checking...

That video gives me an error too

I tried 10 random videos from YOUTUBE and all of them worked except the one you are using, I think there is some restriction. I even tried that video lyrics version and it worked for me

I'll try to figure out what's going on anyways, to see if I can improve the video downloader to avoid restrictions

[latent kettle]

okay. lemme try with other videos

[viscid moss]

if the original one isn't working try looking for lyrics version or re-uploaded versions

[latent kettle]

okay,

[latent kettle]

I have a doubt, will it download audio in best quality?

[viscid moss]

yep, download the audio in the best possible quality and in 32-bit float WAV

[latent kettle]

okay what about sample rate ?

[distant turtle]

-colab

[azure marsh]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• FaceFusion UI, by Nick088 Google Colab
• FaceFusion NO UI, by Nick088 Google Colab
• EasyGUI, by Rejekts Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

More commands: -audio, -hf, -local, -kaggle, -realtime, -rvc, /help

[proven hill]

italiano?

[viscid moss]

It depends on the source, I think most of the times 44100. I'm also not quite sure if all UVR models will work with 48KHz or 32KHz

[latent kettle]

emm okay. thank you for information

[viscid moss]

ur welcome

[proven hill]

ur welcome in my house

[viscid moss]

[distant turtle]

si

[proven hill]

come ti posso aiutare?

[proven hill]

hru eddyyy

[distant turtle]

grazie, nulla mi avevi già aiutato mesi fà, ora so come fare

[proven hill]

ah perfetto 💖

[distant turtle]

era molto comodo il tuo link del nuovo aggiornamento su yt, peccato perchè io che ho un pc lento e uso solo colab, mi sarebbe toranto utile, ora uso applio che però è lento ad aprirsi, e molte volte si blocca e devo riaprirlo, spero ritornerà un tuo aggiornamento su yt

questo solo per testare la voce finita

[proven hill]

non lo so.. nessuno è più interessato ormai

[distant turtle]

ah peccato

[distant turtle]

io lo sono ancora, e anche molto, lo uso per cover e anche per doppiaggi di serie tv fan made

*uso

[proven hill]

carino

[distant turtle]

anche se per fare dei doppiaggi perfetti aspetto che tortoise tts aggiunga l'italiano e che ci sia una versione si colab

[proven hill]

tortoise se non sbaglio non è più supportato

[distant turtle]

ah non sapevo, io ogni giorno controllo su yt se esce una versione italiana di tortoise tts o di una tts decente, avevo già provato eleven labs, ma non puoi decidere le pause nel testo, non puoi scegliere le emozioni, e parla troppo veloce per un doppiaggio

grazie

[proven hill]

di nulla :)

[latent kettle]

@viscid moss I tried it with more videos but it only worked with 2 video. Out of 10

[viscid moss]

Coding stuff :3

[viscid moss]

damn. For now if it doesn't work try downloading from a web page or something like that. I'll check it later, I just finished fixing the UnicodeDecodeError and I'm publishing the update

[latent kettle]

how do i update ??

[viscid moss]

That's related to yt-dlp (audio downloader) not related to separator so dw so much

[latent kettle]

and after updating can i disable that thin after updation

[viscid moss]

running the updater

[viscid moss]

ye

[latent kettle]

now ? or later ?

[viscid moss]

I tested that on my end and seems to be working without that thing

[viscid moss]

later

not published yet

[latent kettle]

okay.. thank you.. im bug reporter for you 😅

[viscid moss]

fr ngl

[latent kettle]

what does it mean ?

[viscid moss]

Certainly. Not gonna lie

[latent kettle]

doing batch conversion and geting this problem again and again it dont process few audios

[viscid moss]

show me ur input and output paths

[latent kettle]

i closed because i got frustated but the paths are input C:\Users\The Beast\Desktop\DIN Output C:\Users\The Beast\Desktop\NoOUT

@viscid moss

[viscid moss]

I think the problem is the audio path, it's outside the UVR5 UI folder. Try using the inputs and outputs folders inside the UVR5 UI folder.

Other folders may require administrator permissions to write files to them

[latent kettle]

Also how can I select stems. Like I don't want noise output or reverb only output ?

[burnt zealot]

when i try running rvc

[viscid moss]

I tried using an audio on the desktop like you did and it worked for me, I recommend you try placing the audios in the inputs folder of UVR5 UI and placing the outputs folder in the output path, it is not necessary to put the full path you can use: ./input and ./outputs. You can also rename the audios to a shorter name.

If it still doesn't work it could be because your Windows user contains a "The Beast" space but I'm not entirely sure.

[knotty moth]

literally invalid address, it should open localhost address, not like that
also for greater good, stop using opera gx

[latent kettle]

i changed it to Admin but idk why it still show The Beast

[viscid moss]

On Output Single Stem on Advanced Settings u can write the stem u want, but i haven't tested with all models. May not work with some of them (the new ones)

[latent kettle]

its dont work on denoise model. sometime

[viscid moss]

Which denoise?

[viscid moss]

Well... u can't change that username 😅

Have u tried the another workaround?

[burnt zealot]

LOL

[latent kettle]

emmm Why ? why i cant im admin of this PC 😅

[burnt zealot]

whats so bad about opera gx

[burnt zealot]

^

[latent kettle]

Mel-Roformer-Denoise-Aufr33

[burnt zealot]

whats the best denoise now?

[latent kettle]

idk.. just using randomly

[burnt zealot]

who else is getting the 5090

[distant turtle]

-colab

[azure marsh]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• FaceFusion UI, by Nick088 Google Colab
• FaceFusion NO UI, by Nick088 Google Colab
• EasyGUI, by Rejekts Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

More commands: -audio, -hf, -local, -kaggle, -realtime, -rvc, /help

[viscid moss]

I mean, yes, you can change it, but it's a bit complicated, and it could cause problems with Windows. So I really don't recommend changing it.

[burnt zealot]

does this actually approve speed still if i am doing a 20-30 min dataset on a 4090

[simple ore]

not much

but with 24GB, why not?

[burnt zealot]

yea true ig

cant wait until 5090

5090 been edging me on its release

[viscid moss]

stems for that model are "dry" as primary stem and "other" as secondary stem

[latent kettle]

i saw a tutorial on yt he changed his name but it my case it shows admin on lock screen but when i open CMD it show old name "The Beast"

[latent kettle]

what ??

[viscid moss]

I mean if u want the No Noise Stem of that model u should put "dry" on the single stem output thing

[latent kettle]

okay. and is it a good model which im using for de noise

[viscid moss]

Lemme check the bible

ye seems like

[latent kettle]

i sent something to you in DM pls check it only for one time..

[young oasis]

everytime i try and use the Ilaria RVC and click convert, it keeps saying Error and GPU task aborted

[proven hill]

its a glitch with huggingface 😩

[young oasis]

oh

rip

[proven hill]

try with a shorter audio

[young oasis]

my audio is 17 seconds

nvm it works now

[toxic ermine]

should i change any settings while man to woman vocal?

[proven hill]

only the pitch

[toxic ermine]

okay zero to what

[proven hill]

12

[toxic ermine]

is -12 enough?

[proven hill]

-12 is female to man

[toxic ermine]

oh positive, okay

thank you

[proven hill]

no probs

[elder pagoda]

My RVC keeps crashing, and I have it set to these settings. Any way I can fix it?

Okay, nvm I fixed the issue, but now I can't hear my voice

[proven hill]

this is ANCIENT

use the new version

-rt

[azure marsh]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[proven hill]

first link

[elder pagoda]

Oh, alright.

[prime panther]

i have problem w voicechanger, i hear my voice but i dont hear it on discord

[proven hill]

#🔍│help-w-okada

[crude mango]

i havent made an ai model since google colab banned gradio is there a good gui i can use

[proven hill]

what gpu do you have?

[crude mango]

nvidia gtx 1650

[proven hill]

okay then no local options

use rvc disconnected

tho it doesnt have a ui

[crude mango]

https://colab.research.google.com/drive/1XIPCP9ken63S7M6b5ui1b36Cs17sP-NS this one?

[proven hill]

yup

[crude mango]

alr
thx

[lone nest]

Preprocess completed in 0.01 seconds on 00:00:00 seconds of audio.```Why am I getting this on Applio Colab

The dataset is a .flac

[proven hill]

uh strange

[median elbow]

I'm looking to buy Colab Pro, but I'm wondering if I can use RVC without running out of resources for a month, has anyone tried it?

[lone nest]

Btw have people considered making cloud port on Github Codespaces? Wonder how it compares to Colab and Kaggle

[tame mica]

i dont think rvc supports flac

eksdee

if it does just change it just incase

[simple ore]

did not provide anything for applio to slice

so it did not do anything

[lone nest]

Why is that

I don't need to change anything on the copied path right?

Did I mess up

[ionic fossil]

-realtime

[azure marsh]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[hallow thistle]

I bought Google Colab Pro for two months in 2023. There will be three GPUs available for Pro: NVIDIA A100 and L4, and now newly added v5e-1 TPU. However, these three GPUs eat more compute units than CPU, T4 and v2-8 TPU, even with no code going on when a Colab notebook is connected. The longer you use one of these GPUs for anything, the more compute units will be used for that.

[pure quiver]

Hey everyone i'm using realtime voice changer for the first time. I can hear my own voice but without any voice changes just my regular voice. How do I fix this

[hallow thistle]

And supposedly, your compute unit count will be shown here on how much of it left before running out.

[hallow thistle]

For W-Okada the voice changer, go to #🔍│help-w-okada. This #✨│ai-help is about RVC.

[pure quiver]

oh so the other one is for the client right?

[hallow thistle]

RVC is the audio conversion program. W-Okada is the realtime voice changer program. The only W-Okada related to RVC is W-Okada uses RVC voice model to interence, not the code and GUI themselves huh.

[pure quiver]

ok thanks!

[hallow thistle]

You're welcome.

[brisk orchid]

how do i make it so i can hear myself?

i put my input as my microphone and output as the vb cable audio input

[hallow thistle]

For W-Okada the voice changer, go to #🔍│help-w-okada. This #✨│ai-help is about RVC.

[cold cave]

Does anyone had this problem?

[dark hawk]

does anybody have the link to download the voice changer ?

[hallow thistle]

-realtime

[azure marsh]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[hallow thistle]

First link.

[brittle wing]

where is the hank hill download

?

[hallow thistle]

#1175430844685484042

Light blue and light purple named users be asking in wrong channel lol.

[brittle wing]

i dont see it

[hallow thistle]

If the voice model doesn't exist, you can request someone to do that at #1159289738314919936.

[knotty moth]

there are cheaper renting alternatives like runpod, vast.ai, etc

[upbeat ridge]

guys windows defender is saying no for rvc gui what do i do

[hallow thistle]

-gui

[azure marsh]

https://cdn.discordapp.com/attachments/1122285248844144733/1203460490475343953/caption.gif?ex=65d12cec&is=65beb7ec&hm=bd2fb8d010006dd7c6e3c1c67d3ae846fd1478e1a3124c544c31b43086fe54aa&

[hallow thistle]

RVC-GUI in 2025? Wild.

[carmine siren]

-collab

-colab

[azure marsh]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• FaceFusion UI, by Nick088 Google Colab
• FaceFusion NO UI, by Nick088 Google Colab
• EasyGUI, by Rejekts Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

More commands: -audio, -hf, -local, -kaggle, -realtime, -rvc, /help

[simple mango]

Does anyone know what happened to the Ilaria huggingface

[proven hill]

?

[simple mango]

Why is it no longer on the google collab?

[proven hill]

use it on huggingface

[simple mango]

is it used online again?

[proven hill]

what?

[simple mango]

sorry, my apologies I think I understand now

[low shard]

The google colab is broken

The huggingface space is better

[simple mango]

I managed to find it thank you!

[silent stratus]

delete it and get applio😭

[carmine siren]

-colab

[azure marsh]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• FaceFusion UI, by Nick088 Google Colab
• FaceFusion NO UI, by Nick088 Google Colab
• EasyGUI, by Rejekts Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

More commands: -audio, -hf, -local, -kaggle, -realtime, -rvc, /help

[tame mica]

https://cdn.discordapp.com/emojis/1311429177840369664.webp?size=48&name=hur

applio kaggle keeps on stopping training for some reason

[proven hill]

kaggle moment

[tame mica]

https://cdn.discordapp.com/attachments/497928974832369664/1019017514556067840/MemeFeedBot-1.gif

si

[proven hill]

si

[simple ore]

forgot to slice audios?

[tame mica]

?

i barely use applio so idk whats going on

[simple ore]

show preprocess and exrract feature step logs

[tame mica]

this?

[simple ore]

how big is the batch size?

[tame mica]

12

[simple ore]

that's way too much for 12 min audio

try with 4

[proven hill]

isnt 8 best?

[simple ore]

2 gpus, so 4x2 = 8

[proven hill]

kaggle is double gpu?

[simple ore]

that's the indication

[low shard]

There's T4x2 and P100

[proven hill]

gotcha

[tame mica]

ah i see

[crude mango]

i'm using rvc disconnected the gpu stopped at 435 out of 500 epochs is there anyway i can get my model without retraining it

also is 500 epochs nescecarry for a 14 minute long dataset if i do need to retrain it

cuz i didnt rlly fuck with the settings

[proven hill]

did you enable the save option?

[crude mango]

I have no idea but i dont think so

I didnt think i'd need to

[proven hill]

then you lost everything

[crude mango]

Well shit

[crude mango]

Question still stands

[proven hill]

i mean, yea but you should also look at the tensorboard

[crude mango]

I have no idea how the tensorboard works

[proven hill]

-tensorboard

[pastel oak]

Rvc disconnected in 2025 smh

[proven hill]

i thought there was a command

[pastel oak]

https://docs.ai-hub.wtf

Somewhere on there is tensorboard

[real sail]

how do i reduce delay bruh

[proven hill]

ciao

[low shard]

https://docs.ai-hub.wtf/rvc/resources/training/#epochs--overtraining

[real sail]

ciao

[proven hill]

che gpu hai?

[low shard]

wokada? usa #🔍│help-w-okada

[real sail]

rtx 3060 ti

[low shard]

manda il link del tut che hai usato

prob hai usato una versione vecchia

[proven hill]

andiamo sull’altro canale

[low shard]

buona

[low shard]

google colab gives max 4 hours, NOT GRANTED, of GPU

Kaggle gives 30 hours weekly granted

Kaggle is better but harder

[low shard]

you have to read the guide, there isn't a right amount of epochs

[crude mango]

harder as in ?

[low shard]

needs a phone number, less user friendly

[opal cobalt]

hello someone now a real time voice changer for amd gpu pls ?

[low shard]

RVC is for inference on pre-recorded audios and training

Wokada is for RVC inference in realtime for calls, and there's the deiteris fork for better performance, tell me your GPU name in #🔍│help-w-okada , this is the wrong channel

[opal cobalt]

oups sorry

[low shard]

it's fine

[peak falcon]

so I'm used with Applio

[proven hill]

what are you using now

[peak falcon]

I think it's a new version

I haven't used it in a while

[peak falcon]

Nothing

I was just experimenting like 1 year ago

[proven hill]

whats your gpu?

[peak falcon]

it was named EasyGUI before?

[proven hill]

yea a long time ago there was the easygui

[peak falcon]

I don't have one, I have an old pc rn, I'm running it on the incorporated GPU

[proven hill]

then you cant use it locally

-cloud

mh

-colab

[azure marsh]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• FaceFusion UI, by Nick088 Google Colab
• FaceFusion NO UI, by Nick088 Google Colab
• EasyGUI, by Rejekts Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

More commands: -audio, -hf, -local, -kaggle, -realtime, -rvc, /help

[proven hill]

first one

[peak falcon]

I don't wanna train models or create

[proven hill]

what you wanna do?

[peak falcon]

AI voice covers, I want to make a producer tag

[proven hill]

you can use applio on colab, ilaria rvc on huggingface or weights.gg

[low shard]

it was an old colab, it's still up a newer version btw

-colab

[azure marsh]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• FaceFusion UI, by Nick088 Google Colab
• FaceFusion NO UI, by Nick088 Google Colab
• EasyGUI, by Rejekts Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

More commands: -audio, -hf, -local, -kaggle, -realtime, -rvc, /help

[low shard]

it's the last one

however it's not much updated

[peak falcon]

Thank you capo

She's your gf? 😅

[proven hill]

no😭😭😭

[peak falcon]

I thought that bcs you're both Italian

[proven hill]

nono we are just friends and hes a lot younger than me

[peak falcon]

How old is he? 😂

[proven hill]

i wont tell unless he does

[peak falcon]

understandable

can you make a producer tag for me?

[proven hill]

its easy to make it yourself, sadly i just turned my pc off

[peak falcon]

I know it's easy, but I wanted something real and a woman's voice

isn't this one for making and training models?

[proven hill]

seems so

use weights.gg

[peak falcon]

Fml

[peak falcon]

never used it, and I don't have the patience to learn how to use it

[proven hill]

🤷‍♀️

[peak falcon]

Whatever

Thanks for trying to help

[proven hill]

no prob

[low shard]

no

[low shard]

16 yo bro

[peak falcon]

Damn bro

You should call me sir😭

[low shard]

it's the easiest between all of them

[peak falcon]

Really?

[low shard]

yeah, I would suggest it

[peak falcon]

Okay

[low shard]

how old are u

[peak falcon]

almost 23

[low shard]

damn

[peak falcon]

bruh😭😭

[bold yarrow]

folks

sorry if this comes off as insensitive

but is there any way to make the ai's pronunciation of S sounds less noticeably artificial

[peak falcon]

wym

[crude flame]

robotic or harsh?

[bold yarrow]

uhhhhhhh

[peak falcon]

if it's robotic try a different model, if it's too harsh you can fix that while mixing the vocals

[crude flame]

robotic = to small of a dataset so rvc started overfitting the sibilants

harsh = all the sibilances in the dataset were very harsh so rvc only knows those

[bold yarrow]

its not my model tho

would it be better if i just pasted it in here

[peak falcon]

yes

[crude flame]

sounds fine imo (i havent played tf2 so idk what he normally sounds like)but you can try de-essing the output

[bold yarrow]

de-essing?

[crude flame]

T de-ess or RX de-ess

They just makes the esses not as harsh

[bold yarrow]

can you pass me a link to them

im an amateur

please

[crude flame]

https://techivation.com/t-de-esser/?srsltid=AfmBOopc5H9unNMz9ZhLZCfqs1CNoaknxdU-VRX99YG8_jUDaI0F2QTX here is T de-ess its free

[peak falcon]

Are those free pulgins?