✨│ai-help

[violet heron]

But most likely your not going to get a good result out of 40k

[molten pecan]

I assumed the sample rate option was bugged 😭

[violet heron]

I think they hired you just to get back at me

[molten pecan]

They didn't hire me for my knowledge on rvc though

It was for other stuff

[opal depot]

https://tenor.com/view/misinformation-me-purposefully-spreading-misinformation-on-the-internet-teddy-bear-typing-misinformation-on-the-internet-gif-27679617

Julia

/joke

[violet heron]

You know that most of the guides on the website skip over stuff and or just don’t clear out everything

[molten pecan]

I didn't read it from the docs website though

But the one by @tranquil cliff

[orchid aurora]

Ok there is an important update. Looks like there has been a successful upload of the voice model but there is a new issue… I can seem to get the vocal file uploaded in it cause it isn’t highlighted when I try to upload it here

Do I need to do something and get 1 min clips? Wav files?

I can’t seem to get it to highlight when I try uploading it

Sorry I didn’t mean to upload that other pic lol

[violet heron]

On the files app drag the file

Close the app and open the website while still holding the file

Drop it on that box

Recorded this a month ago

[orchid aurora]

Niiiiice

Thanks lemme try

[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

[orchid aurora]

Got timed out

I’m guessing I have to let the screen unlocked at all times? Do I need to trim the vocals into shorter clips?

[violet heron]

Unlocked at all times sadly

[orchid aurora]

Got it

[tranquil cliff]

Hi?

40k is best

[violet heron]

Not at audios that math out at lower rates

[tranquil cliff]

In general the 40k is the best (I don't say it but Ilaria's guides etc ...)

[violet heron]

You know the whole point of audio analyzer right

(Even though a spectrogram can be achieved using basically anything)

To check a songs sample rate for training

[tranquil cliff]

Listen ... I give you access to the guide, modify it as you want

[forest sluice]

I simply just asked him whether he did the command, as it's literally in the guides

plus, he's doing it in a container anyway

[sand iris]

Frankenstein'ing python has never been a good idea

And it'll never be.. There's a reason why it's enforced

[forest sluice]

https://voidlinux.org/news/2023/10/python3-external.html yeah, should have adviced him to do these

[orchid aurora]

I am encountering new problems. The voice I got doesn’t sound anything like hank hill 😂 and I already tried 0, -6 and -12. Any tips on got to fix this issue?

I’m forever grateful for all the people helping me here
Shoutout to @molten pecan and @violet heron for helping the most ❤️

[forest sluice]

What's the model you're attempting to get?

[orchid aurora]

I got hank hill from ell0er user

From this user

[forest sluice]

Also, the source audio?

[orchid aurora]

You mean like the literal track or just an example of what it sounds like? ☺️

Its an mp3 file

[plucky steppe]

that means the model is bad

[rare goblet]

Ayo? @plucky steppe level 1 !!!

[orchid aurora]

It worked I just used the wrong one

[old oracle]

how many epochs should I do for a 30 min dataset?

[brittle wing]

why does half the voices sound robotic

[fluid topaz]

i'd try 750 first and then check if the graph can be still going down

[brittle wing]

unusable xDDD

2560 ms on 960

xD

any fixes

[nocturne otter]

[Voice Changer] VoiceChangerManager initializing... done.
[Voice Changer] MMVC_Rest initializing...
[Voice Changer] MMVC_Rest initializing... done.
[Voice Changer] MMVC_SocketIOApp initializing...
[Voice Changer] MMVC_SocketIOApp initializing... done.
[VCClient] wait web server... done 200
[2024-01-08 10:02:48] connet sid : 3kPOB7sSTq1tUDeQAAAB
[2024-01-08 10:02:48] connet sid : BFaoAB5TAk-waNq9AAAD
Generated Strengths: for prev:(4096,), for cur:(4096,)
[Voice Changer] Pipeline is not initialized.
[Voice Changer] Waiting generate pipeline...
[Voice Changer] Pipeline is not initialized.
[Voice Changer] Waiting generate pipeline...
[Voice Changer] Pipeline is not initialized.
[Voice Changer] Waiting generate pipeline...
[Voice Changer] Pipeline is not initialized.
[Voice Changer] Waiting generate pipeline...
[Voice Changer] Pipeline is not initialized.
[Voice Changer] Waiting generate pipeline...
[Voice Changer] Pipeline is not initialized.
[Voice Changer] Waiting generate pipeline...
[Voice Changer] Pipeline is not initialized.
[Voice Changer] Waiting generate pipeline...
[Voice Changer] Pipeline is not initialized.
[Voice Changer] Waiting generate pipeline...
[Voice Changer] Pipeline is not initialized.
[Voice Changer] Waiting generate pipeline...
[Voice Changer] Pipeline is not initialized.
[Voice Changer] Waiting generate pipeline...

[pure scroll]

FIX 1: The path to the file might be too long. Try re-extracting the folder on the root of your main drive (C:\MMVCServerSIO) or directly on your desktop.

FIX 2: Try closing the app, deleting the “pretrain” and the “model_dir” folders and restarting it.

FIX 3: Make sure that your antivirus or any software of this kind isn’t interfering with the voice changer.

[forest sluice]

@sand iris Hmm, IMO we should advise people to install conda, because it's distro agnostic, and Arch isn't the only distro to package Python >3.10

[sand iris]

Probably ? Conda is pretty fat though

python310 (AUR) : 54.67 MiB
miniconda (AUR) : 550MiB

[daring nebula]

God your people probably hate people like me

[rare goblet]

Ayo? @daring nebula level 1 !!!

[daring nebula]

But...how do I dose it?

[umbral crest]

hi

[umbral crest]

wdym

[umbral crest]

how do I get level 1

[daring nebula]

Keep talking but not here go to general this is a place for asking for help

[jolly pendant]

I need level 1
I want level 1

[forest sluice]

Talk more

[jolly pendant]

I need level 1
I want level 1

I need level 1
I want level 1

I need level 1
I want level 1

Talk more

?

level

1

me

get

?

[forest sluice]

Don't spam please

Chat in #🧬│ai-chat like a normal person, or ask for tech support

[half cove]

[sand iris]

What do you mean ?

[daring nebula]

Just how do I make an ai cover

[sand iris]

oh

[daring nebula]

I know you're talking to a cave man here

But I sware my situation is interesting...

[sand iris]

Do you know which GPU your computer has ?

And do you want to do voice conversion on your machine or on a distant server (which doesn't require a big powerful computer)

[daring nebula]

I got a lap top

Also...

I made the voice

That's the fucky part

[sand iris]

Oh you have the voice model ready ?

[daring nebula]

Like full chromatic Gen with every detail of a bitcrushed vocal

Yes

I have his voice operating with and without a song in the back

I digitally Frankensteined a voice

Strictly for singing

@sand iris want dm to see what I mean?

[rare goblet]

Ayo? @daring nebula level 2 !!!

[daring nebula]

Les go

[nocturne otter]

nvm

found them

[median widget]

А я тут один русский?

[daring nebula]

I assume that's a no

[analog python]

bro got ghosted

[rare goblet]

Ayo? @analog python level 1 !!!

[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

[unique igloo]

hey any1 knows how to reduce the voice delay somehow? ik there must be this delay but i feel its just too long

[molten root]

hi 🙂
i followed the instructions to train my dataset but even the first step doesn't work. I don't have any " end pre-process" after pressing process data button ( i turn it locally )
So i suppose it's because it doens't find any audio file to work with, but in the path i entered , i putted a .wav audio file, so idk where's the problem

Any help please ? Thank you ^^

[sand iris]

sorry I'm back

[daring nebula]

I'm in bed but I can still send you the chromatic

Is that cool?

[sand iris]

I don't have audio gear at the moment

But you need to create the model now right ? You have your sample audio to train and you are wondering how to create a model from it from what I understood ?

[daring nebula]

Nah I'm in bed

And yes I'm wondering who I make a modle

Well Tommorow cuase I need sleep

[proven hill]

LAST HOURS E-GIRL VOICE MODEL SALES!!

DM ME FOR DISCOUNT PRICES!

E-Girl, Mommy, V-Tuber, Anime girl and much more!

Wondering what models?
You find them here: https://ko-fi.com/ilariaowo/shop

INCLUDING FREE ONES!! :swaygirl:

I also do commissions if you want aspecific voice!

I accept the following payment methods:

• Ko-Fi
• PayPal
• Steam Gift Cards
• Amazon Gift Cards
• Throne

Grab yours today!!!

[forest sluice]

robux?

[proven hill]

no robux

[plush compass]

What UVR option do you use when creating a dataset? Is there a site I can refer to?

[copper pebble]

https://discord.com/channels/1159260121998827560/1159588254220288121

[woven furnace]

what if it's not?

[half cove]

then other people can use it if its on ur profile

[woven furnace]

ok so if it's listed in a not private one they still can't use it?

[half cove]

doesnt really matter who tf goes on random peoples profiles and goes into ur spaces lmfao

[woven furnace]

so yes?

[half cove]

yes

[woven furnace]

they can or can't?

[half cove]

brother

it doesnt matter

nobody is going to search your name and go on it

[cinder bobcat]

Hey guys, I trained two voice models from different singers locally (50 epochs small Datasets just4fun) and I think this is really bugging. Both voice models sound weird and sound excatly the same (?) although they are two completely different singers. Never had this problem when I used colab

I appreciate any help. Is my GPU not good enough to train?

[plush compass]

When everyone is learning, which option do you use between 40k and 48?

[rare goblet]

Ayo? @plush compass level 6 !!!

[proud elbow]

more than 200 epochs would sound better (recommended to train until 500, then test and pick the best one)

[proud elbow]

depend on sample rate & frequency cutout seen in spectrogram (Spek or any DAW software), though mostly just go 40k

[plush compass]

I'm really sorry, but I have one more question... A frequency that is not in the original song is set to above 17khz. What is this? Did strange noise come in when removing noise?

[plush compass]

Are you saying that learning is conducted at 40k instead of 48k? So, will singers also proceed with 40k?

[proud elbow]

that's the cutout, so may better go 32k

[cinder bobcat]

I know but it sounds like a very strange npc voice. Tried another dataset from another artist and its exactly the same. Basically you would recognize the desired voice in 50 epochs but its very different

[proud elbow]

due to cutout below 20k

[plush compass]

In the song where the reverb was removed with UVR, it is not there, but after removing it with RX10, it is a shadow above 17KHZ.

[proud elbow]

make better quality dataset and try to get rid of mode collapses

[molten pecan]

The mode collapse is a thing from the TB, right?

[proud elbow]

yes

[molten pecan]

How does it affect the training exactly?

I thought with TB you only see and measure the training process

[proud elbow]

https://discord.com/channels/1159260121998827560/1186762171204505801 I followed this guide and no longer got collapses

also no longer got random noise in silence parts

[cinder bobcat]

im very sure that my datasets are decent (used UVR and audacity). Lemme show you 2 short audios from different singers. both 50 epochs and both sound the same.
When I tested a 50 epochs pth file from colab, you could hear some decent voice as desired

[proud elbow]

im not using pc to be able to see the spectrogram and listen through headphones, maybe later

[cinder bobcat]

Oh okay, no problem 🫡 would be happy to receive further advice

[proud elbow]

btw https://discord.com/channels/1159260121998827560/1186762171204505801 may show you better way to process the dataset

[half cove]

here is another one if you want https://rentry.org/RVC-dataset-RX

[plush compass]

Does everyone use de-hum when creating datasets?

[charred jewel]

Yo guys

Can you use rvc on a Macbook Air with an M2?

I think it's the most recent Macbook Air

[brittle wing]

Hello

[molten pecan]

Hai

Cute anime grill 300

[proper shale]

yeah

don't think you can train yet tho

only inference

[charred jewel]

Damn

[edgy tangle]

How i can train with gpu in applio?

[mossy burrow]

Someone have Dana white éve?

Rvc

[plush compass]

Does everyone normalize the dataset and start AI learning? Or do I leave it to rvc preprocessing?

[edgy tangle]

I get a RuntimeError when i use rmvpe_gpu in pitch extraction

[proper shale]

What GPU do you have

[proper shale]

You can leave it with preprocessing, but it should be fine if you decide to do it too

[edgy tangle]

rtx 3050

[proper shale]

what batch_size are you trying to train on

[edgy tangle]

2

i have 4gb vram

[proper shale]

oh

[edgy tangle]

i get an error on pitch extraction with gpu

[proper shale]

I don't mean to say anything but I think it's probably better if you use Colab

[edgy tangle]

RuntimeError: Expected tensor for argument #1 'input' to have the same type as tensor for argument #2 'weight'; but type torch.cuda.FloatTensor does not equal torch.cuda.HalfTensor (while checking arguments for cudnn_batch_norm)

[edgy tangle]

oh

[restive grove]

How can we write names on this channel? What I am going to write is for Turk Vicall and he says to me so that he cannot write names on this channel.

[proper shale]

...what do you mean by this? Do you want to search a model?

[edgy tangle]

well

ill try to use collab again

but

where can i get the link?

[proper shale]

-colab

[azure marsh]

☁️ 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-HFv2, by r3gm Huggingface Spaces
• AICoverGen, by r3gm Huggingface Spaces
• Advanced RVC Inference, by r3gm Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces

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

[edgy tangle]

ooh

thx

Which one do you recommend to start?

[proper shale]

easyGui is a good one to start but if you want more options (like configuring batch size n stuff) you can try RVC disconnected

[merry fulcrum]

is realtime voice changer with google colab still working? cuz i keep getting errors

[edgy tangle]

ooh

thx

ill try RVC disconnected then

[proper shale]

there's a guide for it in #1159513888199540817 in case something confuses you on it

[edgy tangle]

k

[proper shale]

if you're using the official one then no, it has been restricted by Google

You can try the modified one, it's in the pins of #🔍│help-w-okada

[merry fulcrum]

thanks, i'll try that

[edgy tangle]

omg

i finally can enter on colab

thxx

[proper shale]

you're welcome :)

[restive grove]

yes

But when I write it doesn't work

[proper shale]

#🔍│find-models, use the slash commands

[proper shale]

Then the model probably doesn't exist

[restive grove]

thank you

[rare goblet]

Ayo? @restive grove level 1 !!!

[opal depot]

anyone got a good disconnect prevention method for collab?

[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

[opal depot]

my dataset is about 14m in length. How many epoths should I use

[glad harness]

how do i use RVC v2 Disconnected

[half cove]

https://discord.com/channels/1159260121998827560/1160926856548200531

[fluid topaz]

this what ive been looking for!! tysm

[half cove]

[edgy tangle]

for inference is better crepe, mangio-crepe, rmvpe or rmvpe+ ?

i dont know

[half cove]

mangio crepe better than crepe and rmvpe+ better than rmvpe

[edgy tangle]

ooh

thx

so

its better mangio crepe, or rmvpe+?

[half cove]

most people use rmvpe+

[edgy tangle]

k

thx!!

[fluid zephyr]

Hey, I'm using "train" on RVC and for some odd reason I hit one click train twice and it won't run, or I guess it does run but no epochs have run

Started 5 hours ago

I was supposed to release a heavier model today but I guess tomorrow will have to do

[hybrid wave]

Hi what's the best local fork for me to run a rvc v2 (rvmpe) model?

[fluid zephyr]

I'm training a 20 minute set

I'm away from home but using Chrome remote desktop

Nvm, I just made the .index and trained the model separately instead of one click, already on epoch 3

[half cove]

dont use one click 😭

[hybrid wave]

does anyone know, or do I just choose a random one? 😅

[fluid zephyr]

Lmfao yeah, one click on my 8 min model on 250 epoch took a whole 5 hours, I did index separately now on a 20 min model, finished instantly, then now I'm training my model and it took 5 mins to reach epoch 12

Holy crap 1 clip is slow

[half cove]

i dont even know why its so bad

[fluid zephyr]

Click*

[fluid zephyr]

Lmfao yeah

[half cove]

i mean its just 2 buttons instead of 1

just click em lol

[fluid zephyr]

I was wondering the first time when I made my bandiddums model, I was like "huh, 5 hours for 8 mins of clips? Holy shit that's long"

I though 20 mins was gonna be 15 hours

It's going faster now that I clicked two buttons instead of one lmfao

Well, learned mah lesson

For this one I'm making a test on 250 epochs 20 min clips, and then tomorrow I'm going to do 450 epochs when I have more time

Rn I'm afk and using Chrome remote desktop to control like a big boy

[half cove]

[molten pecan]

mangio is great

[open axle]

Link to make inference with RVC v2

[proven hill]

LAST HOURS E-GIRL VOICE MODEL SALES!!

DM ME FOR DISCOUNT PRICES!

E-Girl, Mommy, V-Tuber, Anime girl and much more!

Wondering what models?
You find them here: https://ko-fi.com/ilariaowo/shop

INCLUDING FREE ONES!! :swaygirl:

I also do commissions if you want aspecific voice!

I accept the following payment methods:

• Ko-Fi
• PayPal
• Steam Gift Cards
• Amazon Gift Cards
• Throne

Grab yours today!!!

[brittle wing]

https://cdn.discordapp.com/attachments/1191852221344780368/1194019529408385085/image.png?ex=65aed457&is=659c5f57&hm=6e16009fcd6a37168cf83128521dff44945e083145e69934c9199c4a15c55028&

hi guys

when ever i use the paractice one is workking

but

the other rvc

isnt working

how do i fix this

help

guys please

[violet heron]

What’s your GPU?

[spiral osprey]

my disconnected rvc spoke again using 1157 epouchs it was unable to complete. They say I've already used the usage limit and it's been 24 hours, does anyone know how many epouchs I should use like what's recommended to make a voice without sounding robotic?

[polar siren]

anyone else's rvc working really weird? or is it just me- Im following a tutorial rlyl close but no matter what I do it's so fucking slow and just barely gets past 2 epoches

[spiral osprey]

what is recommended?

[polar siren]

you could just try doing it again but having the save state to every 50 epoches, that way you keep your progress if you need to go higher

unless I didn;t understand the question right, but im pretty sure that;s just the reccomended thing to do if you wanna keep working on a model and the thing crashes or breaks

[plush compass]

When creating a dataset, is there a reason to make it mono instead of stereo? I think I saw somewhere that it had to be made in mono.

[brittle wing]

its not invidia

[rare goblet]

Ayo? @brittle wing level 1 !!!

[violet heron]

AMD?

[proper shale]

Well, RVC only really uses mono and not stereo

[plush compass]

Then, should the singer's songs be converted from stereo to mono before learning?

[proper shale]

Not necessarily?

If you want to prevent errors then maybe

[plush compass]

You can find out exactly by comparing the stereo model and the mono model. Thank you for your reply.

[proper shale]

You're welcome :)

[opal depot]

@molten pecan doing it here because I can't send screencaps in #📑│making-models

[molten pecan]

That's weird. Is the HF repo public?

Or private?

[molten pecan]

When you named the model

Did you put any spaces in between?

[opal depot]

no

[molten pecan]

I think I see the problem

[molten pecan]

The .zip had a special character

í

Or am I wrong?

[opal depot]

that's my local zip

the one on hugging face does not

[molten pecan]

Maybe the zip file is the one causing problems

Try renaming the zip file

With no special characters

[hybrid wave]

Do I need to download Mangio-RVC-v23.7.0_INFER_TRAIN.7z if I just want to run a model?

[rare goblet]

Ayo? @hybrid wave level 1 !!!

[molten pecan]

For that matter download the one that only says "infer"

[hybrid wave]

great, thanks!

[molten pecan]

yw! Have fun

[molten pecan]

That'll work

In HF name the model to something different, like "test"

It sometimes bugs when you put 2 models with the same name

[violet heron]

Change blob to resolve

[opal depot]

that fixed it

[molten pecan]

What do they mean?

[violet heron]

“In the context of web development, a "blob" typically refers to a Binary Large OBject, which is used to store binary data. On the other hand, "resolve" in a link often pertains to the process of determining the actual location or URL of a linked resource.“

[molten pecan]

huh but why does the link sometimes show one thing or the other?

[violet heron]

People copy the links in different ways sometimes

[molten pecan]

they copy the url of the repo instead of the download link?

[opal depot]

god I hope this comes out good I spent 3 hours on it

[violet heron]

No

The repo would link to the file

[opal depot]

i'm probably gonna retrain my model

500 was a lot I feel

might try with 300 idk

do I just hit training again?

i still have the collab open

and will it overwrite the existing model

[hybrid wave]

Thanks! Also do you know what do I run in the folder? Do I need some setup or something? 😅

[molten pecan]

execute the load preprocessed dataset files cell

[opal depot]

ok cool

[molten pecan]

in Import Model from Drive to Notebook put 2333333

[opal depot]

do I install dependencies first cause I switched to chrome

[molten pecan]

yes, except Preprocessing and Feature Extraction

[molten pecan]

go-web.bat

it'll open a command tab

after a moment it will open your default browser with the app ready to use

:3

don't close the cmd tab or it will stop working

[hybrid wave]

alright thank you!

[molten pecan]

yw!

[crystal cloak]

need help with RVC
I am at step 3 but i dont get the link

the button becomes red and I dont get a link

[molten pecan]

which RVC?

[violet heron]

Which method?

[molten pecan]

rvc

[crystal cloak]

Easy GUI for RVC v2 with crepe with improved downloader

[violet heron]

That is very very very outdated

[crystal cloak]

ah damn

[opal depot]

both of them?

[crystal cloak]

which should i use then

[violet heron]

Are you trying to train or make covers

[crystal cloak]

covers

got the vocals split

just need to convert

[violet heron]

I would recommend you to use Ilaria RVC instead of colab for covers

[molten pecan]

introduce 2333333 and then execute the second one

[opal depot]

is this bad?

[violet heron]

https://rentry.org/ilarvc_inf_guide here is the guide for it

[crystal cloak]

https://huggingface.co/spaces/TheStinger/Ilaria_RVC this one?

[molten pecan]

the D and G files are in the rvcdisconnected folders, right?

[violet heron]

Yes

[molten pecan]

@proper shale can you help him out 😭 I need to finish my work

[opal depot]

wait nvm

i had my enviorment named wrong

💀

[crystal cloak]

sorry for ping
But do yall have some instructions that I can follow like the brain dead mf that i am?

[rare goblet]

Ayo? @crystal cloak level 1 !!!

[violet heron]

https://rentry.org/ilarvc_inf_guide

[opal depot]

Do i do -> load preprocessed dataset -> training

[ripe urchin]

@molten pecan can you help meh

im tryna find the voice mod

[opal depot]

or load preprocessed dataset -> load model -> training

[ripe urchin]

of patrick

im using the huggins thing

how do i find the voice mods for huggins

[violet heron]

#1175430844685484042 or search with the @shy zealot bot

[molten pecan]

weights.gg website or #1175430844685484042 . i'm busy right now

[violet heron]

Or weights.gg

[ripe urchin]

i went to weights

its showing thus

this*

what do i do

[violet heron]

Stupid weights.gg

[ripe urchin]

?

[violet heron]

For some reason weights reuploads all their models on under their own link

-upload ignore “Submitting the Model To AI Hub”

[azure marsh]

huggingface.co upload Guides

English guide

Link: https://rentry.org/fdg_guide_newer
Credits: FDG

Spanish Guide

Link: https://rentry.org/AI_Hub_guide_esp
Credits: Julia (ailen2091)

Hindi guide

Link: https://rentry.org/AI_hub_guide_hi
Credits: Enes

Turkish guide

Link: https://rentry.org/AI_hub_guide_tr
Credits: Enes

[ripe urchin]

wait huh

[violet heron]

Weights hosts a lot of their models under their own website

When HuggingFace doesn’t support it

[ripe urchin]

so what i do

[violet heron]

Follow the guide about to reupload the model to hugging face

[ripe urchin]

this?

[violet heron]

Ignore the “Submitting Model To AI Hub” Part

[violet heron]

Yes

[ripe urchin]

okie okie

thank chu

im confused on this part

can i get an explanation

[violet heron]

Ignore that

Download the model off weights

That’s the zip file

[ripe urchin]

okie okie

thank chuuu

[plush compass]

Is it absolutely necessary to leave 0.5 seconds of free space between data set sections? Then, can’t I use Audacity’s silent removal?

[ripe urchin]

i dont see this option

i pressed create

[violet heron]

You are editing the readme

Scroll down and Click cancel

[ripe urchin]

okie i finished that

so what now

[violet heron]

You copied the link?

[violet heron]

Now paste that link on the Ilaria rvc

[ripe urchin]

i uploaded the file

how do i copy the linl

link*

[violet heron]

Right click, copy link address

[ripe urchin]

the website link

[rare goblet]

Ayo? @ripe urchin level 3 !!!

[ripe urchin]

i copy that?

[violet heron]

The model download link

[violet heron]

This

[ripe urchin]

huh

[red vale]

can somebody help? the voice-changer is working for a very short period of time and then stopping.

Also I have this message in Colab: "You may be executing code that is disallowed which may terminate your runtime without warning. Colab prioritizes interactive notebook compute and disallows some types of usage when executing code without compute units as outlined in the FAQ."

And when it stops I get this "Your runtime has been disconnected due to executing code that is disallowed in our free of charge tier. Colab subsidizes millions of users and prioritizes interactive programming sessions while disallowing some types of usage as outlined in the FAQ. If you believe this message is in error, file an appeal. Please include any relevant context about your usage. To connect to a new runtime, click the connect button below."

[ripe urchin]

elaborate

what link i dont see a link

[copper frigate]

I dont see the optional google drive

[violet heron]

Which colab

It’s probably a outdated one

[violet heron]

On the model files download button

Right click it

Press “copy link” or “copy link address”

[crystal cloak]

yo is there some program that I can use instead of the site?

[violet heron]

-local

[crystal cloak]

the site takes way to long imo

[azure marsh]

Suggestions for @crystal cloak

All Working Local Forks 🖥️

• Applio-RVC-Fork, by IA Hispano GitHub
• Mangio-RVC-Fork, by Mangio621 Huggingface
• RVC Studio, by SayanoAI Huggingface
• AICoverGen, by SociallyIneptWeeb GitHub
• Replay, by Replay Team Website
• Original RVC, by the RVC-Project team GitHub

Need some help? 🤔

You can find more info on the #1159513888199540817 channel. If you can't find your answer, feel free to ask for help in #✨│ai-help. Credits to Faze Masta and Antasma for compiling these links.

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

[violet heron]

The voice changer or for covers?

[crystal cloak]

covers

[red vale]

https://colab.research.google.com/github/w-okada/voice-changer/blob/master/Realtime_Voice_Changer_on_Colab.ipynb

[violet heron]

That’s why

https://colab.research.google.com/github/hinabl/voice-changer-colab/blob/master/Hina_Modified_Realtime_Voice_Changer_on_Colab.ipynb

[red vale]

can you give me not outdated pls?

[ripe urchin]

it wont let me select a model

[violet heron]

Just did

[violet heron]

Press refresh

[red vale]

oh i see, sorry

[rare goblet]

Ayo? @red vale level 1 !!!

[ripe urchin]

i did

it still not coming up

[violet heron]

Did you duplicate the space?

And is the model public

[crystal cloak]

m8 which one do you recommend for covers?

[violet heron]

Orginal RVC or one of its forks like Applio

[ripe urchin]

idk how to know if its public or private

[violet heron]

Did you duplicate the space?

[ripe urchin]

wym duplicate

[red vale]

Thank you, btw 🙂

[ripe urchin]

i did this

and went to interface

and nothing comes up

[violet heron]

Try a different name

[ripe urchin]

idk

[violet heron]

3 dots on the top bar, duplicate this space

[half cove]

have you tried clicking refresh models?

oh i think you have

[ripe urchin]

yea

i did it

nothing comes up

its on private

[violet heron]

The model or the space

[ripe urchin]

huh

idk

the space

[violet heron]

If it’s the model, make it public

[ripe urchin]

how you make it public

[violet heron]

3 dots - settings

Scroll down

Make public

[ripe urchin]

its public

[violet heron]

Ok

Can you send a full screenshot of your Ilaria RVC

So I can make sure you duplicated the space

[ripe urchin]

wym

[violet heron]

Send a screenshot

A image

[ripe urchin]

can we hop inna call

[violet heron]

Can’t rn

[ripe urchin]

and i can show you

[violet heron]

Show me using a image

[ripe urchin]

what do i screenshot

[violet heron]

The website

[opal depot]

how is this rn

[ripe urchin]

i pressedit

its not showing anything

when are you available to call

because

like

this is mad confusing

ive been spamming refresh for the last 5 min

[violet heron]

Try redoing the download voice models part

[ripe urchin]

and nothing comes up

i did

its not letting me

[violet heron]

Without the “?download=true”

[ripe urchin]

because it says i have the file already

[violet heron]

And different name

[violet heron]

You don’t need to make a new model

[violet heron]

Try this again

Name the model something

[violet heron]

Remember no spaces

[ripe urchin]

i named it idek

[violet heron]

k

[ripe urchin]

what i press

[violet heron]

Rename the zip file

To have no spaces

Reupload it, commit changes

[ripe urchin]

like the file itstelf

[violet heron]

Yes

The model repository on HuggingFace shouldn’t have any spaces either

After copy the download link and try it again in ilaira RVC

Using a different name no spaces

[ripe urchin]

like this

what now

yeah

nothing still showing up

[violet heron]

Commit changes

[ripe urchin]

i did

[violet heron]

No you have to redownload it on ilaira RVC

[ripe urchin]

im spamming refresh

i did

[violet heron]

Make sure you named it something new

Without spaces

[ripe urchin]

i did

[violet heron]

And it’s not showing?

It should?!

Try a different model

If the other one works it’s the models fault

[ripe urchin]

okie okie

im doing that rn

[rare goblet]

Ayo? @ripe urchin level 4 !!!

[violet heron]

Should work

If it’s public and the model has no spaces

[light bay]

Is it possible to train ai models without having anything installed locally?

[ripe urchin]

okie

[crystal cloak]

how to use the voice models in applio?

[ripe urchin]

yuhhh it worked thank chuu

[half cove]

yes

google colab

#1159513888199540817

you can use local or ila rvc

local means

on your computer

ila rvc is on the cloud

then use ila rvc

[rare goblet]

Ayo? @brittle wing level 1 !!!

[half cove]

check on #1175430844685484042 or weights.gg

https://rentry.org/ilarvc_inf_guide

[half cove]

thats the guide for ila rvc

no

its free

https://rentry.org/ilarvc_inf_guide

[crystal cloak]

how can I make the voice cover sound less jittery

[light bay]

HOw do you do it?

[half cove]

https://discord.com/channels/1159260121998827560/1160926856548200531

[light bay]

Ty

[rare goblet]

Ayo? @light bay level 1 !!!

[half cove]

[copper frigate]

I been clicking on one of the vocie but it's not selecting

can someoen help me

[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

[spiral osprey]

how to save?

[fluid zephyr]

My model's "Ssss" seemed to glitch out sometimes and go brrrrr, checked all of the datasets and all SSSS are clean, Retrain with more epochs?

did a 20 min dataset with only 250 epochs which i thought was low considering my 8 min dataset version had the same value

This is the "sssss" only appears sometimes

wait.

I think it may just be the inputing audio file to be processed lmfao

Ill retrain with more epochs though to make it sound better

[opal depot]

How do i check if g and d are where they are suppose to be in tensorboard

[spiral osprey]

I think mine already saves but how do I continue from where I left off? Or how can I find out which epouchs it is in?

[opal depot]

can I use rvc disconnected with a cpu

[leaden ruin]

Is there a way to search for models made with 48000,768?

[crystal minnow]

Mine got disconnected cuz its a against the law

something like that

[opal depot]

did I do the disconnect prevention right?

[rare goblet]

Ayo? @opal depot level 5 !!!

[copper frigate]

I got a warning "WARNING: Ignoring invalid distribution -cipy (/usr/local/lib/python3.10/dist-packages)"

[rare goblet]

Ayo? @copper frigate level 1 !!!

[copper frigate]

What do i do?

[opal depot]

should I be worried

[molten pecan]

Use the d/total graph

Same with G

Not grad

[opal depot]

?

[molten pecan]

Yes

[opal depot]

yes as in I should be worried or yes that is what I should pay attention to

[molten pecan]

Yeah training is going well

[opal depot]

ok thank god

[young sparrow]

sorry if it's a common question, but what's considered the best rn for running rvc locally? was looking at the applio fork and replay since they both seem to be updated recently

i'm on a silicon mac but also have access to a windows desktop with WSL and gpu

[placid holly]

i'm still using MANGIO FORK RVC July 2023 release, Applio not reliable for locally to many error from doing update

[young sparrow]

ohhh really

i see ty

is that pretty common?

[potent geyser]

I'm sorry I'm a dummy ! Does rvc work on mac? I though it was supposed to but i Can't get any sound out of it. I have a Mac with M1 MAX chip - Monterey OS

[round bane]

does the file have to be .mp3 for the google collab page cus it isnt detecting my audio/vocal file even after refreshing im using .wav

[half cove]

.wav works on it

Have you zipped it properly?

[round bane]

do you have to zip the file?

[rare goblet]

Ayo? @round bane level 1 !!!

[half cove]

Have you read the guide?

[round bane]

i saw a tutorial on yt

they just put the file into the thing

[half cove]

What colab are you using

[round bane]

https://colab.research.google.com/drive/1Gj6UTf2gicndUW_tVheVhTXIIYpFTYc7?usp=sharing#scrollTo=7vh6vphDwO0b

[half cove]

They won't work on a free account anymore

You need to use disconnected

-colab

[azure marsh]

☁️ 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-HFv2, by r3gm Huggingface Spaces
• AICoverGen, by r3gm Huggingface Spaces
• Advanced RVC Inference, by r3gm Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces

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

[half cove]

^

[round bane]

ty

[static gyro]

does applio run for inferance?

no navidia and no amd

[proud elbow]

still possible with CPU-only mode

[unreal bramble]

how i can use Original Rvc ?

[rare goblet]

Ayo? @unreal bramble level 4 !!!

[unreal bramble]

idk how to use fr

[forest sluice]

You're on local?

[unreal bramble]

hey can u help me

in the webui it need Inferencing voice:

how i can add the voice?

[hollow dome]

sorry to interrupt, is there a way to convert an entire folder of files at once? like a batch inference

[rare goblet]

Ayo? @hollow dome level 1 !!!

[forest sluice]

It's similar to Applio/Mangio. You need source audio, either via TTS software, or audios you grabbed w/o noise

[forest sluice]

There is

[unreal bramble]

like hugging face link?

[forest sluice]

No, the original audio. Like, for example, if you want to make an AI Cover, you'll need to separate instrumental out of the vocal, then upload it to the UI

[molten pecan]

yeah! in mainline RVC there is a Batch tab

I don't know about Mangio

[molten pecan]

can you do inference in batches in Mangio?

[unreal bramble]

i already upload the sperate instrumental, should i DMs you?

[forest sluice]

Yes

[molten pecan]

cool

:3

[unreal bramble]

but how to upload it, link ?

bro it's so hard fr

[molten pecan]

@hollow dome your drawings are so cute you are so talented ogm 😭

if you need help or tips ping me !!

[hollow dome]

waah thank you!! 🥺

[molten pecan]

yw have fun!

[hollow dome]

I'm actually unfamiliar with mainline RVC, I really only use gradio (or DDSP lol), is it local?

[molten pecan]

oh so you have been using RVC via the cloud?

[hollow dome]

Yeah

[brittle wing]

where is the collab to make cover please ?

@molten pecan@hollow dome

[hollow dome]

You can find them in #1159513888199540817

[molten pecan]

yeah Mainline is local. Unfortunely I don't know of any cloud-based RVC that can do inference in batches 😭

Mainline is basically the base, unmodified and original version of RVC

its actual name is not Mainline, it was given that name by the community to properly distinguish it from the other forks

[hollow dome]

Ah I see!

Sad! But oh well I'll figure it out

Thank you 🫡

[molten pecan]

yw good luck!

[brittle wing]

hi... I'm trying to make a model of a character but disconnected rvc won't let me make an index file as it says there's no feature and the model sounds very weird I don't know why does this happen

i read somewhere datasets must be long... the thing is my dataset is big wile more than 1000 files but every file does not go further than 10 sec, most of them not even 5

[rare goblet]

Ayo? @brittle wing level 1 !!!

[forest sluice]

Did you make rvcDisconnected folder in your Google Drive, upload your dataset folder in .zip format to that folder?

Plus, your dataset should be formatted like this: filename.zip > Your Dataset Folder > Audios

[molten pecan]

"Audios"?

But you don't need multiple audios for a dataset nowadays, right?

[carmine osprey]

Just zip file

[latent cypress]

i want to try training a model on ov2, what is the ideal epoch for a 10 - 20 minute dataset?

[glad zealot]

Can use mainline on cloud just very annoying to setup for inference

[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

[digital lodge]

Does anyone know why does the voice is different from the original voice versus after training voice? 😦

[glad zealot]

You need to set the tune

[digital lodge]

Is tune same with transpose or nah? I'm using Mangio btw

[glad zealot]

Yup I think so

[plush compass]

Do you all create AI after normalizing audio? Is there a difference? I know normalizing is done during RVC preprocessing, but is it better to do it directly?

[hollow swift]

how i can dowloand models? when i search it goes kits ai

?

[proper shale]

use other model of the same character if possible

kits doesn't allow downloading their models

[hollow swift]

but all models language Turkish in kits 😦

[rare goblet]

Ayo? @hollow swift level 1 !!!

[proper shale]

Search on weights.gg

Or in #1175430844685484042 too

[hollow swift]

Ty so much but i didnt find anything

[proper shale]

Aw man

[hollow swift]

ah i fing weights.gg
from here

nice

[proper shale]

Yaaay

[hollow swift]

tyy

[proper shale]

You're welcome 🤗

[hollow swift]

https://tenor.com/view/hop-on-conan-exiles-get-on-conan-exiles-gif-17311358980559854050

[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

[hot lagoon]

in (Paperspace) Applio is there a way to infec ?

[molten pecan]

How? With paperspace?

[glad zealot]

kaggle :D

[brittle wing]

tried today making longer dataset files by joining them and now the feature was ok

now I'll infer and check how it sounds 🙂

[plush compass]

Should I use rmvpe or mangio-crepe? What are each other's strengths?

[elfin sparrow]

Hello, a quick question. I'm using voice lines extracted from a game as the dataset, which includes some non-dialogue sounds like laughter, shouts, etc.. Would including them be beneficial in any way or should I remove them and only use the dialogues?

[forest sluice]

rmvpe should be used, as:

• It's faster to train and infer
• Not too strict with noises (though you still need to clean it up)

[plush compass]

If the quality of the data you are using is good, it is correct to use rmvpe for other things like mangio_crepe, right?

[forest sluice]

You can just use rmvpe for infer and train. It's not dependent on mangio_crepe

[cobalt chasm]

Do I understand correctly, that RVC requires a GPU that supports Cuda?

[sand iris]

Yesn't

CUDA has full support

ROCm (AMD Radeon GPUs under linux) also has full support

IPEX (intel) has full support from what I have been told ? Haven't checked, and training requires a dedicated intel GPU

the rest is inference only

[dry mulch]

ive downloaded a voice model but why it wont load in rvc? like theyre not there but ive download it

[cobalt chasm]

ah ok. I know i'm in desparate need to update the old laptop. Just one more reason. Still using a 1080ti. No CUDA support till 3000 series.

[sand iris]

Oh that does support CUDA

[cobalt chasm]

not according to their support page.

[sand iris]

Pretty much all GTX nvidia cards support it, but only GTX 1xxx cards and above support a recent enough version

[cobalt chasm]

hmm. well, I suppose it's worth a try then.

[sand iris]

https://developer.nvidia.com/cuda-gpus

that number is unrelated to the cuda version

[sand iris]

What do you mean ?

[cobalt chasm]

I do see it under standard cards, but not under mobile.

[rare goblet]

Ayo? @cobalt chasm level 1 !!!

[sand iris]

Is there a 1080 ti mobile ?

[cobalt chasm]

oh you are probably right. Not on it at the moment, but it's probably just a 1080

[molten root]

hi 🙂
i followed the instructions to train my dataset but even the first step doesn't work. I don't have any " end pre-process" after pressing process data button ( i turn it locally )
So i suppose it's because it doens't find any audio file to work with, but in the path i entered , i putted a .wav audio file, so idk where's the problem

Any help please ? Thank you ^^

[cunning oar]

Hello, How can I set it so that I only hear the voice of the voice changer and no longer my own?

[proud elbow]

make sure the dataset path and wav file names have no spaces and special characters, also try re-export the wav files using Audacity if they might seem corrupted

[molten root]

I'm using Vegas Pro to export my wave file, I don't think it's corrupted, I'll look at the name.

D:\application\AIVoiceModel\Mangio-RVC-v23.7.0_INFER_TRAIN\Mangio-RVC-v23.7.0\datasets

It looks like there's nothing special about the characters, but the names entered are the default names when mangio rvc is installed, so I don't know if changing them will cause any problems. For example, changing the "-" to " _" or removing the ". " . But maybe in the downloaded files, there are specific paths to the default names, so I'm not sure it's a good idea to change these default folder names.

[hybrid wave]

Hi I have a question, when I use an online cloud fork on huggingface, what do I enter in the "Interfaring voice" field?

when I put a drive link it just disappears

[proud elbow]

I have mangio app, though less used than original RVC, but it worked despite having "-" and "." in its path

[woven umbra]

i want voice model download

[proud elbow]

the special characters I mean may be some asian or unusual unicode characters

[hybrid wave]

does anyone know?

[molten root]

I don't have one, nor in my .wav file, so I don't understand. When I press "process data", I see that it starts to load, then I don't know if it goes all the way, but everything reappears as if nothing had been done, and no message appears.

[rare goblet]

Ayo? @molten root level 1 !!!

[proud elbow]

I had no problems on wav files exported by Audacity, unless they have such special characters

[molten root]

ok, i don't have any special character, so ... what now ?

[proud elbow]

can you anyway do feature extraction and the next steps?

[molten root]

same on feature extraction

it begins to load then everything reappears as if nothing had been done

same when train feature index

and when i click on Train, " error " appears

[proud elbow]

do you also have the same issue on another like original RVC or applio?

[molten root]

?

what's that ?

[proud elbow]

-local

[azure marsh]

All Working Local Forks 🖥️

• Applio-RVC-Fork, by IA Hispano GitHub
• Mangio-RVC-Fork, by Mangio621 Huggingface
• RVC Studio, by SayanoAI Huggingface
• AICoverGen, by SociallyIneptWeeb GitHub
• Replay, by Replay Team Website
• Original RVC, by the RVC-Project team GitHub

Need some help? 🤔

You can find more info on the #1159513888199540817 channel. If you can't find your answer, feel free to ask for help in #✨│ai-help. Credits to Faze Masta and Antasma for compiling these links.

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

[molten root]

Do I really have to download RVC studio and Applio RVC fork to test?

the help documents said to use mangio, didn't they?

[proud elbow]

either the first or the last one like I mentioned before

[proud elbow]

^

[tropic tartan]

whenever I try to run tbL to launch tensorboard is says this

'tensorboard' is not recognized as an internal or external command, operable program or batch file.

what should I do?

[eager pelican]

@half cove welcome to the abusing connections to people you know so that you get help just cause you are too lazy to figure it out yourself

https://huggingface.co/ORVC
which one do i install
OUltimate
Snowie_RuPretrain_40k
Ov2Super
or Snowie_RuPretrain_48k

[half cove]

thats the russian one i think

leme check

[eager pelican]

uwu

[half cove]

https://huggingface.co/ORVC/Ov2Super/tree/main

[eager pelican]

so i just install and rename it to the default pretrains

like f0D40k.pth

since the default is
f0Ov2Super40kD.pth

[half cove]

u sure u need to rename it?

i dont think so

[eager pelican]

thats why im asking sir uwu

[half cove]

i think you just put the path of it

let me get screenshots

[eager pelican]

since if there were two 40k models
it'd prob conflict

[half cove]

no

[eager pelican]

how do i make sure it doesnt use the default and uses OV2 then

[half cove]

ill sends screenshot gime sec

[eager pelican]

reasonable

[half cove]

put the pretrains in petrainedv2 then put the name ig

[eager pelican]

so i just change the names from tehre

reasonable

so i can prob just edit the infer-web.py script for it

[half cove]

why do that? 😭

just put it in your folder and put the name

[eager pelican]

cause why bother changing the name every time
when you can have it automated

[eager pelican]

anyways
thanks uwu

[tropic tartan]

how can I run tensorboard?

I dragged both the TBI and tbL files into my RVC and tensorboard wont launch

[glad zealot]

just-

[tropic tartan]

TBI seems to install well but tbL wont work

[red vale]

Is it possible to record changed voice by OBS?

So that I can't hear it but OBS can

[placid holly]

yes

just do that

[sand iris]

I'd suggest converting voice afterwards tbh

rather than doing realtime

unless you stream

just make sure to have your mic's audio in a separate channel to avoid having to isolate the audio manually

[red vale]

sorry for asking but how to convert voice afterwards? I'm using ngrok version btw

[sand iris]

you'd use RVC instead for that

But I'm assuming you don't need modified voice for anything but the recording

[red vale]

Not-realtime changing is probably what I want because realtime lags enormously. Do I need to install some software for that?

My question is probably stupid, but I'm only second day here, do not understand everything yet

[sand iris]

Oh no it ain't, don't worry

RVC is the "main" software associated with the voice models available here

That's what people train new models on, and also make covers with. There's also colab notebooks available for it

[edgy tangle]

How fast is Tesla p4 graphic card for training

[sand iris]

pretty darn good

only beaten by the 4070 and above iirc

[plush compass]

Use this option if you wish to only copy over weights. What option does this do?

[proper shale]

Only copies the usable PTH over to the folder of your model

[plush compass]

thx

[tropic tartan]

would there be an issue if the audio lines I used for training the model only reached up to 15 kHz instead of 16 kHz?

[rare goblet]

Ayo? @tropic tartan level 1 !!!

[proper shale]

eh it should be fine at 32k

[dawn nebula]

ERROR: [Errno 10048] error while attempting to bind on address ('0.0.0.0', 7897): only one usage of each socket address (protocol/network address/port) is normally permitted

wtf

[half cove]

that only happens to me when i have rvc open multiple times

[dawn nebula]

never mind i have 4 background rvc opened

[half cove]

[dawn nebula]

💀

[upper cape]

@molten pecan

it's telling me that it's building up

so now i wait?

until it charges?

[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

[molten pecan]

si. como dije en la guía, esperas unos minutos a que cargue

[upper cape]

okok, grax ✨

oh it already loaded up lol

[lapis hound]

Hello, i have installed the Voice changer client demo. but i cannot choose the f0 det. to crepe. Any help?

[proper shale]

...you're using the AMD version

and crepe doesn't reeeeally use AMD GPUs and resort to using the CPU

[lapis hound]

can i get a link for the right one please 😄

i am with an 1650s

[proper shale]

(plus, rvmpe is just plain better for realtime)

[lapis hound]

not an amd gpu

[proper shale]

https://huggingface.co/wok000/vcclient000/resolve/main/MMVCServerSIO_win_onnxgpu-cuda_v.1.5.3.15.zip

[lapis hound]

W

thanks

[proper shale]

you're welcome :)

[lapis hound]

and for my graphics card you recommend rvmpe?

[proper shale]

hmmm either rvmpe or crepe_tiny will work fine

[lapis hound]

or my card is outdated asf 😄

[proper shale]

however ive heard that RVMPE needs more VRAM than the 1650 has sooo

maybe crepe_tiny might be the best option

[lapis hound]

yeah i have 4gbs

[rare goblet]

Ayo? @lapis hound level 1 !!!

[proper shale]

also 16 series cards have some issues with the voice changer for some reason, hope they dont affect you

[lapis hound]

hope so

i tried when i talk there is a small background noise, but that is from my microphone that catches sound ig

[proper shale]

u can use sup2 to supress those sounds

and increase s thresh which is just mic sensitivity

[tropic tartan]

when I look at my trained voice model in TensorBoard there are no lines. there are many small orange triangles way at the bottom. is that normal?

[proper shale]

are you looking at loss/g/total?

thats probably grad norm 😭

[tropic tartan]

yeah

how do I change it from triangles to lines

[proper shale]

are you in the scalars tab

[rare goblet]

Ayo? @proper shale level 70 !!!

[tropic tartan]

yeah

[proper shale]

actually it'd be better if you send a screenshot cuz idk

i feel like you're not on the right graph

[proud elbow]

nan values in console window, and using mangio?

[tropic tartan]

yes I am using mangio

[proper shale]

oh damn ok this is weid

(even though it's d/total)

[proud elbow]

probably such issue on GTX card, and outdated driver

[tropic tartan]

Ill go update my card and see if it changes

[bronze falcon]

I am very new to this, but if I want to do a parody of a song where I just change a couple of words from the lyrics while keeping the original voice, what would be the guide to follow?

[proper shale]

You need to sing the new lyrics and convert them with the voice model of the original singer

[plush compass]

Is there any article that explains the pitch_extraction_algorithm setting?

Settings like rmvpe crepe mangio-crepe harvest.

[molten pecan]

harvest is obsolete

rmvpe is considered the best one. both for speed and quality

mangio-crepe has its perks too. in my experience is more detailed and you can change its hop_length, which determines the time it takes the voice to jump from note to note (useful for models/audio samples with dramatic note shifts).
only downside is, it's more prone to artifacting. so that's why rmvpe is a lil better

[delicate wadi]

i uploaded a model and cannot hear anything, my cmd says it waiting to generate pipeline then saying it cannot be intialized

[molten pecan]

the rest of the algorithms are basically obsolete since there are better alternatives. the two explained before

[molten pecan]

#✨│ai-help

oh shit nvm

[delicate wadi]

do you know how to fix it

you're good

[molten pecan]

I thought this was gen chat 😭

[molten pecan]

are you using w-okada?

[plush compass]

You are truly a god... Thank you so much for your reply.

[delicate wadi]

I have the local one on my desktop the voice changer client demo

[rare goblet]

Ayo? @delicate wadi level 1 !!!

[tame mica]

[molten pecan]

then try asking on #🔍│help-w-okada

[delicate wadi]

ok thank you

[tame mica]

this deserves to be in starboards

[molten pecan]

😭 i'm used to people asking for help in gen chat

[dawn nebula]

only a 3090 wtf

sth going wrong

[tame mica]

isnt that like 1 epoch per sec

[dawn nebula]

yeah

no 2 epoch per sec

[young sparrow]

sry replying to something kind old, but what's the one-liner fix you need to make for the mainline install script? am on m1 mac

[sand iris]

Depends on if you have python installed already or not ! And especially the version

[young sparrow]

i think my default python version is 3.11 (kinda annoying sometimes) but i also have conda/know how to use it

[rare goblet]

Ayo? @young sparrow level 1 !!!

[sand iris]

Oh, you could go for a manual setup instead then

Not sure if conda messes much with python versions, but the fix I had in mind was just changing the line with python -m venv to use a python version strictly below 3.11

See if you can install 3.10 via conda and give the path to python3.10 in the install script

(or straight up set up a conda env for it which might be better honestly)

[young sparrow]

got it, yeah think will just try with a conda env, thanks 🙏

[sand iris]

Welp I suppose you know how this goes then 👍

just take a look at the tools/dlmodels.sh script in the RVC code, it will download all the pretrained models which are required to run RVC

[lapis hound]

how can i put the voice changer in discord ?

like i want the ai generated voice to go through discord

[opal depot]

did the disconnect method work?

[karmic heron]

you need to download vbcable driver

https://vb-audio.com/Cable/

[prime ermine]

where can i download Tensor board ? link is gone

[molten pecan]

Check the latest pinned message here

tbi is the installer. Execute it anywhere in your computer

tbl is the launcher. You must move it inside RVC's folder and execute it from there

[prime ermine]

ok thank you 👍

The command “tensorboard” is either misspelled or could not be found.

but all requirements are met by the tbi.bat

[rare goblet]

Ayo? @prime ermine level 1 !!!

[gray token]

i got so much res ms on custom voices

how can i fix that

[narrow dove]

Im slightly confused, after training my model in Applio is this how it should look like, if Yes then what file do i need to upload?

[prime ermine]

fixed it by using different path

[proper shale]

The usable PTH is in the weights folder

[placid holly]

so it long since i do train again and i forgot this

reason why choose Mangio or RMVPE for training based on?

[ionic roost]

I got a weird bug

I can't see the save frequency

[molten pecan]

can you type it out?

[ionic roost]

I think

typing changes the length of the slider

[molten pecan]

ah great

[ionic roost]

Would 25 be the middle?

[molten pecan]

I think this one is 10

@half cove was 10 the default save frequency in Mangio?

[half cove]

let me check

[molten pecan]

reinstalling mangio again?

😭😭😭😭

[half cove]

its 5

[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

[molten pecan]

is there a juice wrld model

[half cove]

Prob

[fluid zephyr]

Train index file isnt making a .index and i need one

[proper shale]

How are the logs/output of the train index step?

Does it give a Traceback

[fluid zephyr]

nothing

just absolutely nothing

which is super odd considering i used it like 20 mins ago

and then i made a new model and poof, nothing

[fluid zephyr]

nothing in CMD, nothing in the output info, just straight "..."

[proper shale]

...it feels dumb saying this, but

have you tried a restart

[fluid zephyr]

I have

5 times

[proper shale]

😭

[fluid zephyr]

to no avail

it just doesnt like my input file ig

💀

it worked for EVERYTHING else BUT that

yk what lemme try a different output

input**

if it works it just racist to that one file

[proper shale]

what type of audio file is that😭

[fluid zephyr]

a .mp3

and a .wav

tried both

NOTHING

please

H E L P

.....

A different file worked

its just racist