✨│ai-help

[leaden yacht]

That was on my previous laptop btw, but yeah

I had tried all of the forks that were recommended to me (like 4 or 5 or smth)

nothing worked except mangio

sometimes it worked but extremely slow

and mangio worked, except for training (a coding error easily fixable by changing 2 parts in the python files tho)

[violet heron]

How long ago was it

[leaden yacht]

i'd say 3-4 months

got this new laptop 2 months ago

[turbid meteor]

how to fix no module named dotenv?

[pastel oak]

Your folder is inside OneDrive. It doesnt work inside, so put it somewhere else

[leaden yacht]

pip install python-dotenv

Had that yesterday

[turbid meteor]

okay
ill try

[leaden yacht]

and also what Shad said, you should take it to another folder

outside of onedrive

[turbid meteor]

okay
im moving it to drive d

[leaden yacht]

sure another drive is also good

I have that too

[turbid meteor]

ive uninstalled onedrive yesterday though..

[leaden yacht]

oh then its alright

I also have uninstalled onedrive

and now its js a normal folder

not affected by actual onedrive

if you uninstalled onedrive you should be fine

[turbid meteor]

but i got that error just now xd

[leaden yacht]

I mean you can try another folder

You did install python-dotenv?

[turbid meteor]

i just did, yeah
trying it again

[leaden yacht]

Js ping me if theres smth else like a dependency error, I got a lot of them yesterday lol

[turbid meteor]

still error

[turbid meteor]

where should i run this?

[leaden yacht]

Js in cmd

[leaden yacht]

There

When u can type there

Js type pip install python-dotenv

[turbid meteor]

yes, still same..

[leaden yacht]

Ok so I want you to run python --version

[turbid meteor]

it says 3.12.3

[pastel oak]

what application are you running anyway

[leaden yacht]

Have you installed all of the requirements? If you're unsure just run pip install -r requirements.txt in the rvc folder

[turbid meteor]

from here https://github.com/Blanc-dot/RVC-test/blob/main/README.md
i just reinstall my windows machine so..

[pastel oak]

Are you trying to install RVC realtime voice changer?

[turbid meteor]

is it redownloading everything again?

[turbid meteor]

yes

[leaden yacht]

Its downloading all needed requirements to run rvc

Will possibly fix all of the dependency errors

[pastel oak]

Do you want to install it manually the way you are right now? You could just download a pre package. What GPU do you have

[unborn viper]

I installed mango

[rare goblet]

Ayo? @unborn viper level 3 !!!

[turbid meteor]

its throwing me a bunch of reds xd
ill try what Shad says

[pastel oak]

what gpu do you have so i can send you quick links

[turbid meteor]

oh im doing it manually...

[turbid meteor]

im using amd rx5600xt

[pastel oak]

https://huggingface.co/lj1995/VoiceConversionWebUI/resolve/main/RVC1006AMD_Intel.7z

Download this

Then I recommend you download the (old) dev version as theres a quick trick to reduce some more delay. To do this, heres a quick step by step:
https://huggingface.co/Shadicti/rvc-old-dev/tree/main

Open the zip, open the first folder called Retrieval-based-Voice-Conversion-WebUI-dev and drag all the files into your RVC folder. If it asks you to replace any files, say yes. If it does not ask you to replace any files, you probably dragged the folder Retrieval-based-Voice-Conversion-WebUI-dev in it, but you should drag the 8 folders and 26 data files directly to the RVC1006 folder.

Then you run go-realtime-gui-dml.bat to start the voice changer

On there, make sure to change the default index thats there, just upload any random custom index to prevent crashes

Select input and output of the same driver type (MME) at the end. Is important
Sample length is delay (you should start with ~0.30 and see if you can go up or down. Got no pointers for rx 5XXX yet so let me know what sounds best so i can take note for the future), ignore harvest algorithm, do max fade length and extra you could do full probably aswell else try 4.00 first
Choose rmvpe

[leaden yacht]

Uhh probably more things conflicting, you could js remove all packages and install the requirements if you want. To do so run pip freeze > oldpackages.txt
Then pip uninstall oldpackages.txt
And finally run pip install requirements.txt

Doing this removes ALL of your python packages and installs the correct ones

[turbid meteor]

okay, ill try it

[pastel oak]

Its fine, let exkoi install the prepackage. most people get too confused with python and its scary if they were never confronted with this

[leaden yacht]

Only do that if you get dependency errors with shads instructions

[leaden yacht]

I know

But its for if they have problems with ur method

The dependencies

Python dependencies really are annoying to have to deal with, one time ur on the wrong python version, other time the wrong package version etc

[leaden yacht]

This js ensures it only has the required packages and installs the correct versions of them all

[unborn viper]

i managed to make a bit of progress

[sand iris]

venv

& call it a day

[leaden yacht]

also an option

but might be a little too complicated for them

[sand iris]

Honestly might be for the better. Just like a single line in a .bat file and it's never gonna be a bother ever again

Much better than possibly breaking other software lol

[peak osprey]

Why can't i see tge public URL?
i am only seeing the local one

[rustic oyster]

yeah im getting the same thing

[peak osprey]

What do we do?

[rustic oyster]

no idea

[peak osprey]

@leaden yacht sir?

[leaden yacht]

There's nothing we can do

But seriously, I don't know as I don't use google collab. Sorry

[peak osprey]

https://tenor.com/view/panda-animated-lazy-sleepy-cute-gif-17741860

[rustic oyster]

im trying out a diff google space

[rare goblet]

Ayo? @rustic oyster level 1 !!!

[peak osprey]

Then what do you use?

[rustic oyster]

https://huggingface.co/spaces/Clebersla/RVC_V2_Huggingface_Version

use this

takes like 20x longer but it works lol

[peak osprey]

Oh crap i don't have much time but ok

[rustic oyster]

how long is what ur tryna convert

[leaden yacht]

I host mangio locally as I have good hardware I can run it on

[peak osprey]

20 seconds

[rustic oyster]

oh lmao youll be fine then

[peak osprey]

I don't know that

[graceful obsidian]

if you can, better avoid mangio releases

[peak osprey]

How do i know that my hardware is strong to handle mangio or whatever strong ai cover platform?

[leaden yacht]

why everyone hating on mangio 💀

[graceful obsidian]

trust me, I was the nr1 defender of mangio fork and why it was the best

over the new rvc " rmvpe " hype

but times changed

[leaden yacht]

true

but still I it cuz I never had any problems with it before

[graceful obsidian]

just for the record, it's the 'rvc' you're having in mind by " ai cover platform " and mangio we talk about is just, say, a variant of original rvc

I'd recommend original rvc or my variant of it ( for mangio features but without mangio's bugs and issues )

[rustic oyster]

this shit is a 2 minute audio clip and is taking over 300 seconds bruhh

[graceful obsidian]

as for hardware, you can check in task manager > performance tab
for your graphics card / gpu
As long it's nvidia graphics card + having at least 8 gigs of vram ( saying 8 as minimum cause of the comfort and efficiency reasons. 4 and 6 is hell to work with )

[peak osprey]

Lemme see

[graceful obsidian]

take your time

[rustic oyster]

oml i spent 400 seconds converting this shit for it to sound ass 😭

[peak osprey]

Just curious, What the cover are you converting?

[rare goblet]

Ayo? @peak osprey level 3 !!!

[rustic oyster]

wym

[graceful obsidian]

he talks about input audio ( whatever you infer on

[rustic oyster]

oh im tryna do a song remaster

[peak osprey]

Which character and which song?

[rustic oyster]

converting low quality to studio quality

carti model

[graceful obsidian]

for me it takes 2x the total length of the input

💀

[rustic oyster]

model good i think my audio isnt clear enough

[graceful obsidian]

cpu inference be like

[rustic oyster]

fucking crazy

the regular google space usually does it in like 10 seconds this ones just annoying asf

[graceful obsidian]

true, any good ish gpu takes anywhere from 10 to 20-40 secs, for 3-5 min inputs ( google's most basic one is Tesla T4

[rustic oyster]

why my shit got a lisp for

[rare goblet]

Ayo? @rustic oyster level 2 !!!

[earnest viper]

kinda new to rvc but the output seems horrible. i am using tts and then convert, it does not seem to even slightly change the voice?

[rustic oyster]

must be that bruh

im tryna get on this other rvc

but its not working

[graceful obsidian]

that'd mean the model is crap or the tts is having some issues

[foggy folio]

where do I add the index files for mangio rvc?

applio uses too much storage btw

that is why I am with mangio for now

[graceful obsidian]

for mangio, it's logs/model_dir/here
model_dir = make a folder for your model

[earnest viper]

i tried my own audio as well, still same issue. its a model called carti so no idea

[rare goblet]

Ayo? @earnest viper level 2 !!!

[graceful obsidian]

ye, that'd indicate the model's issues

if it's not the pitch setting, then surely the model

[rustic oyster]

which carti model u using

[foggy folio]

oh btw it is for a downloaded model

not my own

[graceful obsidian]

ps. you can decrease the size even more

oh, unless you use infer-only mangio then ye, that's the same thing I'd recommend in a way ( comes down to removing some files from any rvc, pretty much - those related to training for instance

[rustic oyster]

nah i have noticed that one typically spits out shitty conversions

[earnest viper]

babycarti

[rustic oyster]

same lmao maybe it is the model

[earnest viper]

its on the space, i didnt download a new model. i am gonna try the avril lavigne

[foggy folio]

so any help with this?

[rustic oyster]

that was me lmfao

[peak osprey]

My GPU is intel(R) iris (R) Xe
And CPU 12th gen intel(R) core(TM) i7-1255U

Is it good for a better performance?
Or even speaking in an AI voice?

[foggy folio]

the index file is not showing for me

[graceful obsidian]

That still applies

just make any folder with the name of your choice

in rvc's folder / logs / ur folder of ur choice

and in there, put the index

[foggy folio]

Ah tyy

got it now

[graceful obsidian]

np, it's just for naming / index searching convenience

[rustic oyster]

this shit so annoying

[graceful obsidian]

that's true, it does not cause any of such
but if one ckpt that's meant to be the best is not present due to 5 vs 1 for frequency

you're just screwed pretty much

those epochs " existed " at the time of their creation, just aren't saved
that's all there is to it

:' )

[rare goblet]

Ayo? @brittle wing level 10 !!!

[graceful obsidian]

yes no yes are the options

that first one is about " D " and " G " files

those in folder's model

for discrim. and generator

cause if you were to select no, you'd have tons of em

if it's set to yes, you just only keep the newest one that corresponds to an actual generated epoch ( small weight )

[blissful scaffold]

still not working?

[rustic oyster]

yep

it doesmt even seem to be down tho bruh

whats kaggle

[earnest viper]

kaggle is better version of colab

[rustic oyster]

where can i find that at

[earnest viper]

2 t4 gpus instead of 1 and tpuv3 instead of v2

[earnest viper]

https://www.kaggle.com

[rustic oyster]

is it new notebook

im very confused how this works

[earnest viper]

oh yeah does not allow share links like gradio tho, and for me whenever i use share, it stops the instance. so probably not good

[rustic oyster]

so am i not able to put models in or something

[earnest viper]

its also by google and it has jupyter notebooks as well

[rustic oyster]

im so confused what im looking at

[graceful obsidian]

nono, it works that way;

• Epoch / small weight model / Neural network's ( G + D summarized knowledge ) = small, 50-56~ mb .pth ( pytorch ) model located in weights folder

• ckpt / checkpoint = Generator file and Discriminator file ( G, D each .pth (( pytorch )) ) they're located in model's directory

whenever an epoch is saved / generated, the Generator and Discriminator elements are doing inner-parameters and such updates ( neural network / ai magic ) to move on, learn other stuff and focus on consecutive epochs.

and so:

• if you select " yes " for " save only latest ckpt to save space " the only G / D files that are always kept are up to date with an actual ( at given moment ) epoch.
• If you were to select " no " each G / D is saved.
  the " no " is helpful for people with HUGE af storage
  and if they wanna extract small models from each stage of the training

and in 98% of cases, nobody has such amounts of space

hope that's clear now

yeyyy

hmmm.. I'd say, you almost never go above 500 or 700 anyway
( maybe if one was to use 1-2 hour sets which is a nonsense for rvc, but that's for another day, another topic )

good habit is leaving it at 500 or 1000 total epochs, just for the sake of tensorboard evaluation
even if overtraining in your case would happen at, say, epoch 362 ( cause you can always stop the training, right. )

And again, if you'll need tensorboard eval help, I am here - At least currently

[peak osprey]

Bro this colab is bad bro ):
I wanted to make heavy from TF2 but he sounded like a squirrel

[rustic oyster]

idk what to use then

[graceful obsidian]

that'd mean, you gotta tone down the transpose

[rustic oyster]

cuz the other ones fucked

[graceful obsidian]

do -12

or -6

[peak osprey]

I'll see

[graceful obsidian]

if model sounds squeeky, it means the pitch is too high for one

[peak osprey]

Aren't there any better colabs by the way?

[graceful obsidian]

if it sounds too low, means you gotta up the " transpose ", which is just nothing else but semitones ( 0 = original pitch from your input. 12 means 1 octave up and -12 means 1 octave down )

[rustic oyster]

anything below -4 always fucks up the model for me

idk how this shit works really

[rustic oyster]

im tryna find out too

[graceful obsidian]

we're talking about tts or inference again

cause if it's inference, that's model's fault
if it's tts then uhhh, something's going on under the hood ( tts's at fault

[rustic oyster]

whats inference

[graceful obsidian]

in simple words, when you use your model to make covers

you do inferences ( using your model ) on type of data; speech / singing

[rustic oyster]

oh yeah

inference

[peak osprey]

what is tts?

[rustic oyster]

probs model then yeah

text to speech

i didnt even know u could do tts with this

[graceful obsidian]

tts is just that

but it uses any tts framework or system to generate " input ", so, audio

which is then inferenced

so if inferencing is fine and tts is not, means tts fails in speech generation ( or something else happens in the middle

[graceful obsidian]

with the one u're using ( not sure what it is ) probs not

but other forks of rvc incorporated it

Orrrr, rvc's devs are making " vits gpt " or something like that
Different system, need new models but that's also TTS ( and pure tts )

[rustic oyster]

oh thats pretty cool

[prime forge]

how can ı dowland vrc?

[graceful obsidian]

https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI

here

[rustic oyster]

im just sick of waiting for this shit im sick of waitng 400 seconds a conversion

[blissful scaffold]

is the public url not working

[rustic oyster]

yes

idk whats wrong w it

it doesnt even load anymore

[rare goblet]

Ayo? @rustic oyster level 3 !!!

[blissful scaffold]

not mine etierh

[graceful obsidian]

Specifically, here:
https://huggingface.co/lj1995/VoiceConversionWebUI/tree/main @prime forge

[blissful scaffold]

:8

[slim ember]

help, delay is very big

[peak osprey]

is that the program where you can speak in real-time in an AI voice?

[blissful scaffold]

is everybody usign the same thing

if its down for everybody

hahah

[graceful obsidian]

Nope, that'd be either RVC's " real time voice changer " aka rvc's native voice changer or w-okada voice changer

2 different things, using same ( rvc origin ) models

@prime forge alr, so, how can I help

[peak osprey]

i tried okoda once but didn't work

idk why

[leaden yacht]

is this because of the training I did?

[graceful obsidian]

any particular log or example of how and what didn't work?

[graceful obsidian]

woooo shit

ye. Def G / D files

[blissful scaffold]

is that when you do it locally?

[leaden yacht]

Nah

only when u train locally

normally it wont even exceed like 6-10 gb

[blissful scaffold]

oh okay maeks sense

[leaden yacht]

but I left it running for like 10 hours or smth I forgot

[blissful scaffold]

wtf

LMAO

[leaden yacht]

and it saved every single epoch

[leaden yacht]

nah thats good, it was planned

[blissful scaffold]

oh really

[leaden yacht]

I wanted to train a model and js left it running to later see where it overtrained

[blissful scaffold]

so is it like

[rare goblet]

Ayo? @blissful scaffold level 3 !!!

[blissful scaffold]

best quality now?

[leaden yacht]

the best I could train myself

[graceful obsidian]

not quite, total trained epoch count =/= quality

[rustic oyster]

600 second conversion for a 2 minute audio is crazy

[graceful obsidian]

hmmm...
should be 250-260 ish for me

approx 2 times the audio length

[rustic oyster]

what do u use

[graceful obsidian]

what you use 👀

I mean, I do cpu inference so

[rustic oyster]

im using this https://huggingface.co/spaces/Clebersla/RVC_V2_Huggingface_Version

[graceful obsidian]

i5 10400 / 8 gig of ram

generally, cpu inference is like, at least 2 times the audio input

[rustic oyster]

i got i7 19700k or something like that

16gb ram

[graceful obsidian]

something must be going on then

[rustic oyster]

i7 9700k

[graceful obsidian]

oh

[rustic oyster]

its the space

the space is dogshit

[graceful obsidian]

nah, don't use hf

try locally then, with those specs

[leaden yacht]

not the best quality, probably could always improve the dataset etc, but because I saved every epoch I was able to get the best one though

[rare goblet]

Ayo? @leaden yacht level 13 !!!

[rustic oyster]

how do i do that

[graceful obsidian]

want me to help?

[rustic oyster]

yes please

[blissful scaffold]

is the audio you input the most important thing?

[rustic oyster]

i jsut wasted 600 seconds for it to sound like dog shit

[leaden yacht]

I found the culprit

[blissful scaffold]

like if you train with dogsh*t quality it will be bad result

[leaden yacht]

yeah

[leaden yacht]

mhm

[blissful scaffold]

got it

[leaden yacht]

exactly that

you need to get clear audio

[blissful scaffold]

makes sense so every voice that got bad quality

[leaden yacht]

I js stripped this from the stanley parable ultra deluxe game files lol

[blissful scaffold]

is people inputing bad isolating

[graceful obsidian]

Alright so

[leaden yacht]

The model will be based on the dataset (audio files of the voice) you provide. So if the dataset is really shitty, you will get a shitty model

[blissful scaffold]

makes sense

i tried finding good yeat

[graceful obsidian]

@rustic oyster
take this:
https://github.com/codename0og/hifi_testing_new

[blissful scaffold]

but guess hard to isolate

his voice hahaha

[rustic oyster]

download it?

[graceful obsidian]

1. Getting the fork: 836_1

yep, the zip

[rustic oyster]

easy

does this take up a fuck ton of space btw or na

[graceful obsidian]

other than that, you'll need few files and python ( some other things too )

that's fine with you?

[rustic oyster]

yeah thats fine

as long as its nothing MASSIVE

[graceful obsidian]

oh, nah, maybe 2-3 gig max or less, 1? 2?

[rustic oyster]

thats fine with me

[leaden yacht]

Btw red, it only takes that much space cuz I let it train overnight, and after some time it overtrains

[trail aspen]

So im a bit new... What program do yall recommend for running the models?

[odd shale]

RVC

Or any fork like Mangio.

[trail aspen]

Oh, didnt realize RVC was the name of the program itself, thought it was just what you called the files

[prime forge]

how can ı add model

[leaden yacht]

It got to epoch 1283 and the best was 606, so everything else (1282 files) are completely useless

[blissful scaffold]

really!!

thats interesting

[leaden yacht]

I just deleted them

[graceful obsidian]

2. Path thingy and naming: 836_2

After you dl it, ideally, unpack all the contents to any 1 word named folder on the root of your drive
say:
C/rvc/contents of zip

[blissful scaffold]

got it so too long is useless

[leaden yacht]

wdym

[rustic oyster]

wait so where do i put this

[graceful obsidian]

like so

[blissful scaffold]

like it doesnt make it ebtter

at one point

[graceful obsidian]

ignore extras in mine, that's debug ver

[rustic oyster]

is this right

[graceful obsidian]

take the insides of the folder and put it to rvc folder straight

[rustic oyster]

okay

[leaden yacht]

too long isnt useless, not at all. If you leave it running long enough, it will train as long as possible. So after that you can just look at the graphs in tensorboard, and see where it started to overtrain. Sometimes it overtrains at 50, and other times for some ppl at 1000. Really its just waiting and later on checking

[graceful obsidian]

3. Getting the required files: 836_3

Good, now

you'll need these two files:
Hubert base:
https://huggingface.co/lj1995/VoiceConversionWebUI/resolve/main/hubert_base.pt?download=true
rmvpe:
https://huggingface.co/lj1995/VoiceConversionWebUI/resolve/main/rmvpe.pt?download=true

[rustic oyster]

yup downloading

[leaden yacht]

And if you save every epoch, you can get the exact point before overtraining, and get the best out of your training session

[rustic oyster]

and just drag em in the folder?

[graceful obsidian]

3.1. Getting the required files: 836_3.1

rmvpe ends up in your rvc folder / assets / rmvpe

[blissful scaffold]

oh smart

[leaden yacht]

got that from code lol

[graceful obsidian]

3.2. Getting the required files: 836_3.2

same for hubert, just in assets / hubert

[blissful scaffold]

wow you deep

[rustic oyster]

easy done

[leaden yacht]

but I left it running so long cuz I was doing it on an external 1tb ssd, which really has enough space for it

[graceful obsidian]

Neat, now tell me, do you have python and / or ffmpeg installed by any chance?

[leaden yacht]

I just set the epochs it should reach to the max and let it train, and when I know its overtraining I js stop it

[rustic oyster]

pretty sure

[graceful obsidian]

can you verify the python version?
go into cmd and type in " py "

[rustic oyster]

3.11.1

[leaden yacht]

w-okada uses gpu right?

[graceful obsidian]

cause for best rvc compatibility, I'd suggest switching to 3.10.6 if that's not an issue

[graceful obsidian]

can be both gpu or cpu
( radeons and nvidia )

[leaden yacht]

alr

[rustic oyster]

sure thats fine

where can i find that

just on the website?

[graceful obsidian]

Alrighty. Lemme get the link for ya

[rustic oyster]

easy cheers

[graceful obsidian]

https://www.python.org/ftp/python/3.10.6/python-3.10.6-amd64.exe

but before installation, get rid of any python you might have

so, uninstall it

( in case you were using it for something, tell me right away - we'll do packages list backup

[rustic oyster]

yeah nah

i dont think so

never had to use python to my knowledge so

[graceful obsidian]

Good, then ye. Uninstall your current and install 3.10.6 ( only thing to keep in mind is, the option " add to path " or something like that, keep it checked

[rustic oyster]

the file says amd64

is that fine

[graceful obsidian]

that's fine ye

[rustic oyster]

if im on intel

ok ez

[graceful obsidian]

that's for 64 bit windows ( just a naming like that, amd64. cause amd was the one to implement* x86-64 arch

[rustic oyster]

easy was just making sure

[graceful obsidian]

gotya

[rustic oyster]

setup was successful

[graceful obsidian]

alright, can confirm the version now in cmd and if all's working?

py
in cmd

[rustic oyster]

yup

we good

[graceful obsidian]

alrighty, now you'll need ffmpeg

lemme get a link to a very good and trusted guide

[rustic oyster]

i might already have that

easy

[graceful obsidian]

try to verify it's working then ( if you legit have it + added to path correctly

ffmpeg
in cmd

[rustic oyster]

nvm

[graceful obsidian]

yuh, then follow the guide

after you done, @ 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

[rustic oyster]

wait @graceful obsidian

is this rvc folder supposed to be in my actual c drive

not my downloads

i think i got confused there

[graceful obsidian]

for the sake of avoiding path problems

ye

preferably straight on C ( so, system drive

[rustic oyster]

easy lemme continue

[bleak smelt]

does Ilaria-RVC not work anymore?

[rare goblet]

Ayo? @bleak smelt level 3 !!!

[rustic oyster]

@graceful obsidian ive done all the steps but its not showing up in cmd

[graceful obsidian]

have you added it to the path properly?

[rustic oyster]

im pretty sure yeah

i just followed what the tutorial told me

[graceful obsidian]

try to check if it's present in the system variables

[rustic oyster]

its too long for me to read

[rustic oyster]

yeah thats what i did

[graceful obsidian]

reconfirm it, for the confirmation sake

[graceful obsidian]

and dl'ed the ffmpeg like so:
C:\Program Files\FFmpeg\bin
?

[rustic oyster]

think so?

[rare goblet]

Ayo? @rustic oyster level 4 !!!

[graceful obsidian]

ye, that works too

hmmm, maybe try to restart the pc

and check then

[rustic oyster]

okay

still doesnt show in cmd

[leaden yacht]

would my 4060 be fine if I try to train an hour long dataset?

[graceful obsidian]

that's super weird

bin contains these?

[prime forge]

@graceful obsidian

[rare goblet]

Ayo? @prime forge level 1 !!!

[rustic oyster]

not too sure

[prime forge]

ım stuck in the menu

[rustic oyster]

theres ones with that name

but idk if its that file type

[graceful obsidian]

hmm, that might the issue

lemme repack mine then

[rustic oyster]

okay

[rustic oyster]

whats this

[prime forge]

rvc gui, idk ım just stuck in there idk what ım supposed to do

[graceful obsidian]

once it's uploaded, the file in here, unpack it directly to ffmpeg dir on C

[graceful obsidian]

oh, well

[graceful obsidian]

I've no idea what you're using

what fork is it?

[prime forge]

rvc gui

[graceful obsidian]

seems pretty outdated

not even rmvpe in there

[graceful obsidian]

@rustic oyster but ye, this one

[rustic oyster]

ok lemme try that

[prime forge]

wdym

isnt it lastest version

[graceful obsidian]

nope

[rustic oyster]

idek what im looking at

[graceful obsidian]

Good, means you had something off with your files

[rustic oyster]

okay easy easy

[graceful obsidian]

newest is og / mainline rvc

check " retrieval voice based conversion webui " on github
will have package called " updated 1006v2"

[graceful obsidian]

Alright, now we can move onto the next step

go into rvc folder

[rustic oyster]

yup

[prime forge]

https://huggingface.co/lj1995/VoiceConversionWebUI/commit/88e42f0cb3662ddc0dd263a4814206ce96d53214

this one?

[graceful obsidian]

open up cmd in there

prada ^

[rustic oyster]

yup done

[graceful obsidian]

nope, this:
https://huggingface.co/lj1995/VoiceConversionWebUI/blob/main/RVC1006Nvidia.7z

[prime forge]

okay lemme dowland it

[graceful obsidian]

alright, now type in:

pip install -r requirements.txt

[rustic oyster]

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

[graceful obsidian]

mh, another path issue I see
did you click " add to path " during python installation?

[rustic oyster]

i didnt have that optino come up i think

[graceful obsidian]

how about pip

[rustic oyster]

idek what pip is

[graceful obsidian]

well nevermind, we'll just go n grab it

in other cmd window:

curl https://bootstrap.pypa.io/get-pip.py -o get-pip.py

[rustic oyster]

doner

done

[graceful obsidian]

then, ( same cmd )

python get-pip.py

[rustic oyster]

Python was not found; run without arguments to install from the Microsoft Store, or disable this shortcut from Settings > Manage App Execution Aliases.

bro oml

[graceful obsidian]

uhhh

you installed 3.10.6 and mentioned it's working

when I asked you to verify it ( py in cmd, post installation of 3.10.6 )

[rustic oyster]

it came up

and told me i had 3.10.6

Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.

[graceful obsidian]

so that works in cmd, correct?

[rustic oyster]

yes

[graceful obsidian]

huh, yea, so that makes 2 path configurations for you

first for python

https://datatofish.com/add-python-to-windows-path/

and if all goes well, you won't ( most likely ) need it the same for pip ( similar procedure

[bleak smelt]

public url wont show up when trying to use a webui rvc, it only shows local

[rustic oyster]

ts is so confusing

[rustic oyster]

yeah things cooked

i might try tomorrow

its 3am rn

[bleak smelt]

oh so it's happening to everyone?

[rustic oyster]

@graceful obsidian ill try tmr shits so late cant even understand

[rustic oyster]

yes

[graceful obsidian]

tl;dr

just replace the username with yours

[rustic oyster]

so i just add all of those

[graceful obsidian]

it's in the env variables

yes

those python ones, ye

[bleak smelt]

oh okay, damn, any alternatives? I tried using replay but keep getting an infinite download error

[graceful obsidian]

oh, nono

just helping the guy install rvc

[rustic oyster]

he means the web version

[graceful obsidian]

as for your guys' issue, must be colab ( notebooks are down, kinda)

or other notebooks

but iirc, it should be fixed in a bit

[rustic oyster]

ok done

[graceful obsidian]

verify in cmd, one by one:

py

pip

ffmpeg

just to make sure it's 100% working

if pip isn't, there will be last step for it

[rustic oyster]

Traceback (most recent call last):
File "<stdin>", line 1, in <module>
NameError: name 'pip' is not defined. Did you mean: 'zip'?

[graceful obsidian]

a yea

that one will be easy

curl https://bootstrap.pypa.io/get-pip.py -o get-pip.py

in new cmd

then, same cmd:

python get-pip.py

and done

[rustic oyster]

C:\Users\broiz>python get-pip.py
Python was not found; run without arguments to install from the Microsoft Store, or disable this shortcut from Settings > Manage App Execution Aliases.

broooo

[graceful obsidian]

tf

[fast shadow]

-train

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• 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
• Ilaria RVC, by thestingerx Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab
• 🆕 UVR5 NO UI for Google Colab, by Eddy Google Colab

How to use?

🤗 Hugginface Spaces

• Applio, by IA Hispano Huggingface 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

[rustic oyster]

i might give up for the night

[graceful obsidian]

we'll see what's going on next time then, get some rest

[rustic oyster]

easy cheers anyway

[graceful obsidian]

cheers

have a good one

[graceful obsidian]

ye

[latent wagon]

hi, why does it only let me choose crepe_tiny and rmvpe_onnx for the F0 DET? all of the other options have N/A after them and I cannot select them

[graceful obsidian]

I'd assume you've no nvidia gpu or are an amd user?
/ got the non cuda w-okada?

[latent wagon]

yeah i only have integrated graphics on amd

[graceful obsidian]

yup, so that's that

you gotta get dml w-okada ( which you most likely have

or use rvc's native voice changer in dml mode ( radeon supported )

[peak osprey]

now i'm the last one standiiiiing

[latent wagon]

ohh tysm

[graceful obsidian]

yuh

[prime forge]

ı dowland it

how can ı open it?

[graceful obsidian]

you got python 3.10.6, ffmpeg ( all properly configured )
right?

[prime forge]

idk ı js dowland it

[graceful obsidian]

a

basically, you gotta install python 3.10.6, add it to the " path "
install ffmpeg and add it to the path

without these, rvc can't work properly

[prime forge]

oh k

[graceful obsidian]

want me to give you a briefing?

[prime forge]

u mena

u mean

vc?

[graceful obsidian]

nono, here

uhhh, there, install this:
https://www.python.org/ftp/python/3.10.6/python-3.10.6-amd64.exe

these have to be checked:

[prime forge]

found a match, I will do it later

[latent wagon]

is that the one named onnxdirectML-cuda?

[rare goblet]

Ayo? @latent wagon level 1 !!!

[graceful obsidian]

yes

oh, hold on

the namings are confusing

gonna get you the link

[latent wagon]

since that one is the one i have downloaded rn, but it says n/a for the f0 det

[graceful obsidian]

https://huggingface.co/wok000/vcclient000/blob/main/MMVCServerSIO_win_onnxdirectML-cuda_v.1.5.3.18a.zip

this one

that's the one u have?

[latent wagon]

yes

[graceful obsidian]

your model is .onnx converted?

other than that, can you show me the ui / how ur w-okada is configured? ( screenshot )

[graceful obsidian]

you should use rmvpe really

[graceful obsidian]

rmvpe_onnx ye

normal rmvpe is for pytorch

cuda ( if I am remembering it correctly

[latent wagon]

oh ok

[latent wagon]

but is there a way to select the other f0 det's?

[graceful obsidian]

also I can tell you right away

your chunk and extra most likely aren't right set

I myself don't use w-okada anymore but rvc's native thingy but

for reference

is what I somewhat stably used on my rx 570 4 gig

[latent wagon]

ah i see

[graceful obsidian]

or either that

extra should be either 8192 or 4096
as for chunk: 700 something to perhaps 256 or 192
( chunk is the main one if it's latency-responsible, iirc

[rare goblet]

Ayo? @graceful obsidian level 25 !!!

[latent wagon]

cool tysm

[graceful obsidian]

also, nope, those other ones aren't available for dml

[latent wagon]

oh i see

[graceful obsidian]

there's also fcpe but that's on native rvc's voice changer ( and perhaps nvidia ver of w-okada? not sure )

which is kinda rmvpe but more speech-efficient

but that's really it, for cpu / amd use cases, it's crepe-tiny, rmvpe and fcpe in some scenarios

[graceful obsidian]

best of luck ✨ have fun with vc ~

Imma go back to my work for now, cheers folks ~

[errant zealot]

Why does public url not work?.

[worthy lantern]

good question, i have same problem

[errant zealot]

I think the problem is from the server

It is busy?

Does somebody know more details?

[worthy lantern]

yeah prolly

[viral pewter]

I'm using Voix

I mean Voice.ai

and I can't get it to sound like the character I want it to

[errant zealot]

The server is busy

I know

[viral pewter]

like the discord server?

[errant zealot]

No

Colab

[errant zealot]

This

[viral pewter]

how do I get coins with voice.ai?

[errant zealot]

I have met the same problem

[viral pewter]

what do I do?

[errant zealot]

Yes

With the coins

[crude flame]

Don't use voice.ai it sucks

[viral pewter]

what should I use?

[rare goblet]

Ayo? @viral pewter level 1 !!!

[crude flame]

W_okada or native RVC voice changer

[viral pewter]

is there a link to them in the discord server or do i just look them up?

[graceful obsidian]

voice AI is not supported on this server

[errant zealot]

I use google colabs:https://colab.research.google.com/github/Blane187/RVC-notebook/blob/main/Easy_GUI.ipynb

But the public url isn t working

[crude flame]

Shapow guide link: https://rentry.co/VoiceChangerGuide

[odd shale]

It's because Gradio is down atm.

[errant zealot]

For how many long?

[rare goblet]

Ayo? @errant zealot level 2 !!!

[odd shale]

I don't even know, ask Gradio devs.

https://status.gradio.app/

You can check Gradio's status here.

[nocturne mural]

well, since we are in this situation, I decided to add local tunnel as a temporary replacement to applio's colab.

[peak osprey]

hello umm

i converted my first ai cover successfully (a voiceover for alastor) but he doesn't sound too much like alastor, he sounds like a mixture between me and him, not just him

so it doesn't sound 100% alastor

how do i fix that?

[stark rose]

Help how google colab

[odd shale]

You can't.

[peak osprey]

what do you mean?

[odd shale]

If you didn't make that model, probably the original author of the model didn't clean the dataset properly or didn't train the model properly.

Also, if you use your voice as an input, you'll must play around with the transpose till it sounds like Alastor.

Also, there are some voices that simply don't fit for AI Covers.

[peak osprey]

https://discordapp.com/channels/1159260121998827560/1203627921546289182/1203627921546289182

[odd shale]

Alastor did sing on Hazbin but my theory is that probably you're just using an not-that-good model of Alastor.

[peak osprey]

well that's the voice i choosed, is there a way to know it's not the proper choice before trying it?

[blissful scaffold]

what does this mean

[odd shale]

Nope, you must test it.

that's how RVC works.

It's voice to voice

[peak osprey]

alright, what does epochs mean?

should it be high or low?

[odd shale]

Actually that has no answer.

While you train a model, you must check the tensorboard to know if your model is overtraining or not.

[peak osprey]

ok

https://huggingface.co/spaces/r3gm/RVC_HFv2

i am using this platform, does it affect the process?

[odd shale]

Nope, but it's outdated

[peak osprey]

and are there better rvc's other then that?

[odd shale]

There are colabs you can use like RVC Disconnected and Applio.

[odd shale]

RVC is RVC, there's no better version.

Just different forks.

[peak osprey]

how is it outdated?

[worthy lantern]

is the public url gone? @odd shale

[odd shale]

Gradio is down temporarily, you're forced to wait.

No need to ping staffers.

[odd shale]

That HF space was made months ago.

Also, HF spaces are kinda slow.

I don't recommend you to use that space

[peak osprey]

so rvc's get slow in a term of few months

[odd shale]

😐

[peak osprey]

sorry i mean old

[odd shale]

I think you're just misunderstanding things.

[peak osprey]

wrong word

[odd shale]

Altho that space you showed is a kinda old version of Applio.

[peak osprey]

alright

[rare goblet]

Ayo? @peak osprey level 4 !!!

[peak osprey]

i've tried appolio and that "llila" thing forgot its name but i didn't find the public url

why?

[odd shale]

Applio and Ilaria RVC

[peak osprey]

yeah

[odd shale]

I've already said Gradio is down

You can't do anything beside waiting till it gets up again.

[elder estuary]

issues with gradio at the moment i think

[odd shale]

You're right.

Indeed Gradio is down.

[elder estuary]

https://status.gradio.app/

[peak osprey]

how long do they take usually to come back?

[brittle wing]

F

[peak osprey]

sorry for my annoying questions but i have 0 knowledge in these things

[brittle wing]

i was just about to report the fact it wasnt showing up for me

[elder estuary]

im not sure

[hard sparrow]

can anyone help with okada?

[odd shale]

You all will never know when Gradio gets fixed sadly

All you can do is check gradio status page and wait.

[brittle wing]

that sounds ominous asf

[odd shale]

Because yeah, the only way to know if Gradio goes up again

is checking the page.

[brittle wing]

no i just meant the way you phrased it sounded ominous

[odd shale]

My english gramatic isn't that good, sorry.

It's just my way of comunicating, it's not intentional.

[brittle wing]

lmao ur good dw 👍

[odd shale]

🐢 👍

Tutel is not rude.

[odd shale]

Request for help in the #1192011222023950368 or #🔍│help-w-okada channel

[peak osprey]

what are exactly gradio, what do they do?

[odd shale]

Gradio is the interface that RVC uses.

Without that, you can't do anything.

You can't convert audio neither train models.

If you would like to do that without Gradio, your option is using RVC disconnected colab.

[peak osprey]

alright

i wanted to use a voice changer to speak in real-time just like they do in VRchat but they said you need a powerful RTX or somesort of that, is it true?

i mean it's only for powerful computers and not mid

[odd shale]

Actually:

[fathom carbon]

i do that in colabs aswell

[odd shale]

https://rentry.co/VoiceChangerGuide

Maybe this guide can help you.

[peak osprey]

that's alot to read

[odd shale]

[violet heron]

You would, just without a UI

[peak osprey]

i have iris xe gpu and intel core i7, will these help?

[violet heron]

(Depending on the fork how much it depends on it)

[odd shale]

For that reason i mentioned the usage of RVC Disconnected tangovill

[violet heron]

No

[peak osprey]

[violet heron]

Better off using google colab

[odd shale]

Timbo, please let me talk with him.

[violet heron]

K

[odd shale]

Let me see, what you want to do?

AI Covers? training? realtime?

[peak osprey]

realtime for now

[nocturne mural]

temporarily now to access the gradio you will have to go to the link that appears in the console, insert the password shown in the console and that's it.

[blissful scaffold]

okay and what purpose does that serve, what happend to public url

[odd shale]

Gradio is down.

[odd shale]

Which are your PC specs?

[nocturne mural]

It's supposed to be fixed by now but it seems it's not yet.

[peak osprey]

hol up

its a laptop not a pc

[odd shale]

Vidal, le estaba diciendo a red que era lo que pasaba con Gradio todo el rato.

Dejame ver el status

[nocturne mural]

gradio esta bien

lo q esta caido es lo del share

[odd shale]

SII eso si está caido.

El Share es lo que jode todo por ahora.

[blissful scaffold]

aw must be so much hard work 😭 🙏🏻

[sand iris]

Are y'all 100% sure gradio share links are down

[sand iris]

"Operational"

Is what I'm reading

[odd shale]

Share API is what fucks up everything.

[odd shale]

Nope darocho, i've already tried to open gradio in Paperspace and it's still down.

Don't get fooled by that "Operational"

[sand iris]

Man they're interesting

• doesn't download share link binary, fails because of that
• installing it anyways just leads to a timeout
• status page doesn't reflect that

[odd shale]

"Operational" rn is kinda misleading

[barren acorn]

well

[rare goblet]

Ayo? @barren acorn level 3 !!!

[barren acorn]

i watched vid for the app

but

i didnt find it

1.5.3.5a.zip @acoustic scarab

[odd shale]

It's not necessary to ping helpers.

[barren acorn]

oh mb

so

where to find it

[odd shale]

What are you trying to do?

[barren acorn]

dowloading the voicchnager

1.5.3.5a.zip only one works fine with me

[odd shale]

Which are your PC specs?

I know a helper that will be able to help you.

@pastel oak

[barren acorn]

dm him?

[graceful obsidian]

nah

you can use it just fine

back to when I was using w-okada, I had it just fine. Sure, delay is there but it works just fine
rx 570 4 gb vram, that's a 2017 mid range gpu

[peak osprey]

So are my specs ok for this?

[peak osprey]

.

[graceful obsidian]

iris xe, isn't that intel?

[pastel oak]

What GPU do you have

[peak osprey]

Yeah

[graceful obsidian]

a ye

[wispy lodge]

For just voice chatting - maybe. But vc + gaming - unlikely

[barren acorn]

idk how to see it

[graceful obsidian]

that's true ye

[barren acorn]

oh gtx 1660

[pastel oak]

open task manager - go to performances

ok

[peak osprey]

Yeah i just want to do voice chatting

[pastel oak]

https://huggingface.co/wok000/vcclient000/blob/main/MMVCServerSIO_win_onnxgpu-cuda_v.1.5.3.18a.zip

Download this

Then download this to use it on discord etc
https://software.muzychenko.net/freeware/vac470lite.zip
"setup64.exe"

you open the app with "start_http.bat"

For the rest of the steps setting up the audio, use the guide
https://rentry.co/VoiceChangerGuide#audio-setup
Step 13 (its linked in the url)

[wispy lodge]

Well, in VRChat? Because you mentioned it

[barren acorn]

tym<:

[wispy lodge]

I'm not sure how well it works with integrated GPUs, especially intel ones

[peak osprey]

No it's just an example i don't even play VR chat

[wispy lodge]

But CPU-only inference is also an option, delay is not that high with it on good CPUs

[severe falcon]

i havent did anything ai wise in a bit, is there a new site or do we still use the google colab?

[violet heron]

We still use google colab for free online training

[peak osprey]

Ok how can i DO IT?
how can i chat using ai?

[severe falcon]

well, i try using the huggingface links for the colab, but they dont work. can you send me the link of the one you use?

[violet heron]

The HF model link for the google colab?

[severe falcon]

i jus use a colab that says "Easy GUI (for RVC v2, with crepe) (with improved downloader)", iunno if theres a new one or not

again its been a good minute since i did anything like this

[violet heron]

That one is very outdated

-colab

[severe falcon]

oh shit

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• 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
• Ilaria RVC, by thestingerx Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab
• 🆕 UVR5 NO UI for Google Colab, by Eddy Google Colab

How to use?

🤗 Hugginface Spaces

• Applio, by IA Hispano Huggingface 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

[violet heron]

inference (covers) or training?

[severe falcon]

wyhich one

[severe falcon]

im jus tryna make voice covers

[violet heron]

Ilaria RVC, Applio, or AICoverGen

for ones with ui

[severe falcon]

thank ye

[rare goblet]

Ayo? @severe falcon level 1 !!!

[junior halo]

Is everyone else still having issues with Gradio? Because I am.

[blissful scaffold]

easyGUI not working for me

anyone got something similar?

not used anything else 😂

[odd shale]

Gradio is down.

[proper shale]

easy gui v2 on colab is broken, check https://docs.aihub.wtf for newer colabs that work

[proper shale]

oh really?

[blissful scaffold]

sorry what does gradio mean?

[proper shale]

https://status.gradio.app/ is up, so dw

[proper shale]

it's the thing that runs the UIs

in rvc and other applications

[blissful scaffold]

interesting okay thank you

[odd shale]

I would have to test in paperspace if it's really back up

[junior halo]

How else am I supposed to test my new model without Gradio? ; - ;

[proper shale]

gradio is currently up, so

https://media.discordapp.net/attachments/806002858649452555/997576203189887097/shit_and_die.gif

[blissful scaffold]

public url doesnt show up

[sand iris]

it drunk

[blissful scaffold]

why does everybody want public url

does it make a difference?

[sand iris]

colab doesn't have any open ports

[proper shale]

cuz it doesnt work otherwise

[sand iris]

gotta rely on some sort of mechanism to access the UI; gradio has one

[blissful scaffold]

okay didnt know that thank you

[barren acorn]

how to use it on discord?

[proper shale]

need a virtual cable, check the https://rentry.co/VoiceChangerGuide

[barren acorn]

i did that

[leaden yacht]

What do I choose for training if all of my dataset .wav files are around those numbers? 32k, 40k, or 48k?

[fast shadow]

Hi!
I make AI RVC covers on mobile, but idk why it doesn't give me the Public URL...
What can I do?

[proper shale]

now just set it up correctly, as the guide tells ya how, in this part: https://rentry.co/VoiceChangerGuide#discord-crackle-fixes

ofc set ur output in w-okada as the virtual cable that appears there

[shadow umbra]

Hello

[thorn abyss]

do yall know when is colab fixing the public url problem?

[leaden yacht]

['extract_f0_print.py', 'T:\AI\Mangio-RVC-v23.7.0/logs/Narrator-TSP', '8', 'rmvpe', '64']
f0fail-38-T:\AI\Mangio-RVC-v23.7.0/logs/Narrator-TSP/1_16k_wavs/211_1.wav-Traceback (most recent call last):
File "T:\AI\Mangio-RVC-v23.7.0\extract_f0_print.py", line 383, in go
featur_pit = self.compute_f0(
File "T:\AI\Mangio-RVC-v23.7.0\extract_f0_print.py", line 245, in compute_f0
f0 = self.model_rmvpe.infer_from_audio(x, thred=0.03)
File "T:\AI\Mangio-RVC-v23.7.0\rmvpe.py", line 375, in infer_from_audio
hidden = self.mel2hidden(mel)
File "T:\AI\Mangio-RVC-v23.7.0\rmvpe.py", line 355, in mel2hidden
mel = F.pad(
RuntimeError: Argument #4: Padding size should be less than the corresponding input dimension, but got: padding (0, 21) at dimension 2 of input [1, 128, 11]

f0fail-99-T:\AI\Mangio-RVC-v23.7.0/logs/Narrator-TSP/1_16k_wavs/393_1.wav-Traceback (most recent call last):
File "T:\AI\Mangio-RVC-v23.7.0\extract_f0_print.py", line 383, in go
featur_pit = self.compute_f0(
File "T:\AI\Mangio-RVC-v23.7.0\extract_f0_print.py", line 245, in compute_f0
f0 = self.model_rmvpe.infer_from_audio(x, thred=0.03)
File "T:\AI\Mangio-RVC-v23.7.0\rmvpe.py", line 375, in infer_from_audio
hidden = self.mel2hidden(mel)
File "T:\AI\Mangio-RVC-v23.7.0\rmvpe.py", line 355, in mel2hidden
mel = F.pad(
RuntimeError: Argument #4: Padding size should be less than the corresponding input dimension, but got: padding (0, 21) at dimension 2 of input [1, 128, 11]

When trying feature extraction

[proper shale]

it's not colab, it's gradio

[thorn abyss]

ohh

[proper shale]

https://status.gradio.app/ share link stuff seems to be workin rn tho

but idk... id say, give it some time

might go down again

[thorn abyss]

aight

is there any alternative?

no covergen

[proper shale]

not sure tbf....

[thorn abyss]

its okk

dw

[brittle wing]

It doesn't generate public URL

[thorn abyss]

ui's aint working

[violet heron]

Did you put in your ngrok token

[brittle wing]

Uh I have the exact same problem

[proper shale]

aicg probably dead rn

[brittle wing]

It worked before without ngrok token

[proper shale]

check if ilaria rvc or other stuff works

rn

[brittle wing]

There's no web UI version through

[thorn abyss]

does applio use gradio?

[violet heron]

Last time I checked I think so

[brittle wing]

https://colab.research.google.com/drive/1u1brjK8IZt647UsbZuGYfW29oFM2I4tk?usp=sharing

[thorn abyss]

let me check

[proper shale]

https://colab.research.google.com/github/iahispano/applio/blob/master/assets/Applio_NoUI.ipynb exists

[barren acorn]

where to find good voices

[proper shale]

#1175430844685484042 (listen to the samples) or https://weights.gg

[thorn abyss]

https://voice-models.com/ you can also try looking models in here, some of them don't work tho

[brittle wing]

Ilaria generates noise I don't recommend it

[proper shale]

huh

that's most likely your model, not an entire fork

[brittle wing]

Not my model 🙄

[proper shale]

then that's the model/audio you're inferring on

[brittle wing]

I have used it with a good model and it generates noise in the samples

[brittle wing]

Probably the audio

[proper shale]

yeh that happens

[thorn abyss]

try using noise gate or denoise in uvr5

btw applio with ui aint working

[brittle wing]

I should Denoise my samples bef converting?

That's a good idea but have never faced this issue with Applio

[thorn abyss]

yep

it will sound better

[brittle wing]

https://colab.research.google.com/drive/1u1brjK8IZt647UsbZuGYfW29oFM2I4tk?usp=sharing
That's the same thing as Cover generator without UI

[thorn abyss]

bc applio has a "clean audio" cell

if you run it in colab

ilaria doesnt

[brittle wing]

This probably explains why my model sounded bad after inferring even through the dataset was Denoised

[thorn abyss]

probably

[brittle wing]

I have done it without the clean audio cell and it still sounded noiseless

[thorn abyss]

i would say applio is better at denoise

i don't use ilaria that much

[dusty lava]

It’s gradio still down?

[rare goblet]

Ayo? @dusty lava level 1 !!!

[thorn abyss]

i cant tell you about it

[thorn abyss]

yep

[brittle wing]

Maybe 👀

What causes it

[worthy lantern]

yup

[brittle wing]

Maybe my fault -

[worthy lantern]

have we any clue when gradio will work?

[proper shale]

frankly we don't know

seems to be going up and down

https://status.gradio.app/

[brittle wing]

Where do I set share=true?Where do I type it

[nocturne mural]

enter this link and then enter the password that is also displayed in the console.

[brittle wing]

What's the password.

[brittle wing]

Oh thanks

Wait where is it

Ah I see

[nocturne mural]

on the colab console

xd

[brittle wing]

After the link

Ok

[nocturne mural]

yes

[brittle wing]

WORKED

[nocturne mural]

yeah

[worthy lantern]

[rare goblet]

Ayo? @worthy lantern level 5 !!!

[foggy folio]

how do I make the perfect acapella for ai covers

[rare goblet]

Ayo? @foggy folio level 8 !!!

[foggy folio]

it doesn't need to be totally perfect

it just needs to be clear

[barren acorn]

its so lagy

i cant even change voices

all the time errors

[brittle wing]

https://colab.research.google.com/drive/1X8YR4Ruv7zzY8YAMPTfC7hkxqT_d4Q5d?authuser=3#scrollTo=fMzuTwbESg58
Use this...

[barren acorn]

error

Notebook loading error

[proper shale]

I'd say, use BSRoformer, model with highest SDR (in https://www.mvsep.com - register for lower queues)

Then maybe use DeReverb-DeEcho inside UVR models in case reverb screws up the model

[proper shale]

what chunks r u running?

and what f0 det were u using

[onyx ferry]

🙂

[proper shale]

ain't no way, its mikuuu

[thin stump]

Oh wait it's working

[junior halo]

Where can I submit an idea for the Gradio interface? I'd like there to be a way for me to skip the audio processing step as I often already separate the vocals using a different website.

[rare goblet]

Ayo? @junior halo level 4 !!!

[proper shale]

just use another program, AICoverGen does the separation since it expects u to drop the song with instrumentals n stuff

[analog aspen]

hello

i need help

[proper shale]

explain the issue, give context, any type of log that could help us, if you're running locally, what are ur specs?

[analog aspen]

my spec is gefore rtx 3050 ti

[brittle wing]

I may be dumb but how do I get python to install this

already installed pytorch pip

[proper shale]

most likely that amd/intel version misses that requirements.txt (considering it misses the launch . bat, literally)

also, make sure you're doing the commands on the folder, iirc you should do cd (insert folder name here)

but idk, ive been a Mac user for a while

[brittle wing]

sorry but can I ask where that specific folder should be located at

[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

[proper shale]

that's something you should answer

[mystic marsh]

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• 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
• Ilaria RVC, by thestingerx Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab
• 🆕 UVR5 NO UI for Google Colab, by Eddy Google Colab

How to use?

🤗 Hugginface Spaces

• Applio, by IA Hispano Huggingface 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

[brittle wing]

when i try and use the voice changer on discord its very laggy

like really bad

[nimble sail]

Im having a "JSONDecodeError: Expecting value: line 1 column 1 (char 0)" When I try google collab ai cover. Does anyone know what this is?

[brittle wing]

and I need help trying to fix it

[pastel oak]

outdated colab, do -colab in this channel for the new ones

[nimble sail]

How do I do that?

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• 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
• Ilaria RVC, by thestingerx Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab
• 🆕 UVR5 NO UI for Google Colab, by Eddy Google Colab

How to use?

🤗 Hugginface Spaces

• Applio, by IA Hispano Huggingface 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

[nimble sail]

ok thnx

[pastel oak]

what gpu do you have

[nimble sail]

Amd

[pastel oak]

not you

xd

[nimble sail]

sorry

[brittle wing]

geforce gtx 1050

is it bad?

[pastel oak]

well itll have way higher delay but we can try to optimize it to the best ability

Do you have onnxgpu-cuda version or onnxgpu-DirectML?

on the top right of voice changer

[brittle wing]

cuda

[pastel oak]

Good

do F0 det: rmvpe
Chunk: 192

Extra: 8192

[brittle wing]

ok

[rare goblet]

Ayo? @brittle wing level 1 !!!

[pastel oak]

Now that youre lvl 1 you can just send a screenshot instead

[brittle wing]

oh good