✨│ai-help

[rare goblet]

Ayo? @prime harness level 2 !!!

[half cove]

does it say something like allow program?

[tribal sorrel]

hold on

[pure scroll]

check my 'About Me'

[spiral osprey]

You know when you cover a song and your voice sounds a bit robotic? How do I remove this robot? Is it possible to do this with Audacity? or something like that?

[pure scroll]

[prime harness]

what you mean friend

idk anything about the FLIE

file

and zip

[west mango]

does anyone know of a good helpful tutorial on how to make ai model

[rare goblet]

Ayo? @west mango level 1 !!!

[pure scroll]

[prime harness]

you mean you are selling voie model

[pure scroll]

that's just the voice model's problem not yours dude

[prime harness]

oh

[pure scroll]

[half cove]

it could be the audio

[prime harness]

well you know what

[pure scroll]

[tribal sorrel]

it is stuck on that

[prime harness]

i dont have a bank

haha

[pure scroll]

[prime harness]

I only have a bank card in my country but no foreign bank card

ehe

[marsh jay]

@tribal sorrel go to the help-w-okada

and look a pins

[fast swan]

You remind me of my country

[half cove]

try reset it and if it windows gives anything to click then press accept but i dont really know vc you could ask in #🔍│help-w-okada

[prime harness]

[fast swan]

Hehe my country stated a declaration past year that anyone that is caught dealing in the black market to actually get their money's worth of any foreign currency

[prime harness]

i wish i can make a yuumi voice

[fast swan]

10 years in jail, 5 million pound bail and all possessions taken

[prime harness]

aghhh making voice so hard but for IT is ez

[pure scroll]

which one you want?

like which language?

[prime harness]

english

i like english

i still learning english

[pure scroll]

pay me $30 first

[prime harness]

haha

friend i dont have foreign bank card

[fast swan]

الاسعار ولعت

[pure scroll]

ayo i'm kiddin

[spiral osprey]

of my voice model because I already redid one and it was the same.

Because there are some people who can remove this robotic voice?

[marsh jay]

my voice changer is still not going though

i dont know how to fix it

[pure scroll]

possibilities of robotic voice:

1. not enough dataset
2. trained for very less time

[half cove]

could be overtraining

or a bad dataset

[pure scroll]

yeah thats no. 3

[pure scroll]

that's no 4

[prime harness]

hey friedn, put this thing to this

teach me how to put it in there

[pure scroll]

send you the model in dm's

[velvet robin]

I have a question for a Model Maker. Can you use FNF Chromatic Scales and/or soundfonts for AI Voice Models?

[proper shale]

If they're in audio formats, maybe

[velvet robin]

Ah, very good. Phase 1 is complete. Would you like to make a Voice Model for me?

[shy monolith]

does anyone know how to install onnxruntime on python 3.12

[small umbra]

..how to install dez nuts in yo mouf?*

[shy monolith]

shut up bitch

[spiral osprey]

I trained a 3-minute audio using 1157 epouchs, what do you mean it was a short time?

[proper shale]

Ehhh I can't, maybe #1159289738314919936

[proper shale]

...3 minutes isn't that long

Most models have 10+ mins of dataset

It's recommended for most models

[velvet robin]

Well, That's what I was going to do. Request a model for you. I haven't got it ready yet. I'll let you know when it's ready.

Or do you just don't want to make a voice model I request?

[proper shale]

I don't think I can right now... I hope you understand

[velvet robin]

Well, Who should I ask?

[grand surge]

help

[spiral osprey]

understand,thanks sorry for ping.

[velvet robin]

Would a Model Maker like to make a voice model for me? If so, I'll put it on the requests.

[spiral osprey]

Is there any way I can make a model with 3 songs for example, since it's 10 minutes long?

[proper shale]

...yeah? you can use multiple songs for datasets

[velvet robin]

If only I could put files on my requests. It'll make things a whole lot easier.

[spiral osprey]

oh,thanks.

[spiral osprey]

So if I make a zip with 3 3-minute songs with just vocals, can RVC render the 3 songs to make a dataset/voice model?

[rare goblet]

Ayo? @spiral osprey level 6 !!!

[spiral osprey]

my epouchs can I continue leaving 1157?

[velvet robin]

How do I ask a model maker for a request? I know how to put my request in https://discord.com/channels/1159260121998827560/1159289738314919936 but I want to know how to ask a model maker for it.

[subtle breach]

I'm trying to train a model locally using 1660 super with 4 gb of VRAM and 8 gb of shared memory, but I always get "cuda out of memory" when model training starts. All my VRAM is used but shared memory isn't used at all. Guys in other chat said me that shared memory can be used by rvc, but for some reason it doesn't work for me. Can I get any help, please?

[velvet robin]

I'm being ignored...

[grand surge]

we're all being ignored

[velvet robin]

pain

[grand surge]

stop complaining and just ask later

[sand iris]

Shared memory can't be used

Except on a few driver versions for some reason

And it's stupid slow

You wouldn't get bearable training speeds lol

[subtle breach]

Ok thanks, is there any way to solve the issue without buying another gpu? Console says pytorch reserves only ~ 2.8 gb of total 4 gb

[sand iris]

Unfortunately no, training will not be possible on 4GB cards 😅

[subtle breach]

How many GBs are enough then?

[fast swan]

gotta pay up champ

[crude flame]

so im using the Applio linux installer and i keep getting this error:
File "/home/razer/Downloads/Applio-RVC-Fork/infer-web.py", line 13, in <module> from assets.configs.config import Config File "/home/razer/Downloads/Applio-RVC-Fork/assets/configs/config.py", line 10, in <module> import torch ModuleNotFoundError: No module named 'torch'

Then when i try to update Applio and dependencies i get this error:
Traceback (most recent call last): File "<string>", line 2, in <module> File "<pip-setuptools-caller>", line 34, in <module> File "/tmp/pip-install-y17m35i0/llvmlite_3a4621a27c3d4b3c8a12f8c970ad888f/setup.py", line 55, in <module> _guard_py_ver() File "/tmp/pip-install-y17m35i0/llvmlite_3a4621a27c3d4b3c8a12f8c970ad888f/setup.py", line 52, in _guard_py_ver raise RuntimeError(msg.format(cur_py, min_py, max_py)) RuntimeError: Cannot install on Python version 3.12.1; only versions >=3.7,<3.11 are supported.
Does getting these errors mean i have to downgrade the version of python i have installed? Or is it something else

[rare goblet]

Ayo? @crude flame level 2 !!!

[sand iris]

6

[subtle breach]

Thanks

[fast swan]

I like how my 4050 is barely enough as 6gb vram

[sand iris]

Oh boy

Not sure the linux installer is tested

which distro is that on ?

[crude flame]

Fedora

Nobara to be specific

[sand iris]

Gotta have python 3.10/3.9 or 3.8 for this

I'd recommend setting up manually honestly

It's not that bad to set up, and you'll avoid the script hell

[proper shale]

yeah you can make a voice model with em

[grand surge]

I followed everything on the guide. idk what i'm missing

[crude flame]

how does one do that? Is it just git cloning the repo and running the install script orr

[sand iris]

• Get a supported python version
• Setup a virtual python env to separate the dependencies from the system's (needed as of not too long ago)
• Clone code
• Install python dependencies
• Run

A little bit of tweaking will be needed for AMD GPUs, but nvidia cards should work fine

Feel free to ask if you need the details, can't be bothered to write them right now, but I can lay out the steps if needed

[placid holly]

NEED HELP!!!!!!!!

so i just started do training again (RMVPE) Locally (V2 48k)

1. When i train index (RMVPE) it said lack of improvment inertia and then my index only ended 30mb (i tried lates RVC1006Nvidia locally)

(back then for 30-60 minute data i get 400++mb)

2. i tried back to my previous version Mangio-RVC-v23.7.0 (not deleted yet). it say lack of improvment inertia too

i dont know what happened, that never happen before

look like index train fail somehow that idk why and how

[sand iris]

I recall hearing something along those lines too

when it decides the dataset is too long, it tries to trim it by doing that

there's an abritrary threshold of time where it just murders the index file

[placid holly]

i see, but this is the first time it happen to me for same long data

i think something happen, i will try trim half of it then

[tired scroll]

Hi, i have RTX 2060 what should i set "batch size for gpu" for fastest train.

[placid holly]

which version of memory u have? 6gb?

[tired scroll]

6gb yes

[placid holly]

8 i guess, since my 12gb use 16

[tired scroll]

thanks

[spiral osprey]

oh,thanks

[placid holly]

thank, i think my dataset just a little bit longer several minute than last one i did, and it pass the limit

now it i have no problem anymore

[crude flame]

Now im getting this error:
Traceback (most recent call last): File "/home/razer/Applio-RVC-Fork/infer-web.py", line 3, in <module> from tensorboard import program ModuleNotFoundError: No module named 'tensorboard'

[placid holly]

need to install tensor board from pip

i think tensorboard not included, so need install manually

[crude flame]

Back to this error:
Traceback (most recent call last): File "/home/razer/Applio-RVC-Fork/infer-web.py", line 13, in <module> from assets.configs.config import Config File "/home/razer/Applio-RVC-Fork/assets/configs/config.py", line 10, in <module> import torch ModuleNotFoundError: No module named 'torch'
Im assuming i just need to install torch

[placid holly]

i guess 🥲

[rare goblet]

Ayo? @placid holly level 6 !!!

[hybrid spire]

So I got the msg saying they're not gonna give me a Model Master status.

Is it possible I could ask what could be the reason? Like if it had something to do w/ the model quality?

[rare goblet]

Ayo? @hybrid spire level 6 !!!

[sand iris]

are you still on python 3.12 ?

If you are it will not work at all

[crude flame]

nope

[sand iris]

Ah good

refer to https://pytorch.org/ for torch install options

[grand surge]

i got this error too. idk what exactly to download or what to do with it.

[crude flame]

so i got this error:
No module named 'fairseq'
so i installed it then i got this error:
error: command '/usr/bin/gcc' failed with exit code 1 [end of output] ERROR: Failed building wheel for fairseq Failed to build fairseq ERROR: Could not build wheels for fairseq, which is required to install pyproject.toml-based projects

[sand iris]

Oh you also need the python header package

Not sure about the package names on fedora but it's python-dev or something close to that

[grand surge]

Aight, let's try this again

I'm trying to run applio according to this guide:https://docs.aihub.wtf/vocal-cloning-guide/applio.
I've done everything the guide says up until the launch section.
when trying to launch the program i'm met with this error:

where do I go from here?

[crude flame]

it says its already installed

[sand iris]

Huh, does it give you any useful logs before what you are getting?

I think you're supposed to get compiler output above pip errors

[vapid gust]

i dont think its supposed to take this long

[proper shale]

if you're using Huggingface this is a possibility

[vapid gust]

it is

ilaria

[crude flame]

`AMDGPU 5.7.2 repository ???% | 2.1 KiB/s | 548.0 B | 00m00s

Status code: 404 for https://repo.radeon.com/amdgpu/5.7.2/rhel//main/x86_64/repodata/repomd.xml - https://repo.radeon.com/amdgpu/5.7.2/rhel//main/x86_64/repodata/repomd.xm

Status code: 404 for https://repo.radeon.com/amdgpu/5.7.2/rhel//main/x86_64/repodata/repomd.xml - https://repo.radeon.com/amdgpu/5.7.2/rhel//main/x86_64/repodata/repomd.xm
Status code: 404 for https://repo.radeon.com/amdgpu/5.7.2/rhel//main/x86_64/repodata/repomd.xml - https://repo.radeon.com/amdgpu/5.7.2/rhel//main/x86_64/repodata/repomd.xm
Status code: 404 for https://repo.radeon.com/amdgpu/5.7.2/rhel//main/x86_64/repodata/repomd.xml - https://repo.radeon.com/amdgpu/5.7.2/rhel//main/x86_64/repodata/repomd.xm
Librepo error: Cannot download repomd.xml: Cannot download repodata/repomd.xml: All mirrors were tried
Repositories loaded.
Package "python3-devel-3.12.1-2.fc39.x86_64" is already installed.`

[sand iris]

I meant the log when you encounter issues building fairseq

[crude flame]

ah

sorry am idot

[sand iris]

Nah no worries !

[crude flame]

` gcc -Wno-unused-result -Wsign-compare -DDYNAMIC_ANNOTATIONS_ENABLED=1 -DNDEBUG -fexceptions -fcf-protection -fexceptions -fcf-protection -fexceptions -fcf-protection -fPIC -I/home/razer/myenv/include -I/usr/include/python3.10 -c fairseq/clib/libbleu/libbleu.cpp -o build/temp.linux-x86_64-cpython-310/fairseq/clib/libbleu/libbleu.o -std=c++11 -O3
gcc -Wno-unused-result -Wsign-compare -DDYNAMIC_ANNOTATIONS_ENABLED=1 -DNDEBUG -fexceptions -fcf-protection -fexceptions -fcf-protection -fexceptions -fcf-protection -fPIC -I/home/razer/myenv/include -I/usr/include/python3.10 -c fairseq/clib/libbleu/module.cpp -o build/temp.linux-x86_64-cpython-310/fairseq/clib/libbleu/module.o -std=c++11 -O3
fairseq/clib/libbleu/module.cpp:9:10: fatal error: Python.h: No such file or directory
9 | #include <Python.h>
| ^~~~~~~~~~
compilation terminated.
error: command '/usr/bin/gcc' failed with exit code 1
[end of output]

note: This error originates from a subprocess, and is likely not a problem with pip.
ERROR: Failed building wheel for fairseq
Failed to build fairseq
ERROR: Could not build wheels for fairseq, which is required to install pyproject.toml-based projects`
these?

[rare goblet]

Ayo? @crude flame level 3 !!!

[sand iris]

Might be me doing a dumb

try with the 3.10 devel package

I swear they used to be interoperable

[crude flame]

didnt work

[sand iris]

[vapid gust]

ilaria doesnt work for me

inference

[molten pecan]

Have you updated it?

[sand iris]

I'm all out of ideas if it can't find the python header files despite them being installed

[vapid gust]

when was it updated?

[molten pecan]

Yesterday

[vapid gust]

oh

how do i update it

[molten pecan]

You gotta restart the space

[sand iris]

python3-devel has them

[molten pecan]

https://rentry.org/ilarvc_inf_guide

I explained it at the end of the tutorial

[vapid gust]

got it

yea i restarted it

[gloomy smelt]

When making a model in applio i'm stuck in Reducer buckets have been rebuilt in this iteration.

Is this normal?

[sand iris]

wait a bit

It just started training

[gloomy smelt]

ohh

[sand iris]

takes a while

[gloomy smelt]

Alright, i see my pc lagging now

It's working then

[sand iris]

You'll get updates every few minutes

Or longer if you fed it unholy quantities of audio on subpart hardware

[gloomy smelt]

Still weird this may be the first time this appeared and i already made like 7 models

[sand iris]

Shows up all the time for me

[gloomy smelt]

hmm

What if it's slow when making epochs?

It's never slow for me, that would be weird

[sand iris]

Do I smell the unexplainable slow training bug?

[gloomy smelt]

,.... maybe 😭

Anyways this goin to take an hour 🥹

Like

uhhh

Oh hey

5 hours

I put 300 epochs and each epoch is taking one minute 😭

[vapid gust]

@sand iris not work

[sand iris]

idk

I haven't used hugging face in a long while

[violet heron]

Shortest HuggingFace inference:

[vapid gust]

no other option

@proven hill nice

[proven hill]

why the tag

its HF what do you expect

[crude flame]

Well i thank you for trying

[sand iris]

Look into that darn "header not found" error, it's having trouble with it for some odd reason

[vapid gust]

thought it was "fast"

[proven hill]

nothing is fast on HF

nothing i can do chief

[vapid gust]

swr u were advertising ur space as fast 😂

[proven hill]

on huffinface it is comparing to others, if you cant read its not my fault man

next time try reading better

[vapid gust]

my bad i forgot to read the fine print before using it lmaooo

[vapid gust]

ai hubs attitude after someone leaves a comment 😡😡😡

[proven hill]

man what im supposed to say

[crude flame]

I did something and now im getting an import error:
File "/home/razer/Applio-RVC-Fork/infer-web.py", line 60, in <module> from lib.modules.vc.modules import VC File "/home/razer/Applio-RVC-Fork/lib/modules/vc/modules.py", line 12, in <module> from lib.modules.infer.infer_pack.models import ( File "/home/razer/Applio-RVC-Fork/lib/modules/infer/infer_pack/models.py", line 12, in <module> from torch.nn.utils.parametrizations import spectral_norm, weight_norm ImportError: cannot import name 'weight_norm' from 'torch.nn.utils.parametrizations' (/home/razer/myenv/lib64/python3.10/site-packages/torch/nn/utils/parametrizations.py)

[sand iris]

Gotta check that one out for myself

That's unusual

How'd you install torch? (if you even did it manually to begin with)

[crude flame]

i asked chatty and it said to download this dnf install cmake make gcc-c++ git so i did and it worked

oh

am i idot again

am i answering the right question

[sand iris]

no x)

But that was how you fixed the other issue right ?

[crude flame]

yea

[sand iris]

I'll be back in a bit, but have you done anything related to Pytorch ?

Applio had a few issues really recently, and it's possible that's what is happening

[crude flame]

the only thing somewhat related to that is that i installed this: torch_directml

[sand iris]

Alright.. Anything else related to pip ?

Also are you using an AMD or Nvidia GPU ?

[crude flame]

AMD

i did a lot of pip installs

[sand iris]

which GPU model exactly?

[crude flame]

6700xt

[fiery bramble]

youre not the only person to have it

i also had it yesterday

my pc just wouldnt process the audio so i had to switch to my laptop

and it works there but

i have to do the extra google colab shit

:superangryemoji:

[rare goblet]

Ayo? @fiery bramble level 7 !!!

[sand iris]

Give it a

pip3 install -U torch torchaudio --index-url https://download.pytorch.org/whl/rocm5.6

just to really make sure it's using the proper version for AMD

[crude flame]

it worked, but now im getting this: OSError: libd3d12.so: cannot open shared object file: No such file or directory

[sand iris]

That's from torch_directml

Uninstall it, it's only relevant on windows/windows subsystem for linux

[hazy sedge]

Now that I'm training an rvc v2 model, I have a dataset of 3 hours and 50 minutes (2170 files of approximately 10 seconds), considering that I'm using free tier of Google Colab to train it (tesla t4), how much should I set the batch_size? Approximately, how many epochs should I train to make it sound its best? How do I know when I should stop training?

[proper shale]

i really hate to be that type of person but...

3 hours is way too excessive

anyways, setting your batch_size to 16 would probably be alright

[crude flame]

ok did that. Just to make sure, is this the right way to manually run applio: python infer-web.py --pycmd python --port 5000 --dml

[proper shale]

and go for 500 epochs

train 50 or 25 epochs per session if that's a possibility

[sand iris]

--dml is only for windows

Oh yeah that's the confusing part, directml is only a band aid for windows amd users

On linux you get an equivalent to nvidia's cuda

Which is much better and works the exact same

OH WAIT I FORGOT ONE THING

try with HSA_OVERRIDE_GFX_VERSION=10.3.0 python infer-web.py

Looking into it

[crude flame]

weird my message got deleted

heres the error just in case you still need it
2024-01-03 15:46:31 | WARNING | lib.tools.loadThemes | Error Loading: Base.set() got an unexpected keyword argument 'chatbot_code_background_color' Traceback (most recent call last): File "/home/razer/Applio-RVC-Fork/infer-web.py", line 3113, in <module> app = GradioSetup() File "/home/razer/Applio-RVC-Fork/infer-web.py", line 1910, in GradioSetup record_button = gr.Audio( File "/home/razer/myenv/lib64/python3.10/site-packages/gradio/component_meta.py", line 157, in wrapper return fn(self, **kwargs) TypeError: Audio.__init__() got an unexpected keyword argument 'source'

[half cove]

did you edit it?

if you edit a message after a while a bot deletes it

[crude flame]

to add a line that i didnt copy

[rare goblet]

Ayo? @crude flame level 4 !!!

[crude flame]

ah

ok

[half cove]

like after a few min if you edit it dels it

[crude flame]

understood

[half cove]

if you edit before that it would del

[sand iris]

Took me a while, sorry about that

somehow you are using a more recent of gradio than what applio uses

[crude flame]

its alright

[sand iris]

give it a pip install gradio==3.34.0

hope nothing else also has the wrong version

[crude flame]

IT WORKED

Thank you so much for your time

now how do i open it properly after i close it

[sand iris]

what do you mean ?

[crude flame]

like once i close applio do i just use go-applio.sh to re open it or do i have to do the whole activate the env and do that command

[sand iris]

Oh, gotta activate unfortunately

but you can write a quick script for it

#!/bin/sh

source /path/to/env/bin/activate
HSA_OVERRIDE_GFX_VERSION=10.3.0 python infer-web.py

[viscid moss]

Applio recently updated their version of PyTorch, so recently trained models may not work (They have static instead of speech) in:

• Hugging Face spaces (like Ilaria RVC)
• W Okada Realtime Voice Changer
• Mangio RVC, Mainline RVC and their versions in Colab
  For this reason, if when making an inference you can only listen to static, you can convert it using this Colab:
  https://colab.research.google.com/github/Eddycrack864/PTH-Fixer-for-RVC/blob/main/PTH_Fixer_for_RVC.ipynb
  Btw Applio has added a converter in its new update, but if you don't use Applio you can use this Colab.

[viscid moss]

Engineering team are working on a better solution

[sand iris]

Update your pytorch y'all

[viscid moss]

But I need to update a lot

Well.... tbh I can do that but gimme time

[peak tusk]

pero si son 10 lineas

[viscid moss]

Can u provide the code?

I'm too lazy to check GitHub rn

[hazy sedge]

Yes, I can

[rare goblet]

Ayo? @hazy sedge level 1 !!!

[hazy sedge]

Maybe I could increase the batch size more to train faster? I currently use 16

[proper shale]

16 is nice

[eager pelican]

nuh uh 4 is better imo
same result with ess chance of messing up

[proper shale]

they're already using a 3 hour dataset I don't wanna make em wait so long

but again they can always retrain at 4

[eager pelican]

just cut the dataset shorter then.. higher quality as well i'd bet

[brittle wing]

Pick the best of the best, and cut it down massively

[proper shale]

again, ive already said that to em

¯_(ツ)_/¯

[formal zenith]

at what point should i stop training

should it be at the local minimum or should it be somewhere on the downward line

[proper shale]

are you on g/total

[formal zenith]

yes

[proper shale]

alright hm

[formal zenith]

wait

[proper shale]

i believe you should continue

[formal zenith]

that was grad norm d

[proper shale]

oh.

[formal zenith]

thought that was g/total

[proper shale]

yeah search "total" and check both d and g

[formal zenith]

this is g/total

suppose the full graph would help

is it overtrained or should i keep going

[proper shale]

yea keep going

seems to be trending downward

[formal zenith]

and when it's trending upward that's the point to stop right

[proper shale]

not necessarily

you gotta check d/total, see if it's going down

if it isn't, then that's a yellow flag

[formal zenith]

d/total is going downwards

[rare goblet]

Ayo? @formal zenith level 2 !!!

[proper shale]

then check g/mel, which is model clarity, if that's going down still your model may not be fully cooked

[proper shale]

yeah ur good

[formal zenith]

g/mel is good as well

[proper shale]

yeah ur fine for now

[formal zenith]

alright

thank you!

[proper shale]

you're welcome

feel free to give updates on the graphs

and btw there's also a guide for overtraining too! it helps you understand the graphs as well.

-overtrain

[azure marsh]

How do I know if my model is overtaining?

All-In-One Guide on how to make a good model

This guide explains how the D and G files works and much more: https://rentry.org/RVC_making-models

Credits: LUSBERT

Automated Overtraining Detection (AOD)

Will be available soon in #1159513888199540817

Credits: grvyscale

[brittle wing]

How do fix this and why does Astra labs give me horrible results even when it's the same models

[frosty night]

-rvc

[azure marsh]

Documentation

📚 All-In-One English documentation

https://docs.aihub.wtf/

❔ English FAQ

https://docs.aihub.wtf/faq

🎄 Credits: AI HUB FRANCE

✨ More guides

Full AI Voice Model Training Guide (Local)

Link: YouTube
Credits: Christopher Villanueva

Model training with Mainline RVC

Link: Rentry
credits: Raven (ravencutie21)

AICoverGen Colab Guide

Link: Google Docs
Credits: Eddy (Spanish Helper)

Create a model with RVC disconnected (colab)

Link: Google Docs
Credits: Angetyde

How To Make an AI Cover With Ilaria RVC

Link: Rentry
Credits: 👽 Julia (ailen2091)

[brittle wing]

why is voicemy.ai infinitely uploading the model?

[crude flame]

so uh why are these stuck in queue

[proven hill]

HIGH QUALITY FEMALE VOICE MODELS

Are you searching girl voices like e-girl, mommy, v-tuber and more?
You find them here: https://ko-fi.com/ilariaowo/shop

And for paid items, if you DM me, you will get a 20% discount if you pay with

I also accept Steam gift cards as last resort

Grab yours today!!!

[half cove]

[lunar iron]

Hi! I'm using Original RVC and I'm a little confused about the Vocals/Accompaniment Seperation & Reverbation Removal tab. I want to remove the harmonies from the vocals, which would be HP5, though when I click on convert and get a success message, I don't get any new file. Where would these be stored or does it not work like that?

[narrow dove]

Hey There, can someone tell me if i need to change the model from pyt to onnx if i have an amd gpu

[thin stump]

How it works in UVR is that you're supposed to get two files if you select Ensemble > Instrument/vocals > no checkboxes selected. Idk if that's the same with the built in one

[cerulean dew]

im wondering if there is a problem with this dataset bc the end product was kinda staticky

[velvet crown]

hello

can someone help me

im doing rvc locally and when doing the inference it works but the audio is mumbling and with noise

[half cove]

copt

y

shit i cant type

copyright?

[cerulean dew]

how would copyright effect it?

[half cove]

you cant post copyright on ai hub

[cerulean dew]

fr? its an ai cover server lmao

[half cove]

we got banned last time for copyright

and warned

so no more allowe

d

[thin stump]

Yeah but only youtube videos are allowed on #1159290752195633273

[rare goblet]

Ayo? @thin stump level 7 !!!

[half cove]

did you denoise it?

[cerulean dew]

idk how honestly

i just tried to find an interview with clean audio

[half cove]

do you have a good gpu?

[cerulean dew]

i dont rlly know what is considered good

id say probably average

[half cove]

what gpu

thats easier

[cerulean dew]

i remember this question being asked of me before and i remember it being rlly hard to find and i think they said that since i have a laptop it isnt rlly strong at all and shouldnt even technically work on rvc gui but it still somehow does

[half cove]

go on task manager then performance and screenshot what it says

[half cove]

click on gpu1 and say what it says 😭

it says rtx so im assuming it would be good enough

[half cove]

the name 😭

[cerulean dew]

oh

[half cove]

ok good

download uvr

-uvr

[azure marsh]

Ultimate Vocal Remover

One of the best free and open source vocal and instrumental isolation tool.

🎄 Tip: Use /guides UVR to learn how to use it.

[cerulean dew]

alright thx

[half cove]

for denoise use VR Arc: UVR-DeNoise and DeNoise-Lite

or*

[marsh jay]

hey i dont know whats wrong but my voice wont go though on discord when i have the VB-audio for the input

[pure scroll]

change these discord settings

[spare remnant]

anyone know how to change the cpu to gpu in "real time voice changer client". my gpu dont show up

[rare goblet]

Ayo? @spare remnant level 1 !!!

[marsh jay]

still wont work

[pure scroll]

INPUT - Your Microphone
OUTPUT - Virtual Audio Cable
MONITOR - Headphones if you want to see if working

[marsh jay]

already got that

[violet heron]

Use RVMPE

Turn up out

Gain is too high

[pure scroll]

Make sure that the Nvidia Cuda Toolkit drivers are installed on your PC and up to date.

[spare remnant]

i dont have nividia

[violet heron]

That’s why

You downloaded the wrong version

[pure scroll]

amd gpu?

[spare remnant]

i have 6950 xt

[marsh jay]

still nothing

[violet heron]

AMD users have to install the DirectML version

[violet heron]

Did you do the other 2

And do the other models work?

[marsh jay]

nope

nothing works

[pure scroll]

@spare remnant select “rmvpe_onnx” or “crepe_tiny” in F0 Det.

and then check

gpu

[marsh jay]

discord is not detecting the input

[violet heron]

It’s already the wrong one

The DirectML one shows GPU buttons instead of a drop-down

[spare remnant]

dont work

[marsh jay]

@violet heron

[pure scroll]

you have installed wrong version

install amd one

[spare remnant]

which one

[marsh jay]

this one @spare remnant

[spare remnant]

ok ty

[violet heron]

https://huggingface.co/wok000/vcclient000/resolve/main/MMVCServerSIO_win_onnxdirectML-cuda_v.1.5.3.17b.zip

Direct download

[marsh jay]

hey any idea how to fix this

like is their something in the setting i need to mess with>

?

[spare remnant]

how do i get non anime voice?

[marsh jay]

https://www.youtube.com/watch?v=e8CaftH9olk

try that

[narrow dove]

Hey Guys so i got an onnx model since i got an amd gpu and if i upload it like this and try it out, it doesnt give me any Soundfeedback, am i uploading it wrong?

[spare remnant]

brha

[violet heron]

You opened the old one

Open the new one

[violet heron]

Check weights.gg

[spare remnant]

no but i think i dowloaded the worng one agien i dowloaded from your link now

[rare goblet]

Ayo? @spare remnant level 2 !!!

[spare remnant]

if i just del the old files is it gone then?

[proud elbow]

that's gpu-cuda for Nvidia GPUs, download the directML one for your AMD GPU

[lucid rune]

i have a quick quick question

uh

how do i,, use the ai model

like, not with a website, on my computer locally

[spare remnant]

this? https://huggingface.co/wok000/vcclient000/resolve/main/MMVCServerSIO_win_onnxdirectML-cuda_v.1.5.3.17b.zip

[proud elbow]

yeah, would better get the version 15 for less issues tho
https://huggingface.co/RavenCutie21/OkadaVersion15Backups

[spare remnant]

ok thx, do you know any good not anime models. just realistic?

like the voice thing

it only has anime

[violet heron]

Check weights.gg

[spare remnant]

what do i click?

[violet heron]

Go to task manager

If you dedicated GPU is under GPU 0 you do that one

If it’s under GPU 1 you do that one

[violet heron]

Do gpu0

[spare remnant]

i click it but nothing really happens

and it still use my cpu

oh nm

i just had to select a model mb

[wind saddle]

im having trouble unzipping the file

[rare goblet]

Ayo? @wind saddle level 1 !!!

[wind saddle]

Do i use winrar for it or what?

[narrow dove]

can someone tell me why im not getting any sounds?

[wind saddle]

Try turning up the threshold

[spare remnant]

it works when i try it now but when i use it in discord it dont

[narrow dove]

uh turned it up to 0.0006 but still no sound

[hot lagoon]

when makeing a ai cover is it good to covert the background vocals?

[violet heron]

Download the VAC

lemme get you the download

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

[spare remnant]

it works in discord now. but it dont pick up all of the sound and kinda cuts it off

[narrow dove]

I am still not getting any sounds :_/

[spare remnant]

what setting do you think is good i have no ide, i have a 6950 xt gpu

and amd ryzen 7 7800x3d 8core

[half cove]

https://rentry.co/VoiceChangerGuide should say on the settings part of this guide

[latent kettle]

Batch Size is the number of samples that are processed at once during training. For example, if you have a dataset of 100 samples and a batch size of 4, then the dataset will be divided into 25 batches, each containing 4 samples. During training, the network's parameters are updated after processing each batch of samples. This means that the network will be updated 25 times during one epoch (one pass through the entire dataset).
Due to the nature of GAN each model is unique so some people make multiple models of the same dataset and pick the best one.
VRAM also limits Batch Size, but it’s not a factor.
The smaller the batch size, more randomness (aka, noise) the network has, so, in other words, more randomization + AI guessing.
Sometimes you tend to go for smaller when your dataset is limited (not enough characteristics, data, phonemes, etc.) and hope for AI to figure it out well. Sometimes you have to decrease it to improve the model in general as, model if having too many examples (samples to see) learns it all too fast, and too fast learning = memorizing (overfitting) so, in simple words, it has an easy time recreate them 1:1 (for example during inferences) than manipulating them correctly, so they're more fit for inference.
NOTE: There is no such a thing as perfect batch for every case. Each voice is unique, the dataset is unique and one might work well with 4, some might like 19.
In general, you'd want to use those that are a power of 2, so: 4, 8, 16, 32, 64 etc
but it is not a must. One might try 5, 9, 11 etc.
Generally, you'd want to start from 4 or 8 and expand the range both wise and see which one does the best on graphs.

Can someone explain it in simple terms with relevant examples

[vapid gust]

no dataset will be able to have the same settings as any other dataset during training. each is unique to its voice so in that case it is better to train your model 3 times. First in batch size 5, batch size 9, and batch size 11 and so forth if you wish. Then it is up to you to test your model to figure out which one is better n then that’s the best batch size for your specific dataset

[proud elbow]

https://discord.com/channels/1159260121998827560/1186762171204505801
more tip:

Batch Size, how many clips it grabs, and will learn the similarities between the clips, if you have a bad dataset, ie too lazy to clean a lot out, higher works better, if you have a really good dataset, lower works better.

generally go either 8 or 16, depending on VRAM and dataset length, or somewhat 4 for short dataset as long it's pretty clean

[vapid gust]

if I have a dataset of studio recordings about 10-15 minutes length what batch size would be reccomendd

[hot lagoon]

in paperspace idk why but When I try to train it in the MANGIO-RVC one and I press train modle it dose not work

[hot lagoon]

also when I try to inference APPLIO in paperspace it dose not let me

[hot lagoon]

how to fix
torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 14.00 MiB (GPU 0; 7.79 GiB total capacity; 6.69 GiB already allocated; 8.50 MiB free; 7.54 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
echo "[Terminal 1]"

paperspace ^

[safe leaf]

why doesn't my audio show up in the easy gui? I have refreshed it several times

[raven dagger]

is it normal for it to crash after i put it in live?/

[hidden drum]

how do you remove sfx if you have 5.1 DD show of icarly

[thorny shoal]

SOrry its really early and I'm a dumbass but like I'm trying to get back into doing some AI cover stuff and just having fun and checking out Kalomaze's guide rn and I noticed "advanced settings" thing for algorithms but I couldnt figure out how to actually access that so like how does that work?

[pure scroll]

-guide

[azure marsh]

Suggestion for @thorny shoal

Documentation

📚 All-In-One English documentation

https://docs.aihub.wtf/

❔ English FAQ

https://docs.aihub.wtf/faq

🎄 Credits: AI HUB FRANCE

✨ More guides

Full AI Voice Model Training Guide (Local)

Link: YouTube
Credits: Christopher Villanueva

Model training with Mainline RVC

Link: Rentry
credits: Raven (ravencutie21)

AICoverGen Colab Guide

Link: Google Docs
Credits: Eddy (Spanish Helper)

Create a model with RVC disconnected (colab)

Link: Google Docs
Credits: Angetyde

How To Make an AI Cover With Ilaria RVC

Link: Rentry
Credits: 👽 Julia (ailen2091)

[pure scroll]

follow this guide for making AI Cover

https://docs.aihub.wtf/vocal-cloning-guide/hugging-face

[pure scroll]

if you want to learn more, refer this

[pure scroll]

or follow this

and welcome to AI buddy

[thorny shoal]

Thank you TvT

[old bough]

for best real time quality which one is better, rvc or w-okada?

[pure scroll]

if you have a very good graphics card use RVC
if not W-Okada

both are good

[old bough]

saved up for a good year to get an rtx 4090 :S
I tried w-okada earlier because that guy jareds journey made a vid saying its better

but in the guides section on the discord i read rvc is better so i was wondering :p

[pure scroll]

4090

just use RVC man

or use both

haha

[old bough]

xD

okay, ill try rvc!

thank you c:

[pure scroll]

anytime man

https://docs.aihub.wtf/voice-changer-realtime/rvc-voice-changer-realtime

here's the guide for RVC

[pure scroll]

this also has W-Okada's guide

use what you find best for you

[zealous crane]

Hi, I wanted to download applio but I got an error " Gpu not found". I already reinstalled the cuda drivers, but that didn't help. Pls help me

[pure scroll]

what's your gpu and what's 'chunk' size have you set?

[harsh gust]

RX 7700 XT, tried all chunk sizes and same issue arrises.

[pure scroll]

128 chunk +16384 extra

try these values

[pure scroll]

follow this vid

https://www.youtube.com/watch?v=x7aOq7gCY4s

[harsh gust]

It is still delaying

[rare goblet]

Ayo? @harsh gust level 1 !!!

[harsh gust]

But I'm able to get one initial word in before it gradually climbs back to tens of thousands of MS

[pure scroll]

112 chunk +16384 extra
these?

[zealous crane]

Thanks, I'll see

[harsh gust]

Same occurence

[pure scroll]

you are using an imported voice model right?

[harsh gust]

Yes

[pure scroll]

and this delay is happening on them only?

[harsh gust]

Yup

The default voices work like a charm

I have attempted an assortment of different imported models

[pure scroll]

this fix would work:

You will need to first, change the GPU DML selector to CPU mode, this is to prevent crashes, then press the Export to ONNX button right next to save settings

If it crashes, reboot your system, close out of anything that auto opens and try again. If it doesn't you will now have your model but in .onnx, which you will upload it just like you did for the .pth you started with.

This should fix the issue and its what you would do for any models you upload in the future

follow this

and i'm sure it will be fixed

[harsh gust]

Will try

It is all 0 now and the voice is not doing anything

[harsh gust]

It works now! Thank you so much man.

[pure scroll]

IT DOES?!!!

less goo man

[olive hill]

Any suggestion links for RVC COLLAB?

[stoic cloud]

-colab

[azure marsh]

Suggestions for @olive hill

☁️ Google Colabs

• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab

🤗 Hugginface Spaces

• Ilaria RVC, by thestingerx Hugginface Spaces
• RVC-HF, by r3gf Huggingface Spaces
• AICoverGen, by r3gf Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces
• Advanced RVC Inference, by neuclya Huggingface Spaces

🎄 More commands: -audio, -local, -overtrain, -realtime, -rvc, -help

[hot lagoon]

is it looking good?

[thorny shoal]

I need help with Ilaria RVC T-T
I keep getting an Error whenever I Try to convert anything other than the test_audio.mp3. Not sure what error cuz' idk where I can see it and logs doesnt seem to show anything useful either I think. I tried to google little bit around and tried stuff like reloading and restarting the space but hard to look for anything specific when all I see is just red error after like 5 minutes of converting

[rare goblet]

Ayo? @thorny shoal level 1 !!!

[thorny shoal]

oh caught it saying "connection errored out" this time as a popup

[proud elbow]

looks screwed by resuming training with batch size different from the previous training session

[hot lagoon]

so what you say is a good point?

[proper shale]

Probably didn't duplicate space

https://www.rentry.org/ilarvc_inf_guide

[trim tendon]

hi what license should i use here??

[proper shale]

keep empty, if you cant then just use mit lol

[trim tendon]

is it ok if i use "unknown"??

[proper shale]

ehhhh

sure?

[rapid forge]

oh its not there

here it is

@proper shale should i have placed it in the dataset file?

[rare goblet]

Ayo? @rapid forge level 2 !!!

[proper shale]

Yes... and also removed the space on it

Spaces in model names or in folders can cause shit to break

[pure scroll]

true!

[rapid forge]

alr

[rapid forge]

wow im dumb

i was watching videos for 30 minutes just to try and see what i did wrong

thank you

[proper shale]

Nah you're not! Takes some time to get used to the quirks of RVC...

[rapid forge]

Yeah I've been making AI covers but I wanted to try and make my own model

[proper shale]

Oh nice

Good luck on the model :)

[rapid forge]

thank you, once again, Have a wonderful, day/evening

[proper shale]

You too :)

[thorn bane]

@proper shale This is what i ment with cpu's and gpu,s do i need to select one?

[proper shale]

You do

But again the Task Manager has a tab where it tells you what number the GPU is

For most people it's GPU 0 though

[tardy egret]

Stupid question probably, but I'm having a hard time thinking up the right things to search.

Issue: Downloaded models aren't converting, I just hear a high-pitched whine.

Notes: Example models work fine, I'm on the most recent RVC version and I think pytorch is the right version.

[proper shale]

Probably related to the Applio thing in #📰│dev-updates

[tardy egret]

Confirmed apparently, older models are working. Thanks!

[thorny shoal]

Sorry for late answer but it was duplicated and I followed that guide when making the whole thing in the first place

[proper shale]

Np :)

[proper shale]

Do you have any settings?

Like, pitch extractor

[thorny shoal]

uhh yeah I mean mangio-crepe
changed hop from 120 to 128 and index from I think default was 0.66 to 0.7
but idk if that could even be a problem considering that it worked with the default audio test

[proper shale]

try rvmpe

[thorn bane]

@proper shale So i downloaded the interface but it says that it is not supported is there a other version for AMD?

[proper shale]

For model training, there isn't

You can try Colab like I've linked b4

[proven hill]

HIGH QUALITY FEMALE VOICE MODELS

Are you searching girl voices like e-girl, mommy, v-tuber and more?
You find them here: https://ko-fi.com/ilariaowo/shop

INCLUDING FREE ONES!!

And for paid items, if you DM me, you will get a 20% discount if you pay with

I also accept Steam gift cards as last resort

Grab yours today!!!

[thorn bane]

Someone that knows the methode?

[finite galleon]

Theres a js script you can run on the console that keeps the connection connecting

[thorn bane]

Where?

[finite galleon]

Pretty sure its on the pins on one of the channels

I forgot which tho

[finite galleon]

FOR PEOPLE WHO TRAIN ON COLAB AND HAVE RUNTIME DISCONNECTED ISSUE

to stop colab runtime disconnect, Ctrl+ Shift + i to open inspector view . Then goto console and paste:
function ClickConnect(){
console.log("Working");
document.querySelector("colab-toolbar-button#connect").click()
}
setInterval(ClickConnect,60000)

There

[blazing garnet]

Wait can I send any song as demo?

[blazing garnet]

Clicking that rn

[proper shale]

Just use a TTS, wouldn't recommend using songs

[blazing garnet]

K

[green prairie]

anybody using rvc webui locally?

where exactly do I need to put the models I download?

I go to \assets and then I find few folders

[rare goblet]

Ayo? @green prairie level 1 !!!

[green prairie]

and I have an .index and .pth file

[proper shale]

Add the .pth to the weights folder
Make a folder in logs with your model's name and add the index there

[green prairie]

weights

okay

thank you

[proper shale]

yw :)

[green prairie]

and once I download the one I generated

what is the default download directory?

[proper shale]

the models u generate should be in the same paths?

[green prairie]

what do you mean, same as what?

ah I think I found it

[proper shale]

same paths as the ones you placed,
pth in weights and index in a folder in logs

[thorn bane]

@proper shale I fixed the rcv interface but i only get the index files?

[proper shale]

did you do one click training

[thorn bane]

No

[proper shale]

Oh wait

You can't train stuff on AMD GPUs

[thorn bane]

It did run an it said the training whas done and then i got the index file

[proper shale]

See, the thing is, it doesn't train with AMD cards

Just doesn't have that support rn

[green prairie]

anybody can give me some assistance on how to make the voice sound less like robotic?

😄 Sounds like Cortana lol

[tulip cloak]

One of the main reasons why AMD GPU's don't sell as much

[thorn bane]

So is there another way like a software to train voices and get a Index and Pth file? the colab didn't work for me asswel

[thorn bane]

True haha

[tulip cloak]

Just use colab @thorn bane

[thorn bane]

All im getting is errors

[green prairie]

that should sound like Adele 😄

[tulip cloak]

I don't even know why would anyone use local when you don't have to restart training

[tulip cloak]

Make sure you use a clean dataset

[tulip cloak]

Use rvcDisconnected

[green prairie]

no idea what you just said

[tulip cloak]

Use cleaner dataset

[proper shale]

Use a better TTS probably

[proper shale]

And yeah

[tulip cloak]

Oh

First of all

Delete the existing rvcDisconnected folder

[rare goblet]

Ayo? @green prairie level 2 !!!

[tulip cloak]

Then run the first cell again and keep the drive open

And once u see the folder

[thorn bane]

Google drive folder?

[green prairie]

that's the original, used UVR to take it

[tulip cloak]

Make a folder in your desktop put your dataset file in it and then make the folder a zip then upload it to Google drive rvcDisconnected folder

[tulip cloak]

Which models

[proper shale]

Please don't upload acapellas of songs here

[green prairie]

I used a model from this discord, from Adele

[tulip cloak]

No

I mean which uvr models

[green prairie]

Kim Vocal 2

MDX-Net

[tulip cloak]

Use mdx23c, voc ft, inst hq 3, de reverb, de noise and that should be it

Don't use mdx net

[green prairie]

this is what I have

[tulip cloak]

You need to download the other models

But if your lazy just use mvsep

That's what I use

anyway be right back

[green prairie]

and where exactly can I download those models? I do not see them in here

[proper shale]

you didnt run set training variables

run that step

[thorn bane]

That worked

But now this

[proper shale]

ok did you not run all dependencies steps at once

[thorn bane]

I did

[odd shale]

Denoise has a cutoff of frequency.

So his usage must be only under your own risk

[proper shale]

honestly im not sure

this normally doesnt happen

[thorn bane]

Do you know how i can reset the colabs?

[proper shale]

runtime > disconnect and delete runtime

[stoic cloud]

Yes, go to runtime, disconnect and delete runtime. But you will have to re-run everything

[proper shale]

^

and make sure not to maximize the dependencies step

just run all the dependencies steps at once

[green prairie]

any good places where to download more models, ecxept this discord?

[proper shale]

weights.gg

[thorn bane]

Still

[proper shale]

nah i think you're not running some steps

there is no way it's doing that

[peak whale]

hey

can anyone help?

[proper shale]

what's up

[peak whale]

oh i can't send img

https://ibb.co/hLZ11GF

im traing to process data

and still gettin this error

[proper shale]

yeah so

the name of models cant contain spaces

u have to do all preprocessing n shit again

[stoic cloud]

Change the spaces to underscores or remove them

[peak whale]

aightttt

im gonna try it

[proper shale]

^^^

[proper shale]

should work! lmk if anything happens

[thorn bane]

It worked!

[peak whale]

now is working

[rare goblet]

Ayo? @peak whale level 1 !!!

[thorn bane]

Its training now

[stoic cloud]

Pog

[peak whale]

thanks u so much lmaoooo

[stoic cloud]

Good job

[proper shale]

WOO

[proper shale]

you're welcome :)))

[thorn bane]

Thanks for the help!

Hahahha

[stoic cloud]

Goodluck @thorn bane

[thorn bane]

Ty

[proper shale]

you're welcome too, good luck with the model @thorn bane

[pure scroll]

this thing always works lol

[peak whale]

and btw, sorry for this newbie questions but

which algorithm is the best

[thorn bane]

ty, last question how long does it normally take for a model to train?

[stoic cloud]

Rmvpe

If you mean pitch detection method*

[proper shale]

it takes a while

[pure scroll]

how much epochs you set

[proper shale]

like hours

but you can stop for a while and resume training later

[pure scroll]

well it depends on your gpu too

[peak whale]

yes

[proper shale]

a day after

[proper shale]

Colab

[stoic cloud]

Yeah, that's RMVPE

[pure scroll]

my bad

[peak whale]

thanks u!!!

[stoic cloud]

Np!

[thorn bane]

500

[pure scroll]

on colab?

[thorn bane]

I have a good one

[peak whale]

also as a music producer, i know is better to have the largest sample rate like 48k

[pure scroll]

what's the dataset length?

[thorn bane]

yeah

[peak whale]

but i think is different for training?

[proper shale]

@thorn bane I'd recommend stopping the training after an hour or two and exporting the model

[thorn bane]

80 10 sec clips

[proper shale]

So you don't accidentally lose the model

[proper shale]

800 secs

[pure scroll]

check the sample rates of your audio on spek

[thorn bane]

fast meth there

[proper shale]

around 13 mins

[peak whale]

have some 48k and 44.1k

prolly gonna for 48k

[pure scroll]

would take 3 hours at the most ig

[stoic cloud]

I recommend training 40khz, since there isn't much difference that I can hear myself and that way you also dont make the model learn useless highs

[thorn bane]

How do i do that?

[pure scroll]

check on 'spek' for sample rate of audio bud!

[thorn bane]

Should better run it at night then

[proper shale]

You can stop the step with the same button as you clicked start with

[peak whale]

aight!

[proper shale]

Keep in mind you gotta do this on a non-saving epoch, so if its saving the D/G files wait for the next epoch and stop training

[peak whale]

gonna give a try tho @pure scroll

[proper shale]

Then run the export model step and the train index

You gotta be slightly active on the Colab page tho, check it every 5/10 mins

[thorn bane]

Not sure tho if its working or not

[pure scroll]

it is

don't worry

[proper shale]

hmmmm not sure

[proper shale]

1.7 secs for 13 mins?

yeah no the dataset screwed up

probably said no-feature-todo on feature extraction

[pure scroll]

well you are better in these things so i wouldn't say anything

[proper shale]

who said so bro

you have much more models than I do lmaoooo

[peak whale]

lmao i got this msg rn

[pure scroll]

they are trash - ||a mod||

[proper shale]

oof okay no this is bad

[peak whale]

lmao

[proper shale]

how's the folder?

[peak whale]

DUDE..

im a dumbass

[proper shale]

?

[peak whale]

EMPTY

ahahhahahahaha

[proper shale]

_ _

[peak whale]

rofl

[pure scroll]

_ _

[proper shale]

I thought "what if this is another audio problem aaaagh"

[pure scroll]

ayo thanks for the invisible text

_ _

_ _

[proper shale]

Well at least you know the fix

[peak whale]

cuz i moved the whole folder outside the sub one

[proper shale]

_ _

no

[pure scroll]

_ _

_ _

[proper shale]

ok don't do this please

dumb idea

[pure scroll]

aight

[proper shale]

ohhh

[peak whale]

problems on problems just for this sound

[rare goblet]

Ayo? @peak whale level 2 !!!

[proper shale]

greatest sound of all time

[peak whale]

it would be funny to try a fart model

1 epoch

[proper shale]

nah go for 5000

[proper shale]

me when training fart model to 14,982 epochs

[half cove]

[peak whale]

https://tenor.com/view/bonk-big-smoke-gif-26028132

oh

no gif previews

[proper shale]

no embed perms 😭

[peak whale]

bruh

[half cove]

https://tenor.com/view/thanos-memoji-gif-23490017

[thorn bane]

U sure? still at 0% tho lmao

[peak whale]

blud can

[half cove]

i have perms

[stoic cloud]

This is normal

[peak whale]

pls let me see this masterpiece @half cove

[stoic cloud]

Epochs are increasing

The 0% is how far it is on that specific epoch, it's just not realtime % so.. it just says 0

[half cove]

https://tenor.com/view/bonk-big-smoke-gif-26028132 tf is this 😭

[peak whale]

gonna send u a dm really quick

[half cove]

[pure scroll]

bro check the number written after epochs

don't worry about that 0%

[stoic cloud]

I told im

[pure scroll]

[peak whale]

oh someones have those tensorboard .bat(s) "TBl" something

[thorn bane]

Well it did work until i was at the last step

[pure scroll]

Read the exception?

did you run that?

[crude flame]

Why is this stuck in queue?

[thorn bane]

Yeah thats why im so confused

[rare goblet]

Ayo? @thorn bane level 5 !!!

[thorn bane]

I did read that

[pure scroll]

do this

[thorn bane]

But it will delete it

[pure scroll]

i too know that

look i haven't encounter this error ever

let's see what happens

i would see it too for the first time

[thorn bane]

U wanna see me dying inside that bad hu?😂 😂

Bro imagine you had it running for a few hours

[half cove]

do this then rerun untill feature extraction also make sure you zipped the file correctly and on experiment name make sure its the same as the dataset then run the index

[thorn bane]

Bro what kind of a screenshot is that

Thats so confusing

[pure scroll]

[half cove]

im too lazy to send seperatly

[green prairie]

can somebody help me with those settings?

[thorn bane]

Thats sound amazing i dont get youre tutorial

[brittle wing]

guys im kinda fucked

jk its because i resized a window

its alr bbg

dw

[green prairie]

I am having issues with the vocal being weird and I don't know how to fix it. It should be with those settings

[thorn bane]

Thats looks fire you should but it as youre background

[brittle wing]

sadwhich

[brittle wing]

this is my wallpaper

[thorn bane]

You crossed the line there

[brittle wing]

@silver anvil

[brittle wing]

L

I'm not xenophobic i just dont like french people whats wrong with that

[brittle wing]

[thorn bane]

Please explain it if you are like talking to a 5 year old

[silver anvil]

y am i pinged?

[brittle wing]

sadwhich

[silver anvil]

i saw

[pure scroll]

he's trolling ig

[thorn bane]

I think he is just trolling

[silver anvil]

he likes @pallid thicket's meatballs i see

[pure scroll]

[silver anvil]

i love that its ur desktop wallpaper now

[brittle wing]

yeah ikr ikr

[thorn bane]

@pure scroll Im gonna restart it now

[pure scroll]

do tell me what happens next

[thorn bane]

Fack now everything is bugged

Such a good idea

[quick nest]

Is there another thing like Ilaria RVC I can use because the thing im trying to convert is taking so long that by the time its finished converting the wipe happens and it errors

[pure scroll]

https://docs.google.com/document/d/e/2PACX-1vThk7Qo7yCWNVbxOmahl2R8_Jgi6TFuMBUIi-PWre_HIN0lFTq-dr37Rh5iJlGgYb_vFapXMHt2W8Kp/pub

follow this guide

[thorn bane]

I saw that

@pure scroll Can i send you a dm?

[pure scroll]

why?

[thorn bane]

I need something from you

Ohh okay wait that sounded weird

[thorn bane]

Someone knows what this error means?

[green prairie]

how do you guys fix the high pitch distortions in the EVC WebUI?

all models I try, different songs, when there are highs it sounds like my grandma is trying to sing after 2 weeks of being sick

anybody?

[spiral osprey]

pls help-me

[half cove]

https://discord.com/channels/1159260121998827560/1160164377006194728

[green prairie]

thanks, but that's not the issue, the issue is the conversion itself

in the RVC WebUI

[half cove]

thats usually the models fault

[green prairie]

I am trying lots of models

[half cove]

what uhh pitch extraction are you using

[green prairie]

UVR5

[half cove]

no

like crepe

[spiral osprey]

help-me

[half cove]

rmvpe

[green prairie]

you mean that?

[half cove]

on rvc webui

[green prairie]

(sorry, I am new at this)

ahh

I tried rmvpe

crepe

I find crepe being a little better in some cases

I think it's the settings in here that are making it weird

playing with them and converting, but I dunno

[rare goblet]

Ayo? @green prairie level 3 !!!

[half cove]

what version are you using?

[green prairie]

the last one, for NVIDIA

[half cove]

what number

like this

[green prairie]

RVC1006Nvidia

https://tenor.com/view/absolutely-awful-american-idol-judge-judging-serious-gif-13013491

[spiral osprey]

help-me

[half cove]

have you tried deleting the runtime and redoing it?

[spiral osprey]

where deleted?

[spiral osprey]

mine is limiting the use of the gpu or colab and I used it yesterday to render voice and even so it couldn't finish it, it stopped at 800 or 700 epouchs and couldn't even complete what was missing.

[half cove]

i think you got at the limit that colab provides

you can use after 12 hours of when it says

[spiral osprey]

but yesterday I only used it once and I hadn't planned on rendering the voice even though I only used it once, and then I thought it was strange.

[marsh jay]

this normal??

[karmic heron]

is there a video tutorial for rvc v2 disconnected colab?

[marsh jay]

alright thanks

[marsh jay]

where is it?

[rare goblet]

Ayo? @marsh jay level 4 !!!

[crude flame]

Why whenever i click the refresh button in applio these keep getting stuck in queue

[young halo]

does someone have any dashboard tutorial?

[proven hill]

HIGH QUALITY FEMALE VOICE MODELS

Are you searching girl voices like e-girl, mommy, v-tuber and more?
You find them here: https://ko-fi.com/ilariaowo/shop

INCLUDING FREE ONES!!

And for paid items, if you DM me, you will get a 20% discount if you pay with

I also accept Steam gift cards as last resort

Grab yours today!!!

[brazen plaza]

can anyone tell me why the voice is so choppy?

[rare goblet]

Ayo? @brazen plaza level 1 !!!

[violet heron]

Send a screenshot

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

🇺🇸 English (main)

https://docs.aihub.wtf/

​

Translation by country

🇧🇷 Brasil (PT-BR)

https://docs.aihub.wtf/v/brazil

🇪🇸 Spain

https://docs.aihub.wtf/v/spain/

🇫🇷 France

https://docs.aihub.wtf/v/france

🇩🇪 Germany

https://docs.aihub.wtf/v/germany

🇯🇵 Japan

https://docs.aihub.wtf/v/japan

🇰🇷 Korea

https://docs.aihub.wtf/v/korea

🇵🇱 Poland

https://docs.aihub.wtf/v/poland

🇷🇺 Russia

https://docs.aihub.wtf/v/russia

🇮🇹 Italy

https://docs.aihub.wtf/v/italy

[half cove]

use rmvpe

[brazen plaza]

i tried

[violet heron]

Change crepe to RVMPE

[tawdry spade]

any link for model train by colab google ?

[violet heron]

-colab “RVC Disconnected”

[azure marsh]

Suggestions for @tawdry spade

☁️ Google Colabs

• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab

🤗 Hugginface Spaces

• Ilaria RVC, by thestingerx Hugginface Spaces
• RVC-HF, by r3gf Huggingface Spaces
• AICoverGen, by r3gf Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces
• Advanced RVC Inference, by neuclya Huggingface Spaces

More commands: -audio, -local, -overtrain, -realtime, -rvc, -help

[brazen plaza]

i tried all of them

[proper shale]

check cmd

[brazen plaza]

it all looks fine

[violet heron]

Do that and change your chunk

[proper shale]

Send a screenshot just in case

[brazen plaza]

didnt help

[ivory plank]

what does 20 leaked semaphore objects in colab mean

[vapid gust]

could someone uvr extract for me