✨│ai-help

[fiery raptor]

where my pc blocks blue light

[proven hill]

UH

[fiery raptor]

so i coudlnt read the text

😭

[proven hill]

it says convert

[fiery raptor]

ye i get it now

[simple igloo]

@pastel oak you got a clue? o-o

[fiery raptor]

thanks for the help lmao

[proven hill]

no worries

let me know if i can help any further

[simple igloo]

@slim geyser Sorry for the ping o-o

[proven hill]

please avoid pinging random people

what are you trying to do with ASIO?

follow this guide: https://rentry.co/VoiceChangerGuide

[simple igloo]

I need their help, and Mares/Shad are the only ones that were talking about it in the past day

Mare helped Shad set it up

[proven hill]

oh

then its okay, i didnt know

[pastel oak]

https://github.com/dechamps/FlexASIO/releases/tag/flexasio-1.9

https://github.com/flipswitchingmonkey/FlexASIO_GUI/releases

get these two things then ping me again

[simple igloo]

Done

[pastel oak]

also make sure to get this https://dotnet.microsoft.com/en-us/download/dotnet/6.0

[simple igloo]

alr

installed x64

[pastel oak]

Ok did you install everything aswell

[simple igloo]

yep

should be

[pastel oak]

start the Flexasio GUI app

[simple igloo]

mhm

[rare goblet]

Ayo? @simple igloo level 4 !!!

[pastel oak]

i mean mares already sent screenshots of what to tick since you saw our conversation

copy that, and save to flex toml

[simple igloo]

alr lemme find it

done

just changed the mic to mine for input

copied mare's latency and other settings

[pastel oak]

thats pretty much everything

if youre using realtime gui then select flexasio as input and output, thats it

[simple igloo]

hmmm

gimmie a sec

think my first installation might've went wrong somehow?

[pastel oak]

i didnt know youre using wokada

i never tried it on wokada only rvc

[simple igloo]

Hmm

was trying to make it work with okada

[pastel oak]

im not sure if this is the fix but select audio SERVER and see if its there?

[simple igloo]

I see it there now

Okada's getting errors now though

[pastel oak]

ill check myself too

[simple igloo]

getting this error repeatedly

no idea what it means

[pastel oak]

yea doesnt work for me either

just download rvc mainline its imo better anyway

[simple igloo]

fixed it by changing output to line 1

[fluid bramble]

what does rvc mean

[simple igloo]

the difference changed drastically

Way faster now

😱

im using 256 like its 192

niceeee

[fluid bramble]

is rvc ai voice

[pastel oak]

did you use MME line 1?

[simple igloo]

yep

[pastel oak]

wait youre using 192 chunk on a 4060?

[simple igloo]

I mean I dont got a problem with it

[pastel oak]

im using 80 chunk atm with a 3060 ti

[simple igloo]

gives me the best quality and I can maintain a convo

[pastel oak]

overall its not bad but on rvc i can use max Extra for better quality with barely any cpu usage, on wokada if i choose 32k it goes up to 50% ish already

if youre happy with it, nice 😄

[pastel oak]

retrieval based voice conversion

[simple igloo]

is it bad that i have RVC Quality set on high 😭

[wispy burrow]

retrieval based voice conversion

[pastel oak]

i guess you can sum it up as "ai voice"

[simple igloo]

everybodies telling me to set it to low

but setting it on high gets rid of the cut off

[pastel oak]

its unnecessary resources used for no difference

[simple igloo]

everytime I stop speaking

[pastel oak]

what cut off do you mean?

[simple igloo]

you know how normally when you finish speaking, the ai just eerily goes silent

like cuts off

think its the noise supression

[pastel oak]

noise suppression is deactivated when you use audio server

what you could try is start 0.0 end 1.0 trancate 100

and silencefront off

silencefront always resets to On when you restart the app or change a voice model fyi

[simple igloo]

it helped a little bit I think

[pastel oak]

Also, not sure if you have the issue, but if i play a video game with wokada then it adds another layer of delay. Not sure if this would still be the case with the asio routing inside wokada, havent tried it yet - but its an issue i dont have with realtime. if you have something similar, do let me know if the issue still persists with the new routing and let me know 👀

[simple igloo]

Will do 👍

[muted verge]

What time zone does the app use?

[quartz coral]

question can i have any nvidia gpu for this?

cause it says there is no compabatible gpu or it has tto be aspecific ndvidiagpu?

[red mural]

ayooooooo
what kinda of software should i download for AI voice changer

[pastel oak]

whats your gpu

[azure patio]

i have used some rvcs but which is the best one that gives the best result for you guys

[quartz coral]

since i will mostly likely buy a new cheapest laptop which nvidia gpu is worst that can ne ablle to train like a picture a posted up before or which model is nessecary for it

[paper silo]

yo so the colabs aren't working anymore correct

[red mural]

rtx 2060 super

and cpu ryzen 5700x

[pastel oak]

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

[quartz coral]

i see about of 40+ of nvidia gpus on wikipedia which is the best or worst of the best @shad ?

[pastel oak]

how much money do you got to spend

if you can get one of the rtx gpus you should be fine. ideally 2060 or better

if you want to get technical, one with 8-12gb vram

[red mural]

can you send me pls a video for the best configuration in order to get the best and the realistic voice sir

[rare goblet]

Ayo? @red mural level 1 !!!

[quartz coral]

i mean i got about 1000 euros but also my dad would lend some money soo

[quartz coral]

so basicallz i can get any nvidia gpu for it right?

[pastel oak]

there are no videos, youll have to ✨ read ✨

https://rentry.co/VoiceChangerGuide

go through each step, ignore amd, colab, mac, intel

[pastel oak]

which country are you from

[red mural]

tysm < 3

[quartz coral]

serbia

[pastel oak]

can you search for a " GIGABYTE G5 KF-E3DE313SH " on any tech stores you got

or GIGABYTE G5 KF-E3DE313SD

[quartz coral]

and then what

[pastel oak]

or ASUS TUF Gaming A15 (FA507NV-LP002W)

these should be in the 900 - 1200€ area

well, buy one of them?

[desert niche]

getting this error when running batch ASR in GPT SoVITS

RuntimeError: CUDA failed with error CUDA driver version is insufficient for CUDA runtime version

[quartz coral]

hmm ok or i want buy cheapest or since i have sapphire nitro+ radeon rx580 4gb gddr5 can that gpu do for mangio voice training help me or nah also by putting 2060 on medium laptop can that help me? i have a father friend who knows these type of stuff so can that help me?

[pastel oak]

i mean we already talked about this, i sent you the amd version of voice training to try it out and you ignored it

it might be possible with rx 580 idk

but very slow and maybe not the best outcome

[quartz coral]

sorry for annoyin ya i just suck at this

[quartz coral]

trust m e i tried it and nothing i think

[pastel oak]

ah then unlucky

and im not sure if you can just put a different gpu laptop inside another that easily, but you can ask your tech friend you mentioned

[quartz coral]

actually i have saphire nitro ericx 580 4gb would it help?

[ashen lance]

How do I add my own voice model?

[low shard]

did you already upload the voice model you trained?

[red mural]

should i donwload it in ssd or hdd or its not a prblm ?

[pastel oak]

ssd is always better

[hasty orchid]

Hi guys, beginner here. I was trying to figure out how to make features with two models. My approach was to make two full songs and stitch them together but if the 2 singers in the original sound too diff then it will distort the voice. Does anyone know a different approach?

[red mural]

when i chose a voice it shows me this message

unknwon message

If you clear the information being managed by this app, it may be recoverable

[pastel oak]

i dont understand, can you send screenshot also use #🔍│help-w-okada for voice changer help

[red mural]

done

[strong fulcrum]

Hello, how do I make TTS with RVC have a less monotone and more animated intonation?

[low shard]

well, the integrated tts in ilaria & applio rvc is edge tts which hasnt much emotions, you could to use bark tts #1212368971127590922 and then put the generated audio as an input in tts with the voice model you want

[strong fulcrum]

i see. are there tts models trained on people's intonations?

[low shard]

like, italian, brazil, etc accents? well that depends from tts to tts, by testing, technically rvc should change the tts output to the voice model accent

[strong fulcrum]

well, not exactly accents, but people's intonation and ups and downs in the voice

to make it less flat if that makes sense

[low shard]

you mean make it more real ?

[strong fulcrum]

yeah i guess so

[low shard]

well, edge tts looks real but emotionless, you could try bark tts which has more emotions in it

[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

[strong fulcrum]

i've tried the google colab test notebook for it, but i've noticed it's often inconsistent

like the voice changes a lot, sometimes it just sounds like random noises, etc

[low shard]

be sure you have this checked on and select one of the voice presets

[strong fulcrum]

i don't see that

[strong fulcrum]

op nvm found it

[low shard]

and btw the blue "HERE" redirects you to a page with more info about all the voice presets

there are english ones

brazilian ones

etc

[strong fulcrum]

ok, i don't see "HERE"

[rare goblet]

Ayo? @strong fulcrum level 6 !!!

[strong fulcrum]

this is what i have

!pip install git+https://github.com/suno-ai/bark.git
from bark import SAMPLE_RATE, generate_audio, preload_models
from IPython.display import Audio

preload_models()
text_prompt = """
    I have a silky smooth voice, and today I will tell you 
    about the exercise regimen of the common sloth.
"""
audio_array = generate_audio(text_prompt, history_prompt="v2/en_speaker_1")
Audio(audio_array, rate=SAMPLE_RATE)```

[low shard]

send the google colab you are using

[strong fulcrum]

https://colab.research.google.com/drive/1eJfA2XUa-mXwdMy7DoYKVYHI1iTd9Vkt?usp=sharing#scrollTo=6p153cqEmXLA

are there different demos?

[low shard]

idk where you got that cus it isnt in the ai tts guide

use this one instead its easier https://colab.research.google.com/github/Nick088Official/Easier-Bark-TTS-Google-Colab/blob/main/Easier_Bark_TTS.ipynb#scrollTo=NyYQ--3YksJY

its better if you use only the colabs/hugging face spaces inside of that #1212368971127590922 guide

[strong fulcrum]

I got it from here:
https://github.com/suno-ai/bark?tab=readme-ov-file

[low shard]

thats the official one but its not really user friendly, thats why i made that google colab i sent you which is in the guide

its easier to use

[strong fulcrum]

yeaa that makes sense

[strong fulcrum]

but so i would install it either from this github or pip install it for local right

[low shard]

you wanna do it locally? youd have to look up in their official github guide https://github.com/suno-ai/bark?tab=readme-ov-file#-installation

[strong fulcrum]

and speaking of, there are still random noises even with voice presets. like a random loud beep, random faint music in the back (from which i assume is from it being trained from videos?), and also it often randomly switches voices even with voice preset set

[low shard]

from what i just tested, i checked the use voice presets option and it was always the same voice, there might be some weird noises sometimes tho as it is focused only on emotions and its not still in development anymore

[bright fern]

hiii, i was trying to use a model, but often the link doesn't work, for example this one: #1175500281979605034 message

can anybody tell me if there is something i can do?

[low shard]

what google colab do you use? send it

[bright fern]

sure, ilaria rvc

this one https://colab.research.google.com/drive/16LkwvFZeudTpUOsE_6bMjOq2qkxFo8Hr?usp=sharing#scrollTo=9yQA9h6N2YV7

should i use another one?

[low shard]

Its the right one yea

You dont need to use another one

The problem is the model download

[low shard]

That model download is deleted

So thats why

[strong fulcrum]

oh really? interesting

also yea, idk, seeing how barktts works, it would be a bit slow for local :(

[low shard]

I mean, local depends all on the power of you pc

[strong fulcrum]

not sure how this would fair

colab ^

[low shard]

How long was the text😭

[strong fulcrum]

i inputted an array of sentences from a text, and generating each sentence

lol

[low shard]

If u put very long text ofc its gonna be slow 😭

[strong fulcrum]

tho each generation is only a sentence or two

[low shard]

This all is just for one of the sentence?

[strong fulcrum]

no, each one is a sentence or two

[low shard]

You sure you are connected to the colab gpu btw

Check to the up right and see if theres a "T4" near to ram

[low shard]

Okay good

Well, it all depends on the lenght of the text, considering bark is abbandoned from some months😭

[strong fulcrum]

unfortunately

[bronze maple]

can anyone understand the graph?
I did 400 epoch
Could I have continued training him?

or is the model overtrained?

I used a quality dataset extracted from the games with deverd and denoise

[rare goblet]

Ayo? @bronze maple level 2 !!!

[strong fulcrum]

yea, and it seems the voice bank isn't very balanced. like en voices only 9 males and 1 female, while japanese voices have like 8 female and 2 male

[low shard]

Sorry but i can't really do much about that as i didnt make the tts model😭

[strong fulcrum]

yeye im just sayin

ur good

[bronze maple]

sei italiano ?

[strong fulcrum]

also, are there other tts that have similar goals/functions as bark?

[low shard]

Si

[bronze maple]

Piacere cosa cerchi da fare ?

[low shard]

In che senso che cerco di fare

Sto solamente aiutando alcune persone

[bronze maple]

a capito 😂 mi sembrava che eri tu che ai bisogno

[low shard]

Mmm maybe you could try tortoise tts and xtts2 which are in that guide i sent yoi

[low shard]

Nono stavo parlando del modello di bark ai tts, visto che mi aveva chiesto perché ci sono di meno voci feminili e gli ho detto che non so visto che lha fatto suno ai non io il modello

[strong fulcrum]

I just crashed my computer lmao

[bronze maple]

aaah ok
Alleni anche le voci tu con RVC ?

[low shard]

How did you even manage to do that😭

[low shard]

Si ho trainato alcune voci

[strong fulcrum]

Lmaoo idk

Screen froze, then pink screen appeared, then turned off XD

[low shard]

Are you using the ai locally

[strong fulcrum]

yea lol

[low shard]

Your pc is prolly too old😭

Whats your gpu

[strong fulcrum]

it's a mac m1 16gb lmao

[low shard]

Macs arent well known for running ai things

[strong fulcrum]

well m1 seems to be alright from other ppl

[bronze maple]

E' una settimana che sto cercando di scambiare con qualcuno, da 2 settimane mi sto allenando per allenare le voci, voglio migliorare nonostante le conoscenze che ho appreso

[strong fulcrum]

tho tbf, i have a million tabs open in safari

[low shard]

It also depends on the gpu, for example m1 is very very less powerful than T4 (google colab gpu)

[strong fulcrum]

huh interesting

[strong fulcrum]

tho can't be using google colab all the time

[low shard]

In che senso allenando? Stai trainando una voce al giorno?

[brittle wing]

Where do I find readymade datasets?

[strong fulcrum]

the more u take advantage of it, the more they nerf it for ur account

[low shard]

I mean true, if you want theres also an hf space on the guide for bark tts (it uses T4 gpu paid by suno AI)

[strong fulcrum]

yea i've seen it

[low shard]

Btw bark technically could run even on cpu if im not wrong, not sure why you got that error on mac but honestly not even the newests mac are able to train ai models, there's a difference from an ai task to safari tabs lol

[bronze maple]

Sì ahah 😂 , padroneggio il dataset, ma non capisco il grafico della tensorboard, voglio capire come realizzare una voce nel modo più perfetto possibile

[strong fulcrum]

so the only thing i have on right now is discord, terminal, finder, and textEdit, and i'm not running any AI programms rn (no applio no nothin), and this is my mem

[low shard]

Se vuoi questo ti può aiutare https://docs.aihub.wtf/rvc/resources/epochs--tensorboard/

[bronze maple]

Si o visto fra , ma non capisco Non so quando dovrei interrompere o continuare la formazione, se il mio modello è addestrato meglio o meno

[low shard]

I honestly dont use mac but its using 11gb of ram by just having that apps you said in the message what😭😭

[low shard]

Dovresti cercare "g/total"

Con la g

[hoary zephyr]

how many epocs might i need for a 20 minute dataset?

[low shard]

E monitorare quel grafico

[bronze maple]

sì, è nell'immagine che ho inviato

[low shard]

There isn't a right amount of epochs for training an rvc ai voice model, check this pls https://docs.aihub.wtf/rvc/resources/epochs--tensorboard/

[low shard]

Tecnicamente il modello non sta overtrainando, sembra un po sus quel punto alla fine che si rialza, aspetta un altro po e controlla come va, il punto di overtraining é il punto più basso prima che si rialza tutto, se vedi che overtraina stoppa il training

[strong fulcrum]

I tried closing everything and now it’s using 5gb

[low shard]

You could retry using it locally but i dont know if it will work for mac honestly

[strong fulcrum]

Bro i had like maybe 500 tabs open in safari and closing it made it drop like 4 gigs

[low shard]

😭

[strong fulcrum]

Lolll

[hoary zephyr]

hows this im at 130 epocs and 5.95k steps

[bronze maple]

Ok grazie riprovo allora

[rare goblet]

Ayo? @bronze maple level 3 !!!

[hoary zephyr]

its set to go to 500

[bronze maple]

How many minutes of dataset do you have?

[hoary zephyr]

20 min

[rare goblet]

Ayo? @hoary zephyr level 3 !!!

[bronze maple]

and do you think that model is training correctly and that he is not over-training?

[low shard]

You putted smoothing to max right

[hoary zephyr]

mhm

[hoary zephyr]

idk im new to this

[low shard]

Model looks fine right now dw

[hoary zephyr]

well would you think 500 is ok based on it

[low shard]

Well, that depends model to model, it could be yours overtrains before 500 epochs, or doesnt, 500 epochs is just used to see how the model training is going, like as its much epoch and you will have a detailed graph

[bronze maple]

Which RVC do you use? I have Applio v3 but I don't know if it's good

[hoary zephyr]

its a model of my voice with talking and singing

[low shard]

You mean locally or on google colab?

[low shard]

Oke, just check the graph time to time and see how it goes

[bronze maple]

Local fra

[elder estuary]

i need help with paperspace

[hoary zephyr]

kk thx ill keep yall updated

[low shard]

Ah non ho un buon pc lho faccio in colab io, cmq applio é buono da quante ne so

[low shard]

I might passout later its 3 am lol

[bronze maple]

google collab e una cagata oh un buon pc i9 RTX Studio 4080

[hoary zephyr]

im gonna be here till 3 am mst at this rate lol

[low shard]

Google colab offre una T4 gpu gratuita che è diciamo lo stesso buona per ai, ovviamente non super veloce come una rtx 4080 ma non una cagata per persone che hanno un laptop o vogliono farlo da mobile

[elder estuary]

how do i install rvc on paperspace

[bronze maple]

Aggiungimi in amico se vuoi posso train un modello per te proviamo

[low shard]

Sinceramente non ho nessuna richiesta di modello da creare per ora, tutti i modelli che volevo (40 se non mi ricordo male) gli ho già fatti

[full token]

how do i fix my rvc not creating any .weights file

[bronze maple]

posso vedere un modello che ai fatto bene ?

[delicate dome]

Hi

Can you help me with an error on applio?

[bronze maple]

@low shardlo sai comme si fa per riprendere il train di un modello ? cosi non ricomincio da capo

[low shard]

Mi dispiace ma in locale non so molto, potresti provare a vedere il 5 resuming step qua https://docs.aihub.wtf/rvc/local/applio/#training-

[low shard]

Sinceramente gli ho fatti mesi fa ormai, pero li trovi in #1175430844685484042

[austere ice]

why arent weights showing up

[rare goblet]

Ayo? @austere ice level 1 !!!

[austere ice]

in the folder

[low shard]

Are you using ilaria rvc google colab version?

[full token]

same problem as me too

[low shard]

Send the model download link you are using

Thats the link to open the ilaria rvc ui

[low shard]

Ima delete it i dont think you want others to use it lol

Yup lemme check it

That isn't a rvc model

That's a gpt so vits model (the tts ai tool)

So ye u cant use it in ilaria rvc

Youd have to use gpt so vits which is inside the #1212368971127590922 guide (the docs link one)

Or just find another rvc model

Yw

[austere ice]

when

extracting

weights isnt showing up

[full token]

yo my rvc isnt making a .weights file

ah fuck im just stupid

forgor to look in the assets folder

[hoary zephyr]

hey is there anyway to export a model early with the disconnected Collab its about to reach 250 epocs and i want to save this iteration for use in case of it getting overtrained at 500

[bronze maple]

Look guys, I tried to resume training on my old model, the graph has drifted, is this normal? 😂

[timid olive]

How and how should the graphic look?

[brittle wing]

Type loss/g/total

First go to scalars, then reload data, from the settings icon type loss/g/total set smoothing to 0,987

[strong fulcrum]

Hey man another question for ya, do u use a serverless/cloud computing service for ur personal ml stuff? What are some good ones? Are there any good free ones?

[low shard]

You mean what do i use for using my personal ai projects online ?

Google colab or hugging face spaces

[strong fulcrum]

Yeah

[low shard]

I dont really do anything local as my laptop doesnt let me lol

[strong fulcrum]

O ic. Arent there a lot of limitations tho

Well, not sure about huggingface spaces tho. Can u connect it to ur local program?

[low shard]

I mean for google colab you get free daily gpu and you can just switch google alt accs yk, and some stuff works without taking much time on cpu on hugging face space like #1163571683848900629

[low shard]

Well hugging face spaces work as like docker online, you have only free cpu 24/7 and if you build smt really good you could ask for a community gou grant for always have gpu on your hf space

But i dont think you can connect it to your local pc

[strong fulcrum]

I see

[strong fulcrum]

What about for colab? Can u hook up outputs to ur local program?

[low shard]

Ehh never tried that but i don't think you can hookup local stuff to cloud one

[elder estuary]

how do i fix error 1 in paperspace

[strong fulcrum]

Ic. So all thats rlly possible with those two would be in an enclosed environment then

[rare goblet]

Ayo? @strong fulcrum level 7 !!!

[full token]

in tensorboard what are epoch number are steps equivalent to?

[low shard]

Yup i really think so

[strong fulcrum]

ic, ok. thanks man

[low shard]

Yw

[gusty kestrel]

@coarse roost, I have found 5 results that match your search!

syberg

Meliodas (The Seven Deadly Sins Japanese) [REUPLOAD]

• Uploaded: <t:1697273403:d>
• Likes: 0

Epochs

200

Technology

RVC

Algorithm

Mangio-crepe

Requested by minamishimura · Apr 2, 2024, 4:04 AM

[fossil vigil]

oop

so rvc wont launch

not sure what to do

ill send the log if anyone can help

Running with the runtime Python, Please wait.
Error processing line 1 of C:\Users\myuser\Downloads\RVC-GUI-pkg-20220525-mp3fix\RVC-GUI-pkg\runtime\lib\site-packages\distutils-precedence.pth:

Traceback (most recent call last):
File "site.py", line 169, in addpackage
File "<string>", line 1, in <module>
ModuleNotFoundError: No module named '_distutils_hack'

Remainder of file ignored
Error processing line 1 of C:\Users\myuser\Downloads\RVC-GUI-pkg-20220525-mp3fix\RVC-GUI-pkg\runtime\lib\site-packages\google_auth-2.16.2-py3.9-nspkg.pth:

Traceback (most recent call last):
File "site.py", line 169, in addpackage
File "<string>", line 1, in <module>
File "<frozen importlib._bootstrap>", line 562, in module_from_spec
AttributeError: 'NoneType' object has no attribute 'loader'

Remainder of file ignored
Error processing line 1 of C:\Users\myuser\Downloads\RVC-GUI-pkg-20220525-mp3fix\RVC-GUI-pkg\runtime\lib\site-packages\matplotlib-3.6.2-py3.9-nspkg.pth:

Traceback (most recent call last):
File "site.py", line 169, in addpackage
File "<string>", line 1, in <module>
File "<frozen importlib._bootstrap>", line 562, in module_from_spec
AttributeError: 'NoneType' object has no attribute 'loader'

Remainder of file ignored
Error processing line 7 of C:\Users\myuser\Downloads\RVC-GUI-pkg-20220525-mp3fix\RVC-GUI-pkg\runtime\lib\site-packages\pywin32.pth:

Traceback (most recent call last):
File "site.py", line 169, in addpackage
File "<string>", line 1, in <module>
ModuleNotFoundError: No module named 'pywin32_bootstrap'

Remainder of file ignored
Traceback (most recent call last):
File "C:\Users\myuser\Downloads\RVC-GUI-pkg-20220525-mp3fix\RVC-GUI-pkg\rvcgui.py", line 3, in <module>
from tkinter import filedialog
ModuleNotFoundError: No module named 'tkinter'
Press any key to continue . . .

[gusty kestrel]

@coarse roost, I have found 4 results that match your search!

tjkcreative

Elizabeth Olsen (Scarlet Witch)

• Uploaded: <t:1693353600:d>
• Likes: 0

Epochs

300

Technology

RVC

Algorithm

Rvmpe

Requested by minamishimura · Apr 2, 2024, 4:20 AM

[paper silo]

Hey does anyone know where this error would be coming from?

File "/Users/USERNAME/pinokio/api/rvc.pinokio.git/app/infer-web.py", line 496, in click_train
& set([name.split(".")[0] for name in os.listdir(f0_dir)])
FileNotFoundError: [Errno 2] No such file or directory: '/Users/USERNAME/pinokio/api/rvc.pinokio.git/app/logs/my-test/2a_f0'

I am using pinokio which help me to easily setup RVC on my Macbook Pro M1

[solid path]

has anyone managed to get Applio to run on Mac (m1) locally?

[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

[kindred rain]

are there any angelic, holy, or echoy voices? sorry this is all very new to me

[rare goblet]

Ayo? @kindred rain level 1 !!!

[rugged stirrup]

You could look on weights.gg and type the keywords in and see or use one of the bots or just look in the #1175430844685484042 channel

[kindred rain]

okay! thank you so much!

idk how to use the bots but ill look at that website thank you!

[rugged stirrup]

Np

[frozen quartz]

does anyone know why i'm getting this when i try to run the go-web file in rvc?
(i closed the console by accident while running it the first time, but after deleting and unzipping the whole folder again it still doesnt do anything)

D:\RVC\Mangio-RVC-v23.7.0>runtime\python.exe infer-web.py --pycmd runtime\python.exe --port 7897
'runtime\python.exe' is not recognized as an internal or external command,
operable program or batch file.

i have python 3.12.2-amd64 and pytorch pip and nvidia cuda12.1 installed

[red spire]

guys im kinda new to all this ai stuff

but does anyone know what is python

[frozen quartz]

python is a programming language code developers use, similar to other ones you might have heard of e.g. java, c++, ruby, etc

[red spire]

i was told its needed to run the program

[red spire]

is pytorch good?

[frozen quartz]

pytorch is a library that some apps use for machine learning (basically stuff to do with AI) in tandem with python

[rare goblet]

Ayo? @frozen quartz level 2 !!!

[red spire]

alr

[rare goblet]

Ayo? @red spire level 1 !!!

[latent kettle]

can some tell me how to use conpiled version of applio. What should i have to download. .exe or .zip or both

[sage seal]

i would like help with setting up the jcshlatt link on the google page. it doesn't seem to work

[latent kettle]

try with python 9.8

[sage seal]

idk what that it dawg

im using mac

[latent kettle]

what error do you get?

[sage seal]

uhh

[rare goblet]

Ayo? @sage seal level 1 !!!

[sage seal]

JSONDecodeError Traceback (most recent call last)

<ipython-input-11-20feb5d5be2f> 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)

[storm hatch]

is there at least a way to voice clone with rvc by tts? not ai cover.

[sage seal]

and this is the link for the model: https://huggingface.co/analogspiderweb/analogspiderweb_RVC/resolve/main/JSchlatt-HD.zip

[frozen quartz]

now it just doesnt respond

returns this instead when i try to run as administrator:
C:\WINDOWS\system32>runtime\python.exe infer-web.py --pycmd runtime\python.exe --port 7897
The system cannot find the path specified.

edit: nvm it works now, sorry for ping

[latent kettle]

working? i think you forget to copy python to path. and now you have fixed it ryt ?

[latent kettle]

are you using rvc on mac locally?

[frozen quartz]

working now, installed python 3.9.8 and installed to path, then let go-web run for a while (also copied to path last time with 3.12.2, i think i just didnt let it run long enough)

[plush compass]

CoverGen_NO_UI What do these options mean in Colab? Can anyone tell me?

[graceful obsidian]

you only really wanna play with:

• pitch or pitch all ( not sure how they implemented the transpose in here )
• index rate

rest leave untouched

index = accent, voice gestures of the model and such alike.

• more index, more style of your model you retain but can results in artifacts if the model isn't the best.
• less index, more features you take from input audio ( acapella )

in simple words, it's " do I want it to sound ( vocal features and accent wise ) closer to the model or closer to whoever sings the input / acapella "

0.3, 0.5 and 0.6/0.75 are typically fine. you rarely, if at all, ever go above 0.8.
You can also choose to not use index at all by setting it to 0.
Do 2 renders, one at 1.0 and one at 0.0 index, you'll understand better what it does.

[frozen quartz]

when OT happens is the "step" number on tensorboard the same as number of epochs?
also if i am trying to continue training someone else's model, is there any way can i tell what sample rate their model has been trained at if they havent told me?
(i got the model off applio from the applio bot search and it doesnt seem to mention it anywhere)

[proper shale]

steps compose epochs, so no

[forest sluice]

For sample rates, you can check the output audio

[proper shale]

and you can continue training someone else's model if you have their d/g files and their preprocessed files

just tell them to send the model folder inside logs, that's probably safer

[proper shale]

oor check the .json and .log files, iirc

either f0 extract and preprocessing have the sample rate in the beginning

[cunning prawn]

Hi is there a valorant agent Gekko voice for voice.ai?

[hallow blade]

wich are good e girl voices

[lilac glacier]

hii

Pretrain type- OV3 or RIN_E3 which one best

for 20 Minutes data set

[maiden lynx]

i cant download anything like its just 0 percent all the time

[latent kettle]

which pre trainer is best?

ov2 or rin_3

[latent kettle]

same question

[lilac glacier]

kiddan broo

[hidden saddle]

how do i get the voice changer?

whats the website

[rare goblet]

Ayo? @hidden saddle level 1 !!!

[red spire]

whenever i click start nothing happens

someone help

pls

command prompt pops up then disappears

[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

[proven hill]

use #🔍│help-w-okada

[obtuse stream]

can anyone tell me how the appio overload sensor works? because it's a little strange that i set the workout to 300 epochs, but the sensor can be set to a maximum of 100 and it stops the workout, but it doesn't matter what i set it to, it can be 5 or 10 and it still stops. it wouldn't make sense to set a number for the workout and then the sensor detects if there is an overload and stops it. it's totally unnecessary to set a number

I don't think you should set a number for overtraining, but the program itself should detect when it are overtraining and stop himself.

[red spire]

do you know the problem i have??

[proven hill]

no

[red spire]

okay

[stable hemlock]

how do i fix the voice changer to stop echoin

[pastel oak]

enable "echo" and "sup2"

[stable hemlock]

dint work and also it seems to stop like it just stops

[quartz coral]

@shaq since you told me to search GIGABYTE G5 KF-E3DE313SH " and found it to be 900 dollars or 1100 euros is it good for training voicees and i found it to be rth4060 is it good and enough?

[pastel oak]

yes any choice with rtx 4060 is great.

[quartz coral]

Ok thanks

[rare goblet]

Ayo? @quartz coral level 6 !!!

[latent kettle]

which pre trainer is best?ov2 or rin_3

[half cove]

rin_e3 is super sensitive to noise

ov2 is good for shorter datasets

[latent kettle]

can i use it for longer dataset or for shorter dataset

[half cove]

longer

[half cove]

i think it works by seeing the loss and if it doesnt go down based on the threshold amount it stops training

if your model has mode collapse it prob breaks

[latent kettle]

how to use conpiled version of applio. What should i have to download. .exe or .zip or both

[half cove]

both should work

[latent kettle]

is it Compulsory to download both

or anyone of them will work

[half cove]

nope only one

[latent kettle]

which one is best

[half cove]

both work

doesnt really matter

[mint anvil]

what is the index file and how do i get it

i trained my ai model some time ago and i didnt get the index file

[half cove]

index file is the accent

you get it by clicking train index

[frozen quartz]

anyone know where i can find voice samples for characters?

[low shard]

You mean voice samples for making your dataset to train the ai voice models?

I mean

Like YouTube

[gray ermine]

Where can I try to train a model?

[ripe jacinth]

I went on a side quest to make it (the default takes 44mins for me to install while this takes 4mins)
https://colab.research.google.com/github/Ran-Mewo/voice-changer/blob/main/Voice_Changer.ipynb

[frozen quartz]

yea but they either have special effects or dont have every voiceline and its gonna take ages to splice all the voicelines together

[sand iris]

44mins ?

Seems off to me, unless you're using the old one from the voice-changer repo ?

[low shard]

I think he is

Its not hina one

[frozen quartz]

also if my audio is 44khz sample rate do i set target sample rate as 40 or 48khz

[ripe jacinth]

yeah for me google colab downloads at 2MB/s
idk if it's aussie or something

[rare goblet]

Ayo? @ripe jacinth level 1 !!!

[haughty plaza]

Doesn't this graph converge too quickly for just 1h of training data? Running 32 batches on 4090 so that might be why? https://imgur.com/kNmC0IJ

[low shard]

Btw you dont have to splice all the voice lines in 1 file or use more files, it doesn't really matter as long all of them are in a folder, and you could try to remove the effects via uvr

[haughty plaza]

but 8m for a 1h dataset seems too quick

[ripe jacinth]

• I am paying for google colab so it taking an hour to install is like rip monei

[frozen quartz]

also if i have a model that i want to train further do i just put the same model as model g and model d?

[haughty plaza]

yes just look for these files D_2333333.pth & G_2333333.pth

[frozen quartz]

only 1 model and 1 index though, neither state whether they are d or g

also btw with a 2:30 voice sample how many epochs do i go for? 500?

[rare goblet]

Ayo? @frozen quartz level 3 !!!

[haughty plaza]

check the logs folder, iirc that's where they are

[gusty dagger]

its says i need a smaller image sizw

e

[frozen quartz]

i got the model off #🔍│find-models with the applio bot and it just gave me the model and index

[brave garnet]

RVC Colabs and Spaces

⠀

Google Colabs

⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.

AICoverGen-NoWebUI
Useful for making covers, doesn't inclued a UI, by Ardha, by Eddy, Hina and Gdr.

RVC Disconnected
To train new voice models, by Kit Lemonfoot.

EasyGUI
The OG interface, by Rejects.
⠀

Made with love by Ilaria 💖

[burnt girder]

Is there any free alts for voice.ai or is everything paid

[frozen quartz]

use w-okada instead, also free
check pins in #🔍│help-w-okada for installation guide

[burnt girder]

alright thanks!

[haughty plaza]

Oh I see sorry, that won't work for further training, unless the author shares the D_2333333.pth & G_2333333.pth files from the model's folder, you would need to train from scratch

[rare goblet]

Ayo? @haughty plaza level 1 !!!

[frozen quartz]

ty for help anyways though

actually one more thing, what do i set batch size per gpu to? im using a nvidia gtx 1660 super
oh and also at around how many epochs will the model quality start dropping off if im using a 150s voice sample (was thinking will train to 2k epochs or so, idk)

[haughty plaza]

that's 14 GB VRAM right? I think it's okay to start with 16 and if it errors out just lower it to 8 maybe? Also for 2-5m it should be fine to train for 200-250 epochs

[frozen quartz]

anything over 300 or so epochs and model quality will start going down right

[haughty plaza]

theoretically yes but if you want to be 100% accurate I would install tensorboard and check as it's training #✨│ai-help message

[frozen quartz]

im just gonna start training and go to sleep lol, in that case ill just set for 500 epochs and check tomorrow to find which one is best. thanks!

[haughty plaza]

no worries good luck

[proven hill]

where did you take this

[haughty plaza]

the epochs? #✨│ai-help message

[proven hill]

lol, funny

i made that stuff 1 year ago

and still goes around

[sand iris]

weird

[haughty plaza]

If it's one year old I'm guessing it's obsolete

[proven hill]

its more that..

its really depends

[haughty plaza]

well I'm just running tensorboard as I'm training so yeah thought it's really not a one size fits all thing. this is 1hr dataset converging at 8mins btw

[brave garnet]

RVC Colabs and Spaces

⠀

Google Colabs

⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.

AICoverGen-NoWebUI
Useful for making covers, doesn't include a UI, by Ardha, by Eddy, Hina and Gdr.

RVC Disconnected
To train new voice models, by Kit Lemonfoot.

EasyGUI
The OG interface, by Rejects.
⠀

Made with love by Ilaria 💖

[ripe jacinth]

oh I thought hina's one was outdated my bad

but I just looked at hina's code and it's basically the same without the colab quality of life stuff
like

# Install dependencies that are missing from requirements.txt and pyngrok
!pip install faiss-gpu fairseq pyngrok --quiet
!pip install pyworld --no-build-isolation --quiet
# Install webstuff
import asyncio
import re
!pip install gdown
!pip install torchfcpe
print(f"{Fore.CYAN}> Installing dependencies from requirements.txt...{Style.RESET_ALL}")
!pip install -r requirements.txt --quiet
``` (hina's code)

it's those dependency installing stuff that takes like so long to do
my colab thing has everything preinstalled as a 4.5 GB image

[ionic roost]

Anyone know how I remove echo?

[spare gull]

hi guys,i dont know how to train a voice model,i search on youtube and down a file

i cannot find go web.bat

can anyone give me the file

[brittle wing]

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

[spare gull]

all tutor tell me to down RVC1006Nvida

[spare gull]

thx

lemme try

wait

same thing

[brittle wing]

Did you download the files?

[spare gull]

yes

go web

this?

the 1kb one

my pc even cant open the webui

[rare goblet]

Ayo? @spare gull level 1 !!!

[spare gull]

sad

[brittle wing]

My .PTH file is not creating, but the rest is fine, or I can't find it. Any ideas on how to fix this?

[wispy burrow]

Is that all you downloaded?

[spare gull]

no

i down the whole file

but i wait for it 10 min

nothing happen after the black window cames out

seems my pc hate me

[brittle wing]

I have created .index files "added_IVF923_Flat_nprobe_1_Test_v2" but that's all

[rare goblet]

Ayo? @brittle wing level 1 !!!

[spare gull]

i gonna do something crazy

[gusty kestrel]

@violet pecan, I have found 8 results that match your search!

rainbowzigzag2006

Girl Jordan (The Loud House)

• Uploaded: <t:1692921600:d>
• Likes: 0

Epochs

N/A

Technology

RVC

Algorithm

Rmvpe

Requested by sillyfoo1s · Apr 2, 2024, 3:51 PM

[spare gull]

let my pc become the original mode

[wispy burrow]

Does it display any errors or is it empty?

[brittle wing]

Not there either :/

[spare gull]

keep like this

[wispy burrow]

You did train, right?

[spare gull]

nothing happen after i i open go web

[brittle wing]

yep

[spare gull]

i even wait 10 min

nothing happen

my laptop

[brittle wing]

I can send you screenshots on DM what folders and files I have after training

[wispy burrow]

You can post it here.

[spare gull]

so i can just give up?

sad

@wispy burrow sir,is this mean my laptop too bad

but i can open the rvc voice changer

i have a wav.file,is that possible other ppl help me train

[wispy burrow]

Don't you have anything in your weights folder?
The ModelName.pth goes in there.

[wispy burrow]

Yuo may be missing something.
Do you have Python?

[sand iris]

no need to have python with the runtime releases

nvidia1006.7z has python bundled

[rare goblet]

Ayo? @spare gull level 2 !!!

[sand iris]

just have to unzip it via winrar or 7zip

else it bugs out

[spare gull]

i use the normal unzip

uh

[brittle wing]

Maybe I've messed something up here?

[proper shale]

batch size too high for ur GPU, don't think it'll train like that

[wispy burrow]

Batch size 16?
I have the same GPU.
I can't go higher than Batch size 2.

[wispy burrow]

You very likely ran into an OutOfMemoryError.

[brittle wing]

alr i will change that

[proper shale]

next time you click on "Train model" check cmd, as it can help to identify errors

[proud elbow]

go batch size 4, as I know 1660 super 6 gb is fine on that

[wispy burrow]

It's a GTX 1060 6GB though.
An even older model.

[proud elbow]

same vram but idk if 1060 could throw OOM error on that

[proven hill]

why the hell are you using harvest in 2024

[wispy burrow]

Mine does...
I can't go higher than Batch Size 2

[proven hill]

TRAINING WITH A 1060????????????

man you...... no...........

[wispy burrow]

Yes...

[proven hill]

just no

[proven hill]

i kinda wanna disappear

[wispy burrow]

17 hours for a 15 minute dataset...

[proven hill]

17 hours is a very, very positive est.

1060 is a 3GB Vram, it can barely run okada, let alone training with cuda cores

[haughty moss]

rvc disconected how to create model

[brittle wing]

I'm new, first time using this 🫸 🫷

[proven hill]

use the guides, trust me

[haughty moss]

anybody provide me regulations

[proud elbow]

@brittle wing is using 1060 6 gb

[proven hill]

still a 1060

[rare goblet]

Ayo? @proven hill level 101 !!!

[proper shale]

vram depends on the variant but yes you're right

[proper shale]

https://docs.aihub.wtf/

[proven hill]

even a 24gb 1060 would still be a 1060

[proper shale]

yeah

slow as balls

[wispy burrow]

By the way, now that you're here.
I can't get llaria RVC Mainline beta to work...

[wispy burrow]

Yeah...

[proven hill]

okay lets troubleshoot it together

whats the error?

[brittle wing]

Anyone know how to connect rvc to discord?

[proven hill]

#🔍│help-w-okada

[latent kettle]

follow guides

[latent kettle]

just follow the steps as written

[latent kettle]

https://docs.aihub.wtf/

this is what you actually need

[brittle wing]

okay 🙂

[rare goblet]

Ayo? @brittle wing level 2 !!!

[wispy burrow]

I don't know...
"Ilaria RVC is starting..."
Than after a short while it displays some text but it immediately closes.
I have no idea what it said...

[proven hill]

open it with the cmd

[haughty moss]

Ok

[wispy burrow]

It's the cmd that closes.

[proven hill]

open the cmd inside the folder and paste

Ilaria-RVC-Launcher.bat

[latent kettle]

just prepare a dataset uf you dont know follow the guides

[proper shale]

does it do the instant close thing?

[proper shale]

if so try the video that gdr pinned in #🔍│help-w-okada but instead use ilaria rvc mainline's path

[proven hill]

open it like this

then paste that line

[wispy burrow]

Ah...now it shows ModuleNotFoundError: No module named 'audio_separator'

[proven hill]

use the Assistant to get the update

should fix

if it doesnt

open cmd again

and paste this

pip install audio-separator

after that

pip install audio-separator[gpu]

but again, with the assistant should work

or you can simply do

pip install -r requirements.txt

to fix every dependencies problem

[latent kettle]

sorry for asking but can you please guide me how to start learning Ai development ?

[proven hill]

dont

for the sake of your mind

[latent kettle]

i think i have pinged a wrong person. 😰

[proud elbow]

[proven hill]

who do you wanted to ping

lol

[latent kettle]

you don't want to guide me that's why i said this. i think i have to find someone else. Sorry 😊

[proven hill]

how can i guide you tho

[latent kettle]

i just said that please guide me how to start learning Ai development ?

[haughty moss]

rise issue

Rvc could not find file 😥

[proven hill]

i dont know where to start

[latent kettle]

dont creat a folder inside RVCdisconnect folder

[proud elbow]

put the zip file in rvcDisconnected folder, not in a subfolder

[trim escarp]

Hi, I am trying to use the local version of RVC, but when connecting to the web, I get "ERR_SSL_PROTOCOL_ERROR" in Chrome and the terminal says that "Invalid HTTP request received". Anyone met with same problem?

[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 burrow]

Ok I've done that now.
But now I've got:
ModuleNotFoundError: No module named 'importlib_resources'

[proven hill]

pip install importlib-resources

[wispy burrow]

Alright!
It opened Ilaria RVC.
Oof...Red over Brown is hard to read...

[proven hill]

Red over Brown?

wdym

[wispy burrow]

The brown buttons with red text are pretty hard read.

[proven hill]

theres no brown button and no red text..

[serene hound]

umm... why?

i just wanted to try how blending works..

[toxic scarab]

im struggling to find an avatar for my custom voice can someone help._.

[wispy burrow]

...My text is red, not orange...

[proven hill]

the hell happened

its supposed to be like this

[proven hill]

OHHHHHHHH

ITS LIGHT MODE

enable dark mode in your browser

[wispy burrow]

Ah that solved it.
Looks much better and easier to read.
Thank you.

[gray ermine]

I created a model, but for some reason when I want to make a cover, the site says that the file is not zip, although the model and the index are in the zip folder

[proven hill]

no problem, i need feedback, if you want you can DM me !

[trim escarp]

I really hate that the bot responded with the pages I follow to install it and there is no mention of my problem

[brittle wing]

where do I download rvc

[sand iris]

define RVC

real time voice changer ?

or the "real" rvc to train models with ?

[trim escarp]

The Mainline branch of Original RVC

[rare goblet]

Ayo? @trim escarp level 1 !!!

[sand iris]

make sure to connect via http and not https

[trim escarp]

Yes, both http and https conenctions give the same result

System time is also okay

Oh, seems like a browser issue, it seems to work on Edge

[fiery raptor]

ive been trying to get into the ilaria rvc but whenever i do the run part it just shows me this

[hoary zephyr]

Is rvmpe the best or is that just because it's faster with decent quality?

[brittle wing]

I would like to clarify a question: how to use artificial intelligence to create music using AMD's video cards? I'm using Gradio and it's using my processor to run the model, which is taking too long. Is it possible to use the video card instead of the processor?

[gray ermine]

I created a model, but for some reason when I want to make a cover, the site says that the file is not zip, although the model and the index are in the zip folder. What should I do?

[elder estuary]

what rvc did you use

[rare goblet]

Ayo? @modern marlin level 3 !!!

[elder estuary]

did you do it locally

no did you do it on google colab or using your pcs resources

ok

[rigid jewel]

-help

[azure marsh]

Command has expired

[upper furnace]

Help (ilaria RVC)

https://discord.com/channels/1159260121998827560/1224818776990814239

[rigid jewel]

-help

[azure marsh]

Command has expired

[rigid jewel]

-rvc

[azure marsh]

Documentation

📚 All-In-One English documentation

https://docs.aihub.wtf/

❔ English FAQ

https://docs.aihub.wtf/faq

Credits: AI HUB FRANCE

✨ More guides

Full AI Voice Model Training Guide (Local)

Link: YouTube
Credits: Christopher Villanueva

Model training with Mainline RVC

Link: Rentry
credits: Raven (ravencutie21)

AICoverGen Colab Guide

Link: Google Docs
Credits: Eddy (Spanish Helper)

Create a model with RVC disconnected (colab)

Link: Google Docs
Credits: Angetyde

How To Make an AI Cover With Ilaria RVC

Link: Rentry
Credits: 👽 Julia (ailen2091)

[rugged nymph]

ive been making audiobooks but the time it takes varies wildly from a minute and a half to five minutes similar length text so why is it settings or sumthing

[hazy thunder]

does having a really long clean dataset make the model better or worse

long as in about 30 minutes

[wise wind]

the voice-models tab wont open for me

[rare goblet]

Ayo? @wise wind level 1 !!!

[low shard]

no no it doesnt make things worse as long as its well cleaned, the max id suggest for making a dataset is 1 hour as after that u dont really hear changes

[low shard]

please be more specific

[hazy thunder]

whats the average

length

to make it actually sound good

good as in realistic

[low shard]

i mean 30 mins is good

there isnt really an average tbh

some models can be good even on 10 mins or 50 mins

what mostly depends is quality over quantity

for a good model u require 5 or more mins (i dont mean u have to throw ur dataset, its good u have a 30 min one)

[wise wind]

@low shard if fix e#it nevermind sry

[low shard]

oke dw

[hazy thunder]

does denoise work for keyboard noises in the background?

[pure lance]

why does this not work

[rare goblet]

Ayo? @pure lance level 1 !!!

[low shard]

i already replied to you in https://discord.com/channels/1159260121998827560/1224832529807446056

please check that

[balmy creek]

Hellooo !

Can someone please help me?
Basically when I try to use RVC, there are the normal stats at top left corner (vol, buf, res, rtf) and for some reason I couldn’t hear myself or record anything. Then I noticed the vol thing was at 0.00000 too and idk what to do honestly:)

[pure lance]

srry mb

[low shard]

dw its fine

[balmy creek]

Anyone ? :p

[low shard]

are you using the realtime voice changer wokada?

[balmy creek]

Not sure what you mean by wokada

[rare goblet]

Ayo? @balmy creek level 1 !!!

[balmy creek]

I’m really confused rn I just got it lol

[low shard]

are you using the tool for ai covers or realtime voice changer

[balmy creek]

Realtime

I’m on the realtime voice changer client rn

[low shard]

yea i think what you are using is called wokada, for that please ask in #🔍│help-w-okada or #1192011222023950368

[balmy creek]

Alright

[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

🤗 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

[hazy thunder]

do dataset clips HAVE to be under 10 seconds?

[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

[low shard]

no no, usually for a model is reccomended the dataset is atleast 5 mins for a good one

[hazy thunder]

i mean like

the cut clips

if u have a 5 min dataset does it have to be cut into small 10 second clips

[low shard]

nope, it doesnt matter if you cut your dataset into lil clips or put it all into a single wav file

[hazy thunder]

yea but i dont want silence

[low shard]

its your choice

[hazy thunder]

in it

i have no idea how to make audio slicer filter out the silence

[low shard]

for that you can read this https://docs.aihub.wtf/rvc/resources/datasets/#cleaning

[proper shale]

10 HOURS???

ON M1?????

what the hell

what fork r u using

oh, so original rvc?

damn

😭

damn

honestly i suggest using replay in this case, dunno much about it but people seem to like it, and it's the best one for the Apple Silicon Macs: https://www.tryreplay.io/

Might be RVC issue. M1 can easily run W-Okada so idk

[hazy thunder]

should i train in v2 or v1

[proper shale]

realtime conversion with rvc

[proper shale]

v2

[hazy thunder]

is v2 always better

[proper shale]

yeah

[hoary hare]

What program should I use to play RVC voices?

[hazy thunder]

what batch size should i use (using an rtx 4060 8gb vram on a 21 minute dataset)

[proud elbow]

8

[hazy thunder]

damn i did 7 is that okay

can i just cancel the training once it reaches a checkpoint then change it?

is 500 epochs on a 21 minute dataset overtraining

i think i mightve messed up

[hoary zephyr]

I tried with my voice and it was minimal between 250 and 500 id do 250.

[bronze maple]

yo bandit

[hazy thunder]

im so bad at detecting overtraining

[hoary zephyr]

Mhm

What up

[gusty kestrel]

Translated to English

tu a réussi a creer ton model ?

Translated message

Did you manage to create your model?

Requested by totoz8970 · Apr 3, 2024, 12:57 AM

[thin stump]

Think of it as a strategy game. I don't think you even have to look at mel that much #1213509354343637065 message

only g/total, d/total, kl

[hazy thunder]

im taking a look at my previous models when i didnt know overtraining existed and im just really confused

[thin stump]

the thread inside explains it all. I don't see how it can be confusing since you're looking for graphs that are going up indefinitely

[hoary zephyr]

No 500 was too much and I don't know how to use trained models with colab im not sure how to upload it because it requires a link to huggingface plus it's my own voice which I don't want used

[hazy thunder]

idk how to use d total to detect ot

[bronze maple]

it's not possible I made models with 5 minutes of dataset at 600 epoche and it was good

[wispy burrow]

The shorter the Dataset the more Epochs.
500 is usually too much.

[native scaffold]

can anyone help me switch it so it uses my gpu insted of cpu it just says gpu dml

[rare goblet]

Ayo? @native scaffold level 1 !!!

[bronze maple]

How many minutes of dataset do you think you need for 500 epochs?

what is complicated is that 500 epoche does not mean a quality model...

[hazy thunder]

yo this good as of rn ?

damn bruh the d loss is dipping DOWN

[rare goblet]

Ayo? @hazy thunder level 6 !!!

[hazy thunder]

i dont think thats good

im still at the 95th epoch

still we got a long way to go

[bronze maple]

I'm like you, I can't understand when to stop training so as not to be overtrained.

[odd tide]

anyone can tell me what pitch to use for voice.ai app

[uneven siren]

is it the higher the epochs the better?

[hazy thunder]

no

the shorter ur dataset the more epochs u use

dont be an idiot and make it really high

[uneven siren]

oh

[hazy thunder]

cause im an idiot and i made it really high and now my shit my overtrain

[uneven siren]

ok good to know lol, so lower is better cause more dataset

[hazy thunder]

if u have like a 20 minute dataset

u might go for 300 epoch or 250

[uneven siren]

what about an hour dataset

[hazy thunder]

500 epochs

[uneven siren]

ok

[hazy thunder]

u could do 600 but that might overtrain

[hazy thunder]

is this an OT??

[thin stump]

yeah the step value should tell you like 32.5 for example

[hazy thunder]

if it goes down while its going up

then is it overtrained

like goes down by a bit

[thin stump]

Yes past that red circle is overtraining

[hazy thunder]

it just dipped down again

[thin stump]

can I see?

[hazy thunder]

just by a bit

[thin stump]

no, it's done

[hazy thunder]

alright

how do i turn the checkpoint file into an actual model though

cause ive set it at 500 epochs and now it overtrained at 350

[thin stump]

Look at your step count before you stop the training

[hazy thunder]

yea then?

[thin stump]

then when you stop it, it gets saved in your weights folder under Mangio RVC/fork

[hazy thunder]

i dont even know how stop it lol

do i just close the terminal

[thin stump]

oh you are using local? Idk. I only used rvc disconnected colab so I can't help you there

[hazy thunder]

yea i am

which stop training button

in the rvc web gui?

how do i know

it just says rvc webui

think its the mainline one

alright

am i gonna have to redo the training from the closest checkpoint?

increase or decrease

yea

ok i understand thanks

[fresh mantle]

Help

I try to make ai cover but it just shows ERROR: Could not open requirements file: [Errno 2] No such file or directory: 'requirements.txt'

And I don't understand code

Android

Motorola

😦

Ohh Google

I also tried using my ipad

[fresh mantle]

I do

[strange herald]

Hi, Idk if this has a quick solution, but here is the thing: everything works well, but then, my mic from the headset is picking the desktop sound (so if I play a video while having it activated, the video count as a voice, so I do hear the conversion too besides my voice)

[fresh mantle]

?

[rare goblet]

Ayo? @fresh mantle level 2 !!!

[fresh mantle]

Let me see

Webui

[austere mortar]

where do i use a .ckpt?

[finite galleon]

Bro

[fresh mantle]

WhAt

[finite galleon]

Tell us the name of the google colab

[fresh mantle]

I'm not that experienced with this stuff so give me time

AICoverGen_colab.ipynb

Idk

[finite galleon]

Which link is it

Share

You maybe using an old version

-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

🤗 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

[fresh mantle]

https://colab.research.google.com/github/SociallyIneptWeeb/AICoverGen/blob/main/AICoverGen_colab.ipynb#scrollTo=YYVAKuNBc-X4&uniqifier=1

[finite galleon]

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

Use that one

[fresh mantle]

Ok

[spare gull]

what this mean guys

[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

[spare gull]

The percentage become lower and lower

Is that any problem??

Or it is normal

Pls someone help me

[flint geyser]

Hello, did someone ping me?

[hybrid rivet]

@flint geyser How can I use voice-lines for voice.ai

also hello

[flint geyser]

hi :)

[hybrid rivet]

:)

[hybrid rivet]

How can I use voice-lines for voice.ai

[flint geyser]

Since voice.ai isn't partnered with AI HUB, i do not know, refer to their support instead:
https://support.voice.ai/hc/en-us

or their discord :
https://voice.ai/discord-invite-tb

[south flume]

can somebody help me on w-okada, I only hear a robotic voice
I'm currently using

Ryzen 2600x
Radeon 6600

[hybrid rivet]

https://www.youtube.com/shorts/3jGJYVT6KOE

uhm

this video shows

you can use ai hub x voice.ai

[flint geyser]

oh, you want a custom model

[hybrid rivet]

yh

[rare goblet]

Ayo? @hybrid rivet level 1 !!!

[hybrid rivet]

ig so

Idek anything bout ai rvc etc.

😭

[flint geyser]

#1175430844685484042 or #🔍│find-models

[hybrid rivet]

ik but

which file

.pth

or

other things

cuz none of them work

[flint geyser]

upload the zip 😭

[hybrid rivet]

:O..

mb

ty 😭

[flint geyser]

np

[hybrid rivet]

is there any better free* alternetive for voice.ai

[flint geyser]

w-okada

[hybrid rivet]

how do i download it ;)..

[flint geyser]

https://rentry.co/VoiceChangerGuide

[tawdry ruin]

blud pinged 5 people for that

[flint geyser]

damn sorry

my bad

just copied from the guide 😭

[serene hound]

i just downloaded applio and i have on the drive standalone versions of rvc and rvc for training models, so i saw applio have these functions, these work like them? better or worse?

is the sense of keeping them all?

[sharp osprey]

help me pls

[rare goblet]

Ayo? @sharp osprey level 1 !!!

[sharp osprey]

TypeError Traceback (most recent call last)
<ipython-input-2-2952be55a264> in <cell line: 56>()
56 if param_link == "":
57 paramset = requests.get("https://pastebin.com/raw/SAKwUCt1").text
---> 58 exec(paramset)
59
60 clear_output()

<string> in <module>

TypeError: VoiceChangerParams.init() missing 1 required positional argument: 'whisper_tiny'

[lapis fox]

Can anyone assist me with the VCC? I'm having insane delays and am curious on how fix or speed it up

[rare goblet]

Ayo? @lapis fox level 1 !!!

[proper shale]

yeah

[lament furnace]

Deecho-dereverb is crazy icl

[proud elbow]

what audio length is that?

[fathom reef]

Well, this is... Quite the predicament.

Traceback (most recent call last):
  File "threading.py", line 980, in _bootstrap_inner
  File "threading.py", line 917, in run
  File "C:\Users\[CENSORED]\Downloads\RVC1006AMD_Intel1\gui_v1.py", line 733, in soundinput
    with sd.Stream(
  File "C:\Users\[REDACTED]\Downloads\RVC1006AMD_Intel1\runtime\lib\site-packages\sounddevice.py", line 1800, in __init__
    _StreamBase.__init__(self, kind='duplex', wrap_callback='array',
  File "C:\Users\[UNKNOWN]\Downloads\RVC1006AMD_Intel1\runtime\lib\site-packages\sounddevice.py", line 898, in __init__
    _check(_lib.Pa_OpenStream(self._ptr, iparameters, oparameters,
  File "C:\Users\[DONTHACKMEPLSTHANKYOU]\Downloads\RVC1006AMD_Intel1\runtime\lib\site-packages\sounddevice.py", line 2747, in _check
    raise PortAudioError(errormsg, err)
sounddevice.PortAudioError: Error opening Stream: Illegal combination of I/O devices [PaErrorCode -9993]

It also stopped responding altogether.

Basically crashed.

[sand iris]

Classic RVC realtime moment

[fathom reef]

I was going for a more optimized performance setting because of my graphics card, so that's why the sample length is at minimum.

[sand iris]

So that's the opposite; you want it to the max

[fathom reef]

I was doing it mainly as a test.

Again, I have a terrible GPU, so I set it to minimum as a test.

[sand iris]

Yeah, that has the inverse effect though

[fathom reef]

Uh...

what

But what about the error log?

[sand iris]

Error log, I wouldn't know

The realtime module has never behaved properly on my end, the device management stuff has always been iffy

[fathom reef]

Then again, the log has information that senior software engineers have instead of me.

I will never understand this stuff, so I need someone to explain it to me.

[proper shale]

thats for the index

[fathom reef]

That did absolutely nothing.

[sand iris]

Yeah, error lies elsewhere

It can't access audio peripherals for some random reason

[fathom reef]

Full error (screw censoring my PC name)

C:\Users\Gamer\Downloads\RVC1006AMD_Intel1\runtime\lib\site-packages\torch\nn\utils\weight_norm.py:25: UserWarning: The operator 'aten::_weight_norm_interface' is not currently supported on the DML backend and will fall back to run on the CPU. This may have performance implications. (Triggered internally at D:\a\_work\1\s\pytorch-directml-plugin\torch_directml\csrc\dml\dml_cpu_fallback.cpp:17.)
  return _weight_norm(v, g, self.dim)
Exception in thread Thread-1:
Traceback (most recent call last):
  File "threading.py", line 980, in _bootstrap_inner
  File "threading.py", line 917, in run
  File "C:\Users\Gamer\Downloads\RVC1006AMD_Intel1\gui_v1.py", line 733, in soundinput
    with sd.Stream(
  File "C:\Users\Gamer\Downloads\RVC1006AMD_Intel1\runtime\lib\site-packages\sounddevice.py", line 1800, in __init__
    _StreamBase.__init__(self, kind='duplex', wrap_callback='array',
  File "C:\Users\Gamer\Downloads\RVC1006AMD_Intel1\runtime\lib\site-packages\sounddevice.py", line 898, in __init__
    _check(_lib.Pa_OpenStream(self._ptr, iparameters, oparameters,
  File "C:\Users\Gamer\Downloads\RVC1006AMD_Intel1\runtime\lib\site-packages\sounddevice.py", line 2747, in _check
    raise PortAudioError(errormsg, err)
sounddevice.PortAudioError: Error opening Stream: Illegal combination of I/O devices [PaErrorCode -9993]

[sand iris]

try switching to the other one

might help

[fathom reef]

I can't read Japanese.

[sand iris]

I can't either

I looked at the code, and it seems to be related to how it decides to pick the settings for the audio devices

Perhaps it'll allow it to not screw itself up

[fathom reef]

Nothing changed.

[sand iris]

¯_(ツ)_/¯

[pastel oak]

select MME type on output

has to be matching

[fathom reef]

Oh yeah...

[pastel oak]

start sample length on 0.50 or tbh higher and work your way down after it works aswell

[fathom reef]

Oh, it's perfect!

For a GPU that's 12 years old...

It works!

I am so grateful for your help.

[pastel oak]

Nice

[fathom reef]

So... How would I be able to hear what I say?

Probably can't.

If that's the case, then whatever.

I think I can handle it.

[pastel oak]

you need virtual cable, go to the help wokada channel, pinned message open the rentry guide, go to step 6 download VAC

[fathom reef]

Already have it.

I'm talking about in--

Oh wait.

I think I see.

[pastel oak]

then select Output: Line 1 MME

if you have it

[fathom reef]

I have an AMD card.

[pastel oak]

That has nothing to do with virtual cable

[fathom reef]

I also don't have any lines.

[pastel oak]

Line 1 is the default name of the VAC virtual cable

[pastel oak]

if you dont have it do what i said

[fathom reef]

I do have it.

It's just that it says "CABLE Input" instead.

[pastel oak]

Ah you have the other one

[fathom reef]

Wait...

[pastel oak]

then use that

[fathom reef]

There's two of them?

[pastel oak]

yeah

[fathom reef]

Huh.

[pastel oak]

the one you have causes issues sometimes

if you dont then you can use it

[fathom reef]

Installing it now. I'll have to restart after this, so I'll be back.

Wait, which one do I use?

[pastel oak]

vac doesnt require restart pc but you can

[pastel oak]

64 without the a

[fathom reef]

Alright.

[fathom reef]

I forgot to mention: I meant to say, "Hear myself whilst using it in-game."

[pastel oak]

Ah this rvc doesnt have that option. youd need something like voicemeeter banana

[fathom reef]

Oh, I already have that.

Hang on.