✨│ai-help

[low shard]

Your Nvidia GPU is good enough to do inference (use models) locally (on ur pc), not the best to train (make models) even if still possible

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 daily, not granted, of GPU

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

your PC GPU shouldn't matter for the quality, but matters for the speed

[brittle wing]

thats probably y my models sound bad

[low shard]

it shouldn't depend on your GPU

the speed and some settings and things like Training depends on your GPU

did you train the model on that GPU?

[brittle wing]

yes

ok before i was my time can u make sure this model is good

[low shard]

it's not much suggested, you will be limited on the batch size and dataset lenght

Train (make) RVC Models on cloud:

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

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

3. Be sure to know about the tensorboard

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

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

You can use either:

• Weights.gg: Easiest Possible Ever Automatic
• Ilaria RVC Zero: Fastest free on cloud
• Applio (ui)

[low shard]

I think you confused then?

Training is making, Inference is using

it seems you used another model rather than making one on your gpu

[brittle wing]

ok for training how long does the audio sample have to be

[low shard]

there isn't a strict limit but i'd say atleast 2-5 mins

[brittle wing]

bruh this guy was telling me to make a make hour long models

welp thats why u shouldn't listen to people who don't know what they talking about

[low shard]

who said that

what server is that

No one makes models with over 1 hour of dataset

[brittle wing]

just forget that but ok 2 to 5 minutes

[low shard]

I mean that's the minimum

the maximum would be 1 hour, since more than that has no sense as there's no improvement

[brittle wing]

ok forget that i just need to find a voice model that is already good

[low shard]

just get as much as you can as long as they are good quality

[brittle wing]

i looked here

[low shard]

alr

[brittle wing]

i am looking for any anime chartcdter

or movie charracter

[low shard]

You can search rvc ai voice models at:

• #1175430844685484042
• In #🔍│find-models , Do /find with @earnest musk
• https://weights.gg/ (login required)
• https://huggingface.co/models (but watch out cus in hugging face there arent only rvc ai voice models)
• https://voice-models.com/
• https://thevoicemodels.com/ (for Turkish Models, login required with discord and level 2 on their server)

if there isnt one, you can:

• #1159289738314919936
• #1191429836321849435
• make it yourself with our docs guides https://docs.ai-hub.wtf/essentials/how-to-make-voice-models/

[earnest musk]

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

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

[brittle wing]

or streamer

[brittle wing]

i dont feel like download appilo again

so can show me how to use the gooogle colalb

[low shard]

it's in the guide https://docs.aihub.gg/rvc/cloud/applio-colab/

[brittle wing]

thanks

[low shard]

yw

btw I just found out about it https://store.steampowered.com/app/993090/Lossless_Scaling/

AI FG and Scaling

it can help in game since it basically records and outputs ur screen in realtime with the AI, but would never help other AI program lol

[hallow thistle]

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

[brittle wing]

i just did

[brittle wing]

@hallow thistle

[low shard]

click refresh, and put the index too

[hallow thistle]

Nah, I'm just too slow to respond.

[brittle wing]

[low shard]

Idk gojo, but i guess you gotta try another model

also reminder that you can play with the pitch too, higher is more feminine, deeper is more masculine, just saying

[brittle wing]

welp that was the only model

the other model had a curropted index file

Overtraining or not

@low shard @hallow thistle

[brittle wing]

Can someone tell me if I should stop training or not?

Keeps going up after 150 epochs

[near timber]

anyone here that can send me some kanye text to speech

[storm mica]

any huggin face spaces worht their salt?

[near timber]

? 🙂

[brittle wing]

just buy eleven labs

if you are doing text to speech

because it can match the accent

[near timber]

Its only for a quick voice over for school

I can't find Kanye west his voice on Elevelabs

[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

[brittle wing]

@tight ether was KLM x4 trained mostly on female voices?

[pallid kraken]

Lots of errors on both Mainline RVC and EasyGUI. Cannot get either to run in Colab. Is there a new Colab link? All of them are "ModuleNotFoundError:"

[brittle wing]

How many minutes of speech against 2:05 of rapping?

[brittle wing]

https://youtu.be/Pl_Nf8-9vVk?si=67taQ-5lzjz5VeKu
Voice conversion isn't that new

[tame mica]

more like text to speech but yeah

it's been around for a while

[brittle wing]

Ikr

But it's similar

[next ginkgo]

anyone else getting a build error when duplicating the Ilaria rvc huggingface?

I’ve restarted it a dozen or so times and done a factory rebuild and it still gets an error

[hallow thistle]

The "optimized" Ilaria RVC that is made to run in Hugging Face space?

[next ginkgo]

I'm using this one:
https://huggingface.co/spaces/TheStinger/Ilaria_RVC
the one that's been around for a while

[next ginkgo]

is there another newer one I don't know about?

[hallow thistle]

Yeah, this is definitely the optimized version of Ilaria RVC made to run on Hugging Face space. You can use HuggingFace-hub command in Python to clone the entire repository to your disk, if you know how to do it. However, it is not generally recommended to make it to run locally, because this Ilaria RVC was made to run on Hugging Face especially. But there's another RVC fork program than can run both locally and cloud service.

The process to make this Ilaria RVC to run locally missing some components to run. It requires a batch file that can make venv (virtual environment) to pick up pre-installed Python and its components to run, and two FFmpeg program files that should be inside the folder.

[simple ore]

Why would you even bother installing that locally?

[hallow thistle]

If it's for fun experiments, yes. But seriously? No.

There's another version of Ilaria RVC on GitHub, it is more advanced. But there's no update currently, making it outdated.

[analog obsidian]

bro just download mainline 😭

[simple ore]

[sour sun]

Same, just downloaded the newest version of Applio but I'm also not seeing an "Applio" setting or the 44k sample rate

[analog obsidian]

refinegan still in development, not meant for casual usage yet

if you want to experiment with it check this #1327283027776245810 message

this is highly experimental and the models trained with it dont work in mainline and realtime/wokada

[crude mango]

are there any other options than rvc disconnected that arent local

[vale solar]

how to connect to google drive in Realtime_Voice_Changer_on_Colab.ipynb??

[brittle wing]

How much batch size per minutes

-guides

[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

[jagged berry]

Hi I am looking to install a software that can open the rvc voices who has please

[brittle wing]

How do I filter speech?

[sour sun]

Right; I was attempting to follow this but I guess I'm not doing something right or I simply don't have access

Ah well

[remote steeple]

[unborn grove]

For some reason when I use rvc it doesnt work and the audio is really gliched. I do not know how do I fix it can someone help me?

[azure marsh]

Hey, kspio! 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

[brittle wing]

https://github.com/codename0og/codename-rvc-fork-3

@glacial pollen

[glacial pollen]

@brittle wing ?

[brittle wing]

Isn't it what I think it is?

[glacial pollen]

No I meant, why the fuck you @ me without giving any context

[brittle wing]

Leave that, how do I prepare a dataset for a speaking model also I have

[brittle wing]

Uh I'm sorry

[glacial pollen]

So then, what's the matter?

🤔

[brittle wing]

I just wanted to ask how do I filter data for a speaking model

[glacial pollen]

Filter in what way?

[brittle wing]

Against 2:55 minutes of rap

[glacial pollen]

wdym by " filter " ?

[brittle wing]

Like for example you have someone talking in a vid with background music or just recording walking on the street and you can hear cars and stuff

[glacial pollen]

oh, well

uvr, mvsep

[brittle wing]

Which mvsep model

[glacial pollen]

Vocal / voice isolation pretty much

melroformers and bsroformers

is what I use

try both and see which works better for you

[brittle wing]

There's models for speech

[glacial pollen]

Trash, for me trash

[brittle wing]

I tried with these

They work well but there are cinematic ones for speech

[glacial pollen]

as I said, those are trash for me and I don't use such ( and so, can't give you any opinion on that

[lavish lintel]

Congratulations Razer by Weights!

Your Torterra is now level 39!

New move!

Your Torterra can now learn Synthesis!

‎

‎

‎

‎

[brittle wing]

Damn where's my Empoleon

[brittle wing]

None of the mvsep dnr models is here

[crude flame]

use bandit v2

dont use dnr

[brittle wing]

Do I have to remove "backing vocals" off the speech too?

[crude flame]

if there are "backing vocals" then yes

[brittle wing]

It's a person talking...

[crude flame]

ok?

[brittle wing]

https://youtu.be/D-8BAcelF2E?si=MvUJ7oZnA9k8rOoc

[brittle wing]

Bvs are music stuff

[crude flame]

then why did you ask if you need to remove the backing vocals when there are none

[brittle wing]

Why does Melroformer karaoke filter part of the speech

I mean a silent layer together w the noise

[crude flame]

why did you use mel karaoke when there are no backing vocals?

[brittle wing]

Cause I used bsroformer previously...

[crude flame]

ok? then again why did you use mel karaoke when there are no backing vocals?

[brittle wing]

Uhm so i just have to remove the reverb and noise, nothing else

[void abyss]

i need help guys, where can I use the models?

[brittle wing]

It doesn't damage the speech

[crude flame]

it does

all ai models do

[brittle wing]

What models

[void abyss]

voice model

[brittle wing]

Yes

[void abyss]

with which customer

[brittle wing]

On Colabs

Or GUIs

-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

[brittle wing]

Applio

[crude flame]

rvc or w-okada?

[brittle wing]

https://huggingface.co/spaces/TheStinger/Ilaria_RVC or this site

[brittle wing]

Is this for speaking models

Like for speech conversion.

[crude flame]

yes

thats what rvc is

w-okada is just a version of rvc but for realtime

[brittle wing]

Hmmm applio is for rapping and singing

[crude flame]

no

[brittle wing]

And okada for speech

[crude flame]

applio is a fork of rvc

same thing

just different UI

[brittle wing]

Applio works for speech tho

[crude flame]

they all do

[brittle wing]

Does okada work for singing and rap

[crude flame]

yeah

if you can sing or rap then it can

[brittle wing]

Uh I didn't know so I can inference with that one too?

[crude flame]

technically

yeah

but its a voice changer so

[brittle wing]

Nice

[brittle wing]

Are you the only QC still?

[crude flame]

no

[brittle wing]

Cause someone evaluated one of my models and accepted me

[crude flame]

That was Lyery

[brittle wing]

Nice of them

[knotty moth]

bandit v2 is cleaner, but melrofo also cleaner and gives better quality despite less effective to remove more sfx-es

[brittle wing]

English model right?

[brittle wing]

I know that but a model for speech should be used for speech, that's what it's for.
That's the purpose

[knotty moth]

so games are for kids?

[brittle wing]

What does this have to do with what I'm saying uhm

[knotty moth]

like? "applio may not be for speech?"

[brittle wing]

Can you stop already? 🙏 I'm begging you

[brave ermine]

hello

im training a model

what does this mean

lowest value = best step?

[rustic aurora]

hi guys

where is the colab link

for the training voice clone model

[polar tendon]

That's the lowest loss value, however that's only the loss, i recommend testing the checkpoint models too to see what gives the best result

[polar tendon]

There's Applio (both with WebUI and No UI), RVC v2 Disconnected, and more, however those two are most known to me

[simple ore]

lowest value is the "loss gen" (without loss fm/kl/mel)

it may not be going down while other do

[brittle wing]

If I have 2:55 mins of rapping, how many mins should the speech be?

[narrow nova]

Rvc v2 Disconnected, does anyone know what could be happening here and how to resolve it, this error appeared today on colab.

[peak lake]

How youre voices are so good like this ?

#1328427302681444473 message

mine looks kinda wierd

[teal ridge]

I got this problem as well and I have NO idea how to solve it.

[analog obsidian]

do you mean for realtime/wokada usage?
not all models work good in realtime for multiple factors, but the main are: the model was trained using low quality monotone speech, or low quality singing and/or the model is undertrained/overtrained
for example the model you sent sounds good because it was trained using a clean dataset and i properly choose the right epoch, so thats why it sounds nice but it might not be natural in realtime due to the model being trained using voice lines instead of natural speech
edit: models that were trained with a small dataset are also unrealistic

if you have more questions about realtime/w-okada use the help wokada channel ---> #🔍│help-w-okada

you can't fix a model that sounds robotic/unrealistic

thats a model problem

[knotty moth]

#1159290752195633273 message

[simple ore]

It can be fixed, but... it is time to let it go

[narrow nova]

Would there be a way to fix it? RVC v2 Disconnected has been a great way to train models online

[crude flame]

Use the applio colab, the mainline colab or kaggle or lighting ai

[unique rock]

applio is very slow to train jakska

[unique rock]

what is this lighting ai?

[crude flame]

its not

[crude flame]

another cloud platform like kaggle and collab

[unique rock]

Compared to Rvc Disconnected yes

[unique rock]

and how is it used and what are its GPU limits, do you have to pay?

[crude flame]

no

[crude flame]

for free

[crude flame]

there are creds btw

[unique rock]

I once tried it with a small 15min dataset and it didn't finish, it disconnected, in Rvc disconnected I train models up to 50min

[crude flame]

thats just rng

[unique rock]

22 hours per month?

[crude flame]

depending on the gpu yes

[unique rock]

T4

[crude flame]

then yes

[unique rock]

or are the others better?

[crude flame]

L40 is the best out of that list

[unique rock]

Could you help me use the platform?

[crude flame]

https://docs.aihub.gg/rvc/cloud/applio-lighting-ai/

[rich skiff]

what is the best voice changer for AMD gpus?

[knotty moth]

read the pinned guide in #🔍│help-w-okada

[knotty moth]

only if you use fp32 precision as in the applio settings

[unique rock]

I actually use fp16 but the difference is big

I don't know why but rvc disconnected allows training datasets of almost an hour, mainly every 100 epochs, without any problem.

[knotty moth]

try kaggle, it should be faster

[unique rock]

I use a cell phone and I don't know how it is used in these devices

[small light]

I'm just now using Rvc v2, and I have no idea how to connect the AI audio to discord or OBS.

Any fixes?

[knotty moth]

wdym, rvc inference works offline so it doesn't need to connect the audio shit
if you mean the realtime voice changer, go to #🔍│help-w-okada and read the pinned guide there

[small light]

mb mixed shit up

[woeful cave]

if rvc disconnected doesnt get fixed i will cry

😞

thats my fav way to train

[pallid kraken]

I’ve been getting similar problems with Mainline Colab

[runic thorn]

Hello, Hina_Mod_AICoverGen_colab.ipynb is down (gradio)?
It no longer offers the public (or local, but I don't use it) link.

[median lance]

Trying to use Hina Modified RVC on Colab, but when I try to run the server part I get this error.

mv: cannot stat 'MMVCServerSIO.py': No such file or directory
Timer: 00:03:43Traceback (most recent call last):
File "/content/voice-changer/server/HVoice.py", line 10, in <module>
from downloader.SampleDownloader import downloadInitialSamples
File "/content/voice-changer/server/downloader/SampleDownloader.py", line 12, in <module>
from voice_changer.RVC.RVCModelSlotGenerator import RVCModelSlotGenerator
File "/content/voice-changer/server/voice_changer/RVC/RVCModelSlotGenerator.py", line 5, in <module>
import onnxruntime
ModuleNotFoundError: No module named 'onnxruntime'
Timer: 00:03:44WARNING:pyngrok.process.ngrok:t=2025-01-16T06:32:42+0000 lvl=warn msg="Stopping forwarder" name=http-34787-672780c6-b081-4acf-bb28-1979f50eb8e7 acceptErr="failed to accept connection: Listener closed"
--------- SERVER STOPPED! ---------

[midnight jolt]

Help. I can't upload audio to weights on phone browser, it says file format not supported. Basically all format not supported.

[hallow thistle]

On Google Chrome browser or the Weights "app" itself?

[hallow thistle]

You can cry if you have time.

[midnight jolt]

I use chrome-based browser but not Google from.
I tried using weights app, but it can't create, always error.

[hallow thistle]

Google Chrome and Microsoft Edge are Chromium-based browsers, anyways. Are you sure you have checked the audio file name extension from file manager right? If the file name extension is "mp3" or "mp4", but the file itself is corrupted, this can happen.

Have you tried using Weights that also exists as a phone app?

[timid olive]

what is the best point on the chart?
I wonder if he's overtrained.

[low shard]

@median lance @runic thorn @woeful cave @pallid kraken #📰│dev-updates message

also for realtime voice changer wokada issues, use #🔍│help-w-okada next time

[pastel eagle]

i use 'realtime voice changer client' i wanted to troll sum people on valorant but it doesnt work or i just dont know how to use it

[worthy lantern]

so google colab by hina is down or what?

[worthy lantern]

i have the same issue

[analog obsidian]

rvc does not stand for realtime voice changer
for voice changer questions use help wokada channel -> #🔍│help-w-okada

[pastel eagle]

ohhh im sorry

[knotty moth]

go commission in #1191429836321849435

[low shard]

#1191429836321849435 or #1159289738314919936

[pliant knoll]

Oh so sorry I didn't know

Can i do it myself tho?

Lol

Making pth file of a data

[low shard]

well yeah you can train a model yourself

what's ur pc gpu?

[pliant knoll]

I mean 😂 I have a IMac 2017 idk

🥲

[low shard]

well, technically you could train an RVC model on mac locally, but from what I heard no one sucessfully did that and the speed is not as good as an nvidia GPU

mac isn't the best for AI tbh

[pliant knoll]

Aah nvm🥲

[low shard]

You could either try locally, or do it on cloud (remote good pc)

[pliant knoll]

Locally? Like in pc?

[simple ore]

just no

[low shard]

locally means it runs on ur pc hardware

[brittle wing]

10k

[low shard]

cloud means it runs on a remote good pc hardware

Train (make) RVC Models on cloud:

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

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

3. Be sure to know about the tensorboard

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

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

You can use either:

• Weights.gg: Easiest Possible Ever Automatic
• Ilaria RVC Zero: Fastest free on cloud
• Applio (ui)

[low shard]

is the speed that bad lmfao

[brittle wing]

@timid olive the epochs state at 10k steps

[low shard]

I heard that it's possible but no one did it successfully

[pliant knoll]

@low shard any video on that? Yt

[pliant knoll]

Rude

[low shard]

nope all youtube tuts are outdated

[pliant knoll]

Ikr

I can try for cloud one

But i know I wont be able to do it

[low shard]

most of the things I mentioned are on our docs https://docs.ai-hub.wtf tho, which is also what some of the links I sent redirects to

maybe that's easier to follow

[low shard]

I mean, I think he meant it as mac ain't good for AI rather than being rude

[pliant knoll]

Yea yes jk jk

[low shard]

Alright

[low shard]

I mean, if there's any issues you can ask here

or if you really don't want to spend time and just want it, u can request/pay someone, it's your choice afterall

[simple ore]

it is really not a PC designed for heavy workload

[low shard]

true

[simple ore]

at best you can run inference, but training is definitely no

[low shard]

maybe a latest mac could try? I'm not sure if RVC even supports MPS for training

it does support MPS for inference for sure tho

[simple ore]

we've removed that from Applio because it was just not working properly

and hanging up after 1st inference

[timid olive]

Was it 90e? I shared 40 with 100.

[brittle wing]

How many epochs it was during 10k?

[timid olive]

It's in the picture.

[brittle wing]

You removed the picture

[timid olive]

@brittle wing

s step isn't 10000 step

[brittle wing]

No you sent a pic of your tensorboard

[low shard]

what mac did yall test

was it THATT bad

also I think that should be said since I don't think everyone knows applio removed MPS support

[pliant knoll]

It's not happening for me

[simple ore]

as I said, it took way too much effort and without basically neutering a lot of settings at best you could run 1 inference

after that it just froze and died

[low shard]

please use #1159289738314919936 and #1191429836321849435

[simple ore]

eventually it was narrowed down to OMP number of threads = 1

[low shard]

wtf

May I ask which Mac btw

I knew that mac was bad, but didn't know THIS bad

2k pc for just graphic design and things even a laptop could lol

[simple ore]

Mac Studio m2 Ultra

and some others

"M4 MacBook with latest Sequoia."

[low shard]

smh

[simple ore]

"MacBook Pro 14 M2 pro silicon"

[low shard]

apple gotta be the most overpriced company

[low shard]

wait now that I think about it

@wispy lodge made so u can use Wokada deiteris fork on mac

which is RVC inference

I'm not sure how much good it actually runs

but maybe he could help ?

it sounds weird it crashes after 1 inference on RVC but not wokada

[simple ore]

not crashes, just dies and stops responding

[low shard]

like the PC stops working at all?

[simple ore]

no, the software..

like it makes an inference call and nothing happens

[low shard]

bruh

maybe emojikage can help ?

not sure lol

[simple ore]

pytorch supports mps, but something is fucky there... like OpenMP library or whatever is just not cooperating

and of course you have to enable a bunch of fallbacks that use CPU instead of MPS

anyway, I dont have a mac, i dont know anyone with mac who could test anything, so whatever

you bought it, you deal with it

[low shard]

me neither

I'm not sure if it would even work well on wokada then, last time I saw someone running on m2 they had like 600ms chunk #🔍│help-w-okada message

this sounds pretty weird tho

maybe emojikage can help?

anyways, I don't have a mac

[wispy lodge]

It's not like I have it. I only test if it at least works using a VM

iirc, only PYTORCH_ENABLE_MPS_FALLBACK is required as a workaround and everything else should work fine

[simple ore]

OMP_NUM_THREADS=1

[wispy lodge]

Didn't use OMP_NUM_THREADS since it's not what you normally want to do

[simple ore]

without it the whole thing just hangs

[wispy lodge]

Though there are reports that multithreading with macos openmp is borked so setting it to 1 for mac probably makes sense. But voice changer works more or less fine without it

[pallid kraken]

Ty for the info Nick!

[tidal spear]

I can't train my model pls help

INFO:torch.nn.parallel.distributed:Reducer buckets have been rebuilt in this iteration.

[W CUDAGuardImpl.h:124] Warning: CUDA warning: an illegal memory access was encountered (function destroyEvent)

I did not change the default settings

[young halo]

What site can i use for covers

The hinamod didn't work

[low shard]

did you try checking your PC GPU first? If it's good enough you won't have to rely on sites and use it locally

[low shard]

I'm guessing you're talking abotu hina mod ai cover gen, check #📰│dev-updates

[latent cypress]

should i click run-install.bat again when i tried to update codename fork? or simply just copy paste the rev files?

[brittle wing]

I can't see the epochs

[silver sky]

does anyone know how to set up an RVC on a Mac?

[brittle wedge]

I am working on that now, and i was able to install it but I cant get it to run locally. it only runs through the web UI, and that doesnt work

i cant set the output, it just says "none" and doesnt show my audio devices including the virtual mic

[brittle wedge]

trying to do this in safari if that helps. chrome lets me run it in "client" mode but I still can't get the UI to run locally on my mac, which i would like to do. when i run the program from the MMVCServerSIO it just opens the web UI

[frozen ledge]

-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

[silver sky]

I have the RVCGUI thing, does that not work on mac?

[bleak smelt]

is aicovergen hina working rn?

[grand eagle]

I tried using it today as well
And it’s not providing a link to the generator like
Is usually does when the commands
Are finished and completed
Does anyone have any direction on what to do and help please !

[low shard]

what tutorial link are u using?

@grand eagle @bleak smelt check #📰│dev-updates

@glad zealot btw push the mainline colab fix to ur other colabs, if they use the same code

[glad zealot]

already updated earlier

trying to fix voice changer now but the downgrade method im using doesnt work on some stuff

[low shard]

tell me specifically the other colabs that have been fixed please

[glad zealot]

just mainline for now, voice changer should also be working now but the model upload tab is broken atm

[low shard]

@ me for every colab that has been fixed 🙏

[knotty moth]

#1159290752195633273 message the original one doesn't seem to be fixed yet

[low shard]

good job, will add that to automaze and in #📰│dev-updates , let me know also about who's the original creator and who made that colab

oh jarredou

i don't think he's here

but let him know if u can

[low shard]

@knotty moth added it!

-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
• 🆕 Music Source Separation Training (Inference), by Jarredou & Makidanye 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

[brittle wing]

How many minutes of speech against 2:55 of rapping, so I can balance it all out?

[timid olive]

what makes you say that 10k is the best

[brittle wing]

It's the point before overtraining happened.

[timid olive]

Was there a button showing the epoch?

[timid olive]

How did you figure it out?

I always took the bottom one.

[formal wind]

Yo how do I continue training a model on mainline?

[steep junco]

Hello how do i make oiiaoiia cat cover with ai?

[midnight jolt]

Is there update for this vits? https://huggingface.co/spaces/zomehwh/vits-models

Or any tts like this one that support English

[low shard]

There are different Text To Speech (TTS) AIs:

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

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

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

With RVC Models:

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

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

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

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

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

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

• if you don't wanna use edge tts, you could try another tts ai from our tts index and use the output as an input in rvc

[jade drum]

-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
• 🆕 Music Source Separation Training (Inference), by Jarredou & Makidanye 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

[brittle wing]

-rvc

[azure marsh]

Suggestions for @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

[lone nest]

So if the dataset has 44.1k sample rate, do you choose the 40k or 48k target sample rate?

[simple ore]

48

[brittle wing]

Uh I'm a little experienced in training.

[knotty moth]

https://cdn.discordapp.com/attachments/1159290139609137264/1327962704459337839/image.png?ex=678b90a5&is=678a3f25&hm=dd1c81cc83b9db2ab3b51cb7a9a65c0a9811d52b1ec8a5613b626bce337e145e&

[simple ore]

if you have 48k file, use 48k

yes, the spectrum may be lower, but sampling is higher

[brittle wing]

How many minutes of speech against 2:55 of rapping?

[simple ore]

How many packs of orange juice for 3 bottles of vodka?

it depends

try experimenting

[low shard]

3

-# /j

[rough plinth]

who can help with voice changer? i cant send image here with my issue

[low shard]

realtime voice changer for calls? wrong channel, RVC means Retrieval-based-Voice-Conversion, used for speech to speech inference (use models) on pre-recorded audios and training (making) models, it doesn't mean realtime voice changer

Wokada is RVC but realtime, use #🔍│help-w-okada and specify ur issue more like what tut link and what's ur pc gpu

[fair yoke]

CalledProcessError Traceback (most recent call last)
<ipython-input-2-54e893f947a1> in <cell line: 0>()
19 for package in pip_packages:
20 print(f"Installing {package}...")
---> 21 subprocess.check_call(['pip', 'install', '--upgrade', package])
22
23 # matplotlib version force to fix tensorboard error

/usr/lib/python3.11/subprocess.py in check_call(*popenargs, **kwargs)
411 if cmd is None:
412 cmd = popenargs[0]
--> 413 raise CalledProcessError(retcode, cmd)
414 return 0
415

CalledProcessError: Command '['pip', 'install', '--upgrade', 'faiss-gpu']' returned non-zero exit status 1.

HElp

[timid olive]

now I will make a new model but this time I will do it from local

[brittle wing]

It's just a waste of time but I wish you success

[timid olive]

Is colab better

[brittle wing]

Idk really I'm on a mobile phone

[timid olive]

if your video card is low then colab can be a better advantage

[low shard]

what tutorial link are u using

[fair yoke]

https://colab.research.google.com/drive/1XIPCP9ken63S7M6b5ui1b36Cs17sP-NS#scrollTo=FBpJW3YB5zu1

[low shard]

google colabs are having issues, check #📰│dev-updates

[leaden rune]

how do i rvc for free

[low shard]

explain exactly what u wanna do and what's ur pc gpu

[leaden rune]

im tryna change my voice to sound like drake

1660 ti

[low shard]

realtime voice changer for calls?

[leaden rune]

no

for music

i make music

doesnt have to be local it can be like a colab or something

just dont want to pay

[low shard]

Your Nvidia GPU is good enough to do inference (use models) locally (on ur pc), not the best to train (make models) even if still possible

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 daily, not granted, of GPU

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

[leaden rune]

how long does it usually take compared to a colab

if locally

[low shard]

I'm not sure about that but prob faster than ur gpu

[leaden rune]

once i download applio how do i get the drake voice

i dont wanna train

[low shard]

yeah u got a 1660 super, cloud will be faster

[low shard]

You can search rvc ai voice models at:

• #1175430844685484042
• In #🔍│find-models , Do /find with @earnest musk
• https://weights.gg/ (login required)
• https://huggingface.co/models (but watch out cus in hugging face there arent only rvc ai voice models)
• https://voice-models.com/
• https://thevoicemodels.com/ (for Turkish Models, login required with discord and level 2 on their server)

if there isnt one, you can:

• #1159289738314919936
• #1191429836321849435
• make it yourself with our docs guides https://docs.ai-hub.wtf/essentials/how-to-make-voice-models/

[leaden rune]

it says "#unknown"

huh

[low shard]

oops, edited it

[leaden rune]

ok thanks

[tiny totem]

Methinks that Colab doesnt loke RVC.suddenly

<ipython-input-1-54e893f947a1> in <cell line: 0>()
     19 for package in pip_packages:
     20   print(f"Installing {package}...")
---> 21   subprocess.check_call(['pip', 'install', '--upgrade', package])
     22 
     23 # matplotlib version force to fix tensorboard error

/usr/lib/python3.11/subprocess.py in check_call(*popenargs, **kwargs)
    411         if cmd is None:
    412             cmd = popenargs[0]
--> 413         raise CalledProcessError(retcode, cmd)
    414     return 0
    415 

CalledProcessError: Command '['pip', 'install', '--upgrade', 'faiss-gpu']' returned non-zero exit status 1.```

Is this even the right channel to post an error?

[tiny totem]

oh I'm stupid shoulda pron checked first here

[hallow thistle]

If y'all wondering why RVC and certain Colab notebooks doesn't work, #📰│dev-updates has already said this.

[teal ridge]

Has anyone fixed that damn notebook?

[lone nest]

I fixed it by running this first (credit to Namari)

!pip install faiss-cpu
!sudo apt-get install python3.10
!sudo update-alternatives --install "/usr/bin/python3" python3 "/usr/bin/python3.10" 1
!sudo update-alternatives --install "/usr/bin/python3" python3 "/usr/bin/python3.11" 0
!sudo apt-get install python3-pip
!python3 -m pip install --upgrade pip```Then run Install Dependencies and don't restart the runtime when it prompts you to, then run
```bash
!pip install tensorboard```Afterward run the rest of the notebook as usual

[hallow thistle]

Nah, I'm sorry, but I didn't work for this notebook. #📰│dev-updates explained that Google Colab has updated its main Python version, which caused RVC and certain Colab notebooks to break.

[upper furnace]

Thanks 🥹

[lone nest]

Hm it seems that index training can't be run because it can't read the faiss module, but pip freeze shows faiss-gpu is already installed and installing faiss-cpu doesn't fix it either

[lone nest]

Ugh I give up

[hallow thistle]

Downgrading Python version can work with certain scenarios, but like not everything could work with this method.

[lone nest]

But not being able to import a package when it's already installed is so baffling

At least index training is separate so you can do it elsewhere

[hallow thistle]

Next time maybe try python -m pip install faiss-gpu with Python 3.10 again, unless Colab is still selecting some dependencies from Python 3.11.

[formal wind]

how do i continue/resume training a model on mainline

[lone nest]

Did that, and another try with --force-reinstall too

[timid olive]

I'm going to edit the concert recording of an old Turkish artist who performed in Japan. Would it be an issue if I don't separate echo and reverb effects, as I only plan to isolate the instruments?

Can I also show the epoch here? Is there a setting?

[upper furnace]

help @lone nest

I don't know why it doesn't do index trening?

?

[timid olive]

!pip install faiss

[timid olive]

.

[timid olive]

If you use applio, there may be problems with old colabs because they forget to update them.

-applio

[azure marsh]

Not available yet

[upper furnace]

I have it's not possible that I only have these problems anyway and others don't

@timid olive

[timid olive]

I use applio

https://github.com/IAHispano/Applio

[timid olive]

applio is more up to date

[upper furnace]

ok but i just wanted to solve the problem i have with this part thanks 🙂

[upper furnace]

in the sense he had done everything right but randomly at the last one he failed what balls (I just wanted to fix it to have my model complete)

[timid olive]

did you run the install part at the beginning?

[timid olive]

Is this colab rvc disconnect?

[upper furnace]

yes

[lone nest]

Yeah I still haven't found a solution for that

[upper furnace]

ok

[lone nest]

In the meantime you can try using applio just for the index training, since it's independent from model (.pth) training

[low shard]

@errant ginkgo this is the right channel for making AI Covers, what's ur pc gpu

[upper furnace]

But is it still okay for it to make 0 seconds?

[simple ore]

it is asking Dataset path, not .zip

and not .zip to a model

[upper furnace]

ok

How do I make it work?

?

[lone nest]

I think I've found the solution, start a new runtime and run !pip install faiss-cpu before everything else

Updated my post with that

[unique rock]

Does anyone know if rvc Disconnected is working now?

[low shard]

no it's still broken

allegedly not updated yet

all u can try is #✨│ai-help message

[upper furnace]

and is it normal for it to stay like this? it's been like this for a few minutes now

@timid olive

[formal wind]

Can someone tell me how i'm supposed to resume training in rvc mainline?

[low shard]

Colab: https://docs.aihub.gg/rvc/cloud/mainline-colab/#4-resuming-training
Local: https://docs.aihub.gg/rvc/local/mainline/#resuming-

[nocturne mural]

I don't know if I remember well but the last message there means that maybe the program has stopped, who knows why?

[low shard]

@upper furnace @unique rock #📰│dev-updates message
rvc disconnected is fixed

BUT

the dev said that he's busy irl, he's not interested much in RVC anymore, so he won't be always able to fix it, that's why he also said it's fine if people switch to other notebooks

unfortunately

[unique rock]

In short, we're already screwed.

[proven hill]

@low shard is applio on pinokio updated and stuff orrrr

[unique rock]

Thank you very much, I will take advantage of it.

[low shard]

You're welcome

[low shard]

yeah it won't get much updates, AI changes much, which is why staff suggestes to try other things

[low shard]

idk I don't use pinokio, try asking @simple ore , I used a precompiled

[low shard]

also you missed this message #📑│making-models message

[proven hill]

you should

[low shard]

I heard about it, maybe one day since it seems an easy installer for other AIs

[simple ore]

i do not use 3rd party shims

no pinokio, no stability matrix or any other garbage

[low shard]

but I rather download it myself usually

[low shard]

u do SD too?

I thought u did only RVC

[proven hill]

meh, trying it out for work, doesnt seem bad, it does stuff for you

[simple ore]

SD.Next

[proven hill]

also Forge Flux

[low shard]

Oh lol

I use ComfyUI

How's SD next?

[simple ore]

Comfy is nice, but you really gonna learn how to make workflows and shit

some things are good in Comfy, some are easier in SD Next

[low shard]

Yeah it's hard ngl

[simple ore]

i've never managed to get Flux working in SD Next, but it was fine in Comfy

[low shard]

But I managed to run flux and loras

[low shard]

Yeah tbh I rather best support than simplicity

[low shard]

Btw back in the days I used that colab too, I said that warning not bc I don't like it or smt, I said it just so people understand that some engineers are busy

Many colabs are fixed now that's good

[low shard]

@proven hill so what about #📑│making-models message ?

[proven hill]

whats the error?

i remember the message

im on phone (talking) rn

[low shard]

dependency issues between fairseq and infer-rvc-python

goodluck 😭

[low shard]

the issue is if u don't fix it people won't be able to duplicate it, and also you will have trouble in the future if u reset it or update it

average python moment

[proven hill]

okay but

the issue is different on me

oh no fuck found it

[low shard]

goodluck debugging it

I tried too by downgrading pip in the requirements

but idk it doesn't work

[proven hill]

idk if its an issue i can fix atp

[low shard]

please just let me know about it because I take care of updating the working resources

welp it 'works' as long as it doesn't automatically restart so ig it's fine for now

[proven hill]

oki, idk who to ask for help, because it seems its a package i cant modiffy (?)

and mikus really cant be contacted

[low shard]

well im not sure if there was a package update or they updated python just like how colab did recently honestly

[low shard]

still arguing?

[proven hill]

probably

idk if anyone on the team is willing to give an hand to fix it

maybe @viscid moss (?)

if she isnt busy with UVR UI

[low shard]

I tried also putting

# fix installation
os.system('pip install pip==23.2 omegaconf fairseq==0.12.2 hydra-core==1.0.7 infer-rvc-python==1.1.0')

from infer_rvc_python import BaseLoader

in the app.py instead of the requirements.txt hoping it forced the pip version but unfortunately same issue

[low shard]

wait eddy is a she?

[proven hill]

yes

[low shard]

my bad I didn't check pronouns

[proven hill]

unfortunately i was hired just yesterday so i dont have much time to develop stuff like i used to do and to be VERY honest i dont really know the problem i dont know how to fix that

[low shard]

what did you get hired at?

if u want to answer ofc

else u don't have to

[proven hill]

ai engie

[low shard]

I meant like at building what if u can tell?

[proven hill]

restaurants infrastractures with ai

but mostly, im the guy who use chatgpt to make a facebook post

[low shard]

oh lol

[proven hill]

moments i wish i had an a100 at home

this is for a charity im working for too

i might just port ilaria rvc for pinokio at this point, so people can use it locally(?)

[low shard]

idk what's that

-# the ss not the gpu

[low shard]

cool

[proven hill]

echomimic2 creates avatars that speak for you

[low shard]

oh never heard of it before

[proven hill]

its useful.
since they are a charity they dont have much money influx and havig a studio for recordings would be too much, so a good looking picture can do the trick

[tulip coyote]

can i somehow use rvc models in code? rather than in guis

[storm harbor]

can someone send me the doc link for all the downloads?

[viscid moss]

👀

[proven hill]

hewwo

[viscid moss]

Hewwo

i was checking messages, something isn't working?

[proven hill]

ilaria rvc zero no work

when duplicated and probably next restart will completely die

[viscid moss]

when duplicated? wow that's so weird

lemme try duplicate and see

[proven hill]

the problem is:

infer_rvc_python

[viscid moss]

Ok, I'll try

[proven hill]

ty <3

[low shard]

maybe u can look at https://github.com/blaisewf/rvc-cli

[low shard]

https://docs.aihub.gg

[tulip coyote]

this seems like quite a beefy swiss knife for rvc stuff too, im not interested in training or things like real-time voice change, im moreso looking for ways to load an rvc model and use it right away, for example to generate tts, something like:

model = # somehow load .pth and .index files

# use the model somehow

# save the result to .wav or .mp3 

[low shard]

RVC is STS not TTS natively

u could also host locally Applio and use the gradio api if u want

[tulip coyote]

damn rvc models are all i can find nowadays lol

[low shard]

yup it's the most widespread

There are different Text To Speech (TTS) AIs:

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

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

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

With RVC Models:

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

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

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

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

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

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

• if you don't wanna use edge tts, you could try another tts ai from our tts index and use the output as an input in rvc

you could check this out ^

[tulip coyote]

ty

so because rvc isnt natively tts, it uses a helper to generate speech and then convert that speech to the voice it represents?

[low shard]

u can generate any TTS audio with another TTS, then use it as an input in rvc

that's what Applio does with Edge TTS

[tulip coyote]

i see

ill experiment with those options and see where to go from there

thanks for the help

[proven hill]

@viscid moss any clue?

[low shard]

yw

[viscid moss]

not yet

[proven hill]

gotcha

[unique rock]

I overtrained, right?

[simple ore]

well, something happened.. did you resume training with a different batch size?

[unique rock]

No, it's first training, 80 times

Epochs*

40 minutes of dataset and a batch size of 8

[simple ore]

then it is strange

what's on other charts?

[unique rock]

In the previous one it also went up

I'm going to try 70 eras

[brittle wing]

@glacial pollen how many minutes of speech against 2:55 of rapping?

[glacial pollen]

omfg

Cmon man, how many times you gonna ask everyone this question?

I told you the last time, 1:1 balance

There is no universal rule to anything in rvc that'd work 100% of the time, so asking around won't get you anywhere, in fact, you wasted quite a big portion of time asking than doing some training trials

[proven hill]

@red kayak

[pastel oak]

[latent cypress]

https://tenor.com/view/waiting-bored-mr-bean-walk-taking-so-long-gif-16887080

[unreal ridge]

For the real time voice changer what’s the best settings for NVIDIA GeForce GTX 1070 (8gb)?

I keep lagging

[brittle wing]

Mhm training trials...I don't wanna do that

[glacial pollen]

I'll refrain from commenting it

[brittle wing]

So 2:55 minutes of speech?

It's a waste of time I have other stuff to do too I need a definitive answer

[brittle wing]

Really?

[hallow thistle]

For W-Okada, go to #🔍│help-w-okada. This channel #✨│ai-help here is all about RVC and certain RVC programs.

[glacial pollen]

Yea really. Too many kids, as I like to call them, gettin' butthurt lately

So please, don't waste my time if you can't afford to test what works for you

And so am I so I'll tell you what, stop wasting others' time if you aren't willing to give in a lil from yourself

[brittle wing]

It's just that I have a very little data of the person rapping

[glacial pollen]

yeah but I've been literally telling you few times on what to do

[brittle wing]

And I wanted to know how much speech will be good to have balance

[brittle wing]

Yes but simple terms

[glacial pollen]

excuse me man but, how the fuck am I supposed to know?

[brittle wing]

2:55 of talking

[glacial pollen]

1:1 ratio is a golden rule

[hallow thistle]

https://cdn.discordapp.com/emojis/1015964749470650400.webp?size=48

[glacial pollen]

what's so hard in that?

[brittle wing]

Fr that's my personal info I shouldn't be giving it out here

[glacial pollen]

🤦‍♂️

[hallow thistle]

Imagine getting a model maker role, but still have no idea how to train a voice model.

[glacial pollen]

I can't take people incapable of reading no more

[hallow thistle]

[glacial pollen]

cmon

[brittle wing]

But I chose some samples of singing and rapping in another model and the model still turned out fine that aren't the same length in comparison and it still worked

[glacial pollen]

that's why I literally told you to test it on your own

[hallow thistle]

Please let's be real.

!howtoask

[glacial pollen]

Come the heck on

[patent trellis]

How To Troubleshoot

__**GIVE CONTEXT.**__ 📝

• Don't simply mention your issue, like "my rvc is not working".
• Describe the step you are on, what you're trying to do, the RVC you're using, a screenshot, etc.
• The more context, the better.

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

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

__**BE PRODUCTIVE.**__ 🤝

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

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

[brittle wing]

?

[glacial pollen]

once it'll turn out fine, once it'll suck

1:1 is a good rule to follow but there are exceptions

and it's in your hands to figure it out

not mine, at 3 am

[brittle wing]

That's literally a discord server where you're supposed to ask for help

[glacial pollen]

you've wasted over a week now? maybe couple days asking

rather than doing 3 stupid runs each taking an hour

for a simple test

[brittle wing]

Uh really why are you up late

[hallow thistle]

Nahh, do you think anyone here have time to help you? What?

[glacial pollen]

it's not really like, purely about time but rather ignorance and laziness of people

[brittle wing]

I prefer to be done with my dataset and train once it's done

[hallow thistle]

Not everyone is in the mood to help you either.

[glacial pollen]

Rather than learn and experiment, people expect us engineers or staff to have 100% accurate perfect formula for everything

[hallow thistle]

[glacial pollen]

cause lazy or " wuee don't want to "

[brittle wing]

Laziness huh

[glacial pollen]

yeah and isn't it

you said you don't want to do trials

[hallow thistle]

Don't try keep asking for the same question, it's annoying.

[brittle wing]

And y'all always have time to be active on here so...

[hallow thistle]

So?

[brittle wing]

Okay I won't anymore if that bothers some of you

[glacial pollen]

just because I happen to be online and see notification doesn't mean I'm obligated to help

sure, I may respond but that's that

[brittle wing]

So someone could see my question and answer

[glacial pollen]

• I already told you to what to do

[hallow thistle]

Is me being active a problem?

[glacial pollen]

and you keep on bragging and bragging on

[brittle wing]

Not for me

[glacial pollen]

haven't you already figured out why nobody wants to answer you?

[brittle wing]

I'm sorry

[glacial pollen]

because there's no damn fucking answer man

[brittle wing]

What-

[glacial pollen]

if you can't test it and see what works best for you

we don't know better

[hallow thistle]

[brittle wing]

Uh

[glacial pollen]

it's the same as if I asked you how much of moan data to keep vs yelling samples

and how you gonna estimate it?

do you see the point now?

[brittle wing]

I thought you know better than I do and that you're an expert in that

[hallow thistle]

And you'll be wondering why no one is helping you. You made yourself, you made thing yourself huh.

[brittle wing]

Okay

[glacial pollen]

being and expert or knowledgeable doesn't equal knowing or predicting every single training outcome

[hallow thistle]

Shit. Not everyone could know what you were doing.

[glacial pollen]

that's out of our human boundaries

[brittle wing]

Uh that makes sense

[glacial pollen]

even creators themselves don't really know the fuck is going on in neurons themselves

nobody does

[brittle wing]

Uhm yes

[glacial pollen]

you can only predict or estimate, more or less, results based on previous trials or empirical testings

[glacial pollen]

which you haven't done nor intend to

[brittle wing]

My fault

[glacial pollen]

happens to everyone really

[tawdry rampart]

Hi

[hallow thistle]

Wouldn't be my fault.

[glacial pollen]

but yea, like I said, if you don't wanna do tests, go for 1:1 or 70:30 ratio

70 is what you prioritize

yet I'll insist on 1:1 as AI doesn't like any biases towards any kind of data, that's just against generalization

[brittle wing]

Reading books is my passion wdym

[brittle wing]

Absolutely not

[glacial pollen]

yeah you already showcased it with delayed response + not catching up on what I meant

my guy, now kindly do as you please

[hallow thistle]

Are you slow?

[glacial pollen]

I already said all I can in this topic
and am back to my stuff

[brittle wing]

Nope I just have a mind of my own.

[hallow thistle]

If you're here just to troll around, please get out from here.

[glacial pollen]

alr goofball, quit it

[brittle wing]

No I'm not.

[brittle wing]

Okay I'm ending it

[glacial pollen]

any provocations or induced conflicts won't be tolerated well in here

[hallow thistle]

Get the fuck out from here now, and try do the thing by yourself.

[glacial pollen]

both in terms of mocking or @ s

[brittle wing]

I'm not here for that purpose & no, I won't leave just because you're telling me to.

[hallow thistle]

https://cdn.discordapp.com/emojis/1015964749470650400.webp?size=48

[brittle wing]

Okay end of conversation

[glacial pollen]

in simple words, don't test patience of people you don't know. end of the topic

[brittle wing]

Yeah well said

[glacial pollen]

as I said, don't @ me. any further rejection of this request I'll take as provo

✨

[brittle wing]

OK I'M SORRY

[hallow thistle]

If you don't end, we can't end it either. So please stop already, I told you.

[glacial pollen]

^

[brittle wing]

I stopped-

[hallow thistle]

If you keep replying to me, it doesn't make this to end easily.

[brittle wing]

I said already

[cloud finch]

Is it possible for splitting the audio of a recording to mess up the timing of it?

[brittle wing]

Be more specific

Splitting audio in seconds?

[cloud finch]

Im referring to Applio's splitting audio feature

The base recorded audio is timed perfectly with the speaking images whenever they talk

but when i run it through ai

its off timed

[brittle wing]

Uh I use Applio only for inference sorry

[cloud finch]

?

im

talking about inference though

inference is making the voice ai right?

like

you put audio in

and it comes out ai

[brittle wing]

Converting it into another

[cloud finch]

i assumed thats what it was, yes?

[brittle wing]

Yes

[cloud finch]

ok yeah then, yes

im talking aout inference

[brittle wing]

Then what's the problem you're having

[cloud finch]

Actually

ill see if i can show you hol up

[brittle wing]

Take a screenshot

[cloud finch]

I will

ok

Top is normal audio, bottom is ai

their both timed

now if i go further into the recording...

they desync

I think this is because of the split audio feature, but im not sure

[analog obsidian]

Yes, audio split causes delay
This is already fixed in applio main branch

but not in the compiled version

[cloud finch]

How do i get the main branch

[simple ore]

this should be fixed in next release

[cloud finch]

do i use visual studio

and compile it or sum?

[simple ore]

or you can grab the updated splitter

[cloud finch]

Do you know how i would do that?

[simple ore]

https://github.com/IAHispano/Applio/commit/75371088d41175b1a2c1a832eab102722542fb5d#diff-abc51b32e100779efcd001432e0e02914f3f51a6951be8ce0ae990dcaafe745b

those two files

[cloud finch]

these two?

so i just replace them

in my build?

[simple ore]

yeah

[cloud finch]

awesome sauuuuuuuuce

[simple ore]

hopefully that does not blow up 🙂

save the copy

[cloud finch]

bro??? 😭

[simple ore]

i mean there were other changes between 3.2.8 and this

[brittle wing]

How much batch size for short datasets? where's the guide?

[cloud finch]

ik this question is kinda baby but i dont use github like that, where do i download ts

[simple ore]

it would be safe to make a change to infer.py manually

just that chunks part

[cloud finch]

oh like

[simple ore]

and then download the split_audio

[cloud finch]

edit the file??

im askin here right uh

how do i

download it

githubs wierd

[simple ore]

https://github.com/IAHispano/Applio/blob/75371088d41175b1a2c1a832eab102722542fb5d/rvc/lib/tools/split_audio.py

[brittle wing]

-guides

[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

[cloud finch]

oh there we go

i see a download now

[simple ore]

so replace that file in your local build

and then add chunks,

[cloud finch]

im gonna assume i did it right here

[simple ore]

yes

[cloud finch]

alright

ill see if it works...

omg

it worked

tysmmmmmmmmmmmmmmmm

❤️

[hidden laurel]

chat

[raw mango]

help, rvc not open

[hallow thistle]

!howtoask

[patent trellis]

How To Troubleshoot

__**GIVE CONTEXT.**__ 📝

• Don't simply mention your issue, like "my rvc is not working".
• Describe the step you are on, what you're trying to do, the RVC you're using, a screenshot, etc.
• The more context, the better.

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

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

__**BE PRODUCTIVE.**__ 🤝

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

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

[hallow thistle]

Please be more specific on which problem you're encountering while using RVC.

[hidden laurel]

i have a 24 minute .wav audio file, 522 mb file, full vocal range + talking, what settings should i do?

[echo totem]

i downloaded rvc on my new pcand using the same voice models as before, but when i try them theres extreme delay and splitting happening which didnt happen on my old pc.
i would show my rvc config if i had image perms.

also yes my correct gpu is being used and i messed with the chunk and extra options.

@tired kraken

[brittle wing]

-guides

[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

[tropic plank]

hi

which one do you recommend

[polar tendon]

Inst HQ 4, the most accurate as of now

[timid olive]

https://github.com/Eddycrack864/UVR5-UI

[timid olive]

It's a thing of the past

this already tells you which is the best model

[tropic plank]

is this a suggestion

[timid olive]

yes personal

because the best model is here.

It works well when I separate Turkish songs and Japanese songs.
Imagine how it is in English.

[tropic plank]

right ?

helper olan enesmisin sen

[timid olive]

yeah

[tropic plank]

niye rolün değişti

[timid olive]

Do you speak Turkish?

[timid olive]

istifa ettim

saldım

[tropic plank]

ah be

[timid olive]

alın dedim rolümü

bu arada dedğim uygulamayı yükle

sil onu

[tropic plank]

linuxtayım

[timid olive]

hem türkçe hemde iyi ses ayırma modeline sahhip

[tropic plank]

çalışırmı o

linuxta

[timid olive]

onada indirilebiliyor

apple da dahil

[tropic plank]

indirelim

neyse bunuda deneyelim

[timid olive]

içine install bat var ama sen apple ve lenux lar için UVR5-UI-installer.sh mi ne var onu çaıştıracaksın

[tropic plank]

gördm varmış

[timid olive]

bat win için

[tropic plank]

UVR5-UI-installer.sh

[timid olive]

https://youtu.be/ZhQgaH_oATw

bunu izle tek adım UVR5-UI-installer.sh i çalıştıracaksın

[tropic plank]

indiriyor

bekleyelim bakalm

[timid olive]

appleden müzik indirince daha iyi oluyor

benim sonuca göre

[tropic plank]

appledan kastın macbookmu

[timid olive]

yok apple music den müzik indirince kaliteli

yt den

[tropic plank]

ordan indiririz

sen napıyorsun niye bıraktın

tek helper sendin neredeyse

tr için

yapay zekayımı saldın tamamen

[timid olive]

yoo yapay zeka yı salmadım

fazla aktif olmuyorum

[timid olive]

liderlik tablosundan en iyi model e bakarsın

bunda düet ayırmada var

pek iyi değil ama bazı şarkılarda iyi sonuç

olabiliyor

[tropic plank]

fazla detaya gerek yok ya bazen lazım oluyor vocalremoverdan ayırınca

çöp ediyor

sağol

[timid olive]

lenux da ffmpeg nasıl indiriliyor bilmiyorum

ffmpeg lazım olabilir

[timid olive]

ben değildim tek

[tropic plank]

Aktif olarak yardimci olan*

[timid olive]

@pastel oak buda vardı türk almanyadan (etiket için sorry)

[tropic plank]

Var zaten bende

Ama eklenmis zaten sanirim kendiliginden

[timid olive]

@brittle wing Buddy, now the graphic keeps going down, aren't we taking the bottom part?

You can tell how it's going into overtraining, the graphic hasn't gone up at all.

[primal bridge]

Hi guys

I want to start trolling my friends with a Voice changer which has every audio in them and is also kinda cheap

[timid olive]

this is reset here lol

[primal bridge]

Can ya'll help me by suggesting the best voice changers and settings , thanks guys 🤝

[timid olive]

For real-time?

[primal bridge]

yes

[timid olive]

#🔍│help-w-okada

Write here.

[primal bridge]

btw can I do it without a graphics card? \

[timid olive]

This channel is for model training and model output generation, not for real-time help.

[primal bridge]

That doesn't answer my question but alr

Also how to my own voice cloning to any voice I need?

can u help me in that pls?

Like I can record my voice and then change it to any voice I desire

@timid olive ?

[tropic plank]

Auto translate falanmi kullaniyorsun

Buraya yazilan herseyi tryr ceviren

[timid olive]

yok

chatgpt

yazım düzltmesi yaptırıyom

az inglizcem var

[tropic plank]

türkçe yazdığımdada ingilizce yazınca öyle düşündüm

[formal wind]

What files do I have to transfer to a new account to continue training in 1 day instead of waiting for my colab to regain its time>

[crimson depot]

@viscid moss

Chech this repo https://colab.research.google.com/drive/1qezaBvCn4eSG1U2L4a64tEHTynbIu8eN?usp=sharing

ERROR: Cannot install ort-nightly-gpu==1.17.0.dev20231205004, ort-nightly-gpu==1.17.0.dev20240118002 and ort-nightly-gpu==1.19.0.dev20240531001 because these package versions have conflicting dependencies.

[civic dirge]

Yo guys i tried using vc in valorant but i can‘t hear the ingame sound while using the vc can someone help me?

[azure marsh]

Hey, 38baby🐍! 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

[knotty moth]

I don't think the rvc application can do that thing, or go to #🔍│help-w-okada if you mean the voice changer and read the pinned guide there

[lavish lintel]

Congratulations okepeko!

Your Fuecoco is now level 4!

[low shard]

I think you'd have to transfer the whole dataset and model folder if u didn't even lose them

There's a risk that u did

Why don't you use Kaggle?

It gives more time

Kaggle >>>>>> Colab

[formal wind]

I gave up trynna use kaggle. Plus yeah I had to transfer the whole folder over.

[crimson depot]

Can anyone tell me a working colab that doesn't have a UI?

[low shard]

Why no UI and colab?

UIs are encrypted to not get detected

And there's Kaggle that has more GPU time

[low shard]

You can ask here for help if there's smt wrong, but anyways ur choice

[low shard]

Also, colab for what?

Colab is a cloud computing service for people who got a bad pc

[crimson depot]

Is it annoying to have to download my GD model and re-upload it to the Gradio websites

[low shard]

So u need a colab to make ai covers?

Check #📰│dev-updates

Colab are having issues

The only no UI one is applio no ui

[formal wind]

no im just stupid and struggle to follow instructions that are formed by written words. My brain works with visualization most of the time

[snow halo]

https://discord.com/channels/1159260121998827560/1323752992729600010 this one does not have a .index when i download, can someone help?

[azure marsh]

Hey, Shoebill! 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

[snow halo]

!howtoask

[patent trellis]

How To Troubleshoot

__**GIVE CONTEXT.**__ 📝

• Don't simply mention your issue, like "my rvc is not working".
• Describe the step you are on, what you're trying to do, the RVC you're using, a screenshot, etc.
• The more context, the better.

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

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

__**BE PRODUCTIVE.**__ 🤝

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

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

[brittle wing]

If it's going down, then it still hasn't overtrained

[viscid moss]

currently working on a fix. U can try UI meanwhile or HF space

[hallow thistle]

Try Ilaria RVC on Hugging Face.

[viscid moss]

https://huggingface.co/spaces/r3gm/AICoverGen