✨│ai-help

[static gyro]

also is looking for a zip i believe

[upbeat ice]

yeah, that's pretty much a pain in the ass from that

and i guess if it's not a zip file then i won't be able to get it at all right?

[static gyro]

i wonder if cloud convert can make .zip files

[upbeat ice]

hmmm that can be a good idea

[static gyro]

https://www.ezyzip.com

i test in kits.ai but every one is different. you can also use huggingface

[rare goblet]

@static gyro has reached level 2.

[upbeat ice]

right, I've downloaded the zip file made by the ezyzip, what do i do next?

[static gyro]

uplad it to hugging face some how?

[plush tree]

I need help guys, when i generate an ai cover on huggingface, it takes too long, how to fix it?

[static gyro]

get a human upgrade on patience 😎

[upbeat ice]

wow, it actually worked

[plush tree]

How?

[static gyro]

sorry thats beyond my skill set

[plush tree]

Oh

[echo creek]

I trained with Disconnected, but where can I obtain the .index file?

[static gyro]

press the button to train index

[upbeat ice]

where do i delete voice models on aicovergen?

[modern snow]

Welp I listened to the model and it doesn't sound like what I thought was going to. Makes sense as I had only included one wav file for reference when training. Has anyone played scp containment breach? You know the announcements you hear when encountering mtf? Well I was trying to make a model for that voice. And it does not sound like it. Firstly the cool effects are completely gone, secondly it sounds monotone and boring, and 3... I think he is British for some reason. Next time I am going to build a zip file containing(see what I did there) the entire voice lines for him. And hopefully that will get me better results. And @quiet haven thanks for the help!

[quiet haven]

Your welcome

[echo creek]

thank you😂

[reef bison]

anyone running into the issue of rvc gui this system cannot find the path specified for the rvc gui app?

[bold ermine]

A stupid question guys 😅 but can you tell me where can i put pretrained model file in so it become available in inference tab

[rare goblet]

@bold ermine has reached level 2.

[remote grove]

I paid for colab and finished learning quickly.
If you are suffering from hours of wasted work, please deposit a little into colab and enjoy heaven.

[rare goblet]

@remote grove has reached level 1.

[echo creek]

I heard i can prevent overfitting by referencing the graph. But how can I use the graph to determine when to stop?

[glad zealot]

Go to the scalars on the tabs of the tensorboard

On the search bar search for g/total

The line should be going down

When it starts going up way too high that's when you stop

I may be wrong

[sterile onyx]

How to fix "'NoneType' object has no attribute 'tobytes'"

[echo creek]

Ah, if the graph keeps going down and then starts going up, that's when I should stop, right?

[rare goblet]

@echo creek has reached level 1.

[glad zealot]

There's should be a slider on the left

Set that to the max (0.999)

Then you watch that graph

[sterile onyx]

@glad zealot

[echo creek]

Okay, thank you.

[glad zealot]

No idea

[sterile onyx]

local

[glad zealot]

Yup no idea

I made my models online

[sterile onyx]

but you can't use colab now right

[rare goblet]

@sterile onyx has reached level 1.

[glad zealot]

You can use rvcDisconnected

-colab

[azure marsh]

Google Colabs

RVC

• AI-Cover-Gen-No-UI (inference only)
• RVC v2 Disconnected (training only)

🌐 Translated colabs

• PT-BR: -colabs_br

Audio Separation/Isolation

• MDX Net Colab
• UVR arch colab 5.0.2/c1.04
• KaraFan colab

[sterile onyx]

training only?

[glad zealot]

Ya

Specifically training only

[sterile onyx]

i'm making a ai cover i dont need to train

[glad zealot]

Ahh you can also

Lemme get it

It's on pins

[proud elbow]

elaborate, maybe missing input audio, model file, corrupt/invalid files?

[sterile onyx]

seems fine

[glad zealot]

This one if you wanna make covergen with ui

Using colab

[reef bison]

hey so im new to this, does anyone have a basic guide on how to use the ai cover rvc link in pinned?

[glad zealot]

With that one you can just login to your Google account

Then press the circles with the arrow inside

From top to bottom

[reef bison]

oh bet, thanks

[rare goblet]

@reef bison has reached level 1.

[glad zealot]

Then at the bottom wait for it to finish and it will give you another link that you can use to make covers

The stuff there are self explanatory and you can just follow what it says

Also make sure to check #1159513888199540817 for additional stuff

[sterile onyx]

ok

[reef bison]

i just click on all the circle with the arrow inside and i dont need to insert in my ai model or seperated vocals yet right

[glad zealot]

Yup you will do that later when it gives you the link

[reef bison]

okay lit thanks for the help

[rose kindle]

dataset of 15 minutes and passed through uvr, is it good?

[sterile onyx]

@glad zealot what the hell

[glad zealot]

You need to import a zip not pth

[sterile onyx]

oh bruh

[glad zealot]

Wtf

Ye not sure

You are uploading the zip file and adding the name too right?

[sterile onyx]

no i drag it in "RVC models" folder and it sees it

[rare goblet]

@sterile onyx has reached level 2.

[glad zealot]

Ah that's why

Wait actually

???

On the gui theres an upload tab right?

You know what..

@reef bison you made yours work so... XD

[reef bison]

whats up?

[sterile onyx]

@reef bison do you know how to fix this?

oh

[glad zealot]

My phone didn't like that XD

[sterile onyx]

why is it so complicated

[reef bison]

where did u get ur rvc file

[glad zealot]

On the top there's something called "upload"

[sterile onyx]

made it

[sterile onyx]

yeah i got it

[reef bison]

looking at the error it looks like you didnt make a directory for some of ur files maybe

[glad zealot]

That's where you drag your zip file then add the name of your model

[reef bison]

download a random voice models and see the files that they have #1159289700490694666

try to mirror it

[sterile onyx]

yes thanks

[glad zealot]

@reef bison this is what he was doing

[sterile onyx]

goblin you helped me

[reef bison]

ahhh

my bad lol

yeah go to the upload section

[sterile onyx]

it works

now

[glad zealot]

Can't remember take screenshots of my own atm since I'm on ma phone

[lime ferry]

what is happening here?

Connecting to models.weights.gg (models.weights.gg)|172.66.43.45|:443... connected.
HTTP request sent, awaiting response... 403 Forbidden
2023-10-11 08:52:08 ERROR 403: Forbidden.

i could use it yesterday with no problem

[glad zealot]

What that for?

[lime ferry]

downloading model using rvc v2 huggingface version

was told to use models linked from weights.gg

[sterile onyx]

MAN THIS GRADIO STUFF SOUNDS SO GOOD

[glad zealot]

It's mostly banned on colab tho

[glad zealot]

Ahh not sure then

I have a mental list of the models I use that I just search on hugging face manually

[lime ferry]

it's fine, i figured it out - using gdrive works as well

[rare goblet]

@lime ferry has reached level 1.

[meager quarry]

I can still not train on RVC Disconnected V2 Colab due to "error: credential propagation was unsuccessful" when trying step "Mount Drive"

[glad zealot]

Are you using the latest version?

[meager quarry]

The 0.21 version?

[glad zealot]

Idk what version it is

The one in #1159949278270193734

[meager quarry]

just checked, it's the same one

[rare goblet]

@meager quarry has reached level 4.

[glad zealot]

If your drive full or something?

Oh wait are you trying to connect a drive from a different account than what's being used in the colab?

[meager quarry]

nope, plenty of space there

[meager quarry]

What do you mean by that? I am trying to connect it with a Google account

[glad zealot]

Ye but are you trying to connect it to a different google account than what is being used to run the colab?

[meager quarry]

No. This error happens when I click on the prompt that asks me to connect to Drive. So I can't even choose what account I'd want to connect in the first place.

[glad zealot]

Ahhh

Also there's no account selection anyways.. it just shows 1 account so you can't even select a different one

But ye not sure why it doesn't work

Try a different acc?

[meager quarry]

Yeah the error happens right as that menu pops up

or is supposed to show up

[glad zealot]

Nope it should shows you the account selection screen

But only 1 acc shows up

SO WHY EVEN GIVE A SELECTION bruh google

[meager quarry]

same error

[glad zealot]

Ye not sure

But it's the same link right?

[meager quarry]

yes

[glad zealot]

That's the one in info is using

[meager quarry]

yep, same one

[glad zealot]

Ye idk then

Not sure what the problem is

[lament eagle]

Anyone know what does target sample rate do in training

[glad zealot]

Not me

[meager quarry]

when I reload the runtime the error changes to "fatal: destination path 'Mangio-RVC-Tweaks' already exists and is not an empty directory"

[glad zealot]

Ye not sure what's going on there

It's not supposed to even be RVC-Tweaks

It should get renamed

[meager quarry]

that happens during step "Clone Repositiories' instead

[glad zealot]

You deleted the runtime?

Try fresh start

[meager quarry]

ao ctrl + m

*so

[glad zealot]

No no

Like fully stop everything

There's an down arrow on the top righr

There should be something called "stop and delete runtime"

[meager quarry]

trying the error just changed to "error: credential propagation was unsuccessful" on step "Mount Drive" again

[glad zealot]

Ye idk

There's no popup right?

Asking "you want to connect to Google drive" or something

[meager quarry]

That popup appears, but confirming it doesn't bring me to account selection but instead results in that error

[glad zealot]

Ye idk then sorry

That error haven't happend to me ye so not really sure how to fix it

[meager quarry]

Does kitlemonfoot at least know that errors like these have been happening to some people

[glad zealot]

I guess so since they made it

[dull raptor]

it doesnt let me login

help

[loud berry]

how many epochs recommended for 10 min dataset?

[brittle wing]

500-350

[rare goblet]

@brittle wing has reached level 4.

[brittle wing]

anyone knows how to fix this?

Traceback (most recent call last):
File "/Users/kai/Desktop/Applio-RVC-Fork-main/Applio-RVC-Fork/infer-web.py", line 16, in <module>
import ffmpeg as ffmpeg
ModuleNotFoundError: No module named 'ffmpeg'

i already did install ffmpeg hundred times.. it doesn't fix it

[bold ermine]

A stupid question guys 😅 but can you tell me where can i put pretrained model file in so it become available in inference tab

And also a pretrained file has 3 files g_ , d_ and modelname.pth

So where should I place all three

[proper shale]

even ffmpy?

[upbeat ice]

anyone?

[proper shale]

Delete the .pth in weights and the logs folder with ur models name

[rare goblet]

@cyan crane has reached level 1.

[upbeat ice]

weights and logs folder, where do i find that?

[proper shale]

inside the AICoverGen foldee

[proper shale]

and btw, i should've said

"the folder inside logs"

so dont delete the logs folder, delete the folder inside logs

💀

[upbeat ice]

i see, welp I'm trying to find it

[rare goblet]

@upbeat ice has reached level 3.

[proper shale]

oh

well uh

you can't really delete a model there

[upbeat ice]

oh

[proper shale]

yeah, sorry

[upbeat ice]

dang

[cyan crane]

After I downloaded it and click refresh voice model my voice not there

[proper shale]

...did you even click download

[cyan crane]

Yes

[proper shale]

i hh

dont think you have

output wouldve said something

but anyway

after u click refresh on generate, click on the dropdown for model

ur model should be there

[proud elbow]

login using your discord account

[stark grove]

HUH?!

this shit is a scam

[ionic pilot]

how many epoch

[rare goblet]

@ionic pilot has reached level 3.

[proud elbow]

find yourmodel_eXX_s12550.pth in the weights folder, e is the epoch

[meager quarry]

are you using rvc disconnected v2 colab?

[stark grove]

is covergen

oops

[meager quarry]

then nvm

I hope the colab is fixed soon so there is no more of this "mount drive" credential login fail error shit

[stark grove]

oh

[plush valley]

how many minutes of audio do I need from someone to train an AI from them?

[glad zealot]

at least 5 of clear audio i guess

[plush valley]

would 30 minutes of clear audio make a good AI?

[bold ermine]

Don't know why I am the only not getting reply here asked many times 😔

[glad zealot]

yup

[plush valley]

or is there a loading limit?

[glad zealot]

not sure

i personally never even used a 10 mins audio

since im too lazy to make datasets

[rare goblet]

@brittle wing has reached level 1.

[meager quarry]

yeah, it can take quite some time to make one

[glad zealot]

thats why i save every time i do anything

cause the thing i use crashes very often

[viscid moss]

That collab was banned today, we fixed it, it should work fine now
https://colab.research.google.com/drive/1qezaBvCn4eSG1U2L4a64tEHTynbIu8eN?usp=sharing

Test it guys

[glad zealot]

the webui covergen might still work

lemme test that

[viscid moss]

the original has a warning

let me show u

-colab

[azure marsh]

Google Colabs

RVC

• AI-Cover-Gen-No-UI (inference only)
• RVC v2 Disconnected (training only)

🌐 Translated colabs

• PT-BR: -colabs_br

Audio Separation/Isolation

• MDX Net Colab
• UVR arch colab 5.0.2/c1.04
• KaraFan colab

[glad zealot]

no the web ui version i mean

[glad zealot]

thats the original web ui right?

[viscid moss]

This AI Cover gen

[glad zealot]

ahhh

theres also this

[viscid moss]

I will talk to RayTracer to modify it for my collab

[glad zealot]

a working(?) version of the webui

made it like 2 weeks ago

[viscid moss]

Yesterday a person told me that it was disconnected but I tried it for 40 minutes and it works

I think it still works

[glad zealot]

most people are using the original still

it looks the same at first glance on the colab so ig thats why they dont know

[viscid moss]

Yes, but it disconnects in 5 minutes 💀

[glad zealot]

one key thing is this is at the bottom of the modified lol

really wish there is a way to install requirements faster

[viscid moss]

Your collab is set in #🌍│español by me. Thank you for your work

[glad zealot]

is gud

you also have the moded voice changer?

[viscid moss]

I only use voice changer sometimes using collab, first with the original and now with your collab

[glad zealot]

ye

[rare goblet]

@glad zealot has reached level 16.

[viscid moss]

But people always come asking for help

[glad zealot]

surprised that still works ngl

[glad zealot]

cover gen already passed 6 mins les gooo

[brittle wing]

[Errno 2] No such file or directory: '.npz'
can someone please help me, i get this while trying to create my own tts for my module.

this is what i used to do it

[glad zealot]

ahh tts havent touched that yet

[stark grove]

Oh nice

[viscid moss]

try edge tts

[viscid moss]

omg

[brittle wing]

i don't have tts model, i want my own

[viscid moss]

For now you can only use RVC models with edge or bark infering the voice output

[brittle wing]

ok, why i get this then?

[Errno 2] No such file or directory: '.npz'

[meager quarry]

Is there an alternative to the disconnected RVC V2 colab?

[meager quarry]

There has to be one!

I just had this same issue credential login failure error on the 9/17 training colab

[potent root]

could you screenshot the issue?

[wispy burrow]

You didn't select a TTS Model.

[glad zealot]

yup tensorboard seems to be not showing on the colab anymoroe

but it does still train

[meager quarry]

This shows up on step "Mount Drive"

restarting the runtime results in this error instead on Step "Cloning Repositories"

[potent root]

what happens if you continue?

[meager quarry]

it can't train since nothing is mounted to drive or installed there so it also can't pull any zips from there

[rare goblet]

@meager quarry has reached level 5.

[glad zealot]

i just did it on mine and it works so idk whats wrong with yours

[meager quarry]

I fixed it? The issue was that appearently I wasn't using Chrome

[glad zealot]

ahh

wait it should work tho

[meager quarry]

maybe it also was the adblockers I had, idk

[glad zealot]

might be

[meager quarry]

I'm gonna try training, then I'll report back

[glad zealot]

imma train a scuffed shylily model overnight

[meager quarry]

question: batch_size should be set at 16 right

[glad zealot]

i have no idea

i have mine on 10

[meager quarry]

also tensorboard is broken for me as well

[glad zealot]

yup seems to be

[viscid moss]

Works only in Firefox

[glad zealot]

wtf why so specific

[viscid moss]

idk

[ember ermine]

for those who want this to work on chrome there's still a solution

right-click the 403 window and click inspect element.

you will get a highlighted link on the side after src:

some url that ends with ?tensorboardColab=true or something

copy that url(you can edit the attribute to be able to copy the url)

open it in another tab but change ?tensorboardColab=true into ?polling=true

it will show properly in the new tab

if there's a mod around maybe u can pin this and help others having this issue

[patent stirrup]

Any other way to launch the WebGui?

[short ridge]

raise BadZipFile("File is not a zip file")
zipfile.BadZipFile: File is not a zip file

[torn vale]

what is "epochs" mean?

[mental spruce]

can someone tell me how to fix whaiting web server ?

[viscid moss]

https://colab.research.google.com/github/hinabl/AICoverGen-Colab/blob/main/Hina_Mod_AICoverGen_colab.ipynb

[spare fog]

Works great! Thanks so much

[patent stirrup]

But isn't gradio banned?

[rare goblet]

@patent stirrup has reached level 2.

[exotic tendon]

Um you're right.

[rare goblet]

@exotic tendon has reached level 10.

[west hazel]

use brave

[sand iris]

https://cdn.discordapp.com/emojis/662668346063585311.webp?size=48&name=hmmmm~1&quality=lossless

[slender lotus]

can someone link me a working colab or huggingface? all the ones in #1159949278270193734 dont seem to work 😢 thanks in advance!

it just keeps circling..... using https://huggingface.co/spaces/Clebersla/RVC_V2_Huggingface_Version

[median ivy]

Using RVC v2 Disconnected, just finished the training and i didn't get the files?

i have the index file but not the other

wait nvm

[nocturne ferry]

should i use the google collab thingy or just my own pc?

what would be better/faster

i have a 2070 super as my graphics card

[proper shale]

Your PC would probably be better

[candid furnace]

How long does rvc take?

It's been like a whole hour

[sage matrix]

it depends on a lot

[half cove]

Yep

[half cove]

It might take a while how long was the audio?

[brittle wing]

How do I make a model?

[proper shale]

#1159513888199540817

[brittle wing]

Thnx

[modern snow]

Welp I have really good news! I reran the ai trainer with the specific changes and got something that sounds almost exactly like what I was making bad audio quality and all. I would love to post this voice models but I can't. How can I get the roles to get access?

[stoic rampart]

does anyone know if local rvc has an update

[proper shale]

#1159514067187277865

[modern snow]

What extraction method does disconnected use?

[proper shale]

harvest, crepe, mangio-crepe and rvmpe

i think

[modern snow]

Oh ok yhanks

[half cove]

Rmvpe 🔛 🔝

[glad zealot]

Doesn't work

Oh wait no that's just the voice changer that doesn't work

[west hazel]

colab is specific with its browsers

[glad zealot]

The voice changer doesn't work on Firefox for me

Meh

[west cipher]

wait so in the box, do put the link to the dataset itself, or the audio in the dataset?

[quiet haven]

What?

[west cipher]

in your collab, in the section that says to put the huggingface link, do i put the link to the dataset, or the wav file thats in the dataset

[rare goblet]

@west cipher has reached level 6.

[quiet haven]

The one wav file that the colab uses?

The original wav file

[west cipher]

ok heres the thing

when i import it using the huggingface link it works, but after i import it, then can i process the audio for classification or do I have to do the other steps first

sorry if your getting confused btw, im kinda new to this stuff

[candid furnace]

does how fast ur voice gets completed depend mostly on internet speed or your pc power?

[nocturne python]

when i launch the https_start file it will briefly open consol and then disappear how i get the rvc to actually launch. (dont tell me to reinstall ive done it 8 times)

[west cipher]

Okay I finally figured it out

The reason why it wasn’t working was because I was skipping all the other parts

After I went through them again correctly I finally got the zip file for the new dataset

[quiet haven]

Colab or locally? If locally depend of your PC power

[west cipher]

So now it’s onto the training part

[remote grove]

Looks like someone just wasted hours of their time.
I put some money into Colab and succeeded in machine learning the day before yesterday.
The fastest way for people with little knowledge who can't even install it is to first succeed with Colab and then try local as a sub.

Follow behind my success and taste the heavens

[glad zealot]

Pay for colab? Naaaa

[remote crane]

Hey I ran step 3 and the runtime always disconnects while in gradio

[glad zealot]

If you're using colab and using the trainer with web ui it won't work

Unless you pay for colab pro

[ripe harness]

someone slide the colab link so i could make sum models rq

[glad zealot]

Uhh -colab

-colab

[azure marsh]

Google Colabs

RVC

• AI-Cover-Gen-No-UI (inference only)
• RVC v2 Disconnected (training only)

🌐 Translated colabs

• PT-BR: -colabs_br

Audio Separation/Isolation

• MDX Net Colab
• UVR arch colab 5.0.2/c1.04
• KaraFan colab

[glad zealot]

The rvcdisconnected one

[upbeat pine]

Could someone teach me how to separate vocals from ad-libs?

[gloomy smelt]

Upload widget is only available when the cell has been executed in the current browser session. Please rerun this cell to enable.

im getting so tired of this error

[main patio]

Following this tutorial. What do I do next from here?

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

[elder kayak]

Hello I need some help

[modern snow]

How do you clean a data set

[golden elm]

bros im running in an issue where once i hit the train button the only thing that happens is that the console outputs this

python train_nsf_sim_cache_sid_load_pretrain.py -e C.A.S.S.I.E -sr 40k -f0 1 -bs 4 -g 1 -te 500 -se 50 -pg pretrained/f0G40k.pth -pd pretrained/f0D40k.pth -l 1 -c 1 -sw 0 -v v1

after that nothing

[stoic cloud]

Can you send what the console says?

CMD

Or is that what it says?

[rare goblet]

@brittle wing has reached level 2.

[golden elm]

its literally it

[rare goblet]

@golden elm has reached level 1.

[golden elm]

well

write filelist done
use gpus: 0
python train_nsf_sim_cache_sid_load_pretrain.py -e C.A.S.S.I.E -sr 40k -f0 1 -bs 4 -g 1 -te 500 -se 50 -pg pretrained/f0G40k.pth -pd pretrained/f0D40k.pth -l 1 -c 1 -sw 0 -v v1

@stoic cloud

here's the full log

[stoic cloud]

So it's running the command, and then nothing

[golden elm]

exactly

also some aditional info that you may find useful

i had to use pyenv to switch my version to 3.10.5 because 3.11.5 was too new

[stoic cloud]

Are you using the gui

[golden elm]

you mean gradio?

yeah

[stoic cloud]

Try reinstalling it if possible

[golden elm]

reinstalling what

the rvc program?

[loud robin]

having some problems

can't send picture however

[golden elm]

i think you gotta chat here for some time

to hit level 1

[loud robin]

huh

damn

basically it's saying failed the reset the quota database

[golden elm]

then you get embed perms

[loud robin]

what do i do

the entire client is failing

[stoic cloud]

Yes

The gui

[golden elm]

do i just redownload it

then

because i dont get wdym by GUI

i run the gradio by running infer-web.py

[loud robin]

nvm restarted my pc and got it fixed

[golden elm]

gg

[rare goblet]

@golden elm has reached level 2.

[loud robin]

might because i kept my pc running for 2 days

[golden elm]

ouch

[loud robin]

[stoic cloud]

That's not how you run it, are you on windows?

Just run the run-windows.bat

[golden elm]

oh

yeah im on windows

lemme try

[stoic cloud]

Okay, do that, you should get an interface then

Use that

[golden elm]

there is no run-windows.bat

maybe this?

ill run it and see what happens

yep i get ths

yeah but i cant use that for training

im trying to train an ai model not use one

wait i think i solved the issue

[stoic cloud]

Oh sorry

You need to run go-web.bat

That's where you can train

[golden elm]

yea i figured it out

thanks

[stoic cloud]

No probs, goodluck with training!

[golden elm]

how much does training take for an RTX 3060

just curious

i calculated it myself

apparently 1:47 hrs

[stoic cloud]

Uhh I've not tried, but normally it's around a few hours, 1-3 usually

But it depends on your dataset and batch size the modt

Epochs as well

[golden elm]

can i increase the batch size

for my gpu

[stoic cloud]

Yes, what is it currently at

[golden elm]

its 6 rn

gpu is rtx 3060

[stoic cloud]

Try 8 or 10

[golden elm]

alr ill go with 10

im fine with my gpu working at 90%

[lament eagle]

Guys wtf why is it fall flat??

is it overtrained now

[glad zealot]

nope that seems to be fine

[glad zealot]

damn thats around 145 epochs

that seems normal on that actually.. at least on my models

[lament eagle]

60 lol 💀

the dataset is 26 minutes though

[glad zealot]

tf

welp i still have no idea how to use the step counts then lol

[glad zealot]

or do you have it save every big number?

[lament eagle]

wdym?

[lament eagle]

from a yt video btw

[glad zealot]

ye

[lament eagle]

epoch 60 is the closest

[glad zealot]

on my models for some reason step cound seems very the same everytime

[lament eagle]

Honestly dude

if it sounds EXACTLY like the person

it's a good model

[glad zealot]

gud dataset then

great job

[lament eagle]

LOL fr

[stoic cloud]

That's what's supposed to happen

[lament eagle]

Okay yeah but like

the fact that it falls flat early after around 60+ epoch is baffling

I never had this before lol

it sounds good though ngl

I should've do this with mangio instead of rmvpe training lol

[upbeat pine]

Could someone teach me how to separate the vocals from the ad-libs?

[wheat apex]

how to make realistic hq female voice i have rtx750ti and 10s audio

[brittle wing]

so like this happens bc i need to download some c++ thing idrk but like idk exactly what i'm supposed to download to fix this

this isn't really an rvc issue but it also kinda is bc i can't use a gui i usually use

text incase u can't read what's on the screenshot

OSError: [WinError 126] SPecified module cannot be found. Error loading "D:\rvc gui\RVC-GUI-pkg\runtime\lib\site-packages\torch\lib\c10.dll" or one of its dependencies.
press any key to continue...

[rare goblet]

@brittle wing has reached level 2.

[brittle wing]

nvm i did it

[upper charm]

guys my gpu is nvida but its still doesnt let me to choose

any idea?

[red kayak]

dw about it

but when u see such drops that means that the model migth be mode collapsing

basically it's reading the silence u have in the dataset

[lament eagle]

Does this count as that "drop"

wait no nvm you mean something flat

srry lol

[red kayak]

this seems to have a mode collapse

[lament eagle]

mode?

[red kayak]

but i would need to ask u to change ur graph smoothing

[lament eagle]

huh are you saying there are many variations of overtraining

[upper charm]

its been 20 min and still converting and doesnt let me use gpu plshelp

[red kayak]

no no

[red kayak]

this gui is old and outdated

it's only used for so vits models

[upper charm]

can u send me new version

[red kayak]

and doenst support v2 if im not mistaken

[red kayak]

-rvc

[azure marsh]

Hey, @upper charm!

👇 Here are some useful resources to help you learn how to make ai covers

Documentation

🇺🇸 English Documentation

https://us.aihubfrance.fr/

Credits: AI HUB France

🇺🇸 English FAQ

https://us.aihubfrance.fr/faq

Credits: AI HUB France

Guides

🍏 Applio-RVC-Fork: Guide

Link: Google Docs

Credits: IA Hispano

How to install RVC on windows

Link: YouTube

Credits: iroaK

RVC v2 Tutorial for Paperspace

Link: Google Docs

Credits: LollenApe

🌐 Translations

• 🇫🇷 French Documentation
• 🇮🇹 Italian Guides or -it_g
• 🇷🇺 Russian FAQ for RVC/Voice changer
• 🇧🇷 RVC Guides in PT-BR (-rvc_br)

[upper charm]

whic one should ı choose

[brittle wing]

what gpu do u have

[red kayak]

go for the "How to install RVC on windows"

[upper charm]

gt630m bro

[brittle wing]

you need gtx750 or higher

[red kayak]

not really

[brittle wing]

fr?

[red kayak]

u can still infer with cpu

and gpu

[brittle wing]

oh

[red kayak]

u cannot train though

[brittle wing]

ik

[upper charm]

everbody says diff sht lol

[brittle wing]

i used to have a gtx650 on my pc and the gui didn't even boot up

[upper charm]

so interface is same on new version?

[red kayak]

no ur actually going to use the webui

which is much better than the one this old ver has

[brittle wing]

i don't use the colabs anymore it's too condusing without gradio

[upper charm]

wait rvc also was software of yours right

or you guys just tryna promote your own software

😄

[brittle wing]

i use it for rvc bc it's like the only somewhat good gui i could find

local gui i mean

idk if there is any local guis i can use to train tho

bc i haven't made a model since google banned gradio

[red kayak]

RVC is the vanilla build and the engineers made a fork

[upper charm]

also https://github.com/IAHispano/Applio-RVC-Fork this enough ?

[red kayak]

use local rvc mangio

[upper charm]

should i set up someting else

[red kayak]

yeah it is

although i cannot guide u further than this cuz i run rvc on a cloud

[upper charm]

thanks man

i will watch video for rest

ım too lazy off

[candid dust]

guys with applio local
how i get the model to recognise?
i have it in weights folder is it different for applio?

[red kayak]

werent u a model maker?

[candid dust]

yeah

verified user and the rest haha

just never used applio

[red kayak]

u still like rvc v1?

[candid dust]

im slowly moving into rvc v2

[red kayak]

ok well i remember u so here u go

[candid dust]

thankyou 🙂

[rare goblet]

@candid dust has reached level 5.

[red kayak]

and sadly i cant help u i do stuff on the cloud

but keep in mind applio is kinda buggy

[candid dust]

its ok i just fixed it but i appreciate you brother

[red kayak]

ohh good thing lmao

[candid dust]

haha its better than mangio local tbh that had bug where u couldnt disable pitch guidance

[red kayak]

ohh yeah i forgot they also fixed that

[candid dust]

yeah latest colab notebook i believe

[red kayak]

[upper charm]

what is total file size of apollo

[rare goblet]

@upper charm has reached level 3.

[red kayak]

like 6 gb or 5 gb

[upper charm]

aight i giv up

[red kayak]

probably more if u add the uvr and the mdx models

[upper charm]

mani just wanna do ai voice covers

whats the other stuffs for

[upper charm]

mdx and uvr

etc

[red kayak]

it's a whole interface with a bunch of features to enchance ur stuff

for the mangio for

fork then

[deep tulip]

I'm training the model through colab, tied a google drive to it, put 300 epochs on the training. The training reached 300 and did not stop, and for some reason there is no folder with pth and index on the disk

[proper shale]

You could just stop the step and export model to drive if you're using RVC Disconnected

[brittle wing]

Is it possible to use a ZIP file from Google Drive in colab? because it says access denied

[proper shale]

Probably the sharing settings are wrong

Make sure everyone w link can see

[brittle wing]

how do i do this?

I've been trying all day

[proper shale]

Right click ---> Share

And then a screen should pop up right?

[brittle wing]

yes

[brittle wing]

It works, just a little old compared to Applio

Can I have a screenshot please?

okay okay

[brittle wing]

it may take a little time the first time

It was the developers of Applio who decided like that. But it's true that it's strange

[wispy burrow]

I think it's 20 GB or something when finished.
Mangio is only 8GB.

[brittle wing]

does anyone know why i have a really high res or something ms

[rare goblet]

@brittle wing has reached level 1.

[wispy burrow]

Applio can really be a pain to install.
Quite a lot of of people had issues installing it.

[proud elbow]

download & install python 3.9.8 manually from its official site

[rare goblet]

@proud elbow has reached level 3.

[wispy burrow]

What version was the installer?

[rare goblet]

@wispy burrow has reached level 9.

[brittle wing]

please can someone help

[wispy burrow]

I meant applio installer.

[proud elbow]

I had no issues doing the 4.0.1 installer, or you can try installing from https://github.com/IAHispano/Applio-RVC-Fork/releases/tag/v2.0.0 (actually it downloads the same version as the 4.0.1 installer, only without prerequisites)

[brittle wing]

https://us.aihubfrance.fr/vocal-cloning-guide/applio

There are also links to the latest official version

[rare goblet]

@brittle wing has reached level 3.

[wispy burrow]

Those are pretrained Models.
Those are needed if you want to train your own models.

[astral perch]

Do you need to train the index after or before training the model??

[wispy burrow]

pip install torch_directml should solve that I think.

[wispy burrow]

Doesn't matter.
I do it before so I don't forget it after along training session for the model.

[astral perch]

Thank you!

Also what is the difference between the trained files and added

[wispy burrow]

No, It is trained/created separately.

[astral perch]

What do you mean by training on your pc?

[wispy burrow]

Honestly, I don't know.
But the added_IVF* is the one we need for the model.

[astral perch]

Okay thank you

Index controls accent strength right?

[wispy burrow]

Mine RVC folders are in my documents folders, so that shouldn't be the problem...

Yeah, I had to pip install so many things in order to get Applio to work.
I took more than an hour...

[brittle wing]

hi, I'd like help for installing applio RVC on linux for voice chat, me and a friend have been trying to install for some days now since switching to linux (we are new user), but no matter what i seem to do I always get depenency errors, the installer always creates a ffmpeg not found error when attempting to launch applio, and then beyond that it complains about faiss but faiss-gpu does not compile, what am I doing wrong? I also tried to install an older version of python, but the installer did not recognise it

[wispy burrow]

For an installer that was supposed to download & install everything (except git)...yeah...

[calm kite]

I want to make a model what is the optimal way of doing it

[wispy burrow]

Did try pip installing those missing modules?

[humble minnow]

Can someone give me the tutorial for how to do the ai voice changer real time i need to send it to a friend

[brittle wing]

why is the delay for the ai voice changer over 10 seconds

[brittle wing]

yes, as I described it will install ffmpeg-python, but then complain about faiss being missing and refuse to install faiss past that

it fails to compile faiss

[wispy burrow]

You should have enabled "set as PATH" when you nstalled 3.9.8.

[brittle wing]

what should i have this set as

[humble minnow]

?

[brittle wing]

just noticed im in the wrong channel lol

[wispy burrow]

'pip install faiss_cpu'
not just "faiss".

[brittle wing]

oh i've been trying to install faiss-gpu

[rare goblet]

@brittle wing has reached level 1.

[brittle wing]

i'll try that

[rare goblet]

@brittle wing has reached level 4.

[brittle wing]

okay so that worked, but now its giving me this error on fairseq

the model im using only changes the pitch of my voice

all models do

[wispy burrow]

It's a bit difficult for me to read...
But I see you're using python 3.11...

[brittle wing]

the installer wont work if I install python39 from the aur (which seems to be the only way to install a version of python like that on arch linux easily)

even if python3.11 is uninstalled, and 3.9 is installed and added to the .bashrc, it doesn't recognise and says it isn't installed

[wispy burrow]

User mode? Advanced? Sounds like they changed a few things...

So they did change a few things.

[hazy apex]

How can I send an image here?

I'm having issues with the web (www.weights.gg)

[brittle wing]

@wispy burrow help all models only change the pitch of my voice

[wispy burrow]

Double click the + icon left of the the text field to upload an image.
Drag&Dropping the image in the text field should work too.

[hazy apex]

I only have "create a thread" and "use apps" 😦

[rare goblet]

@hazy apex has reached level 1.

[hazy apex]

ohhh, now it appears, let me show you thank you

I don't know what's that

"funnyfortnite24"

[wispy burrow]

I can't help with W-Okada.
Haven't used it yet.

[glossy grove]

what cursed hellscape is that?

[hazy apex]

I can't download the models in the web because of that thing 😦

I don't know what's that to be honest

[wispy burrow]

Must be a glitch (I don't have that), try refreshing the page.

[hazy apex]

I've already tried, even with other browser but I'm still with the glitch

let me try something else

it's working now only on edge. What a weird thing...

Wait, now it's the same glitch but with other model (edge)
The glitch appears when starting scrolling down for models

[wispy burrow]

Odd, I use Edge as well and I don't have that issue...

Don't know...

[wispy burrow]

Never mind...
Just got it as well...

[low owl]

Isnt this RVC 1?

Or is it 2

[viscid moss]

use RVC, is AICoverGen

for inference only

[low owl]

¿Es RVC 1 o RVC 2?

Is it RVC 1 or RVC 2?

[viscid moss]

RVC V2

[low owl]

Gracias.

[viscid moss]

works with RVC 1 too

De nada

[smoky umbra]

What is it with people being able to train models so fast? I've got a dataset of a few thousand files, all clean, a batch size of 32, with a 4090 to power it all but I get like 3-5 minutes per epoch.

[low owl]

Huge batch file and huge dataset = longer training time

[smoky umbra]

So would a smaller batch size speed it up?

[woven atlas]

some one know why when i oppning tensorboard it dont concet to my logs file
?

[rare goblet]

@woven atlas has reached level 3.

[low owl]

Yeah, aswell a smaller dataset

[rare goblet]

@low owl has reached level 5.

[woven atlas]

I'm getting error like this

[woven atlas]

.

[outer crescent]

What is a feature retrieval rate exactly?

[astral perch]

What is the difference between RVC 1 and 2

[proper shale]

Usage of the index file - index helps with accents and stuff like that

[proper shale]

RVC v2 is faster, has better pretrained files, etc

[astral perch]

What are pretrained files

[rare goblet]

@astral perch has reached level 2.

[smoky umbra]

This is bullshit

[rare goblet]

@smoky umbra has reached level 1.

[smoky umbra]

I just shrunk the dataset, set the batch size to 2 and im now at 5 minutes per epoch

What the fuck happened

[outer crescent]

Ah, so a higher number would make the voice model utilize more of it's accent?

[rare goblet]

@outer crescent has reached level 1.

[stoic cloud]

What do you use to train

[smoky umbra]

rvc v2 on a 4090 and 7950x3d

[stoic cloud]

How much data?

[smoky umbra]

wdym

[stoic cloud]

How much data are you using?

[half cove]

Dataset :0

[smoky umbra]

at first i tried 3 hours of audio split into 4000 files, which was slow, so then i just tried the raw audio, which is the same, but split into 12 15 minute files

[stoic cloud]

Okay no. You don't need 3 hours, try 30 minutes and split it into 10-35 second files

[smoky umbra]

okay sick

before that message i was already starting training with just 1 hour

now im getting 22 seconds per epoch

phew

[smoky umbra]

so will having less audio altogether affect the quality of my rvc voice?

[stoic cloud]

Barely reallly

Really

[modern oxide]

Even 5-10 minutes is enough to make a good model, people make bigger datasets if a person, whose voice they're making, has a high range voice. Batch size depends on your GPU memory, but you have 4090, so values between 20-40 should be fine.

[tulip spire]

hello guys how to update applio it is already installed

[proper shale]

Files used for training

[proper shale]

Yes

[upbeat ice]

so how do i get voice models from kits.ai?

[wispy burrow]

You don't.

[upbeat ice]

really?

[wispy burrow]

The models kits models can only be used on kits.

No way to download them.

[upbeat ice]

or even to link them somewhere right?

[wispy burrow]

You can find kits models on weights.gg

And some in #1159289700490694666

[rare goblet]

@wispy burrow has reached level 10.

[light horizon]

Hey guys, could someone tell me which 'fork' of RVC is recommended to install, and also a link to that if there's one out there? Thanks

[rare goblet]

@light horizon has reached level 1.

[proper shale]

Both Applio or Mangio will work great imo

#1159949278270193734 should have some of them

[light horizon]

Is there a guide anywhere explaining the differences between the two?

[light horizon]

Thanks

[elder kayak]

So my issue I’m having is whenever I paste the AI MODEL link in colab in part 2 it gives me a error idk why

[proper shale]

what colab r you using

[mellow phoenix]

who can help? my res.. its sad
30k+ res

[rare goblet]

@mellow phoenix has reached level 1.

[wispy burrow]

Res? What's that?

[proper shale]

must be getting some settings wrong then

#🔍│help-w-okada

[proper shale]

W-Okada stuff... cant understand how ppl still confuse the channels

[elder kayak]

The google colab

[west cipher]

i need help with the rvc-disconnected collab

the dataset loads and it makes the model but when i get to the index training part, i keep getting an error, no matter how much times i restart the runtime the error keeps happening

idk if im doing something wrong so can someone help me with that

[frigid wasp]

Is anyone else getting error 500 on HuggingFace?

[west cipher]

can someone please help me with this, ive been having this issue for a while

.....

this was when i did the index training in rvc-disconnected

[wispy burrow]

You forgot to do 'Feature Extraction'? (Step 2b)

[safe jackal]

what gpu should i use my rtx 2070 super or quadro m4000

[rare goblet]

@safe jackal has reached level 1.

[safe jackal]

or should i buy the tesla p4

[astral perch]

Can you add more audip to dataset on rvc local and still continue from the latest checkpoint??

[wispy burrow]

Yes, but you'll need to 'Process Data' & 'Feature Extraction' again before training.
Personally I'd start over instead of continuing though.

[astral perch]

Why would you start over

[proper shale]

Won't be able to retrain actually

You can't add more stuff

[astral perch]

Oh

That's a bummer

So you need to start over if you add more

[proper shale]

Yeah

[wispy burrow]

Hmm...I could have sworn that was possible...

[sonic tapir]

does the colab work with the weights.gg site?

cause i can't get it to work

[proper shale]

the models do work

or well

at least some of them

links can get privated or something

[sonic tapir]

i figured it out. im on mobile so it wasn't showing me the entire page. 😭😭 but thanks

[rare goblet]

@sonic tapir has reached level 1.

[proper shale]

Oh alrighty. No problem! Good luck on your model

oh wait shi

i confused u with the person on the other channel haha

😭😭😭

Anyways good luck on your cover

[elder kayak]

@proper shale but yeah on my end the ai model when I past it to google colab it doesn’t work

[rare goblet]

@elder kayak has reached level 2.

[proper shale]

Again, what Colab notebook are you using?

[high night]

-local

-rvc

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

🇺🇸 English Documentation

https://us.aihubfrance.fr/

Credits: AI HUB France

🇺🇸 English FAQ

https://us.aihubfrance.fr/faq

Credits: AI HUB France

Guides

🍏 Applio-RVC-Fork: Guide

Link: Google Docs

Credits: IA Hispano

How to install RVC on windows

Link: YouTube

Credits: iroaK

RVC v2 Tutorial for Paperspace

Link: Google Docs

Credits: LollenApe

🌐 Translations

• 🇫🇷 French Documentation
• 🇮🇹 Italian Guides or -it_g
• 🇷🇺 Russian FAQ for RVC/Voice changer
• 🇧🇷 RVC Guides in PT-BR (-rvc_br)

[candid goblet]

anybody know how i can get all the recent features i only have these

pm
harvest
dio

im using lj1995/VoiceConversionWebUI

[violet heron]

Wow

Old

Use the new ones

Use applio if your not that tech savvy

[timid sage]

what does http not found mean with models

a bot i used doesnt work now

not a bot but yk

[hazy apex]

Can I ask you something? I've been using this version of RVC for 2 months already, just small issues but overall good results, but, do you know if there's something better now for a real time voice changer?
(23.7.0) Thank you in advance

[violet heron]

Okwada

For real time

[hazy apex]

that's the first time I hear that name, I'll search it now in the server, thank you

[rare goblet]

@hazy apex has reached level 2.

[hazy apex]

ohhh, it's "okada "

[dreamy lantern]

i need help i get res 1000ms

[brittle wing]

how can i make it so the ai reaches high pitches

[glad zealot]

make a model that includes high pitches i guess?

[vale horizon]

Anyone use a Minimac M1 or M2 with okada real time voice changer? Is the voice response fast or has latency?

[sand iris]

There aren't that many mac users around here lol.

From what I've seen, it works fine, but I don't remember the exact numbers

[vale horizon]

Thank you 👍

[modern oxide]

Use high range voice dataset

[snow scarab]

Hello, my PC completety crashes while converting with local, can somebody help me?

[stoic cloud]

I know one setting can cause a blue screen of death, can you show your training settings

[lean oak]

any guides on the advanced settings in Applio? I downloaded a model and it sounds pretty real but theres some bits that get wonky and was wondering if theres more info on the advanced settings to know what im messing with when I do?

[rare goblet]

@lean oak has reached level 1.

[remote grove]

I tried AMD with applio but all my efforts were in vain. Don't follow me.

[lean oak]

do yall not recommend applio for local? is there something better?

is replay more geared towards songs or can I drop normal text recordings in it and would still work?

[round marsh]

Is there a way I can determine the amount of epochs to use depending on the length of the dataset

[potent root]

i recommend using 750 and just relying upon the lowest point of the graph

[astral perch]

My training won't contunue from the checkpoint

On rvc local eve tho I used the same name and dataset

[daring verge]

tensorboard

[round marsh]

I use brave browser and tensorboard doesn't work

[daring verge]

probably it got restricted in the settings

[pine grail]

i wanna cry

[zinc anchor]

Looks like a connection error on the other end

Check if that download_checks.json file that it's trying to download is actually accessible

[brittle wing]

it gave me module not found errors when i launched with applio and when i manually installed the modules it gave me another error:

[remote grove]

I have had success with Colab by paying a small amount of money. Beginners should first succeed with Colab and then slowly try local in a sub environment.

[zinc anchor]

https://github.com/TRvlvr Should be here, it's public tho so idk what to say there, I'm not that knowledgeable

[zinc anchor]

Maybe you tried already, but I'd check if you have that file in that folder

[brittle wing]

i dont know if you're being sarcastic but it tried to download that file and it didnt show up in the directory i guess? idk much about all this.

[zinc anchor]

Yeah that's weird

[brittle wing]

nevermind

i think i fixed it

i have no brain lmao

nevermind once again i swore that was it

[zinc anchor]

It's good that you fixed it

You do have a brain

[brittle wing]

nah lmao

i think i figured it out

its because i was starting applio from the installer exe

for some reason it doesnt download anything when you start it from there

[meager quarry]

is tensorboard still broken on colab

[proper shale]

yup

[astral perch]

Please help my voice wont continue from its checkpoint

[proper shale]

you got the wrong settings then

make sure batch_size is sae

same

[astral perch]

It is

[proper shale]

• rvc version, sample rate

[astral perch]

I used all of the same settings

[proper shale]

did it start from the first epoch?

[astral perch]

Yes

I was already at 247 and nownit starts at 1

[proper shale]

yeah then it probably didnt save G and D files properly

guess you'll have to start all over really

[astral perch]

I have a lot of those files in the logs folder

[pine grail]

ermm.. wha

[spare pond]

yo, I recently got a new GPU and want to train locally now, but when I interact with literally anything inside the Mango WebUI it just goes into queue: 1/1 forever
Does anyone know why? 😭

[astral perch]

Because evreytime I wanna paise training i just alt f4

Is there a way to exit training in a more safe way where everything saves correctly

[stoic cloud]

Yes, the start training button becomes a stop training button, you can press that

[spring chasm]

where can I find the G and D files?

[astral perch]

For me it still says start training model

[potent root]

in the colab file explorer, just go to drive, RVC_Backup, then click on your model's name

you will see the G and D pth files

same process with local training i assume

[stoic cloud]

Press it again to stop, it hopefully should stop

[torn seal]

Hey guys

Is disconnect broken?

A friend said it was broke

[still crystal]

its there anyway to train models with colab?

[quasi monolith]

how do i adjust the pitch when using the colab?

[sand iris]

Shouldn't be

[sand iris]

Yes !

-colab

[azure marsh]

Google Colabs

RVC

• AI-Cover-Gen-No-UI (inference only)
• RVC v2 Disconnected (training only)

🌐 Translated colabs

• PT-BR: -colabs_br

Audio Separation/Isolation

• MDX Net Colab
• UVR arch colab 5.0.2/c1.04
• KaraFan colab

[sand iris]

"RVC v2 Disconnected" should be working

[still crystal]

ok thanks

[rare goblet]

@still crystal has reached level 1.

[slender laurel]

guys i cant post a model on huggingface it stuck on hashing

[rare goblet]

@slender laurel has reached level 1.

[torn seal]

Oh cuz makaveli ai said it was broken

🤷‍♂️

[sand iris]

Apparently it's broken for some users, somehow

Tried it myself, it worked, and I've also seen people use it recently so I'm not sure what's up with it

[glossy grove]

any idea if rolling the pytorch version back to 1.13 will break anything? Trying to get applio running on a tesla K40c

the k40c only supports Cuda 11.3. Sadly pytorch 2 only suports 11.7 and newer

[sand iris]

Can you try it lol

I'm curious as well. Don't think it'll work but it's worth a shot

[glossy grove]

lol

[stoic cloud]

Hi

Oh wrong chat sorry

[glossy grove]

2023-10-13 13:30:01 | INFO | assets.configs.config | Found GPU NVIDIA GeForce GTX 1070, force to fp32
2023-10-13 13:30:01 | WARNING | bark.generation | torch version does not support flash attention. You will get faster inference speed by upgrade torch to newest nightly version.
2023-10-13 13:30:02 | INFO | assets.i18n.i18n | Using Language: en_US
2023-10-13 13:30:02 | INFO | lib.infer.modules.vc.modules | Get sid: logs/weights/MikuDiva.pth
2023-10-13 13:30:02 | INFO | lib.infer.modules.vc.modules | Loading...
2023-10-13 13:30:05 | INFO | lib.infer.modules.vc.modules | Select index: logs/MikuDiva/added_IVF234_Flat_nprobe_1_MikuDiva_v2.index
Running on local URL:  http://0.0.0.0:7865

well this is promising

[sand iris]

https://cdn.discordapp.com/emojis/924763610197147678.webp?size=48&name=hmmm&quality=lossless

[glossy grove]

now to test on the k40c

Barktts is busted anyways fyi

[sand iris]

Quick thing : if you try model training, the GPU listing will not show your card as supported

But you can still try to specify it's CUDA device index and it might just work

(For some reason the code related to displaying """supported""" GPUs is beyond shit, and is pretty much hardcoded to match against most cards that are known to work)

[torn seal]

I'm thinking his dataset might be to big 2 hours 🤔 idk 🤷‍♂️ who knows, all I know is I'm going to train with a bigger dataset thanks to medleyvox

[glossy grove]

atully it just stright up told my that torch didnt know what to do with the k40 when i tried it last night with torch 2

[sand iris]

Yeah expected

"no kernel image was found" or something ?

[glossy grove]

yes

2023-10-13 13:53:20 | INFO | assets.configs.config | Found GPU Tesla K40c
2023-10-13 13:53:21 | WARNING | bark.generation | torch version does not support flash attention. You will get faster inference speed by upgrade torch to newest nightly version.
Using Language: en_US
2023-10-13 13:53:22 | INFO | lib.infer.modules.vc.modules | Get sid: logs/weights/MikuDiva.pth
2023-10-13 13:53:22 | INFO | lib.infer.modules.vc.modules | Loading...
2023-10-13 13:53:28 | INFO | lib.infer.modules.vc.modules | Select index: logs/MikuDiva/added_IVF234_Flat_nprobe_1_MikuDiva_v2.index
``` promising

[round marsh]

I did here tensorboard works with firefox

[glossy grove]

terminate called recursively
Aborted (core dumped)
(app) [wispers@scrapboox002 Applio-RVC-Fork-2.0.0]$ 

Interesting

It attempted to load the gpu

[barren moat]

i keep getting error

[glossy grove]

Seems faiss uses a different biuld for older Cuda versions

Poor folks having issues getting it running and I'm over here trying to get it running on a 10 year old data center gpu

[glossy grove]

What error?

[rare goblet]

@glossy grove has reached level 3.

[barren moat]

dms as i cant send image

[glossy grove]

K

[sand iris]

Might need to replace some auxiliary libraries 'version as well

[glossy grove]

Already had to do that just to get the ui up

A lot of torch depends were too nee

Lucky for me that faiss version supports python 3.9.

If I do get this going I might push my version to github as a legacy hardware version

The tesla k40c has about as much power as my gtx1070, but with 12gb of vram instead of my 1070s 8

I can already trian on my 1070

But larger models give me trouble

[west cipher]

I did do that but it still failed

[brittle wing]

Is there anywhere else I can download models other than weights.

[rare goblet]

@brittle wing has reached level 1.

[frail cave]

https://colab.research.google.com/drive/1r4IRL0UA7JEoZ0ZK8PKfMyTIBHKpyhcw?authuser=2#scrollTo=Sb5fzhzEXK8X

Can anyone help me with how to use this colab with already made models?

[wispy burrow]

#1159289700490694666
voice-models.com

weights.gg has most of the models though.

[astral perch]

Wich target sample rate should you use?

[rare goblet]

@astral perch has reached level 3.

[brittle wing]

-colab

[azure marsh]

Google Colabs

RVC

• AI-Cover-Gen-No-UI (inference only)
• RVC v2 Disconnected (training only)

🌐 Translated colabs

• PT-BR: -colabs_br

Audio Separation/Isolation

• MDX Net Colab
• UVR arch colab 5.0.2/c1.04
• KaraFan colab

[vivid rock]

Is there is guid how to use rvc locally on CPU ?

[round fog]

local cpu is rough

takes a long time

[solid thorn]

decided to download since i found the colab annoying to work with and i keep getting this error every time i try to launch :/

shit posted the wrong image

this is the error

[vivid rock]

I know.I want to know how long it be.

[vivid rock]

So i want to test to find out

[round fog]

install it using this

[solid thorn]

i'll give it a shot! thanks!

no dice, briefly shows this window before closing

[analog oxide]

hi guys, can anyone tell me how can i make something like my voice in some music

like using my voice in edd sheeran perfect?

[reef marten]

how can i remove reverb from an audio 😭

i'm trying to do it with audacity but i just can't

[solid thorn]

if you're able to, download "ultimate vocal remover", it has options to remove reverb pretty well

[reef marten]

wait, it does?

[rare goblet]

@reef marten has reached level 1.

[reef marten]

which option is it?

[solid thorn]

here are the settings for it

[rare goblet]

@solid thorn has reached level 2.

[reef marten]

ooh

[brittle wing]

help

[solid thorn]

oh yeah just click on download more models, click on vr architecture and go into the drop down until you find the correct model (in this case UVR-Deecho-Dereverb)

[reef marten]

alright, ty

[reef marten]

bro thank you

it works like a dream

[solid thorn]

of course! really changed everything when i learned about it hehe

[reef marten]

yeah lol

it is so crispy 🔥

[astral perch]

@solid thorn does it remove background noise. My dataset has a lot of sfx in the background and I haven't found anything yet that could remove it

[solid thorn]

can't say, haven't really tried it with background noise, i wouldn't be surprised if there's a model for that though

[floral heath]

-rvc

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

🇺🇸 English Documentation

https://us.aihubfrance.fr/

Credits: AI HUB France

🇺🇸 English FAQ

https://us.aihubfrance.fr/faq

Credits: AI HUB France

Guides

🍏 Applio-RVC-Fork: Guide

Link: Google Docs

Credits: IA Hispano

How to install RVC on windows

Link: YouTube

Credits: iroaK

RVC v2 Tutorial for Paperspace

Link: Google Docs

Credits: LollenApe

🌐 Translations

• 🇫🇷 French Documentation
• 🇮🇹 Italian Guides or -it_g
• 🇷🇺 Russian FAQ for RVC/Voice changer
• 🇧🇷 RVC Guides in PT-BR (-rvc_br)

[astral perch]

It didn't stop

What should I do

[stoic cloud]

What version is it?

[astral perch]

Where can I see that

[stoic cloud]

the folder name

[astral perch]

RVC-beta0717