✨│ai-help

[lone lance]

so i used https://colab.research.google.com/drive/16LkwvFZeudTpUOsE_6bMjOq2qkxFo8Hr?usp=sharing#scrollTo=9yQA9h6N2YV7 and I did all the steps in the guide https://docs.aihub.wtf/rvc/cloud/ilaria-rvc/ but then when I clicked convert errors showed up

[low shard]

what model did u use? do u got the download link

[lone lance]

uh yes

let me get it wait

I also used other models to test if those work as well but they just showed error.

Here’s the link https://discord.com/channels/1159260121998827560/1219038257136926800

[proven hill]

what error?

[lone lance]

it was just a text in red called error

I tried multiple times

[low shard]

the model files look correct, can u retry and show the error?

[lone lance]

alright give me 5 minutes.

wait now it works

great

! 😃

[teal stag]

how do I upload and share my model on huggingface, like I did it but when I ran the RVC (the Ilaria version), go to the "download voice models" section and paste the hugging face link for the model it won't work

sorry in advance bcuz idk if i was describing it clearly, just starting out the thing recently

[swift cradle]

same i got red "!"

on the second step

[teal stag]

the files of the voice model i trained is pretty fine, i guess, i did it correctly as the tutorials but when it comes to uploading it on huggingface things get odd

i mean i actually uploaded it but can't use it in any RVC version

[low shard]

Be sure that u dont close the google colab or that it doesn't get disconnected

That prolly was the problem

[lone lance]

okay 👍

thank you

[low shard]

What google colab are u using? And wdym with !

Yw

[low shard]

Share the model download link

[swift cradle]

easy gui

[low shard]

Share link u may be using the old one

[swift cradle]

https://huggingface.co/lioooooooooo/v2/resolve/main/cartiv2.zip?download=true the link im using

[teal stag]

https://huggingface.co/Damo2910/NTANCVA/resolve/main/DAAN.zip?download=true here, idk if it's right

thank you for helping btw

[low shard]

The model files are correct, i meant send the google colab link not the model

[low shard]

Did u put it public or nah

[teal stag]

i thought i did

lemme check

yea i think i did

is it

[swift cradle]

https://colab.research.google.com/drive/1Gj6UTf2gicndUW_tVheVhTXIIYpFTYc7?usp=sharing sorry haha im new to this here is the colab

[rare goblet]

Ayo? @swift cradle level 1 !!!

[low shard]

Outdated colab thats why, please check the ones in our docs https://docs.aihub.wtf also dw

[low shard]

Mmm kinda weird lemme retry

[low shard]

U putted it public but people need access to see the files, which is why it doesn't work

[teal stag]

oooh okay

[low shard]

Try to change license to openrail or mit

[teal stag]

thanks, i'll try it now

[rare goblet]

Ayo? @teal stag level 2 !!!

[teal stag]

sorry if im asking again but i changed everything as you said and the thing says the model was downloaded, but when i get back to the inference section and start spamming refresh for a bit it still doesn't appear

[low shard]

The model got a weird macos folder and doesn't have the added.index

Also the pth file should be .pth without the copy

[teal stag]

ok thank you, that explains

[clear lily]

yo so im trying to isolate vocals but i get this error. any fix?]

hello?

[odd shale]

Isolating vocals with which program?

UVR?

[prisma igloo]

So I'm using Rvc v2 disconnected and my model was at 150 epochs and it just disconnected, how would I go to continue training on a different account

[pulsar coral]

i need help every time i try open the voice thing it closes instantly

[pastel oak]

wokada?

[pastel oak]

#🔍│help-w-okada message

Do this

[tough fiber]

should i continue to training?

this graphic weird lol

idk is it overtraining or not

[pulsar coral]

do uk how to help?

[wispy burrow]

I would train it a little more.
It could be that G/Total will still go down.
But if it rises, it's overtraining.

[proper shale]

yeah this is fine

still decreasing

[tough fiber]

i know but

i cant understand on this graphic

is it overtraining or not

well i guess it started overtraining

[proper shale]

nooope not yet

i wouldnt say so

[tough fiber]

its increasing so

[proper shale]

well, here's the part that people don't get (because the tutorials sadly haven't really described this well)

g/total is just your average model (well, technically, generator average, but eh it's your model anyway)

just because g/total rises doesn't necessarily mean it's overtraining

[pastel oak]

Litsa shouted at me when I said this same thing. I've not recovered from that and dont say this to anyone anymore 😭

[proper shale]

Wow

That's not very Litsa like

I think he was in a bad day

[pastel oak]

(shouting might be an interpretation here. I dont remember, i fully supressed that from my memory)

[proper shale]

😭

[pastel oak]

(Litsa has been nothing but helpful with understanding the theory behind everything so cant talk bad for learning)

hey btw do you know who wrote this guide

[proper shale]

Litsa is truly based

[pastel oak]

https://rentry.co/RVC_making-models

[proper shale]

Which one

[proper shale]

LUSBERT

our favorite iranian

sadly tho, it's filled with misinfo

...some parts of it, at least

overtraining shit is completely incorrect (like, d rising, g down= overtrain is madness)

but

it has useful stuff like graph logs n shit

[pastel oak]

Pretty much, what was said in this guide is

If g and D are going up, check if mel and kl are going down - if yes, its still good

[rare goblet]

Ayo? @pastel oak level 38 !!!

[pastel oak]

and I applied that logic but that opinion seems to be divided from other people on the server and I dont know fully how to judge it. Like if mel, kl and d are heavily going down while g is going up, id still consider picking one of those after the lowest point of gtotal

I still love your suggestion about just combining the best average model with the best mel/kl point. I wanna try that sometime lol

you ever tried it?

[proper shale]

sadly no

but ik codename does it, and they make good models so

¯_(ツ)_/¯

[tough fiber]

so xD which graphic should i check

my mind confused rn

[proper shale]

thats where i got the idea from 😭

[proper shale]

check g/mel, g/kl

and g/total

maaaaybe even d/total but eh it won't matter that much tbf

[tough fiber]

so all of them should be decrease right

[proper shale]

yeah, if mel and kl start going up for a sustained period of time it might not be improving, so probably in overtraining

[tough fiber]

so means it isnt overtraining

[tough fiber]

i hope i understand correct

[tough fiber]

everyone gone ? xd

thats a weird graphic for g/t

[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

[thin stump]

it started overtraining after 30k

[wispy burrow]

The latest model I trained has it's G/Total and G/Kl are basically flatlining for a while. (G/Kl is going down very very slightly)
D/Total and G/Mel are still going down.
I'm already at 500 Epochs (74k Steps)!
Don't tell me I still should train some more...
Ugh...this takes forever...

[tough fiber]

y

they said its not overtraining so

[thin stump]

you do want kl and mel to keep going down but if g/total doesn't want to go down, then it's done

it's rising up after 33k

[tough fiber]

this

thing fukn annoying

i cant use model inference

sometimes this happen and sometimes not

idk WHY

[wispy burrow]

I think your path of the audio file could be wrong.

[proper shale]

yeh random error

[proper shale]

yup, still going

[tough fiber]

i didnt change anything

[wispy burrow]

From my experience it's usually that or spaces/special characters in the path (or filename).
Doesn't mean it couldn't be something else though...

[tough fiber]

i mean it was work

after i refresh page

it gives error

[plush shell]

where can i buy a good egirl voice model

that sounds legit

[dark helm]

how do we turn our model into huggingface url?

[hoary lark]

whad does epochs mean?

[prisma igloo]

Send your model to huggingface and make it downloadable and copy the URL for the Pth file

[prisma igloo]

It's one complete pass of the training dataset through the learning algorithm. In other words, when all the data samples have been exposed to the neural network for learning patterns, one epoch is, as said, complete.

[dark helm]

Step 2 : Download Model - why do I keep getting JSONdecode error?

[rare goblet]

Ayo? @dark helm level 1 !!!

[dim cape]

guys where do i put the share=true in applio? Like, in what line of what py script?

[frail bolt]

around a month ago I submitted a free model but no one ever did it

[analog olive]

hey guys, I'm having some trouble here. When I try to do step 2 the "No module named 'mega'" appears, I then redownload the module and the problem dissapears, but then it says that Mega cannot be recalled from mega, so I don't know what to do anymore, I've been trying to tackle the problem for a while now

[viscid moss]

Submit the model again on #1159514067187277865

[viscid moss]

I'm pretty sure that u re using an outdated colab

Try the new ones

-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

[viscid moss]

U can found guides here:
https://docs.aihub.wtf/

[frail bolt]

but what if you want the model made by someone else? ive never made a model before and wanted one created

[viscid moss]

You can make a post at #1159289738314919936.

Free model request or commission model request. Or you can visit #1191429836321849435 to explore our creators (commissioned)

[frail bolt]

thats the thing, I did request one at #1159289738314919936 like a month ago but no one ever made it

ill make another one ig

[viscid moss]

Well.... you know. People will take a paid request more seriously than a free request 😅

[frail bolt]

what ill do is wait for someone to do the one free rquest and then if I wanna do another one ill pay them

[teal stag]

hi, can somebody explain why it keeps saying error, idk if it's the voice model because yesterday everything was so fine and i didn't make any modification

[viscid moss]

Have u searched everywhere for the model you are looking for?

[viscid moss]

Ur audio file is still in uploading state

wait a bit and try again

[frail bolt]

mhm

[teal stag]

oh yeah silly me

[teal stag]

um yeah the thing is it's still uploading and it's been like 7 minutes

i canceled and try to upload it again but nothing changed

i also try changing the file format from mo3 to wav

[viscid moss]

Is that Colab or Hugging Face?

[teal stag]

huggingface

[teal stag]

where i got the voice model, if thats what u asking

[viscid moss]

I mean... where are you using it?

The website

[teal stag]

oh

colab

my bad sorry

[viscid moss]

Colab is running ? Maybe it got disconnected

[teal stag]

okay thank you

[viscid moss]

Can u send a pic of the colab page?

[rotund pivot]

im trying to do spongebob and whenever i paste the link into the box and click the play button, the play button turns red and says error occured "# file exists, proceed with creation of creds and client"

[viscid moss]

Hmmm can u send a pic?

[rotund pivot]

wont let me send images in here

[rare goblet]

Ayo? @rotund pivot level 1 !!!

[rotund pivot]

unless you mean in dms or something

[viscid moss]

U re now lvl 1

[viscid moss]

oh

That's an outdated colab

[rotund pivot]

so what do i do?

[viscid moss]

Try the new ones

-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

[viscid moss]

U can found guides here:
https://docs.aihub.wtf/

[rotund pivot]

where do i find the new ones?

[viscid moss]

Here @rotund pivot

I recommend Ilaria RVC https://colab.research.google.com/drive/16LkwvFZeudTpUOsE_6bMjOq2qkxFo8Hr?usp=sharing

[rotund pivot]

okay thank you

i just click play button twice then what?

[viscid moss]

https://youtu.be/wh4cLMLLkj0?si=v57FOcR8FalF9yCL

[teal stag]

here, dont know if this is what you ask for

[rare goblet]

Ayo? @teal stag level 3 !!!

[teal stag]

thank you for helping btw

[viscid moss]

Hmmmm your GPU says connecting, it looks like it disconnected for a moment and that caused an error.

Pls restart the runtime.

go to runtime > disconnect and delete runtime

And start again

[teal stag]

ok thankies

[jaunty flame]

im confused

[teal stag]

it's still said "uploading"

the audio file

[jaunty flame]

i keep getting this error JSONDecodeError

[viscid moss]

damn.

Try Applio colab instead

[teal stag]

awwh

okay thank you

[viscid moss]

This seems good, but ya, ur audio is not being uploaded for some reason

[viscid moss]

I'm pretty sure that u re using an outdated colab

-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

[jaunty flame]

oooooh

i was using Easy GUI (for RVC v2, with crepe) (with improved downloader)

[viscid moss]

yep, that's outdated

[teal stag]

i tryna upload the audio file but this is what it said

[viscid moss]

hmmm what's the audio format?

[teal stag]

.wav

i did

nvm i'll try again later

[brittle wing]

How do I add train ai models from weights gg to ai hub? Or astra labs bot

[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

[frank gazelle]

What is the best RVC to use?

[brittle wing]

I wonder the same, what's the best for locally?

[brittle wing]

That's sus

Overtraining

Should I stop training already?

[thin stump]

@tiny elm I thought no one would ever ask. You can make a post in #1159289738314919936. I might do it but do you want korean or english?

[tiny elm]

english

pls

[thin stump]

I only read the manwha so i haven't watched it yet

[tiny elm]

oh

how to get his voice

[rare goblet]

Ayo? @tiny elm level 1 !!!

[thin stump]

I forgot but there should be episodes I can download off a site

[tiny elm]

will you join vc will you explain me how to download it and use it?

[thin stump]

Read up on our guides here https://docs.aihub.wtf/essentials/voice-models/#voice-model-files

I said I might do it because I don't remember the site on top of my head

[tiny elm]

oh

[low shard]

#1191429836321849435 or #1159289738314919936

[thin stump]

Juju island right out the first episode?? I'm so confused

[tiny elm]

its not that hard also

[thin stump]

kk just thought I skipped a season

[tiny elm]

just read it thrice you will understand

[tiny elm]

oh

[craggy schooner]

Has anyone tried to use okada collab in discord from an android?

[tight ether]

학습에서 생성된 마지막 모델 데이터만 저장할건지 묻는거에요.

[pastel oak]

The file to continue training the voice model. If you select yes, you save one D and G path to continue training. If you select no, you will save as many D and G paths as you selected "save frequency".

Its easier to use Yes here in my opinion

[tight ether]

yup

[pastel oak]

You get the latest D and G version. Not model-epoch version.

[tulip cloak]

you can commission me if you ever want, i take commissions aswell

[frail bolt]

Do you do free models?

[tulip cloak]

nope

[frail bolt]

Well then

Cuz I ain't got money right now

I'll figure something out

[tulip cloak]

no worry

[brittle wing]

Which RVC are you guys using?

locally?

[lofty geode]

Is there a way to use someone else's ai voice and what do I do after downloading it

[thin stump]

Since you left the server, I'm assuming you don't want it. You've never been told about weights.gg yet

[carmine hearth]

index 학습은 feature extraction 뒤에 아무 타이밍에 하시면 되세요. 모델 학습은 무조건 feature extraction 뒤니까 모델 학습 뒤에 하시는 것도 가능합니다.

You can train the index at any time after feature extraction. Since model training is always done after feature extraction, it is possible to train index after model training.

[pastel oak]

if you already have the voice model PTH file of the person you want to use, there are local and cloud options, depending on if you have a good GPU or not. I will link you the online version. If you have a strong GPU like an RTX 2060, I can link you the local version aswell

Cloud: https://docs.aihub.wtf/rvc/cloud/ilaria-rvc/

Read up on "Inference", this is what you will be doing. There is a video tutorial attached aswell

[lofty geode]

I have a Rtx 3060 so should work

[pastel oak]

https://docs.aihub.wtf/rvc/local/mainline/

[lofty geode]

My friend told me they shut this down lol jokes on him

Thanks for the help!

[mighty laurel]

@low shard https://huggingface.co/dvrajput/VoiceModel/blob/main/ChiragVoice_ep360.zip

[lofty geode]

After I have all of this do I have to find a program that will output the AI voice?

[low shard]

the model files are correct, u sure u didnt close the google colab or that it wasnt disconnected?

[pastel oak]

No, this IS the program to use it. The guide I linked explains how to make it work

[mighty laurel]

yep, it was showing error in colab, make a cover, you'll get error

[lofty geode]

And how could I search for pre-made ai voices? For example I want ana from overwatch

[low shard]

show a screenshot of the error

[mighty laurel]

closed colab

you plz try if you've connected runtime

[low shard]

do u get the error while downloading the model or making the ai cover (clikcing convert)

bc im at school rn

[mighty laurel]

first got error on download, but on 3rd tried it showing error but downloaded on server when i refreshed

[rare goblet]

Ayo? @mighty laurel level 2 !!!

[mighty laurel]

then I tried to convert & i got error

which not working on 3rd try also

[low shard]

sorry but i cant really do that as im in school, pls when u can retry the colab show a screenshot here

[mighty laurel]

sure

[dim cape]

ty i just made it run the app.py with --share anyways

[pastel oak]

Several options:

https://discord.com/channels/1159260121998827560/1175430844685484042
https://weights.gg
https://voice-models.com

[wispy umbra]

Hey so I am trying to setup voice changing with collab(https://colab.research.google.com/github/w-okada/voice-changer/blob/master/Realtime_Voice_Changer_on_Colab.ipynb), but it keeps giving me this on any action with models:

Error
unhandledrejection
no error stack
TypeError: (0 , Zz.removeDB) is not a function
TypeError: (0 , Zz.removeDB) is not a function

What am I doing wrong here?

[proper shale]

Might be a browser issue... try using another browser if you're not on Chrome

Also, do you have Colab Pro? That Colab notebook will not work (it will probably disconnect in 3-5 minutes) for free users.

[wispy umbra]

I am on Firefox, thanks for the tip.
I am not on a Pro unfortunately(

[rare goblet]

Ayo? @wispy umbra level 1 !!!

[proper shale]

Yeah I'd recommend you use this then:

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

Hope it works, fingers crossed

[wispy umbra]

Thanks!

[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

[wispy umbra]

Is it normal for download sample catalog to take 30 min?

[still chasm]

@low shard

[low shard]

show the google colab output

[fluid matrix]

why is the audio file uploaded twice

remove it from there and click refresh

then convert again

[low shard]

kinda weird, click the x on the audio u uploaded and try to click refresh on the audios dropdown, and then re select it

[fluid matrix]

yeah do that

[still chasm]

maybe i waited too long ?

[brittle wing]

What is this I'm confused

Overtraining under training????

Is it creating a new point

[pastel oak]

It could, but it might also not. If it starts going up again, you might have your best average model at ~47k ready

[brittle wing]

60k is going down 💀

Is

I mean yes

I have the model at 400 epochs before that low drop

Wait maybe 410?

[iron latch]

guy, when i use applio, i got this error

[brittle wing]

This is 410 epochs

Seems like it's creating a new point

[low shard]

Show the colab output again

[iron latch]

guy ? help

[brittle wing]

Ah no it's going down

It keeps training

[low shard]

Mmm thats very weird, try to show what inputs u putted and also send model download link

[iron latch]

i try reinstall it now

how i clear it to the beginning ?

[brittle wing]

Is it possible that my model can generate over 50k steps???

[low shard]

I meant show me the interface to see what u selected

[iron latch]

wait

[iron latch]

ah i fix this

just reinstall it

[low shard]

Oh lol

[mighty laurel]

@low shard https://pastebin.com/M3KbxvxE

here is whole error

[low shard]

Weird lol, can u retry to click both refres buttons and re select everything? Like model and audio

[mighty laurel]

bro closed 😂

[low shard]

😭

[mighty laurel]

you tey if not in school

*try

[wispy umbra]

Hey
Now I get
TypeError: Cannot read properties of null (reading 'modelSlots')
Any way to fix it? I switched to chrome

[wispy umbra]

Logs: [Voice Changer] post_update_settings ex: [ONNXRuntimeError] : 7 : INVALID_PROTOBUF : Load model from pretrain/rmvpe.onnx failed:Protobuf parsing failed

[proper shale]

Hm... did you skip any steps?

[wispy umbra]

Nope

[proper shale]

Maybe the download failed... thats weird

Only happens on the local version

or well, only happened*

[brittle wing]

Can you explain what you meant should I keep training or stop cause it's gonna mess up

[proper shale]

Can you run it locally perhaps?

[brittle wing]

Like this is after 430

[proper shale]

still going

[brittle wing]

I'll continue when I get home then cause I carried the graph at 430 epochs

[wispy umbra]

Already, it works but is laggy even with all settings at the max

[proper shale]

If you have a dedicated graphics card it could probably work locally... depending on how good it is

[wispy umbra]

It's laggy

[brittle wing]

Should I stop training or it can continue cause as far I see the graph keeps generating points

[proper shale]

AMD GPUs have this lag sometimes, mainly because people often use the wrong version and don't do the appropriate steps

[proper shale]

I mean... it's still going so

You do whatever you want

[wispy umbra]

Like, idk why, the settings are on max, I have RTX 2060, but the voice lines have this robotic bumps

[wispy umbra]

Nvidia

[proper shale]

Oh

You can definitely run it locally

[brave garnet]

Voicechanger Download (Okada)

⠀
Download for Nvidia GPUs
Version 18a cuda

Download for AMD GPUs
Version 18a directml

Download for Intel GPUs
Version 18a directml

Download for Mac
Version 17b Mac
⠀

Made with love by Ilaria 💖

[proper shale]

Just download the right version and configure it correctly

[brave garnet]

Voicechanger Settings (Okada)

⠀

Settings for Nvidia GPUs

F0 Det.: rmvpe (suggested for all series)

RTX 40-series: 80-96 chunk | +16384 extra
RTX 30-series: 96-112 chunk | +16384 extra
RTX 20-series: 112-128 chunk | +16384 extra
GTX 16-series: 128-192 chunk | +8192 extra
GTX 10-series: 128-192 chunk | +8192 extra

Advanced Settings

Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low

⠀

Made with love by Ilaria 💖

[proper shale]

These settings will definitely make it a good experience

[brittle wing]

Okay

[brittle wing]

But if you look at the graph carefully you already see It overtrained at 51k steps

[low shard]

#📰│dev-updates message btw

[proper shale]

...it isn't overtraining

the graph clearly went down afterwards

[kindred ledge]

anyone can help why i cant open my vc ?

[analog olive]

yeah thx, that worked

[olive urchin]

hi, I got myself an ARC A770 and my question is if anyone knows an easy way to run RVC on it?

I tried traditional RVC with no luck

RuntimeError: mat1 and mat2 must have the same dtype, but got Float and Half

this is the error I'm getting

and I was following this tutorial i'm using WSL + oneAPI https://www.reddit.com/r/IntelArc/comments/16e35fw/rvc_webui_and_training_on_intel_arc/

[pastel oak]

Huggingface.co is currently down. This means the voice changer doesnt work

[olive urchin]

I also tried Applio, which completely doesn't support arc and this weird applio I found https://github.com/SayanoAI/Applio-RVC-Fork/tree/main

[rare goblet]

Ayo? @olive urchin level 1 !!!

[tepid heart]

could it be possible to make it not reliant on servers, might be a dumb question ik

[candid wind]

using a colab version of rvc and its throwing up this error and im not sure how to fix it

[rare goblet]

Ayo? @candid wind level 2 !!!

[proper shale]

local servers you mean?

because you can run it on your pc if you're referring to colab

[proper shale]

what colab notebook is that

[candid wind]

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

this one

[proper shale]

hm, that's kinda weird

[tepid heart]

im not using collab atm, but since hugging face is down all this client is doing is this showing 'wait web server...' , understandably so

[rare goblet]

Ayo? @tepid heart level 1 !!!

[proper shale]

id use ilaria rvc for the time being ig

https://colab.research.google.com/drive/16LkwvFZeudTpUOsE_6bMjOq2qkxFo8Hr?usp=sharing

[tepid heart]

alright, cheers mj

[proper shale]

oh that

well, it isn't technically 100% reliable on servers

[proper shale]

nononoo wait thats for somebody elseeee

[tepid heart]

ohh

lmaooo

[proper shale]

let me explain

so

[tepid heart]

alrighty

[proper shale]

there's some part of it

that freaks out when Huggingface is down

the entire thing runs just fine by itself, but it freaks out when Huggingface is down

iirc it was the noise suppression algorithms?

so yeah

[tepid heart]

any way to trick it?

[proper shale]

ig if you change the code, somehow

i dont remember what exactly causes the thing to freak out soooo yeah id just wait for now

@sand iris do you remember what freaks W-Okada out?

😭

[sand iris]

can't do that on binary releases

only if you have the raw code

Yeeted it off my drive, can't test anything

[tepid heart]

all good thanks for answering friend

[proper shale]

oh

[sand iris]

wherever it's parsing the .jsons in the code, just return early

[tepid heart]

im too dumb to understand code lmfao

so i'll just sit n wait

[proper shale]

i think it was the samples part right???

now i remember

[tepid heart]

yeah it is

its trying to access it, and its just looping

[proper shale]

yw, wish i could help more but sadly it's beyond me atp

[tepid heart]

all guud

[proper shale]

h

@candid wind nvm, Huggingface is down, nothing will work rn

[candid wind]

ah i see

[dusk token]

Please, how can I solve this?

[mighty laurel]

@low shard any other working colab ? I tried again but not working

[low shard]

#📰│dev-updates

Hugging face is down rn

that is the reason why

[vital bluff]

how to use google colab

can someone tell or show me

[proper shale]

https://docs.aihub.wtf/

has all the main colabs

and guides

[vital bluff]

oke

[proper shale]

rn tho none of em work, Huggingface is down lol

but i think it's coming back

[vital bluff]

k

[rare goblet]

Ayo? @vital bluff level 1 !!!

[mighty laurel]

I've downloaded model with google drive, but error is coming in conversation

as I showed in above error code I sent

[proper shale]

none are working rn

Huggingface is down

[mighty laurel]

@proper shale see, model downloaded

with drive, error in conversation

[vital bluff]

is weight.gg also down

[proper shale]

shouldn't be, as far as i know

[vital bluff]

ok

[proper shale]

your audio might be screwed up ig

id try re-exporting as a wav

in audacity

[mighty laurel]

tried 2 different mp3

[vital bluff]

how long does it take for the song to be completed?

[proper shale]

it'll take a while probably

never used their conversion part so

[vital bluff]

how do you get better quality

[mighty laurel]

any other colab @proper shale I was using mainline repo, but I don't know where to upload my model in that server, if you could tell

[proper shale]

get a better acapella somewhere, and use a better model ig?

[proper shale]

theres applio:

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

ila rvc mainline:
https://colab.research.google.com/drive/1mHKTGH5e3SAyDSBss1KtiYRbDdQzwSMs

normal ila rvc: https://colab.research.google.com/drive/16LkwvFZeudTpUOsE_6bMjOq2qkxFo8Hr?usp=sharing

[brittle wing]

What would you recommend for local use only?

[proper shale]

https://docs.aihub.wtf/rvc/local/mainline-rvc/

i hope the link works 😭

[brittle wing]

I got a 404 error when loading the link

[olive urchin]

anyone knows what's with that?

[proper shale]

https://docs.aihub.wtf/rvc/local/mainline/

[proper shale]

so, abt this

I've got no clue, but it might be bcuz Intel GPU

[olive urchin]

I see

[proper shale]

I believe you've sent the training tutorial but uh

[olive urchin]

okay I guess I'll look into code myself lol

[proper shale]

Yeah this stuff is built for NVIDIA and it's complicated to get anything else working

[proper shale]

If you are able to fix this then damn congratulations

😭

I have no understanding of this stuff

[olive urchin]

I don't promise anything but I'll see

what I can do

I managed to get llama.cpp working on it tho

and stable diffusion webui

[lost relic]

guys can remcomment setting for converting voice in rvc for deep emotion epspecially when shout

[brittle wing]

I hope I understood.

You recommend Mainline?

Is this the mainline in that guide?

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

[proper shale]

yep

[lost relic]

i don't know anything i can only ask here i hope all senior can help me

[proper shale]

get a good model for shouting

and get good shouting audio ig

mess around with the other settings tho

[brittle wing]

Awesome!

Is there a way to use Tensorboard with Mainline locally?

[rare goblet]

Ayo? @brittle wing level 6 !!!

[lost relic]

oh so it's base on sample

[vital bluff]

why cant i download from it

[cosmic girder]

Guys

How i make ai covers on hugging face

I ONLY WANT TO FIND AI COVERS ITS HARD TO MAKE ONE IM GOING TO LEAVE

[proper shale]

Yeah, just install the .bat that is in the TensorBoard guide, in the Docs

[cosmic girder]

BUT IF IM MOBILE

[proper shale]

https://docs.aihub.wtf/

ya dont need to use Huggingface

heck even use weights.gg

[proper shale]

yeah

[proper shale]

?

[lost relic]

thank you senior

[proper shale]

yw

:)

[proper yarrow]

what is RVC_DIRNAME

[cosmic girder]

FCK THIS SERVER IM GOOING TO LEAVE THIS FUCKING TRASH SERVER MAKING AI COVERS IS VERY HARD GOODBYE.

[vital bluff]

from weights.gg when i tap the download button nothing happens

[proper shale]

hm? what are you using

[proper shale]

if ya can't read that's on you, sorry

[olive urchin]

okay I see the issue, basically Intel updated their torch package which broke the code and I see that devs are doing something so either:

• I install some old version of oneAPI
• literally modify torch
• wait for fix that makes variables work properly again

[proper yarrow]

CoverGen_NO_UI_v2_en.ipynb

[proper shale]

oh WOW

[proper shale]

should be the model's name

[proper yarrow]

if i want travis scott just write travis scott?

[rare goblet]

Ayo? @proper yarrow level 1 !!!

[proper shale]

you should write the model's file name ig

so if it's travisscott.pth then write that

[proper yarrow]

where i can find that?

[proper shale]

colab file explorer should have it

[olive urchin]

import torch.nn.functional as F

# Assuming these are your tensors
mat1 = ...  # Float tensor
mat2 = ...  # Half tensor

# Cast mat2 to Float tensor to match the data type of mat1
mat2 = mat2.float()

# Perform the operation
attn_output = F.linear(mat1, mat2)

some shit like this should fix it but like I said it's modyfing packages and specificaly multihead_attention.py

[proper shale]

it might be on a "models" folder or "weights" folder...

[proper yarrow]

it doesent work

[proper shale]

yeah it was an example, idk the actual model name

[proper yarrow]

where can i find it can you tell me?

[proper shale]

_ _

[proper yarrow]

its in the server?

[proper shale]

no, on the colab file explorer

no wait you want a model

?

[proper yarrow]

yeah

[proper shale]

im being dumb sorry 😭

well u can search on #1175430844685484042

and then copy the link and paste it where it tells you to download models in the colab

[proper yarrow]

if i write https://huggingface.co/Bruh110/BITTSMV2/resolve/main/TRAVISBITTSMV2.zip?download=true on the RVC_DIRNAME itll work?

[proper shale]

dont think so

actually id just recommend using something else i never used the non web ui stuff

[proper yarrow]

which one?

[proper shale]

i think you could use this maybe

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

not like aicg, since you need to filter out the vocals

but it might have a dedicated uvr part so you can filter it there iirc

[proper yarrow]

thanks i just click install and start?

[buoyant skiff]

can someone please help me with google colab? why do I get this error on the last step and how do I fix it? it says:

SystemError Traceback (most recent call last)
<ipython-input-58-d1e0a46a13b4> in <cell line: 15>()
13
14 import torch
---> 15 from inference import infer_tool
16 from inference import slicer
17 from inference.infer_tool import Svc

10 frames
/usr/local/lib/python3.10/dist-packages/numba/np/ufunc/decorators.py in <module>
1 import inspect
2
----> 3 from numba.np.ufunc import _internal
4 from numba.np.ufunc.parallel import ParallelUFuncBuilder, ParallelGUFuncBuilder
5

SystemError: initialization of _internal failed without raising an exception

[proper shale]

yeah, run those steps

[proper shale]

what colab is that?

[buoyant skiff]

it's on google colab, the name is "so-vits-svc 4.0 inference FIXED thanks goat YeezyBeaver.ipynb"

[proper shale]

yeah... that's like... outdated

[buoyant skiff]

oh really? is there an updated version?

I remember using it a few months ago and it worked fine

[rare goblet]

Ayo? @buoyant skiff level 1 !!!

[proper shale]

see, sovits isn't really supported anymore

so uh

yeah

you'll have to use RVC

[buoyant skiff]

this is the first i've heard of it, is that a website or do you have a link for it?

[proper shale]

https://docs.aihub.wtf/

has everything you need to know about rvc, how to use it and colab links/guides

[buoyant skiff]

thanks!

[proper shale]

yw :)

[olive urchin]

yeah alr entire RVC code needs to be adjusted

[tender nimbus]

is still couldnt figure out about how i install the voice thingys

[rare goblet]

Ayo? @tender nimbus level 1 !!!

[tender nimbus]

i need to know what software to use and all that

is github good?

[pastel oak]

What are you trying to download?

• RVC
• Voice Changer
• TTS?

[tender nimbus]

voice changer for real tie discord calls

time*

[pastel oak]

Whats your GPU name

[tender nimbus]

uhd graphics 620

[night hinge]

For some reason im getting a error when i click convert?

[pastel oak]

Converting what?

[pastel oak]

That is not good enough to run the voice changer unfortunately

[night hinge]

im at the stage where i, Inferencing voice

[pastel oak]

Ah okay. Is there any error code shown on the command window?

[night hinge]

AttributeError: 'NoneType' object has no attribute 'dtype'

[pastel oak]

This usually happens when the path to the audio is wrong. can you send a screenshot of your inference paths and all

[tender nimbus]

is there any other solution to iit it?

[hearty barn]

can someone help me in #🔍│help-w-okada ??

[night hinge]

Can you direct me on which path i should put it?

[pastel oak]

It doesnt really matter but I created a new folder in my RVC folder called "audios" and put it in there tbh

but send a screenshot just to see if you linked the path correctly

[night hinge]

Next to asset or just in there somewhere?

[pastel oak]

you can put it in assets instead yea

[pastel oak]

#🔍│help-w-okada message

[night hinge]

Okay got the path right from what you told me. Still says error

[pastel oak]

can you paste in what you put in the audio slot on the infering

the path name

[night hinge]

Sorrt if i sound slow.. But do you mean this? "C:\Users\smoky\Desktop\RVC1006Nvidia\RVC1006Nvidia\assets\Audios"?

[pastel oak]

Yes, except you have to select the audio directly

Like:

D:\Desktop\skre6.wav

with no quotations " and the audio selected like that

[night hinge]

Would it work if i just remove the "?

[rare goblet]

Ayo? @night hinge level 3 !!!

[pastel oak]

yes

and you have to select the mp3 or wav or whatevr

[night hinge]

Wait what?

[pastel oak]

I dont know how to be more clear 😭

literally my example

D:\Desktop\skre6.wav

in your case

C:\Users\smoky\Desktop\RVC1006Nvidia\RVC1006Nvidia\assets\Audios\blablabla.mp3

you have to paste that in the infering audio path

It would have been so much easier if you had just sent a screenshot 🫠 ill send one instead

[night hinge]

Sorry im just so new to this.

[pastel oak]

the mp3 or wav or whatever has to be in that

in the name of the path

[night hinge]

Do i just rename it? or do i have to go in where that folder is and do something?

[pastel oak]

Lets break it down.

What is the name of your audio

[night hinge]

Well i just put a new folder in assest and named it audio.

[pastel oak]

Not the folder name

your audio that you want to infer (conert) with

You need an audio file of someone speaking or singing or whatever to convert that into your choice of voice model

[night hinge]

Inferencing voice?

Where i put my model in the wight folder?

[pastel oak]

can someone else help

[night hinge]

God im stupid..

I send ya the screen shot if that helps?

[pastel oak]

yes

[night hinge]

Feel really dumb..

[pastel oak]

Ok now show me a screenshot of the folder where your audio file is

the assets\Audios

just send a screenshot of whats inside the folder

[night hinge]

Umm its empty..

[pastel oak]

Let me take a step back

Infering means having an audio file, like someone speaking, and converting that into a new voice with the voice model

Do you have an audio that you wish to convert

[night hinge]

I do have one but the person is singing does that still work?

[pastel oak]

Yes

Put that file into that folder with assets\Audios

[night hinge]

Okay.

Okay. I put it in the audio folder

[pastel oak]

What is the name of the audio?

[night hinge]

The name of the file i put in is. KafkaEN_From_The_Start got it from here voice-models section

[pastel oak]

I meant the audio of the person singing

Look I mean this

In my audios folder i have a file called dreamsofan.wav this is someone singing

What is the name of your audio with the person singing

[pastel oak]

Ok my bad I just assumed you told me the name of the voice model

[orchid glade]

sorry for the dumb question but where can I download rvc

[pastel oak]

On RVC on the "Enter the path of the audio file to be processed (default is the correct format example):

You will put in the path of that audio. In this case:

C:\Users\smoky\Desktop\RVC1006Nvidia\RVC1006Nvidia\assets\Audios\KafkaEN_From_The_Start.mp3

I just assumed its a mp3, I cant tell from your screenshot. If the ending of your file is something else like .wav or .flac or anything, then put that instead

Do you understand the principle of how it works now?

[pastel oak]

nvidia and amd links are on here on the top

https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/releases/tag/updated1006v2

[orchid glade]

thanks my guy

[night hinge]

Ah its a wav

[pastel oak]

C:\Users\smoky\Desktop\RVC1006Nvidia\RVC1006Nvidia\assets\Audios\KafkaEN_From_The_Start.wav

[night hinge]

I i really just copy past what you just put?

[pastel oak]

yes

This is how it works

if you ever put in a new file, you have to replace the "kafkaEN" thing with the new name

does that make sense

[night hinge]

Yes.

[pastel oak]

Great 😁

[night hinge]

It works!

I converted it.

[pastel oak]

Few more things to note

[night hinge]

Thank you so much for being so patient with me.

[pastel oak]

Transpose is the pitch. So if your voice model is lets say a female, but your singing is male, then you have to change that

Some "rough" numbers to give you an understanding
Female singing to male voice model -> -12
Male singing to female voice model -> +12

So plus makes it go higher pitch, minus makes it go lower

You will most likely not need to use the exact number 12, but you can use it for orientation. If the converted voice is too high pitch, you know to lower that number a little bit, vice versa

And dont forget to select the index file aswell. It should be the last option of those on the left side

You should put your index file in the "logs" folder, just toss it in there and it will show up as a selection

[night hinge]

Okay so got lost. Now that i have converted it. I should do?

[night hinge]

Should i keep following the guide?

[pastel oak]

Right click on the grey area and then you can select "save audio as"

[pastel oak]

Yea it has some explanations there if you get lost or dont know what to do

[night hinge]

Okay now that i have done that. Should i leave it at 0 if its a female to female?

[rare goblet]

Ayo? @night hinge level 4 !!!

[pastel oak]

Yes.

then apply the same logic: if one of them is noticably higher or lower pitch, youll have to change transpose

[night hinge]

So i have done that. Converted it. clicking play on the song for example everything sounds fine. What do i do now?

[pastel oak]

this to save

[night hinge]

I dont see the 3 dots to download as the guide says

[pastel oak]

Did you read what i said

[night hinge]

Got it

[pastel oak]

right click on grey area to select the save audio as

on the screenshot

the 3 dots dont exist idk why so you right click to save

[night hinge]

Got it saved.

[night hinge]

Now that i got it saved finally. Again sorry for being slow. What my next step. The guide says to train it or something?

[pastel oak]

Nah youre done now

Training is a whole different thing

[night hinge]

So that it?

[night hinge]

So lets see if i got the right. If i where to record my own voice just simple lines. I can use this to convert my voice to a ai voice insted?

[pastel oak]

training is to train your own voice model.
theres section on the guide that talks about it. you need a clean dataset and all then train it

im outside so cant explain atm

[night hinge]

Its okay. Honestly you went far and above already. Thank you very much for everything!

[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

[worthy flax]

i just finished installing ilaria rcv but i dont know where to put the link to use the voice (i have no idea what im doing i was used to the old kalomaze rcv)

[orchid glade]

why does rvc sound bad on discord

[worthy flax]

and i need help cuz idk how to use it

[zenith walrus]

Right... My plan is to convert like voice models to L4D2, so instead of like Nick's voice, it's like Zero's voice instead. It's like those RVC AI mods that people converted and uploaded to workshop before, but I have zero clue how to start with it. I have the voice models in .pth, I'm guessing Python is required? (as I downloaded it yesterday very late).

They are entitled like Voice Packs, etc. I would like to convert Nick > Zero, Francis > X, Bill > Solid Snake.

[brittle wing]

Should I keep training

[worthy flax]

is the easy gui for rcv v2 no longer working? is there any other easy rcv to use for making aicovers? i already have the vocals and the model i just want to convert the vocals using the rcv plz

[soft stratus]

Ilaria RVC said error when i convert my voice
@low shard

[low shard]

Send google colab

[low shard]

Sned model download link

[soft stratus]

https://drive.google.com/file/d/1Hz4hothv80K2YvwinWx4Rz4S3d4_0Rv0/view?usp=sharing

[worthy flax]

i already closed it but i was using the cover gen no ui but i didnt know how it works

[soft stratus]

Here it @low shard

[rare goblet]

Ayo? @worthy flax level 1 !!!

[worthy flax]

https://drive.google.com/file/d/1n1bOAkQOaGS62HGFMxo9D0B5DrFyZc1y/view

here is the model i wanna use

[low shard]

Did it give you error during download? Or did it show in the voice list

[soft stratus]

Idk

[low shard]

Did it say downloaded successfully ?

[soft stratus]

Yes

[low shard]

Mmm the model files are correct, show a screenshot of the exact error

[low shard]

Click the 2 refresh button and re select everything, if it still doesnt work, show me a screenshot of the output colab

[worthy flax]

ah no, sorry i didnt get any error its just i dont understand quitewel how to use it. thats why i was asking for a easier rcv link.. causecover gen is tricky i have never used it i was used to the kalomaze one you know the easy gui. the cover gen asks me to put a youtube link and such but i dont wanna use a youtube link cause i already have the vocal and instrumental files made i just want to insert the vocals archive and convert it to the model

is the first time im using the cover gen rcv and i have no idea of any of this hehe..

[low shard]

Show the google colab output

[low shard]

Sure, you can use the other colabs in our docs https://docs.aihub.wtf , i could suggest you either applio or ilaria rvc for inference (using models)

[worthy flax]

how do you use ilaria? is there a tutorial?

[zenith walrus]

I'm using Jo's Bizzare Voicemodding as a Steam Guide but I don't think this is the step... I need to replace ALL the voicelines for each Survivor with the .pth instead... Oh man.

[low shard]

Yes, check https://docs.aihub.wtf/rvc/cloud/ilaria-rvc/ for ilaria, or https://docs.aihub.wtf/rvc/cloud/applio-colab/ for applio

[worthy flax]

ahh thankies thankies nwn

[low shard]

Yw

[soft stratus]

That's it output

[low shard]

You could check our docs for more info btw https://docs.aihub.wtf

[low shard]

No, go back to the google colab page you used, you didnt close it right?

You need to show the info coming out of cell you runned on google colab

Thats the output

[soft stratus]

It's not colab... It's hugging face

[low shard]

Oh thought u were using the colab as it uses the same ui, anyways, did you duplicate the space?

[soft stratus]

No

[low shard]

Also, for your information the hugging face space is wayyy slower than the google colab bc hugging face spaces offers only free cpu, while googlr colab t4 gpu so its faster

[low shard]

Click the 3 dots at the top, and duplicate it your own, i suggest you to use the colab version if u want smt faster tho

[zenith walrus]

Seems like I need to use it cloud, since I don't have a REALLY good GPU, it's a UHD instead. But yeah, this might take a while...

[rare goblet]

Ayo? @zenith walrus level 1 !!!

[soft stratus]

It works in my duplicate

[low shard]

Its all fixed now?

[soft stratus]

Yes

[low shard]

Oke im glad

[zenith walrus]

Nick's L4D2 voicelines is about 2.5k+... I have to convert them 1 by 1?...

I'm using Ilaria RVC btw.

Oh nevermind, I got it.

[orchid glade]

Why is line 1 showing on input

[rare goblet]

Ayo? @orchid glade level 3 !!!

[brittle wing]

It's probably going to 70k

It did

Is that overatrin

[pastel oak]

I got no clue why not lol. Did you restart pc to see if it fixes it?

[orchid glade]

yes

[pastel oak]

maybe reinstall virtual cable

[sly tulip]

Ive got a question, whats the programm/ai name where i can create voice lines with so text 2 speech

[orchid glade]

already tried that

it only shows on MME

[sly tulip]

Ive got a question, whats the programm/ai name where i can create voice lines with so text 2 speech @pastel oak

[pastel oak]

Theres a few different ones, check https://docs.aihub.wtf/tts/tts-tools/

[sly tulip]

danke dir, kann gtpsovits rvc models usen?

und Ov2

[pastel oak]

Ich glaube ja, aber ergebnisse werden mit sovit modelle besser sein als mit rvc

ov2 weiß ich nicht

[sly tulip]

ok

[orchid glade]

do I just use vb-cable ?

[sly tulip]

mit welchem tool ist ov2 gut möglich?

[brittle wing]

Did it overtrain

[pastel oak]

I would not use VB-CABLE personally.. but if it works then sure

[pastel oak]

Damit habe ich leider keine Erfahrungen 😅

[pastel oak]

Probably.

[brittle wing]

Does it look like it?

[pastel oak]

Yes

[brittle wing]

I have it saved at 400 epochs

[pastel oak]

The best average model is at 43k based on the graph

[brittle wing]

It wasn't as much bad looking

[pastel oak]

Wherever that low point is.

[brittle wing]

I don't have the 43k one

[pastel oak]

Then use 47k

[brittle wing]

It was yesterday I finished 400

[brittle wing]

I have 400 epochs

It sounds normal

[pastel oak]

I dont know what steps your 400 epochs is btw.

[brittle wing]

IT sounds normal idk

[pastel oak]

Ok then use it

[brittle wing]

I should quit training right

[pastel oak]

There is nothing wrong with using a model past the low point. It is only the best average model, anything after can still sound more dynamic.

[brittle wing]

And stay with the 400 epochs one right

[pastel oak]

Without going too far into a late model. Just use 47k

I still dont know how many steps your 400 epochs are so I can not answer that¯_(ツ)_/¯

[brittle wing]

The 400 epochs one wasn't that farther

[pastel oak]

Then use it.

[brittle wing]

Anyways I should quit training and stop wasting my time

[pastel oak]

Yes.

[brittle wing]

I found the older capture of 400

What the graph looked like

I should quit to avoid further overtraining it seems

[broken birch]

Are there any general recommendations on the relationship between batch size and VRAM? Like, if you have 8 GB VRAM, start with this batch size, that kind of thing.

[brittle wing]

It kept going up on 70k

[orchid glade]

Made A line 2 and it shows up does it still work ?

[pastel oak]

Probably will work yea

[zenith walrus]

Aaand encountered a error. All I did, was give out the .wav file from L4D2 direction, and with Zero.pth, it gave this.

[pastel oak]

Unsure, most people do your X gb vram - 2 for batch size. However, me on my rtx 3060ti 8gb vram still used 8 batch size and worked just fine

[broken birch]

Thanks, that's what I was looking for!

[pastel oak]

Can you show a screenshot of the entire inference screen, maybe the path file is somewhere where it shouldnt be?

[rare goblet]

Ayo? @pastel oak level 39 !!!

[zenith walrus]

Yeah gimme a second.

So, I got all the voicelines from C:\Program Files (x86)\Steam\steamapps\common\Left 4 Dead 2\left4dead2\sound\player\survivor\voice\gambler, right. I got them in a seperate folder on the desktop entitled "gambler voices"

I downloaded the Megaman X4 Zero .pth so it could convert Nick's voice into Zero. Instead, it gives out a error.

[pastel oak]

Okay. Where did you get the Zero voice model from? Perhaps its not compatible though I wouldnt know why itd be that case

[zenith walrus]

From this server, in #1175430844685484042

Or voice-models.

Perhaps it might be gone?

[pastel oak]

Have you tried out a different voice model to see if the issue persists?

[zenith walrus]

They are almost the same unfortunately.

[rare goblet]

Ayo? @zenith walrus level 2 !!!

[zenith walrus]

Uh, I could try with... Solid Snake. lemme check.

Nope. Same error.

Just keep in mind, I don't wanna make it like a vocal song replacement, it's more of a replacement of voiceline so they call out zombies, grabbing a weapon, talking to survivors, et cetera.

[simple yoke]

I'm getting really really high cpu usage with RVC, like 50% Any advice?

[zenith walrus]

Alright, I used a different RVC tool, and it now magically works. Ilaria does not seem to cooperate with me, so it should be smooth sailing from here.

[orchid glade]

it's crashing

[fluid topaz]

check command window

[pastel oak]

Did you change the default index file? That somehow, even if its not related to asio driver, fixed the issue for me

[orchid glade]

it works

thanks : )

how to I use it tho if the input and output are both FlexAsio

[pastel oak]

On discord and games you select Line 1 as input

[orchid glade]

ah I see

I don't hear anything

[pastel oak]

Or Line 2 sorry

[thorny pulsar]

how im adding a modle?

[zenith walrus]

Okay, everything works, checked the seperate audio files. Now here's another problem - If I want to replace every single Nick voiceline, which is about 2.9k audio files, do I have to separately upload them ONE by ONE, and not upload it as a folder? I'm pretty sure converting 2.9k voicelines one by one could take days, weeks...

[lunar talon]

Whenever I try to use a voice it comes back with a error on step 2

[pastel oak]

Try to use a Program where you can put all audio files into one track line and render out the voicelines as one file

Though I believe that size might be so big itll take ages to convert. else just bulking in like 10 separate audios

I know sony vegas pro, if you drag multiple clips into it, puts them all next to each other. idk about other softwares

[zenith walrus]

Alright, I'll keep that in mind. Appreciate it.

[orchid glade]

what's the best value for this

does this seem okay

[lunar talon]

I just get this when trying to get any voice
JSONDecodeError Traceback (most recent call last)
<ipython-input-10-b114748cdea6> in <cell line: 31>()
31 if os.path.exists(config_path):
32 # File exists, proceed with creation of creds and client
---> 33 creds = Credentials.from_service_account_file(config_path, scopes=scope)
34 client = gspread.authorize(creds)
35 else:

5 frames
/usr/lib/python3.10/json/decoder.py in raw_decode(self, s, idx)
353 obj, end = self.scan_once(s, idx)
354 except StopIteration as err:
--> 355 raise JSONDecodeError("Expecting value", s, err.value) from None
356 return obj, end

JSONDecodeError: Expecting value: line 1 column 1 (char 0)

[pastel oak]

outdated colab, use one of the newer ones

-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

[pastel oak]

Ignore this. You dont need it because we use rmvpe. Any value for it is fine

Did you manage to make it work btw? Did Line 2 work as input?

[zenith walrus]

Funny coincidence: It appears there's a section called Batch. Huh.

I might try this out.

[orchid glade]

is there any program that can create filters like noise suppression on input device

[pastel oak]

Lol perfect

[pastel oak]

Steelseries sonar has advanced noise suppression

[orchid glade]

thanks

[zenith walrus]

Uh oh.

[orchid glade]

do you need to pay for it ?

[rare goblet]

Ayo? @orchid glade level 4 !!!

[pastel oak]

No

[orchid glade]

oh so it's the virtual audio cable saying trial again and again xD

[zenith walrus]

TypeError: VC.vc_single() missing 7 required positional arguments: 'protect', 'crepe_hop_length', 'f0_min', 'note_min', 'f0_max', 'note_max', and 'f0_autotune'

This is weird... it's from the Batch option.

[orchid glade]

man VAC is premium rip

is there any good free alternatives

[brittle wing]

how do you run tensorboard on Paperspace?

[rustic aurora]

hi guys which one is better

[rare goblet]

Ayo? @rustic aurora level 1 !!!

[foggy prism]

Does anyone know how to hear yourself while you use virtual audio cable for rvc for discord or any game? Once you use virtual audio cable you can’t hear yourself

[mellow token]

what’s the problem

[rustic aurora]

which one should i prefer

for the best voice model

training

[rustic aurora]

.

[mellow token]

i mean i use rvc disconnected

i can’t say anything about the other options but rvc disconnected is good

[rustic aurora]

what is the difference between these

[mellow token]

i’m not sure i don’t use the others

[rare goblet]

Ayo? @mellow token level 46 !!!

[rustic aurora]

okay thx

[mellow token]

can probably look in #📰│dev-updates to see what some of the features has on whichever

or look in the guides that those are in

[foggy prism]

@mellow token not to bother you but do you know a solution to the problem I’m having, posted it above

[mellow token]

hmm

i’m not sure but try asking in #🔍│help-w-okada

[foggy prism]

Okay

[potent verge]

excuse me, what happened to rvcv2?

[spring orchid]

anyone having issues with KLMv7s pretrain of rvc disconnected v2?

[sturdy ridge]

So how do I make ai voice? Like if I want Taylor Swift to do a cover of a Harry Styles song?

[mellow token]

#1188551486620303442

[rustic aurora]

bro there is no colab link in the rvc disconnected

how can i enter it

[spring orchid]

-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

[rustic aurora]

thanks

and also i want to ask a question

which colab is better

for the model training

[steel lake]

@viscid moss AICoverGen-NoWebUI [English] is having issues.

[rare goblet]

Ayo? @steel lake level 2 !!!

[spring orchid]

idk i think they are all the same, but applio and ilaria are good, i use rvc disconnected

[rustic aurora]

ok thx

[rare goblet]

Ayo? @rustic aurora level 2 !!!

[rustic aurora]

and also should i which algorithm select

rmvpe, crepe, mangio-crepe or what

[spring orchid]

rmvpe

[rustic aurora]

i have 15 minutes dataset

how much will be crepe hop length

[spring orchid]

doesnt matter since you use rmvpe

crepe hop length is for mangio crepe

[viscid moss]

[spring orchid]

15min dataset use original or snowiev3 pretrain

[viscid moss]

Elaborate pls

[spring orchid]

8 batchsize save every 25epochs

[rustic aurora]

what is these

[spring orchid]

on rvc disconnected

[rare goblet]

Ayo? @spring orchid level 9 !!!

[steel lake]

It gave me an error message while generating an AI cover.

[viscid moss]

Can u send a pic?

[rustic aurora]

oh okay

[viscid moss]

Oh, the problem is the model.

Where did you get the model from?

[steel lake]

huggingface.co

[viscid moss]

send me the link pls

here or in dms

[steel lake]

https://huggingface.co/StefanArnaout/Nicolae_Ceausescu/resolve/main/Nicolae Ceausescu.zip

[viscid moss]

lemme check, gimme a sec

[vapid lava]

I have a question, but how do I send pictures to show you my problem

[viscid moss]

Yep, the model is the problem.

Do u have a Hugging Face account? To fix this u need to dowload the model, delete some files into the .zip and upload the model again

[viscid moss]

U need lvl 1 to send pics

just chat a little bit more

[vapid lava]

thanks!

[viscid moss]

like 3 messages

what's the problem?

Colab doesn't works?

[vapid lava]

I have a model, index and audio wav file as the base, but i get an error not sure why

...another message to level up

[viscid moss]

I see. What app are u using to convert?

ya

ig

[vapid lava]

RVC Web UI

[rare goblet]

Ayo? @vapid lava level 1 !!!

[viscid moss]

now u can

send me a pic

[viscid moss]

show me the logs into the cmd

[vapid lava]

im sorry where do i find that?

[viscid moss]

The window like this

You should have something like this open along with the program

[vapid lava]

ahhh I see what you mean

[half cove]

snowie is trained on russian voices only

[viscid moss]

U re running RVC on Linux... not sure if it works correctly on Linux

[half cove]

swap infer/lib/audio.py with this

also thats the amd_intel version i which is only realtime right now iirc

[vapid lava]

ok I have switched the audio.py file with yours

ill restart it

and it now works!!

thank you!!!

So, my audio.py was not the proper file for AMD?

[half cove]

write binary and read binary sometimes doesnt work

[vapid lava]

THANK YOU BOTH!

[brave garnet]

Voicechanger Settings (Okada)

⠀

Settings for AMD GPUs

Don't forget that your models needs to be converted in ONNX!

F0 Det.: rmvpe_onnx (suggested for all series)

7xxx XT cards: 112-128 chunk | +16384 extra
6xxx XT cards: 128-192 chunk | +16384 extra
5xxx XT cards: 192-256 chunk | +8192 extra

RX 580: 192-256 chunk | +8192 extra
RX 570: 192-256 chunk | +8192 extra
RX 560: 256-384 chunk | +8192 extra

Advanced Settings

Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low

⠀

Made with love by Ilaria 💖

[steel lake]

I don't have a Hugging Face account, but it let me download the model. Which files do I delete?

[viscid moss]

Leave only these 2 files, delete the others

[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

[forest briar]

Rafadan tayfa hayri link please

[lost wasp]

Can you make voice models on your phone?

[pastel oak]

You will have to prepare the dataset in some way or another, which might require a PC. The Training process however can be managed through phone

[brave garnet]

Voicechanger Settings (Okada)

⠀

Settings for Nvidia GPUs

F0 Det.: rmvpe (suggested for all series)

RTX 40-series: 80-96 chunk | +16384 extra
RTX 30-series: 96-112 chunk | +16384 extra
RTX 20-series: 112-128 chunk | +16384 extra
GTX 16-series: 128-192 chunk | +8192 extra
GTX 10-series: 128-192 chunk | +8192 extra

Advanced Settings

Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low

⠀

Made with love by Ilaria 💖

[vague pelican]

does anyone know what is wrong when adding this model

paramDict {'voiceChangerType': 'RVC', 'slot': 12, 'isSampleMode': False, 'sampleId': None, 'files': [{'name': 'Apple_Scam_Voice_e15_s300.pth', 'kind': 'rvcModel', 'dir': ''}], 'params': {}}
RVC:: slotInfo.modelFile Apple_Scam_Voice_e15_s300.pth
[Voice Changer] post_load_model ex: No module named 'utils'
Traceback (most recent call last):
File "restapi\MMVC_Rest_Fileuploader.py", line 92, in post_load_model
File "voice_changer\VoiceChangerManager.py", line 170, in loadModel
File "voice_changer\RVC\RVCModelSlotGenerator.py", line 37, in loadModel
File "voice_changer\RVC\RVCModelSlotGenerator.py", line 42, in _setInfoByPytorch
File "torch\serialization.py", line 809, in load
return _load(opened_zipfile, map_location, pickle_module, **pickle_load_args)
File "torch\serialization.py", line 1172, in _load
result = unpickler.load()
File "torch\serialization.py", line 1165, in find_class
return super().find_class(mod_name, name)
ModuleNotFoundError: No module named 'utils'

i upload the pth and nothing happens

its the most recently added models in the voice model channel

apple scam voices

[crystal niche]

how do i use it on discord??

[queen kraken]

try pip install utils in the command prompt window and see if it works

[queen kraken]

I just checked the model. if you're referring to the model named gpt-sovits, no, you can't add gpt-sovits model on a voice changer. it's for TTS

[rare goblet]

Ayo? @dusky zealot level 1 !!!

[dusky zealot]

Does anyone know of any >>LOCAL<< RVC Gui that has this "record from microphone" function?

[proven hill]

ilaria rvc mainline

[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

[boreal mango]

the mainline colab i not making the index file

[proven hill]

where did you search for it

[boreal mango]

through discord

[proven hill]

no, i mean

the index file

where are you looking for it, wich folder

[boreal mango]

through rot13

[boreal mango]

ummm holdon

[proven hill]

what?

[boreal mango]

it not in logs folder

[boreal mango]

the mainline refuses to creaate the index file what am i doing wrong?

[proven hill]

its in the model folder

[proven hill]

what colab are you using

[boreal mango]

https://colab.research.google.com/drive/1mHKTGH5e3SAyDSBss1KtiYRbDdQzwSMs#scrollTo=6KB2d3y_aKIE

[proven hill]

search in the models folder

[boreal mango]

okay

[boreal mango]

where can i find that?

[proven hill]

wdym? instead of the logs folder find the models folder

[boreal mango]

yea its not there im so sad

[proven hill]

did you click on "train index"?