✨│ai-help

[vapid gust]

damn

ill just stop it then

[graceful obsidian]

ex. 1 ckpt is 20 steps 2nd is 40, 3rd is 60

etc

but tensor logs every idk, 200 steps ( spikes / points on graphs )

and then, how you gonna know where is where and which one is which

• you don't see the real value per ckpt

so really, you gotta sync the log interval, lusbert might help you

iirc I explained him how it's done

[vapid gust]

@eager pelican

[eager pelican]

hm?

uhh

i never knew how to even sync interval logs
since it was not even useful since the new forks remove it

atleast i was told so

@vapid gust just tell me whats your problem
i aint reading all those messages and figure it out

[potent root]

sorry for being late but you're not supposed to play those cells again after you've already ran them

[vapid gust]

just need to identify when it’s going to overtrain

it’s like a 3 minute studio stem dataset

it went 500 epochs without any rise (overtraining)

[eager pelican]

does the model sound good or what

test it

[vapid gust]

it’s super hq

it’s studio stems bro

straight recordings

[eager pelican]

cause in "SOME RARE" cases its actually better to overtrain

im not saying this the same for you but for me its like this for good datasets

[vapid gust]

i let it train to 300 epochs

im gonna test it just now

[eager pelican]

imo train it to 1000 if you want
cause from my testing that gets the hisses and engine noises at generally 500 epochs "removed" so less frequent
but also can turn out bad

so yea i just train to 1000 epoch and pick the best one based on sounds

[vapid gust]

that’s gonna take so long bro

[eager pelican]

well thats why i said "IF YOU WANT"

for me its fine so i do it

[vapid gust]

ah

also

I can train like a 40minute studio recording dataset

what u think the epoch limit should be

[eager pelican]

40 minute is too long imo

at max i'd do 20 minute

[vapid gust]

ah

alr

[eager pelican]

1 epoch is a whole iteration of the dataset
so it doesnt really say:

for this dataset size use this epoch
since it depends for each time you train and for each dataset

since it has a seed and basically each model is random

some train like 4 models of the same dataset

[vapid gust]

ive trained a model

of like 2 minutes

to like 200 epochs

and it came out like

really accurate

[eager pelican]

well maybe use that then

if its already good dont train longer

[vapid gust]

theyre two different people

just an example

[eager pelican]

it would either ruin it or overkill the quality which isnt worth the power bills

[vapid gust]

ah

[eager pelican]

RVC works on pitch so doesnt matter who it is
its just data to the engine

[vapid gust]

ohh

alr

im about to test the 300 epoch model

[eager pelican]

well good luck

-overtrain if you want read these as well

[azure marsh]

How do I know if my model is overtaining?

All-In-One Guide on how to make a good model

This guide explains how the D and G files works and much more: https://rentry.org/RVC_making-models

Credits: LUSBERT

Automated Overtraining Detection (AOD)

Will be available soon in #1159513888199540817

Credits: grvyscale

[vapid gust]

@eager pelican ITS ACCURATE

AF

[rare goblet]

Ayo? @vapid gust level 10 !!!

[weak stratus]

tried looking up help for this but am still really lost... any help is really appreciated! keep trying to test out on a fresh install and my web-ui keeps crashing when i try to run it. gives me the following error: C:\Users\brade\OneDrive\Desktop\code2Win\AI\RVC1006AMD_Intel>runtime\python.exe infer-web.py --pycmd runtime\python.exe --port 7897 --dml
2023-10-29 08:01:54 | INFO | configs.config | No supported Nvidia GPU found
overwrite preprocess and configs.json
2023-10-29 08:01:54 | INFO | configs.config | Use DirectML instead
is_half:False, device:privateuseone:0
2023-10-29 08:01:55 | INFO | main | Use Language: en_US
Running on local URL: http://0.0.0.0:7897
2023-10-29 08:02:04 | INFO | infer.modules.vc.modules | Get sid: guanguanV1.pth
2023-10-29 08:02:04 | INFO | infer.modules.vc.modules | Loading: assets/weights/guanguanV1.pth
2023-10-29 08:02:04 | INFO | infer.modules.vc.modules | Select index: logs\guanguanV1.index
2023-10-29 08:02:41 | INFO | infer.modules.vc.pipeline | Loading rmvpe model,assets/rmvpe/rmvpe.pt
2023-10-29 08:02:41.7569477 [W:onnxruntime:, session_state.cc:1169 onnxruntime::VerifyEachNodeIsAssignedToAnEp] Some nodes were not assigned to the preferred execution providers which may or may not have an negative impact on performance. e.g. ORT explicitly assigns shape related ops to CPU to improve perf.
2023-10-29 08:02:41.7605571 [W:onnxruntime:, session_state.cc:1171 onnxruntime::VerifyEachNodeIsAssignedToAnEp] Rerunning with verbose output on a non-minimal build will show node assignments.

C:\Users\brade\OneDrive\Desktop\code2Win\AI\RVC1006AMD_Intel>pause

i can't seem to figure out the issue it doesn't seem to really be giving me a proper debug error... the connection just times out any time i try to run a conversion..

[rich yoke]

i just realized that I have been using the official version of rvc, do the forks really offer anything to warrant using them instead?

[rare goblet]

Ayo? @rich yoke level 1 !!!

[rich yoke]

the only thing that really bothers me about it is that the window for it isn't able to be resized bigger

[weak stratus]

I think maybe it is trying to use my dedicated gpu instead? how can i configure it to use the correct device/ driver? if anybody can help i will love you long time ❤️

are there any other common issues that cause a connetion time out that i can try? i can't figure out what is going on here... its not giving me an error!

[weak stratus]

is there a bug with trying to run locally right now??? I AM SO CONFUSED. just installed on ubuntu and am getting the same exact issue wtf... should i just run in colab? i would prefer local...

[sand iris]

DirectML on ubuntu ?

Are you on WSL ?

[astral jungle]

Yoo

[weak stratus]

WSL sorry.

[sand iris]

I'm really not sure what could be happening

I see you are under the "Onedrive" directory though. Always a good idea to move away from that

[astral jungle]

EASY has been fixed

[late prairie]

hey, im using an AMD GPU,
which one is the right version for me?

[proper shale]

directml cuda

last directml cuda

[sand iris]

onnxdirectml

[late prairie]

Thanks!

[brittle wing]

I'm using the Mangio RVC GUI locally to try train a model. Everything goes well until the "Train model" part. I don't get my pth and index files and the error specifies that the 'pretrained/f0G40k.pth' file is missing. (That's what it automatically puts in as "Load pre-trained base model G/D path:" in the GUI). Here's the output:

Traceback (most recent call last):
  File "C:\Users\Mihael\Desktop\rvc\Mangio-RVC-v23.7.0\train_nsf_sim_cache_sid_load_pretrain.py", line 181, in run
    utils.latest_checkpoint_path(hps.model_dir, "D_*.pth"), net_d, optim_d
  File "C:\Users\Mihael\Desktop\rvc\Mangio-RVC-v23.7.0\train\utils.py", line 206, in latest_checkpoint_path
    x = f_list[-1]
IndexError: list index out of range

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "multiprocessing\process.py", line 315, in _bootstrap
  File "multiprocessing\process.py", line 108, in run
  File "C:\Users\Mihael\Desktop\rvc\Mangio-RVC-v23.7.0\train_nsf_sim_cache_sid_load_pretrain.py", line 201, in run
    torch.load(hps.pretrainG, map_location="cpu")["model"]
  File "C:\Users\Mihael\Desktop\rvc\Mangio-RVC-v23.7.0\runtime\lib\site-packages\torch\serialization.py", line 791, in load
    with _open_file_like(f, 'rb') as opened_file:
  File "C:\Users\Mihael\Desktop\rvc\Mangio-RVC-v23.7.0\runtime\lib\site-packages\torch\serialization.py", line 271, in _open_file_like
    return _open_file(name_or_buffer, mode)
  File "C:\Users\Mihael\Desktop\rvc\Mangio-RVC-v23.7.0\runtime\lib\site-packages\torch\serialization.py", line 252, in __init__
    super().__init__(open(name, mode))
FileNotFoundError: [Errno 2] No such file or directory: 'pretrained/f0G40k.pth'

[low shard]

wtf is that colab

[dense dirge]

hi

im having an issue in applio

attributeerror: 'nonetype' object has no attribute 'tobytes'

[rare goblet]

Ayo? @dense dirge level 1 !!!

[dense dirge]

this appears

[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

🇫🇷 France

https://docs.aihub.wtf/v/france

🇩🇪 Germany

https://docs.aihub.wtf/v/germany

🇮🇹 Italy

https://docs.aihub.wtf/v/italy

🇯🇵 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

🎃 Spooky contest is active till 31st, check the announcement channel for further details.

[olive hill]

ANY ALTERNATIVE PLATFORM TO TRAIN MODELS???

[spring turtle]

applio tensorboard wont open, know how to fix anyone?

[late prairie]

hey, i converted an .pth rvc model to onnx. what do i need to do to use it in the Voice-Changer ?

[rare goblet]

Ayo? @late prairie level 1 !!!

[late prairie]

whats the right version for onnx?

[sand iris]

rvc

just upload the .onnx instead of the .pth

[late prairie]

ahh thanks again!

[uneven holly]

uh how do i make my own ai voice?

[thin ravine]

rvc gui causes blue screen

What the FUCK !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

[weak stratus]

dude i can't get this working for the life of me. trying on just a normal fresh windows install and still nopeeee. just crashes with no error message. 7900xtx and a 7950x3d

[uneven holly]

how do i train a voice

[hoary zephyr]

hey does the google colab still work because i cant see me paying for one of those subscriptions

just for running the voices no training

[weak stratus]

yo is ther a known bug right now for windows or amd cards? seriously gonna lose my mind man 😦

[rare goblet]

Ayo? @weak stratus level 3 !!!

[weak stratus]

im not trying to train

right now at least. i would use linux lol would be a lot easier for me tbh cant get it working on wsl tho. shit is so confusing to me it wont even give me an error when i try and run a conversion... just connection timed out

literally installed a fresh windows on a new partition to check if im crazy

why i always seem to try shit when its broken just never ceases to amaze me man

does literally everybody here just run this on colab so thats why yall don't know what im talking about? guess i could use local runtime via colab but thats such a pain in the ass with my google drive allready being full of other stuff atm

[sand iris]

Nah, a lot run locally

But few run on amd/intel, and there's also more forks/ways to install on top fo that...

[weak stratus]

do you by any chance know a better fork for amd? dissapointed man i got a 7900xtx...

[sand iris]

I really don't know, directml is wack

The "error" you are seeing isn't even an error as far as I'm aware

[weak stratus]

i mean i dont care if the performance isnt as good maybe but it should WORK still

[sand iris]

Shit's unstable

[weak stratus]

yeah I know, it isn't giving me one at all which is why im getting so frustrated tbh. it just doesn't work ;/ should i try mangio ?

[sand iris]

mangio has no directml support

[weak stratus]

https://tenor.com/view/funni-meme-gif-22841383

[sand iris]

lmao

[weak stratus]

well let me ask you this, have you heard of ANYBODY running on a 7900xtx?

cuz looks like the discord server is newer from an other one that got banned or something it sounds like? so maybe theres more history that i obciously can't search through unfortunately

[mossy ridge]

can someone help mangio rvc?

[brittle wing]

why does the ilaria rvc huggingface interface keep saying "connection errored out"

does it even work?

[weak stratus]

you got an amd card?

[rare goblet]

Ayo? @weak stratus level 4 !!!

[brittle wing]

what does that have anything to do with it?

but no, i don't

[sand iris]

hugging face isn't local either 😅

[weak stratus]

its like google colab no? you can use local run time i thought

man idk, im just trying to get this working on my rig so im reaching for everything i can lol

[brittle wing]

since when can you do local runtime on colab, isn't it all on google's gpu's?

[rare goblet]

Ayo? @brittle wing level 1 !!!

[weak stratus]

its been a while, but i used to do it on some old projects you definitely can yes.

[brittle wing]

i've just been stuck not being able to use anything besides those monthly 15 minutes kits gives just because i don't have a good enough gpu and i'm broke af

[weak stratus]

lmao well dont sweat it, my $1000 is more worthless than yours rn

[brittle wing]

can't fuckin do any covers anymore ever since google decided to be a bitch and ban the rvc colab

[weak stratus]

ill run 5 covers for you if you find somebody that got this running on a 7900xtx 🤣

[brittle wing]

damn unfortunately i'm not knowledged in hardware related stuff sorry

[sand iris]

I recall seeing that you can self host colab runtimes

Or it was named weirdly and made it sound like you could

[weak stratus]

does anybody know if there is anything else you need to to do fully uninstall rvc v2 before trying to reinstall? thought i can just delete the folder... just want to make sure.

[sand iris]

Should be the folder only

[weak stratus]

appreciate u man ty

at this point i am a little bogged down but i just cant let it go, i also have ubuntu i know this is gonna sound dumb but is it supported on there? only see windows and macOS insturctions , id assume something would run better on linux anywho correct? or just not supported at this time?

[vital sun]

how can i find model link? idont want download i need just link

[worn echo]

Does anyone know how to make laughing not sound distorted?

[sand iris]

Not sure on WSL support but bare metal linux works great on AMD

[weak stratus]

for rvc2?

[sand iris]

Yeah

setup is just a bit manual

[weak stratus]

are there any instructions to follow on how to do it somewhere? i can't seem to find anything for somer reason i suspect its right in front of me.

im good at following instructions and googling generally dont have this much trouble xd

[sand iris]

There aren't really any docs out there because they expect you to know stuff already... But I do have the steps on hand

For Ubuntu at least

Just gotta gather them and write them down

[weak stratus]

well i feel like im really tired and am making myself think i dont know how but i do haha

i mean i have been using linux for a few years, no expert tho

[sand iris]

Oh shouldn't be an issue then

[weak stratus]

i got stable diffusion up and running

[sand iris]

Oh that should speeds things up tremendously

With ROCm?

[weak stratus]

the rocm was "working" but very slow, seemed to be a bug with shark which is what i was trying. vulkan was giving me like 20 it/s but yeah i mean some of the stuff is still over my head the rocm drivers seemed to be set up right it was just weird i haven't played with it enough

i built a much nicer pc recently and have really been trying to educate myself more its just hard when you get hung up on something specific, usually goes back to fundamentals i guess

[sand iris]

Mind running /opt/rocm/bin/rocm-smi just to attest it's working?

Also which distribution is this on ? Steps might differ slightly

[weak stratus]

22.04

Sorry just booting up rn

[sand iris]

Great, should be painless then

[weak stratus]

Downloading discord on my linux rn but hmm…. Actually this was the output

[sand iris]

Great it's detected

But there's two GPUs which it might have trouble with

seems like the 7900xtx is the first one though so that could help

Setup goes like this :

• get the python-venv ubuntu package if it isn't installed already
• python -m venv /an/empty/folder
• source /VENV_FOLDER/bin/activate
• clone the RVC repository at https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI
• Rename the folder to something shorter like RVC-WebUI
• Go into the repository's folder
• pip install -r requirements-amd.txt
• Install pytorch for rocm as described at https://pytorch.org/
• python infer-web.py

[weak stratus]

Bro… thank you so much. I think I actually would have been able to get all the way to the requirements part and would have got hung up. Will follow up, appreciate you!!!

[brittle wing]

Is their a way to do, Ai Vocals Live? Like converting my mic into my ingame lobbies?

[weak stratus]

route it through vb audio cable

then use the vb audio cable for your in game input

[brittle wing]

Ok, then how do I not keep on pressing Convert, I'm trying to do Ai Live

if thats possible?

[weak stratus]

sorry i dont know a ton either tbhj just trying to help whil i install my module myself 🙂

theres a different program for doing the live conversion

[gentle widget]

I keep getting this error AttributeError: 'NoneType' object has no attribute 'tobytes'

[weak stratus]

@sand iris i have ALMOST got it up and running thanks to all your help 🙂 realized that the documents just needed to be translated cuz it defaulted to chinese or whatever.... lol. but i am now getting an error saying that the huber_base.pt model isnt being found. trying to troubleshoot, figured it was worth a shot

[sand iris]

Oh yeah my bad I forgot about that last step

get the aria2 ubuntu package and run the tools/dlmodel.sh script in the repo (run it while being at the root of the repo)

[weak stratus]

on it, thank you !

[sand iris]

corrected the script path

[weak stratus]

i understand like everything i did step by step there except that part, do you mind me picking your brain on that? i downloaded a stupid model to test, its just cartmen from southpark and i figured out how to import it and all that, but what is this dl model? idk if its too technical haha

[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

🇫🇷 France

https://docs.aihub.wtf/v/france

🇩🇪 Germany

https://docs.aihub.wtf/v/germany

🇮🇹 Italy

https://docs.aihub.wtf/v/italy

🇯🇵 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

🎃 Spooky contest is active till 31st, check the announcement channel for further details.

[sand iris]

there's other stuff required

The model you import isn't the one and only part of it

[weak stratus]

i was gonna say... still missing that hubert base.pt

[sand iris]

from where did you execute the script ?

[weak stratus]

root of the rvc dir like you said

[sand iris]

Yeah alright, it should've worked then 🤔

[weak stratus]

ah shit

[sand iris]

also did you restart RVC in between

[weak stratus]

i fucking missed that part i guesss... ran it from tools lol

[sand iris]

Oh lol

[weak stratus]

lmao so so dumb and tired i know im close now though because of you

[rare goblet]

Ayo? @weak stratus level 5 !!!

[weak stratus]

literally installed ALL them assets into tools fml haha, fair enough

[dense dirge]

hi

i trained a model

and now the index doesnt appear

where is it?

[viscid moss]

check logs/"name of the model"/addedxxxxxxx.pth

[dense dirge]

it doesnt appear

[weak stratus]

@sand iris holy crap not only worked, absolutely BLAZED. seriously can't thank you enough man, now just super excited to start playing with tweaks and really getting into it!

[viscid moss]

train only the index

[sand iris]

Have fun !

[dense dirge]

also this button where is it

[sand iris]

Heck you can even train models now

[dense dirge]

how

[sand iris]

Don't mind it telling you there are no GPUs supported, just feed it index 0 and call it a day

[viscid moss]

in the "Entrenar" page

set the same name

[dense dirge]

ok

[rare goblet]

Ayo? @dense dirge level 2 !!!

[dense dirge]

name set

[viscid moss]

ok, only set that and train index

[dense dirge]

do u know spanish?

i prefer talkin in it

like this?

the name

and where i click?

im new sorry

[viscid moss]

native lol

[dense dirge]

vale perfecto

en español mejor

pongo el nombre

ya esta puesto le puse el mismo

y ahora donde voy

[viscid moss]

solo eso debes marcar y ve al final

ya no hagas los otros 2 pasos

[dense dirge]

pero todos los ajustes que ya estaban puestos

[viscid moss]

porque ya los hiciste

[dense dirge]

los borro

es que ya los tenia puestos del modelo

[viscid moss]

ah, mejor

solo ve al final

[dense dirge]

voy

donde le doy

[viscid moss]

Extraccion de caracteristicas se llama

en español

[dense dirge]

aqui no sale

es ese?

[viscid moss]

indice de caracteristicas perdon

es que tengo en inglés yo eso

[dense dirge]

le doy y no pasa nada

[viscid moss]

entrenaste con esa PC?

[dense dirge]

si

[viscid moss]

Por que sale que no tienes GPU compatible

[dense dirge]

se queda asi

[viscid moss]

y en el CMD? que sale

[dense dirge]

mira

[viscid moss]

a ver

[viscid moss]

mmmmm pero eso es lo del anterior, osea de la extraccion de caracteristicas

[dense dirge]

es q no me sale nada nuevo

aunque le pulse

a indice de caracteristicas

[viscid moss]

mmmmmm, guarda todos los archivos de ese modelo en otro lugar y has los 2 pasos otra vez

Pero sin entrenar,solo entrenas el index

[dense dirge]

vale

[viscid moss]

ya lo otro si lo tienes

el pth no?

[dense dirge]

la carpeta la guardo en el escritorio x ejemplo

[viscid moss]

en weights?

[dense dirge]

si

tengo el pth

[viscid moss]

has un zip de todo y muevelo al escritorio

[viscid moss]

litno está

💀

[dense dirge]

que no esta

el index

[viscid moss]

el index

[dense dirge]

ya

[viscid moss]

👀

[dense dirge]

la carpeta se esat comprimien

ya esta comprimida

[viscid moss]

osea comprime desde la carpeta de afuera

[dense dirge]

si

[viscid moss]

nice

ahora puedes borrar los otros archivos

[dense dirge]

ahora borro la que hay ahi?

[viscid moss]

si

borra toda la carpeta

y empieza again

[dense dirge]

ok

[viscid moss]

pon un nombre que puede ser el mismo, y los mismos datos

[dense dirge]

es a lo que voy

[viscid moss]

y en vez de entrenar modelos pones entrenar indice

[dense dirge]

vale

un momento que lo hago

ya esta

le doy a lo del index y nada

como que empieza a hacer algo y se corta

se pone asi y se corta

y vuelve a ponerse asi

[viscid moss]

y en cmd?

[dense dirge]

esto es solo si uso el rvmpe gpu no?

[viscid moss]

dice?

[viscid moss]

mmmmm

[viscid moss]

de hecho por eso preguntaba

que GPU tienes?

Porque alli dice que no te detecta una

[dense dirge]

intel core i5

[viscid moss]

eso es CPU

👀

[dense dirge]

xd

gpu cual es

jajaja lo siento

[viscid moss]

https://tenor.com/view/rtx3090-rtx3090ti-wow-wow-meme-wow-gif-gif-24535283

De esas

[dense dirge]

a

intel

iris

xe

[viscid moss]

jesus

mmmmmm pues no sé como entrenaste xD

en teoria no es compatible

pero igual algo debio entrenar si tiuenes un pth

y tengo curiosidad

dejame pensar que hacer

[dense dirge]

las epochs me las hacia

lo he ido viendo

me hizo las 400

20 segundos cada una

[viscid moss]

dataset de cuanto?

[dense dirge]

5 mins

mas o menos

[viscid moss]

mmmmm

y si....

haces el index con un colab?

es lo que se me ocurre

[dense dirge]

como se hace eso?

[viscid moss]

asi podras probarlo

y ver que paso

porque en teoria no es posible entrenar

lo que me cuentas ha pasado 2 veces antes

[dense dirge]

y se resolvio

?

[viscid moss]

ven a #🌍│español

[dense dirge]

voy

[viscid moss]

no supe

[dense dirge]

a

[ember crow]

ah fuck

[astral jungle]

EASY

[astral perch]

Does screams and laughing in the dataset result into a worse sounding model

[vagrant minnow]

i have a probem with the client, when it oppens, just appear a white window, with File, display and window, what i can do?

[umbral grove]

screams and laughter in a dataset can result in a worse sounding model

[astral perch]

Thank you!

[dire vapor]

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

[rare goblet]

Ayo? @dire vapor level 1 !!!

[dire vapor]

help pls

[visual pumice]

always record datasets in a soundproofed room

trust me the quality difference is unmatched

[strong belfry]

Its not easy to get a soundproofed room

although a quiet place with some bedsheets hanged on the walls is doable

[violet heron]

Rather use UVR thank you

[visual pumice]

or that

[visual pumice]

closet, corner of your room, blanket around you and the microphone

anyways im still right

[brittle jay]

can you guys please quit it with the rude reactions? there's an L and 3 middle fingers on my model

[brittle jay]

well its bothering me, when i see those reactions, im being told "you suck. Your model is stupid. GAAYYYYY" thats what it tells me, it doesnt tell me, "I dont really like this one, havent seen the show"

like, edit the post and take them off?

[alpine bolt]

how to make ai voice models?

[brittle wing]

how to lower ping? my ping is 20-30k

[green birch]

https://cdn.discordapp.com/attachments/1159290161683767298/1168443512790974564/image.png?ex=6551c8d2&is=653f53d2&hm=bfff5bfb325395b6006faee08777f051f9439bff4c64f740fd2da6378680e67f& im trying to do a voice for real time voice
i cant pass 1000 epocs and i keep getting 2 types of pth what am i doing wrong ?

also there is NO index can i get some help pls reply or @ me im gong to bed for now i'll check tmw

[dense crane]

Ok holy shit 1 upgrade your RVC

2 don't use 1-Click

And 3 your model files show up in the "weights" folder

[green birch]

i keep asking questions i was told to use 1 click im using the offline version
so im not sure what else id love to get some info from you so i can just finish this up im passing out tho almost 1 am can u shoot me a pm how to or w/e clearly info i have is old id pref not to use google if i can as i think my 3090 would be faster but im good with w/e

[dense crane]

Okay so like

Don't use 1-Click

Whoever told you to use it is wrong

And second use local Mangio

[thin ravine]

is my cpu too shit of why does rvc gui bluescreen for me when i try to make cover, i have amd gpu so it cant use it

[dense crane]

Okay if it's bluescreening your PC that's probably a driver issue

Go update your driver's

And try reinstalling Mangio

[thin ravine]

cheers

[rare goblet]

Ayo? @thin ravine level 1 !!!

[pallid pecan]

Im using the real time RVC

What does the Index rate slider change?

[silk root]

i tried villager model and is not working

[rare goblet]

Ayo? @silk root level 1 !!!

[brittle wing]

i cant find my microphone on the output section when i try to use rvc on discord but i find it on the app itself is there a way to add or fix that ?

[ember crow]

this happens

[plush compass]

How long does it take to use all 100 compute units? Is Colab Pro worth using?

[frigid sapphire]

i would say get a $300 trial for Google Cloud and connect your colab to GCP.

ref blog: https://black.dubverse.ai/p/running-rvc-models-on-the-easy-gui

[plush compass]

Is there no need to buy Colab Pro?

[frigid sapphire]

no - you can use your credit card to get a free 300usd trial. Do make sure you keep the machines switched off when not in use. DM if you need help setting it up.

I am the author of the blog I shared (:

[plush compass]

I'm Korean so I'm not good at English... I'll try using the translator one by one and if I get stuck, I'll tag you. Thank you.

[rare goblet]

Ayo? @plush compass level 2 !!!

[storm prairie]

hello, does anyone know how to use the index files? I just got the local thingy

[frigid sapphire]

https://research.google.com/colaboratory/intl/ko/marketplace.html

this is Google's official guide to connect colab with GCP. They have a Korean translation readily available.

Cheers!

[plush compass]

Thank you. Have a good day

[dense crane]

Index is basically accent. The higher the index the more the accent is going to show through

[loud pelican]

¿Best page or program for downloading YouTube Audio?

[dense crane]

Program: YouTube-dl
Webpage: https://tuberipper.com/16/

Webpage is entirely by opinion

[earnest scarab]

does anyone have a gradio to train on?

[loud pelican]

👍

[rare goblet]

Ayo? @loud pelican level 1 !!!

[elder falcon]

anyone know what's going on here? trying to install Applio locally and the install keeps giving me this

[grim bay]

Does Epoch higher makes it better?

[copper pebble]

you mean high epochs

well it depends

heres a tab

dataset min to epochs

[violet heron]

Not always

An epoch is a training cycle, basically how much times it’s looking over the dataset

There is something called overtraining

[green birch]

i cant get it working

i always get this if i use mangio-rvc-form-main

[dense crane]

Install Python and add it to PATH

[green birch]

i got it working somehow i was getting a fhile with no phython i also have phython on my pc

so weird

so next question is it better to have 1 long audio or many small ones ?

[dense crane]

Depends

I'd say multiple smaller audios but one long audio works too

[green birch]

i have 1 long 16 min one but i can chop it

[proper shale]

one long audio is better

technically performs better :)

[winter seal]

is there any other ai cover colab that doesnt crash

[glad zealot]

the one on the pins on this channel

[green birch]

ya it opens but its back to errors xD

[rare goblet]

Ayo? @green birch level 5 !!!

[dense crane]

What GPU do you have

[green birch]

nvidia 3090

[wispy burrow]

Don't install EasyGUI for Mangio!
It breaks Mangio.

[green birch]

no s*RE#

[wispy burrow]

Get it from the Hugging Face.
https://huggingface.co/MangioRVC/Mangio-RVC-Huggingface/tree/main
Download, extract and run go-web.bat.

[green birch]

easy and in english is welcome im getting to many contradictory instructions.

so i got mangio from hugging face . local. (NOT for singing)

ok

[robust wind]

Where can i concert my .mp3 file with my own voice into a ai modeled voice file ? (i got the model i want to convert it to already)

[green birch]

that seems to fixed some of it

[wispy burrow]

Pretty much any RVC should do.
Except RVC Disconnected.

[wispy burrow]

But not everything?

[green birch]

do i do the model then the index?

[green birch]

lemme make sure its not idiot user error first xD

seems to be going

i'll do this then index i guess

[wispy burrow]

For training the order doesn't matter.
I train index first as that doesn't take long at all.

[dense crane]

Same

I train index first so I can listen to the model at random intervals during training

[green birch]

when i did index it didnt do it so i'll see what happens

[dense crane]

Did you preprocess and feature extract

[green birch]

its also still doing 2 paths

[dense crane]

Ok lemme see the whole screen

Okay could you crop it to center to I can read the text

[wispy burrow]

Why Harvest? RMVPE & Mangio-Crepe are better...

[dense crane]

I'm on mobile

[green birch]

u can zoom in ... ok

so what specificly do u want

[green birch]

it says righ on the thing harvest is better

[dense crane]

It's outdated descriptions

[wispy burrow]

That line comes from regular RVC that didn't have those options.

[dense crane]

I'm working on a fork with better updated descriptions

[green birch]

fair enough

[wispy burrow]

For Mangio?

[dense crane]

Yes

It also has an semi-automatic model packager

[wispy burrow]

so many 'No-f0-todo'.
That can't be good.

[green birch]

i dont even know what it means

[dense crane]

That means that they probably don't have a dataset audio in there

[wispy burrow]

Will it have the updated RMVPE model?

[dense crane]

Sadly not

Not sure how to add that

And also that was deleted alongside RVC2.5

[wispy burrow]

Did you remove all the silence?

[dense crane]

Please enable file extensions

🙏

[wispy burrow]

It's an mp3.

[dense crane]

[green birch]

i was told mp3 and long was ok is that wrong to ?

[dense crane]

Convert it to a WAV

And second

[proper shale]

eh mp3 can error

wouldnt recommend it

[green birch]

ok

its what my sound thing had but i can do wav gonna take a bit

[dense crane]

Make a new folder in "Datasets" and name it to the name of your model (no spaces) and then set that under that dataset path thing

[green birch]

ok

[wispy burrow]

That won't remove silence.
You need something like Audacity to remove silence.

[astral jungle]

Helo

[wispy burrow]

Only used it for removing silence.
Removing reverb and echoes I use UVR.

[green birch]

ok i'll use uvr suggestion on profile to use ? also gotta afk about 10 min

[wispy burrow]

I'd use VR-DeEcho-DeReverb.

The local version of UVR (5.6) has more Models avaiable.

[green birch]

i have uvr i'll use that

[wide parcel]

should i rely on this? or is it something i should avoid using

[wispy burrow]

I have never seen that option.
What are you using?
Seems to me like it's a setting to prevent overtraining...

[wide parcel]

applio

latest version

[wispy burrow]

Guess I'll need to update it again...

[proper shale]

dont think u need it

[green birch]

ok i used vr arch. uvr de echo normal
then used de echo / de reverb

so i think im set

its also a wav file

[wispy burrow]

You should be good then.

[green birch]

made a folder called bt
put file in sed folder bt
lets see if it works

[wispy burrow]

That looks good.
Now make sure it's finished before going to the next step.

[wide parcel]

am i doin something wrong 😭
whys my graph going backwards (ive never used tensorboard)

[green birch]

still wont make index ......

[rare goblet]

Ayo? @green birch level 6 !!!

[wispy burrow]

What do you mean with backwards?

[sonic tapir]

is the NoUI colab also disconnecting for anyone else?

[robust wind]

can you send me a website

[wispy burrow]

Index is made at step 3 (train feature index).
After 'Process Data' (step 2a) you do 'Feature Extraction' (step 2b)

[green birch]

if i take off pitch it does not even give me that option

[stiff bone]

Hello,

I'm new to this and i got this error while trying to swap voices.
If anyone know about this. (i use rvc webui)

Thank you

i cant send image lol

[green birch]

print screeen control v

[stiff bone]

i don't have rights to upload it here apparently

[wispy burrow]

You need level 1.

[stiff bone]

oh ok

https://cdn.discordapp.com/attachments/844975263192580147/1168607191914250461/image.png?ex=65526142&is=653fec42&hm=31ee7798966cf128bb7e88e701268651ee28bdf7901e5bb6576b4e59285e2b1c&

[robust wind]

can some1 send me a rvc website to convert my mp3 into ai

[rare goblet]

Ayo? @robust wind level 1 !!!

[green birch]

use adacity

[sonic tapir]

is there an alternative to Google Colab where i can use RVC? cause all of them disconnects now. even the no WebUI on

[wispy burrow]

Local or Hugging Face Space.

[violet heron]

-colab

[azure marsh]

Colab suggestions for @sonic tapir

RVC

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

🌐 Translated colabs

• PT-BR: -colabs_br

Audio Separation/Isolation

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

🎃 Spooky contest is active till 31st, check the announcement channel for further details.

[violet heron]

Are you using this one?

If your pc is strong enough you can always do local

[brittle wing]

Hey.. bit of a dumb question but I've never made an ai model before, and wanted to quickly ask.

What do epoch amounts mean exactly? Is it just like more = higher quality or?

[sonic tapir]

any tutorials? I've tried it on hf but don't know how to load a model in or upload a song onto it

[wispy burrow]

It's how many time it went through the Dataset during training.

[stiff bone]

https://cdn.discordapp.com/attachments/844975263192580147/1168607191914250461/image.png?ex=65526142&is=653fec42&hm=31ee7798966cf128bb7e88e701268651ee28bdf7901e5bb6576b4e59285e2b1c&

[wide parcel]

the blue line went backwards at ~150

[brittle wing]

Ah, so then I assume higher Epoch just means higher quality?

[wispy burrow]

Not necessarily. It can be overtrained, making the model worse.

[sonic tapir]

im on android actually... an i used the no webui one. but Google baned that one to

[wispy burrow]

Oh, I see it now.
Probably a glitch I guess.

[wispy burrow]

I've seen worse...

[sonic tapir]

if anyone has any alternatives please tell me. i miss making Ai covers 😭😭

[wispy burrow]

HuggingFace Spaces?

[floral shore]

anyone know how to open it

cause im dumb

[wispy burrow]

Open what?

[floral shore]

rvc

cause i cant find it in the folder

[wispy burrow]

Run go-web.bat.

[floral shore]

k

u helped me

thanks

[violet heron]

An epoch is a training cycle, basically how much the ai looks over the dataset

More does not mean better

I think a helper made a chart for dataset length : epochs

Here it is

[brittle wing]

Oh, that's pretty helpful

I'm training an AI voice off of DrSbaitso's voice (SCP-079), and the training data is about 5 minutes in length, tho I set it to 300 epochs

[green birch]

so it varys on pc video card and so forth

[brittle wing]

Ah, I'm training it using my laptop instead of my pc cause I'm at school right now.. and uh

[rare goblet]

Ayo? @brittle wing level 1 !!!

[brittle wing]

well my laptop's card is bascially non-existant lmao

[green birch]

so it could be days vs hours then

[brittle wing]

well it should be about an hour from what I can see

it's been going for 20 minutes and it's at Epoch 177

[green birch]

i guess it idd D / G this time i dont get what that is but w/e idc

no more errors (nothing was changed so dont ask )

[wispy burrow]

The Added_IVF*.index is your Index file.
D & G are used for training.

[green birch]

so i use ADDED when its time ok gotcha wellp imma check back in a few hours i guess

[wispy burrow]

That looks good.

[green birch]

check back in some hours

[green birch]

i just started andim already passing 200 so ya it differ massivly

[wispy burrow]

Renaming is fine.
Just avoid spaces or special characters.

[brittle wing]

My gpu on my laptop says "AMD Radeon(TM) Graphics" soooo :P

[green birch]

i just wanna make it 2 letters vs a paragraph XD

[spring turtle]

through powershell?

[sand iris]

You sure it's actually training?

[brittle wing]

Yeah, it finished about 10 minutes ago

[sand iris]

There's no possible way you can do 177epochs on a 5minute dataset in 20mins

[brittle wing]

it sounds decent

[sand iris]

Did you use colab or something

[brittle wing]

skill issue then ig

[brittle wing]

yeah

[sand iris]

Oh fair enough

Local hardware has nothing to do with colab lol

[brittle wing]

ah, alright then

[sonic tapir]

i duplicated the space. how do I do the AI cover. like upload a model and song?

[low shard]

https://discord.com/channels/1159260121998827560/1167844377046040769

[sand path]

anybody know how much batches i should use with a 12gb gpu (rtx 3060)

is this right? found it on another dc 4 GB RAM = 3 Batch Size
6 GB RAM = 4 - 5 Batch Size
8 GB RAM = 6 - 7 Batch Size
10 GB RAM = 7 - 9 Batch Size
12 GB RAM = 9 - 11 Batch Size

[modest spruce]

local's not working
can someone help me?

[royal jacinth]

does anyone know how to fix the error on applio amd when running 5 2023-10-30 13:01:07 | INFO | assets.configs.config | No supported Nvidia GPU found overwrite preprocess and configs.json 2023-10-30 13:01:07 | INFO | assets.configs.config | Use DirectML instead Traceback (most recent call last): File "C:\Users\Picky\Desktop\rvc\Applio-RVC-Fork\infer-web.py", line 55, in <module> from lib.infer.modules.vc.modules import VC File "C:\Users\Picky\Desktop\rvc\Applio-RVC-Fork\lib\infer\modules\vc\modules.py", line 22, in <module> import tabs.merge as merge File "C:\Users\Picky\Desktop\rvc\Applio-RVC-Fork\tabs\merge.py", line 15, in <module> import tabs.resources as resources File "C:\Users\Picky\Desktop\rvc\Applio-RVC-Fork\tabs\resources.py", line 50, in <module> config = Config() File "C:\Users\Picky\Desktop\rvc\Applio-RVC-Fork\assets\configs\config.py", line 82, in wrapper wrapper.instance = func(*args, **kwargs) File "C:\Users\Picky\Desktop\rvc\Applio-RVC-Fork\assets\configs\config.py", line 110, in __init__ self.x_pad, self.x_query, self.x_center, self.x_max = self.device_config() File "C:\Users\Picky\Desktop\rvc\Applio-RVC-Fork\assets\configs\config.py", line 300, in device_config import torch_directml File "C:\Users\Picky\AppData\Local\Programs\Python\Python39\lib\site-packages\torch_directml\__init__.py", line 16, in <module> import torch_directml_native ImportError: DLL load failed while importing torch_directml_native: The specified procedure could not be found. Press any key to continue . . .

[sand iris]

are there spaces in the file names

[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

🇫🇷 France

https://docs.aihub.wtf/v/france

🇩🇪 Germany

https://docs.aihub.wtf/v/germany

🇮🇹 Italy

https://docs.aihub.wtf/v/italy

🇯🇵 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

🎃 Spooky contest is active till 31st, check the announcement channel for further details.

[fickle gate]

I think I made a mistake. I trained my ai up to 400 epoch while only using around 4 minutes of clips. Is it wrong ? should I start all over ?

[proper shale]

Not really?

If you can find more voice clips then I definitely recommend doing it again for more quality

[real quartz]

guys i made my model but can't we do covers now ? i don't see anything anywhere ?

[vapid gust]

i got a question with dataset

[proper shale]

#1159949278270193734 has colabs n stuff that u can use to make covers

[vapid gust]

hey

[proper shale]

?

[vapid gust]

i got a question with making a dataset

these are officla studio stems

theres a 2 hook parts 1 verse 1 bcak vocals

could i use the hook

twice

for the dataset

or just once

[proper shale]

if its the same thing then yeah i dont recommend letting it there again

[vapid gust]

what about the vocal stabs

can i add those ?

[proper shale]

if they're not comp'd together, yeah sure

[vapid gust]

comp"d meaning?

[proper shale]

a bunch of vocals crammed together in one track

[silent gale]

It keeps shouting trial at me what do i do

[proper shale]

what

oh r you using the muzychenko virtual audio cable

[silent gale]

yeah i just figured it out

is there any free ones

[proper shale]

vb audio cable

[silent gale]

thanks babyboy mwah

[proper shale]

np ;)

[fickle gate]

Well I found a 5 minute voice clip and i'm trying to find a few more to add to those 4 minutes.

[rare goblet]

Ayo? @fickle gate level 1 !!!

[brittle wing]

Can anyone please send me a full tutorial video of local rvc?

[analog fulcrum]

where can i find the most recent working version of local rvc

[river wraith]

what is the ai program called?

[edgy dew]

I actually wonder. In RVC v2 Disconnect, after loading the dataset, do I have to click training to resume?

[viscid moss]

https://youtu.be/ixB9oalT3cQ?si=LnlxDYvisyB_VJc5

[viscid moss]

RVC V2

[edgy dew]

Hello?

[brittle wing]

Thanks!

[viscid moss]

It's not necessary that's for resuming

[jagged flax]

I wanted an "ai Cover" tutorial for Android

[viscid moss]

Use Hugging Face or Colab for mobile

[edgy dew]

Well I'm trying to resume what couldn't be finished.

[jagged flax]

Oh thanks, do you have a tutorial? Sorry to ask if it's not okay I'll come back

[rare goblet]

Ayo? @jagged flax level 2 !!!

[edgy dew]

So if I click training, it'll resume the epochs process?

[viscid moss]

hmmmm then try this steps

and then train

[viscid moss]

https://discord.com/channels/1159260121998827560/1167844377046040769

[vapid gust]

ive got a question

whats the max amount of dataset should i use

[edgy dew]

Did those 2 already. So I click training and it'll resume the epoch process, right?

[vapid gust]

ive got studio recordings

about 50mins worth

how much should i use

to get a good model

[viscid moss]

I'm not too sure but it should work

[viscid moss]

20 - 30 It's enough.

[vapid gust]

ill just use 30 mins then

[viscid moss]

I have not tried with more datasets, you must have a good GPU and patience for 50 minutes, that will take a long time

[vapid gust]

colab

[viscid moss]

[viscid moss]

💀

[vapid gust]

thats fine

ill just keep on overnight

[viscid moss]

maybe 2 sessions

[vapid gust]

na

[edgy dew]

Hmmm... it's not working...

[vapid gust]

ive finished 500 epochs in 2 hours

[viscid moss]

how long dataset?

[vapid gust]

like 5 mins 💀

[viscid moss]

maybe you should have run this last time

i'm not totally sure

[viscid moss]

haha

With colab 25 min take 5:30 or 6 hours

It's too slow

[vapid gust]

i think last time i did 30 min dataset

i let it go to 300 epochs

maybe took 2 hurs

[viscid moss]

I see

[vapid gust]

its straight studio recordings

[edgy dew]

My training didn't finish though. GPU ran out of time.

[viscid moss]

for example, watch this

[vapid gust]

so still if i let it trrain less

it will still be accuratew

i just want each pronounciation

to be correct

[viscid moss]

An RTX 4090 takes like 3 hours for 700 epochs. 15s per epoch and 30 minutes of dataset

[vapid gust]

oh wow

i dont even have a gpu so

nor a pc

[viscid moss]

I see, how long dataset?

[viscid moss]

If the dataset is very short, the pronunciation sometimes fails

[vapid gust]

true

ive gotten away with 2 minute

studio recording

high quality model

[edgy dew]

1019 files. And I used 250 epochs. mangio-crepe.

[viscid moss]

why mangio? use rmvpe

it's better

and faster

[edgy dew]

Because I recall the crepe versions having good quality. Even voices with special effects.

[rare goblet]

Ayo? @edgy dew level 4 !!!

[viscid moss]

hmmmmmm

but anyways...

If you did not execute "Export model" you will not be able to continue the training

[vapid gust]

rmvpe better

[edgy dew]

Ohhhh, okay. And how do I find the Step Count and Epock count?

[viscid moss]

For train u can set, 25, 400, 20

[viscid moss]

in the pth files. Like.... e50s700 "e" means epoch and "s" means steps

[sage ember]

Hi! Can somente help me pls? Its the first time im doing a rvc v2

Just the index training isnt work

How can i fix it, pls aa

[rare goblet]

Ayo? @sage ember level 1 !!!

[edgy dew]

Ah, okay. but because it's only for manual parts, I don't have to worry about the numbers, do I? I can just execute the notebook leaving and step number and epoch number at 000?

[viscid moss]

show me ur CMD

an error appears?

[viscid moss]

leave it to 000

[edgy dew]

Alright

Hmmm... that didn't work.

And all I have is a rvclogs zip

[sage ember]

thats the error

i saw people saying to try the index first, is it really works?

[plush compass]

When using uvr, can I use something like voc ft on multiple files at once?

[crimson basalt]

Yes.

[edgy dew]

Looks like I'm just gonna have to start all over again with training the voice with a different google account.

[daring verge]

your supposed to run the feature extraction cell before running that

[brittle wing]

??????? Easy 10/25 why this only and train dont work

[sage ember]

Okay! I'll try tomorrow! Thks

[daring verge]

wdym train doesn't work

[brittle wing]

just said that

[vapid gust]

how do I stop training to continue it some time else

[daring verge]

are you talking about tensorboard error or not training problem?

[brittle wing]

not training

[daring verge]

lemme see the full cmd

[brittle wing]

@daring verge

[daring verge]

hmm that never happened before, try running it again

[brittle wing]

this thered try

[grand willow]

how to easy train online now and see tensoboard at the same time ????

[brittle wing]

RVC disconnect

[grand willow]

i already use it but no tensoboard

[brittle wing]

RVC disconnect
why dont save g and d file 32222

[grand willow]

u should save them manually if u plain to retrain

[brittle wing]

how

[wide parcel]

2023-10-30 21:19:54 | WARNING | libav.mp3 | Estimating duration from bitrate, this may be inaccurate
RVC libraries are still loading. Please try again in a few seconds.
cannot unpack non-iterable NoneType object

trying to use tts on applio

[grand willow]

just download the files

[brittle wing]

shit its large

[grand willow]

yes

[brittle wing]

https://cdn.discordapp.com/emojis/1133009099857080412.webp?size=160&name=devastated&quality=lossless

[grand willow]

https://colab.research.google.com/drive/1r4IRL0UA7JEoZ0ZK8PKfMyTIBHKpyhcw
the easy colab was great but now i think it not work

[sage ember]

its sad theres no videos about it, id like to get into it more. Kinda Hard when you are a "new programmer"

[brazen wraith]

"C:\Users\Kamra\Downloads\Mangio-RVC-v23.7.0_INFER\Mangio-RVC-v23.7.0>C:\Users\Kamra\Downloads\Mangio-RVC-v23.7.0_INFER\Mangio-RVC-v23.7.0\runtime\python.exe EasierGUI.py --pycmd C:\Users\Kamra\Downloads\Mangio-RVC-v23.7.0_INFER\Mangio-RVC-v23.7.0\runtime\python.exe --port 7897
Traceback (most recent call last):
File "C:\Users\Kamra\Downloads\Mangio-RVC-v23.7.0_INFER\Mangio-RVC-v23.7.0\EasierGUI.py", line 28, in <module>
from my_utils import load_audio, CSVutil
ImportError: cannot import name 'CSVutil' from 'my_utils' (C:\Users\Kamra\Downloads\Mangio-RVC-v23.7.0_INFER\Mangio-RVC-v23.7.0\my_utils.py)

C:\Users\Kamra\Downloads\Mangio-RVC-v23.7.0_INFER\Mangio-RVC-v23.7.0>pause
Press any key to continue . . ."

What do I do?

[brittle wing]

what happin

[grand willow]

it stops importing the wave files to train

[brittle wing]

strange

[grand willow]

so they told me to re use the disconnected version

[brittle wing]

disconnected its great but not savig its big issue

[grand willow]

yes

انت عربى

[brittle wing]

wait that mean i can use mp3

[brittle wing]

نعم

[grand willow]

هبعتلك صداقة

[toxic trail]

I have the Applio RVC Fork downloaded on my PC, is there any way to manually add a model instead of downloading it? The reason being I have a model that I want to add to my Applio, but it's a drive link and my build doesn't take drive links, and I don't want to reupload the model onto somewhere like Huggingface for use because I don't want to be accused of stealing other peoples' models.

[wide parcel]

is there additional setup for applio or is it just meant to work out of the box?

[wanton falcon]

guys

is there a way to stop the covergen from saying runtime disconnected

[brazen wraith]

so wait does easyGUI striaght up NOT work outside colab?

I'm confused

[grand willow]

i think it stopped working

[grand willow]

it works online to use google video cards

[brazen wraith]

im trying to run easyGUI locally so I don't have to shell money to google

[rare goblet]

Ayo? @brazen wraith level 1 !!!

[grand willow]

without card???

to train???

[brazen wraith]

I have a 1070

no not for training

[grand willow]

just download RVC 2 for training

sorry for changing only

it works well

Mangio-RVC-v23.7.0

[rare goblet]

Ayo? @grand willow level 4 !!!

[edgy dew]

Is it possible to locally train models while using data from an external hard drive instead of a C drive?

[grand willow]

i think the data kept in the local hard with Mangio-RVC-v23.7.0
and not the C drive at all

[edgy dew]

Oh.

[grand willow]

waaaaaaaaaw gooood news
the easycolab reworked again loooooool
https://colab.research.google.com/drive/1r4IRL0UA7JEoZ0ZK8PKfMyTIBHKpyhcw#scrollTo=FFfC9x239kC1

[radiant sparrow]

Can anyone tell me how to create a new voice model using acapella from the performer I would like to create the model from?

[radiant sparrow]

what is this for?

[grand willow]

easiest online training and can use tensoboard

[radiant sparrow]

can you guide me through this? can it be used to create a voice model?

[rare goblet]

Ayo? @radiant sparrow level 1 !!!

[radiant sparrow]

everyone has to start from level 1 no?

[brittle wing]

i can't post models or type in the voice models channel

[rigid scarab]

..

[runic inlet]

Current recommended method for TTS with an existing voice model?

[high night]

use ilaria rvc hf space

[runic inlet]

Thanks! Where can I find that one? Not seeing it in the info channel

Wait this one? https://huggingface.co/spaces/r3gm/RVC_HF

[high night]

no not that

[rare goblet]

Ayo? @high night level 28 !!!

[high night]

https://huggingface.co/spaces/TheStinger/Ilaria_RVC

this one

[left perch]

is rvc meant to be a batch file?

[austere pivot]

I was installing Applio and something happened, I can't send images to explain it, where can I send it?

[rare goblet]

Ayo? @austere pivot level 1 !!!

[austere pivot]

"here is the code, or what the box says:"

Cloning the GitHub repository into the temporary directory...
Cloning into 'temp_torchcrepe'...
remote: Enumerating objects: 442, done.
remote: Counting objects: 100% (437/437), done.
remote: Compressing objects: 100% (194/194), done.
remote: Total 442 (delta 253), reused 395 (delta 228), pack-reused 5
MiB/s
Receiving objects: 100% (442/442), 72.20 MiB | 6.94 MiB/s, done.
Resolving deltas: 100% (253/253), done.
Copying the torchcrepe folder...
Removing the temporary directory...
Windows
Downloading https://huggingface.co/lj1995/VoiceConversionWebUI/resolve/main/pretrained/D32k.pth to assets/pretrained/D32k.pth...
Traceback (most recent call last):
File "C:\Users\PC\Desktop\Applio-RVC-Fork\Applio-RVC-Fork\infer-web.py", line 14, in <module>
import lib.tools.model_fetcher as model_fetcher
File "C:\Users\PC\Desktop\Applio-RVC-Fork\Applio-RVC-Fork\lib\tools\model_fetcher.py", line 85, in <module>
download_file_with_progress(url, destination_path) # Use the function tdqm
File "C:\Users\PC\Desktop\Applio-RVC-Fork\Applio-RVC-Fork\lib\tools\model_fetcher.py", line 47, in download_file_with_progress
with open(destination_path, 'wb') as file, tqdm(
FileNotFoundError: [Errno 2] No such file or directory: 'assets/pretrained/D32k.pth'

C:\Users\PC\Desktop\Applio-RVC-Fork>

[wispy burrow]

You can now.

[austere pivot]

Thank you. Right after I installed Applio using the Applio installer, this happened

Can someone please help me fix it? 🙏

[wispy burrow]

EasyGUI does not seem to work for Mangio.
Installing it breaks Mangio.
Infinite queuing is usually a sign that someone tried to install EasyGUI.
Seen it quite often in the old AI Hub.

[green birch]

ya i can confirm that it was my major issue

[wispy burrow]

Can't really help, but I'm just curious.
What Installer version did you use?
V2? V4? Something else?

[austere pivot]

This is the one I used

[wispy burrow]

That's the latest one I guess.
Really no idea what went wrong.

[austere pivot]

I'll try to find that file manually and put it there, to see of that helps

[fierce delta]

Oh yo

I had the same issue

I deleted the entire folder to reinstall it

I used the install_Applio.bat method instead of the .exe one and now it works

[austere pivot]

Where did you downloaded it?

[fierce delta]

https://github.com/IAHispano/Applio-RVC-Fork/releases/tag/v2.0.0

[austere pivot]

I'll install it with that one, Thank you!

[fierce delta]

Hope it works for you 🙏

[austere pivot]

Oof, it says:

Starting Applio with runtime for Nvidia support (you must have it installed)...
The system can not find the path specified.
Press a key to continue . . .

I clicked "go-applio" after the install and that happened

[fierce delta]

Did you select your GPU in this step?

oh shoot i cant sent images

https://docs.aihub.wtf/vocal-cloning-guide/applio

[austere pivot]

yes

[rare goblet]

Ayo? @austere pivot level 2 !!!

[austere pivot]

I selected "NVIDIA GPU"

[fierce delta]

Oh im not sure then

[rare goblet]

Ayo? @fierce delta level 1 !!!

[fierce delta]

You could try run the "go-applio-manager.bat"

to try install the nvidia runtime again

[austere pivot]

I'll try that

That one?

[fierce delta]

yep

[austere pivot]

k let me download all the stuff

Done, now should I go and run "go-applio"?

[fierce delta]

yup

[austere pivot]

I think it's working, it's downloading some stuff, TYSM!

[fierce delta]

[austere pivot]

Yes, it worked! Thank you guys, a lot! ❤️

How many "Batch size" should I use if I have a 4060 ti (16GB)?

[high night]

60

16*

[austere pivot]

perfect, thanks!

[brittle wing]

RVC issue: Suddenly-long Epoch Training generation (not from the get-go) (0 NVIDIA GeForce GTX 1050 Ti)

I've been having an issue in RVC where I've suddenly been having long epoch training generations when I'm making new voice models... And it didn't even happen from the get-go. Beforehand I was able to train models just fine!

Here's a more detailed recap of the whole situation with pictures and a few videos - https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/1278

Someone already suggested replacing the GPU or renting a GPU service.
But since the former would require asking my father (which I’ve chosen as a last minute decision) and the latter is a new to me due to being sheltered...

But honestly I'd rather have any other solution :/

———-

I will be willing to offer more information or screenshots.

[plush compass]

What is cash_all_training_sets?

[brittle wing]

Hey umm, is this a reply to my issue or are you asking for help on your own issue? ^^;

[rare goblet]

Ayo? @brittle wing level 1 !!!

[plush compass]

I'm looking for someone who can answer my questions.

[brittle wing]

Ah okay I see… thanks for explaining ^^;

[glad zealot]

Not sure how to fix it but I guess temporarily until someone knows how to, you can try training using Google colab with rvc v2 disconnected

-colab

[azure marsh]

Google Colabs

RVC

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

🌐 Translated colabs

• PT-BR: -colabs_br

Audio Separation/Isolation

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

🎃 Spooky contest is active till 31st, check the announcement channel for further details.

[brittle wing]

Google Collab? You mean the one that relies on it to be connected with Google Drive?

[glad zealot]

Yup

I mean technically you don't need it to connect to google drive

But colab is made by google so of it's privacy you're worrying about, that doesn't matter since it's made by google

[brittle wing]

Huh, I never actually knew that! Either way it could be convenient since my Google Drive is kind of full ^^;

So…know any video tutorials I can follow?

[glad zealot]

Don't know any video but I think this would work
#1160926856548200531

[brittle wing]

Hmm I see! I’ll go take a look! And I’ll let you know how it goes!

[glad zealot]

Welp imma sleep

[brittle wing]

@rigid scarab ?

[edgy dew]

I just started over with a different account and made the voice with 220 epochs. I just gave up on the resume training thing, for I didn't have the pth files on the other account.

[rare goblet]

Ayo? @edgy dew level 5 !!!

[grizzled moon]

Applio local installation guide?

[austere pivot]

https://github.com/IAHispano/Applio-RVC-Fork/releases/tag/v2.0.0 Try to install it with this, just like I was told

If you get some errors try to look up the messages and help I received

[grizzled moon]

Where can i get an visual guide

[austere pivot]

Hmm, I don't know if there's a tutorial/visual guide. Even though, is not that hard set it up, maybe "🍏 Applio-RVC-Fork: Guide" in #1159513888199540817 can help.

[rare goblet]

Ayo? @austere pivot level 3 !!!

[pale finch]

Hi, I have deployed Applio-rvc-fork on paperspace, and I am getting this error when I try to run TTS. Does anyone have any idea ?
Starting inference for 'tts_out_7.wav' Attempting to load assets/audios/audio-outputs/tts_out_7.wav.... 2023-10-31 09:10:01 | WARNING | libav.mp3 | Estimating duration from bitrate, this may be inaccurate RVC libraries are still loading. Please try again in a few seconds. cannot unpack non-iterable NoneType object

[brittle wing]

Hey if you’re awake.. I have to tell you that..

I read the instructions on one of the links you gave me.. and even though you told me I there is an option where I don’t need to connect Collab it to Google Drive..

The instructions in the doc only showed the process that requires connecting to Google Drive..

[grizzled moon]

I keep getting trojan warnings from applio

[rare goblet]

Ayo? @grizzled moon level 1 !!!

[grizzled moon]

Even if its a false positive i wont even bother downloading it anymore

[austere pivot]

It's normal, I also got those alerts/warnings

I guess you can try using it with Collab

[grizzled moon]

Ive always sticked with colab

Since the last time i tried to install local

[austere pivot]

yep, you can try that

[grizzled moon]

Im literally running an “disinfection” right now

[austere pivot]

This is the second time installing Applio locally, and I've never had any problem

[grizzled moon]

Its so slow

[austere pivot]

I wonder why...

[plush compass]

Is there a colab that automatically saves? The Colab I'm currently using doesn't allow automatic saving, and I don't know how to save it...

[grizzled moon]

I think i fucked up

Im literally using a colab that kills its runtime with no warning

and i gotta be quick with setting it up and downloading the results before it does

[austere pivot]

Yeah, that happen sometimes

I have no idea how to help you with that, I'm sorry

[grizzled moon]

Eh its ok

[rare goblet]

Ayo? @grizzled moon level 2 !!!

[grizzled moon]

I dont do ai much nowadays since the colab ban

[austere pivot]

You can try installing some other like Mangio

Hopefully that one doesn't give you that false positive alert

[grizzled moon]

Ay what the hell

I had to terminate my microsoft edge during to a breakpoint

WTF

[daring verge]

bruh what colab is that

[high night]

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

Use this @grizzled moon

[daring verge]

for rvcDisconnected you just have to run the cell save preprocess files

[high night]

[daring verge]

[high night]

-realtime

[azure marsh]

🎙️ Realtime Voice Changer

• (Local) W-Okada Voice Changer

This one has more options and is somewhat more convenient to use than Real-time RVC.
Credits: ravencutie21

Link: Guide for W-Okada's RealTimeVoiceChangerClient

• (Online) W-Okada - Colab Version

A free, online version of W-Okada that allows you to run code on Google’s powerful GPUs.
Credits: Antasma245

Link: How to use W-Okada's Voice Changer

• (Local) Real-time conversion of RVC

More accurate than W-Okada's, but requires a Nvidia RTX card.
Credits: teas.hi

Link: How to use realtime conversion with RVC

• Russian guide for voice changing

Credits: redoverflow, blackcat76it and tntmaxim

Link: ultimate-voice-models-FAQ/Voice-Changer

• Text to speech:

RVC-TTS-Guide

Note: If you need further help setting up/using the voice changers, you can ask for help in #🔍│help-w-okada or check the channel's pinned comments.

🎃 Spooky contest is active till 31st, check the announcement channel for further details.

[grizzled moon]

What about inference

[high night]

u can do it in ur lcoal pc

or use 9/17

or just use ilaria rvc

[plush compass]

Why can't I see that cell? Can you give me a link to what Colab it is? Or can you just explain it with a picture or something? Please...ㅠㅠ

[daring verge]

first, what colab are you using

[grizzled moon]

link?

[plush compass]

https://colab.research.google.com/drive/1XIPCP9ken63S7M6b5ui1b36Cs17sP-NS
Is this colab?

[daring verge]

https://discord.com/channels/1159260121998827560/1167844377046040769

[grizzled moon]

Does it run well

[daring verge]

its below the training section

[daring verge]

its pretty good, simple to use

recommend using huggingface space

[pale finch]

Hi, I have deployed Applio-rvc-fork on paperspace, and I am getting this error when I try to run TTS. Does anyone have any idea ?
Starting inference for 'tts_out_7.wav' Attempting to load assets/audios/audio-outputs/tts_out_7.wav.... 2023-10-31 09:10:01 | WARNING | libav.mp3 | Estimating duration from bitrate, this may be inaccurate RVC libraries are still loading. Please try again in a few seconds. cannot unpack non-iterable NoneType object

Can someone help me figure out this ? I am stuck here for 2 days. Used this tutorial https://docs.google.com/document/d/1lIAK4Y0ylash_1M2UTTL_tfA3_mEzP0D2kjX2A3rfSY/

[grizzled moon]

Ah ok bet Thanks

[grizzled moon]

Cover gen something

[plush compass]

Is auto-save impossible? And when activating the cell used for saving, do I have to stop model learning?

[daring verge]

you have to run it in order

[plush compass]

First of all, thank you, I will try it.

[silk root]

villager models dont work for me idk why

[grim hull]

does it support amd gpus?

[sand iris]

Which software and to do what ?

[grim hull]

the rvc

[sand iris]

Realtime ?

[grim hull]

sorry im rly unexperienced

[sand iris]

Realtime works
Audio conversion works
Making models doesn't (not on windows)

[grim hull]

when i try to convert a voice it only gives the cpu option the gpu option is greyed out

[rare goblet]

Ayo? @grim hull level 1 !!!

[sand iris]

Mind screenshoting the app ?

There's multiple ones and I can't really know which one

[sand iris]

Okay I've never seen this one

😆

[grim hull]

Damn

[sand iris]

That one supports nvidia only apparently

[jolly wraith]

indian

[grim hull]

Sad

Thanks tho

[daring verge]

#📑│making-models message
use this as a reference

[prime sparrow]

i downloaded applio manually

when i press 4

its says :strating applio with nvdia support

what should i do

[brittle wing]

@glad zealot Can we get back to this? - #✨│ai-help message

[glad zealot]

you only need to do that for resuming iirc

oh ye you need to connect it first for the dataset of the model you are making

[prime sparrow]

can somebody help me

🤡

[brittle wing]

Even though my Google Drive’s storage is full? Wouldn’t GD need space favor this??

[rare goblet]

Ayo? @brittle wing level 2 !!!

[glad zealot]

nah as long as your database is in your drive it wont add anything

you need to add your dataset ZIP File to a folder on google drive called rvcDisconnected

[high night]

-howtoask

[azure marsh]

How to troubleshoot

• Give context: Throwing your error/issue alone will not help. Tell what step are you on, what were you doing before issue, your settings, your terminal, et cetera.
• Be relevant: You come to ask for help, not to complain. You can't ask for RVC help in voice changer help and vice versa.
• Be polite: DONT DEMAND FOR HELP OR I WILL FUCKING MURDER YOU. Ask politely, the server was not made for you.
• Don't be an attention whore: If people are ignoring your help, kindly ping them. If they are ignoring you, either they don't know how to solve it, they are busy, or you are an asshole.

WITH ALL DUE RESPECT, DO NOT BE AN ASSHOLE AND HAVE COMMON SENSE.

[brittle wing]

I see! Alright I’ll try it out…(tried making a folder to test it out) Woah! Huh,…I can still make folders despite full space! O O Let’s hope the ol’gal it can hold a few more inside ^^

I’ll let you know how it goes!

[plucky storm]

where can I download TensorBoard (RVC)?

[forest mortar]

when i installed the program, i remember hearing from someone that you had to install python something? i forget the name of it (sorry, not a great memory lol), i was curious what the name of it was again, and if its needed, or was just an optional thing to make things work better/smoother? (sorry if this is an obvious question or you get this question a lot)

[molten pecan]

Yeah depending on the type of software you use you'll need to install Python

You can just Google it

It seems like you are using Applio locally. If you want to make things less tedious you can use the huggingface space that one of the mods did

[plush compass]

https://colab.research.google.com/drive/1XIPCP9ken63S7M6b5ui1b36Cs17sP-NS
If I use this Colab, will the Epoch be automatically saved when the Colab connection is lost?

[brittle wing]

what the heck this theme gonna kill my eyes, when i go to app, it become light mode but it turn dark mode when i go to others

any solutions?

[hard gust]

Is there a way to downgrade Applio’s version???

[molten pecan]

I use a dark mode extension

[molten pecan]

You could use other versions of RVC if Applio is not working for you

Like the huggingface spaces

[plush compass]

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

[rare goblet]

Ayo? @plush compass level 3 !!!

[proper shale]

It doesn't save the training files (D/G). This is an option if you don't wanna retrain the model and/or wanna save space on your Google Drive.

[brazen schooner]

Hey does anyone know why i get the preprocess.log error when i try to process my dataset on mangio rvc interface? im trying to train a model.

im so stuck on what to do. i cant get past the part where you put in the path of the dataset and process it

[brittle wing]

Hello. Whenever I use the program's default models it works normally, but when I try to download an external one the voice becomes extremely robotic and cutting. What can I do?

[rare goblet]

Ayo? @brittle wing level 1 !!!